Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by...
-
Upload
juan-correa -
Category
Documents
-
view
213 -
download
1
Transcript of Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by...
![Page 1: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/1.jpg)
Optics Communications 242 (2004) 267–278
www.elsevier.com/locate/optcom
Correlations between intensity fluctuations within stimulatedBrillouin waveforms generated by scattering of
Q-switched pulses in optical fiber
Juan Correa, Enrique Manzano, Ryan Tracy, John R. Thompson *
Department of Physics, DePaul University, 2219 North Kenmore Avenue, Chicago, IL 60614, USA
Received 2 June 2004; received in revised form 9 July 2004; accepted 4 August 2004
Abstract
We present a detailed experimental study of the statistical properties of pulsed stimulated Brillouin scattering in opti-
cal fiber using single-mode pump pulses with durations comparable to the phonon lifetime but much shorter than the
fiber transit time. Under these conditions the temporal dynamics of the scattered light is stochastic and results in the
generation of scattered intensity waveforms with complex structures. We provide experimental evidence and simple
theoretical arguments to interpret the energy fluctuations when varying fractions of the scattered waveforms are inte-
grated and statistically analyzed. Two-point intensity correlation functions are used to determine to what extent the
complex intensity waveforms represent sequences of statistically independent scattering events.
� 2004 Elsevier B.V. All rights reserved.
Keywords: Brillouin scattering; Nonlinear optics; Statistical optics
1. Introduction and background
The Central Limit Theorem implies that a ran-
domly fluctuating physical variable that represents
the sum of many statistically independent contri-
butions will have a probability distribution that
0030-4018/$ - see front matter � 2004 Elsevier B.V. All rights reserv
doi:10.1016/j.optcom.2004.08.008
* Corresponding author. Tel.: +1 773 3251375; fax: +1 773
3257334.
E-mail address: [email protected] (J.R. Thompson).
is approximately Gaussian in shape. The widthof this probability distribution will become pro-
gressively narrower in relation to its mean as the
number of independent contributions increase [1].
In this article, we discuss experiments that investi-
gate the pulse energy fluctuations associated with
spontaneously initiated stimulated Brillouin scat-
tering (SBS) in optical fiber when varying fractions
of the complex SBS waveforms are integrated andstatistically analyzed. We use measured intensity
correlations and a simple statistical model to probe
ed.
![Page 2: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/2.jpg)
268 J. Correa et al. / Optics Communications 242 (2004) 267–278
the independence of intensity fluctuations in differ-
ent sections of the SBS waveforms.
Brillouin scattering of optical pulses was first re-
ported for bulk crystals of quartz and sapphire just
a few years after the birth of the laser and, almosta decade later, in low loss glass optical fiber [2,3].
Since that time SBS has enjoyed a long history
of research by many different groups for both ap-
plied and fundamental reasons. SBS has been used
or proposed for pulse compression and beam
clean-up of high power pulsed lasers [4–8], to pro-
duce narrow-band optical filters for communica-
tion systems [9], for signal conditioning ofmicrowave frequency communication signals
[10,11], and for fiber sensors [12]. SBS has also
been extensively studied because of its complex
stochastic and deterministic dynamics [13–20].
This article focuses on fundamental aspects of
the statistical properties of spontaneously initiated
SBS in low loss optical fiber.
Spontaneous Brillouin scattering occurs whenan intense pump wave scatters off of high fre-
quency acoustic waves that are excited by thermal
or quantum fluctuations of the medium. The scat-
tered light travels in the backward direction with
respect to the pump due to phase matching con-
straints, and the scattered light is down-shifted in
frequency from the pump by an amount equal to
the acoustic frequency in order to conserve energy.For glass optical fibers the acoustic wave fre-
quency is typically in the 10–20 GHz range for
near infrared pump light so that at room tempera-
ture the average thermal energy available for a
vibrational mode is much greater than the phonon
energy. Therefore, the random acoustic distur-
bances that initiate Brillouin scattering in glass fi-
ber at room temperature are thermally excited[13,14,21]. The thermally generated acoustic waves
can be represented as a zero-mean Gaussian ran-
dom process with fluctuations that are uncorre-
lated in space and time. For this reason the
waveforms produced by spontaneously initiated
SBS can have exceedingly complex temporal struc-
ture and display large scale intensity fluctuations.
Theoretical and experimental work on SBS usingcontinuous-wave (cw) laser light in the absence
of feedback, has shown that these large scale inten-
sity fluctuations can persist even under conditions
of significant pump depletion [13–15,18]. When
feedback comes into play cw SBS shows temporal
patterns that can be explained in terms of the lon-
gitudinal modes of the fiber resonator, but there
can also be more complex dynamical behavior thatappears to be deterministic chaos. In fact there has
been substantial discussion in the research litera-
ture about the conditions under which the tempo-
ral dynamics are stochastic or deterministic and
about the conditions required for deterministic
chaos to be observed. The very careful experimen-
tal and theoretical work of Dammig et al. seems to
have largely settled these questions. Their workshows that SBS is very sensitive to feedback so
that relatively small amounts of it are sufficient
to change the dynamics from purely stochastic to
largely deterministic, and behavior that appears
to be deterministic chaos can result from thermal
drift of the fiber resonator [13–20]. Feedback does
not have to be ‘‘designed’’ into the system to be
important; it can come from Fresnel reflectionsoff the fiber end faces or, as a couple of recent
studies have shown, by repeated Rayleigh scatter-
ing in high loss fibers used as cw laser resonators
[22,23].
There have been a number of studies on dynam-
ical aspects of pulsed SBS in bulk media and in
waveguides, with limited attention given to charac-
terizing the statistical properties of the scatteredlight [24–35]. Some of the early studies of pulsed
SBS in optical fiber focused on understanding
the temporal structure of noise initiated scattering.
Much of this work explored the relationship be-
tween the time scale of the intensity modulations
of the scattered light and the phonon lifetime for
various scattering strengths [24–29]. Recent theo-
retical and experimental work by Ogusu and Lihave looked at the temporal dynamics of pulsed
SBS and the interplay of SBS with Kerr nonlinea-
rities for propagation through glass optical fiber,
propagation through fiber Bragg gratings, and
propagation through fiber ring resonators. Much
of their work looks at the temporal dynamics un-
der circumstances where the temporal width of
the optical pulse is longer than the transit timethrough the fiber [30–32]. There is a limited
amount of early experimental and theoretical work
addressing the statistical properties of transient
![Page 3: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/3.jpg)
Q-SwitchedPump Pulses
Pump PulseDiagnostics
ND Filter
1/2 Wave Plate
Linear Polarizer
20X Objective
40X Objective
71 m SMPMOptical Fiber
Etalon
Lens
Output BeamDiagnostics
InGaAs PD
Isolater
Beamsplitter
Beamsplitter
Mirror
(a)
Q-SwitchedPump Pulses
InGaAs PD
ND Filter
Aperture 20X Objective
Lens
(b)
J. Correa et al. / Optics Communications 242 (2004) 267–278 269
SBS in gas and liquid filled cells. This work inves-
tigated phase fluctuations in SBS by using two
identical Brillouin cells as phase-conjugate mirrors
in an interferometer which converted phase fluctu-
ations into energy fluctuations recorded by calo-rimeters. This work also investigated localized
temporal fluctuations in the transverse beam pro-
file of pulses generated by SBS in gas or liquid
filled cells [33–35].
In this article we report an experimental study
of the energy statistics of complex intensity wave-
forms produced by transient SBS in low-loss silica
glass optical fiber when the duration of the excitinglaser pulse is much less than the single-pass transit
time of the fiber so that the effects of feedback can
be ignored. In particular we focus on how the en-
ergy statistics evolve as we integrate a larger frac-
tion of the complex SBS waveforms. We will use
the measured ‘‘slice’’ energy statistics, simple theo-
retical arguments, and measured intensity correla-
tion functions to show that the observed statisticalbehavior provides a good example from nonlinear
optics of the Central Limit Theorem. To the best
of our knowledge, such a study has not been car-
ried out by other groups working on pulsed SBS
in optical fiber.
40X Objective
DispersingPrisms
71 m SMPMOptical Fiber
Fig. 1. Experimental setup: ND, neutral density; PD, photo-
diode; SMPM, single mode polarization maintaining. (a) Setup
for integrating and collecting SBS intensity waveforms. (b)
Setup for collecting transmitted pump waveforms.
2. Experiment apparatus and procedures
The experimental setup for this work is shown
in Fig. 1. For clarity the setup is represented by
two schematics: part (a) for the portion of theapparatus used to detect the backward propagat-
ing SBS waveforms; and part (b) for the portion
of the apparatus used to study the depletion of
the transmitted pump pulses by SBS. There were
three types of data collected: large sets of SBS slice
energies for statistical analysis, SBS waveform
ensembles to estimate intensity correlation func-
tions, and transmitted pump pulse temporal enve-lopes to study pump depletion by SBS. The setup
and procedures for each type of data will be dis-
cussed below after detailed information about
the pump laser is given.
The pump laser used to drive the SBS process is
a Q-switched, diode-pumped Nd:YAG laser. The
laser operates on a single transverse and longitudi-
nal mode at a center wavelength of 1064 nm. The
pulse repetition rate is 1 kHz, and the pulses have
smooth Gaussian shaped envelopes with a full-width at half-maximum of approximately 30 ns.
The pulses are also linearly polarized since one
face of the Nd:YAG crystal is Brewster cut. The
energy fluctuations of the Q-switched pulses are
generally quite small with a standard deviation
that is less than one percent of the mean pulse en-
ergy. Single longitudinal mode operation of the la-
ser is obtained by using a combination of twomode selection techniques. First, a 2 mm uncoated
![Page 4: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/4.jpg)
270 J. Correa et al. / Optics Communications 242 (2004) 267–278
optical flat is inserted in the cavity to reduce the
number of lasing modes in each pulse to a few
adjacent modes. The second technique, known as
prelasing, involves turning down the losses on
the acousto-optic Q-switch until there is a weakcontinuous wave background between the pulses.
For a modest range of losses this continuous wave
background is single mode and acts as a seed to
lock the forming pulses onto a single mode [36].
Between 85% and 90% of the total optical energy
was contained in the pulses for all of the reported
data. This energy fraction was determined by mak-
ing average pump power measurements with theAOM operated in two different modes: a mode
where the acoustic waves are periodically shut off
to allow pulses to form, and a mode where the
acoustic waves were continuously on so that only
the weak cw background was present. For the por-
tion of the pump beam coupled into the fiber, the
average cw background power was always less
than 0.1 mW while the peak power of the pumppulses were between 6 and 20 W. When the etalon
and Q-switch losses were properly adjusted, we
could obtain stable single-mode oscillation for up
to 2 h at a time before bimodal operation and
mode hopping occurred. Single longitudinal mode
operation of the laser during data collection was
verified by continuously monitoring the output
of a pulsed laser spectrum analyzer. One couldalso detect periods of multimode operation by
monitoring the input pump pulse temporal enve-
lope, which was smooth when the laser oscillated
on a single mode but deeply modulated in multi-
mode operation, with a beat frequency of approx-
imately 1.25 GHz. Finally, it should be noted that
the efficiency of SBS is very sensitive to the mode
structure of the pump pulses. The SBS process ismost efficient when high-quality single-mode
pump pulses are used, and the presence of rela-
tively small side modes resulted in a substantial
reduction in the generated SBS signal strength.
This fact provided yet another way to monitor
the pump pulse mode structure during data
collection.
As mentioned in the Section 1, the presence offeedback can have dramatic effects on the tempo-
ral dynamics of SBS. We believe that the effects
of optical feedback are not important in our exper-
iments since the physical length of the pump pulse
in the fiber (csp/n � 6 m for our 30 ns pulses) is
roughly 10 times smaller than the fiber length
(71 m). In addition, the pump and SBS pulses
are only coupled when the relationship betweentheir propagation directions are correct. For these
reasons we did not make any special efforts to
cleave the fiber ends at a large angle to discourage
reflections that were coupled into a guided mode.
We used the diffuse Fresnel reflection of the pump
pulse off the fiber input to determine when the SBS
scattering occurred in relation to the entry of the
peak of the input pump pulse into the fiber. Thisbrings up yet another reason why we can ignore
the effects of feedback: all of the measurements
on individual SBS pulses were made during a time
interval that was less than a single round trip time
of the pump pulse within the 71 m fiber.
The SBS slice energy statistics were measured
using the apparatus shown in Fig. 1(a). The optical
isolator prevented back-scattered light from enter-ing the laser cavity. The half-wave plate and pola-
rizer were used to vary the input power to the fiber.
The polarizer transmission axis was fixed at an an-
gle that was parallel to one of the principal axes
of the elliptical core polarization maintaining fiber.
The half-wave plate could then be used to rotate
the incident pump polarization and vary the
amount of light that was transmitted through thepolarizer and subsequently coupled into the fiber.
The 71 m optical fiber supported a single trans-
verse mode at the pump and SBS wavelengths
and was polarization maintaining. The fiber loss
coefficient was between 2 and 3 dB/km at the
pump wavelength, as specified by the manufac-
turer. The effective mode area was approximately
7.6 lm2, which is slightly larger than the area ofthe elliptical core. This estimate of the effective
area is based on earlier stimulated Raman scatter-
ing and parametric four-wave mixing experiments
and simulations. The SBS gain coefficient for our
fiber should be close to the published value for sil-
ica glass which is 5 · 10�11 m/W [21]. The back-
scattered SBS waveform exited the fiber input
and a portion of it was directed to the detectionsystem by the beamsplitter between the input cou-
pler and the linear polarizer. A neutral density fil-
ter prevented saturation or damage of the InGaAs
![Page 5: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/5.jpg)
-50 0 50 100 150 200 250 3000.0
0.2
0.4
0.6
0.8
1.0(a)
Pulse Time (ns)
Pulse Time (ns)
-50 0 50 100 150 200 250 3000.0
0.2
0.4
0.6
0.8
1.0(b)
PulseTime (ns)-50 0 50 100 150 200 250 300
Inte
nsity
(A
rb.U
nits
)In
tens
ity (
Arb
.Uni
ts)
Inte
nsity
(A
rb.U
nits
)
0.0
0.2
0.4
0.6
0.8
1.0 (c)
Fig. 2. Sample SBS waveforms for three values of input pump
peak power: (a) 6.26 W, (b) 12.5 W, and (c) 18.8 W. Each
waveform is the scattering that resulted from a single pump
pulse. The zero on the time scale marks the time when the peak
of the input pump pulse entered the fiber.
J. Correa et al. / Optics Communications 242 (2004) 267–278 271
photodiode used to detect the SBS waveform. An
etalon (4 mm thick, 90% reflectance both sides)
was used to pass the SBS pulse and block any
pump light that resulted from Fresnel reflections
off the fiber ends. A 100 mm focal length lenswas used to image the scattered light onto the
100 lm diameter active area of a high speed InG-
aAs photodiode (400 ps rise time). The output of
this photodiode was connected to either a high
speed digital oscilloscope (up to 10 GSa/s and 3
GHz analog bandwidth) or a fast gated integrator
with computer interface.
The SBS pulses were initiated by spontaneousscattering of the pump pulses off of thermally gen-
erated acoustic waves in the optical fiber. The
newly generated Brillouin pulse counter-propa-
gated through the pump pulse and was amplified
by stimulated scattering while the pump and Bril-
louin pulses overlapped in the fiber. The resulting
SBS intensity waveform that exits the fiber input
has a complex shape even though the pump pulseis quite smooth and regular. Sample SBS wave-
forms generated by scattering from individual
pump pulses are shown in Fig. 2 for three different
input pump powers. The detailed structure of these
waveforms changed randomly from shot to shot,
and there was substantial shot to shot timing jitter
in the leading edge of the SBS waveform. At the
lowest input pump power the shot-to-shot timingjitter in the leading edge of the SBS waveform
was roughly 40 ns for a waveform with a total
duration of hundreds of nanoseconds. At the high-
est input pump power the shot-to-shot timing jitter
in the leading edge of the SBS waveform was
approximately 4–5 ns for a waveform with a total
duration of tens of nanoseconds. It is clear that the
SBS waveforms do not reproduce the input pumppulse shape and are not simply a compressed ver-
sion of the input pump pulse. They appear to be
a series of intensity spikes that represent different
scattering episodes from the same input pump
pulse as it propagates through the fiber. Sections
of the first two sample SBS waveforms appear to
consist of trains of intensity spikes at roughly reg-
ular spacing. The average temporal spacing be-tween intensity spikes is most likely determined
by a combination of the phonon damping time
and the pump intensity profile [24–26,28]. As the
pump sweeps through a section of fiber where
there is a thermally excited acoustic wave of the
right frequency to initiate Brillouin scattering,
scattering occurs until the damping has erased
![Page 6: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/6.jpg)
272 J. Correa et al. / Optics Communications 242 (2004) 267–278
the organized motion of the acoustic wave or until
the pump intensity drops below the threshold to
produce detectable scattering.
As a first attempt at understanding whether or
not the sequence of intensity spikes in an SBSwaveform were statistically independent, we stud-
ied the pulse energy statistics as different fractions
of the SBS waveforms were integrated. The inte-
gration period ranged from 5 ns, the width of the
narrowest observed intensity spikes, to a duration
that was comparable to the total duration of the
SBS waveform. An electronic gate pulse generated
by the gated integrator defined the integrationinterval, and its leading edge was kept fixed at a
position that excluded the shot to shot timing jitter
of the leading edge of the SBS waveform. The
trailing edge of the gate pulse was progressively ex-
tended to later times until most of the SBS wave-
form was integrated or, for the weakest
scattering data, the slice energy noise showed little
change as the gate width was increased.We also collected ensembles of 1000 SBS wave-
forms at different input pump powers to use for
estimating the intensity correlation function. In
this work, the intensity correlation function is de-
fined as the normalized correlation coefficient be-
tween the intensity fluctuations at two different
times in the SBS waveform as a function of the
time delay between the two points:
CIðt0; sÞ ¼Iðt0Þ � Iðt0Þh ið Þ Iðt0 þ sÞ � Iðt0 þ sÞh ið Þh i
r0rs;
where I(t0) is the intensity at the ‘‘initial’’ time t0, sis the delay from the initial time, r0 is the standarddeviation of the SBS intensity at the initial time, rsis the standard deviation at the delayed time, and
the angle brackets denote averaging over theensemble of collected waveforms [37]. The position
of t0 was fixed at a time near the leading edge of
the SBS waveforms, consistent with the placement
of the leading edge of the gate pulse in the slice en-
ergy noise measurements. CI(t0,s) was calculated
from an ensemble of waveforms for various
choices of the delay s. The waveform ensembles
were collected using the apparatus shown in Fig.1(a). The output of the InGaAs photodiode was
connected to a fast digital oscilloscope, and digi-
tized SBS waveforms were transferred to the com-
puter via a GPIB interface. A set of 1000 SBS
waveforms, each one the result of scattering from
a single pump pulse, was collected for several input
pump powers. The digital oscilloscope samplingrates for these experiments was either 2.5 or 5
GSa/s so that the temporal resolution was more
than adequate to capture the fine structure of the
SBS waveform. The pump pulse was continuously
monitored during waveform collection to ensure
that there were no lapses into multimode
operation.
Finally, we collected average transmitted pumpwaveforms for different input pump power levels
to determine the amount and nature of pump
depletion due to SBS. This data was collected
using the apparatus shown in Fig. 1(b). The dis-
persing prisms allow us to separate and detect
any scattered light due to stimulated Raman scat-
tering in the forward direction. The aperture is
positioned to select the transmitted pump pulseand block the Stokes pulse generated by Raman
scattering. The neutral density filter ensures that
the InGaAs photodiode is not saturated, and a
250 mm focal length lens is used to image the
transmitted pump beam onto the active area of
the photodiode. The photodiode has a rise time
of 200 ps and is connected to the same digital oscil-
loscope used to monitor the input pump pulse andSBS waveforms. Each transmitted pump wave-
form was averaged over 500 shots of the laser.
As in the earlier measurements, the pump pulse
was continuously monitored to ensure stable sin-
gle-mode operation during periods of data
collection.
3. Experiment results and interpretation
The relative noise of the energy contained in a
temporal slice of the SBS waveform as a functionof the duration of the slice is shown in Fig. 3 for
three different input pump powers. Measured rela-
tive noises and relative noises obtained from a sim-
ple statistical model (described below) are shown
on each graph. The relative noise for each data
point is the ratio of the standard deviation of the
slice energy to the mean slice energy, expressed
![Page 7: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/7.jpg)
GateWidth (ns)
GateWidth (ns)
GateWidth (ns)
0 30 60 90 120 150 180
Rel
ativ
e N
oise
(%
)R
elat
ive
Noi
se (
%)
Rel
ativ
e N
oise
(%
)
0
20
40
60
80
100(a)
0 20 40 60 80 1000
10
20
30
40
50
60(b)
0 20 40 60 80 1005
10
15
20
25
30
35(c)
Fig. 3. Slice energy noise versus slice duration for three values
of input pump peak power: (a) 6.14 W, (b) 12.5 W, and (c) 18.4
W. Triangles mark the measured data points, and crosses mark
the slice noises calculated using the statistical model described
in the text.
J. Correa et al. / Optics Communications 242 (2004) 267–278 273
as a percentage in the figure, for an ensemble of
7200 slice energies collected at a fixed input pump
power and gate width. The lowest input pump
power corresponds to very weak Brillouin scatter-
ing for which there is negligible pump depletion.
The highest input pump power corresponds to
strong Brillouin scattering for which there is signif-
icant pump depletion and the onset of stimulated
Raman scattering in the forward direction. The
basic trend displayed in each of the graphs inFig. 3 is the same: the energy fluctuations are larg-
est for the narrowest slice and steadily decrease as
more of the SBS waveform is integrated. Looking
at the structure of the sample SBS waveforms in
Fig. 2, one can see that increasing the slice dura-
tion corresponds to adding up the energy in more
of the intensity spikes in the total SBS waveform.
We can represent the total SBS waveform as acomposite of energy units that correspond to indi-
vidual intensity spikes. If the fluctuations of differ-
ent energy units are statistically independent then,
by virtue of the Central Limit Theorem, we expect
that their sum will become progressively less noisy
as more units are included in the sum.
One way of determining whether or not the en-
ergy fluctuations of different sections of the SBSwaveform are statistically independent is to com-
pare the measured relative noise data to a simple
statistical model. In this model we imagine break-
ing up each SBS waveform into a series of 10 ns
slices, a width comparable to the typical phonon
decay time. The slice energies are assumed to be
independently fluctuating random variables with
identical means and standard deviations. The ratioof the standard deviation to the mean energy for
an individual slice is assigned a value that matches
the measured value for a 10 ns slice. The ratio for
the sum of N slices is then calculated using the fol-
lowing relationship:
rN
lN¼ 1ffiffiffiffi
Np r0
l0
� �;
where r0/l0 is the measured ratio of the standard
deviation to the mean for a 10 ns slice. This simple
relationship follows directly from the assumption
of statistical independence [38]. The relative noises
calculated using this model are shown in Fig. 3
along with the measured data. Although there
are some differences between the model and meas-ured values, the level of agreement is quite good gi-
ven the simplicity of the model and supports the
idea that the energy fluctuations in different sec-
tions of the SBS waveform are statistically
![Page 8: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/8.jpg)
274 J. Correa et al. / Optics Communications 242 (2004) 267–278
independent or perhaps weakly correlated. Later
in this section we will present measured intensity
correlation functions that provide possible reasons
for the differences between the statistical model
and measurements seen in Fig. 3.We turn now to the slice energy histograms
shown in Fig. 4, which are derived from the same
data sets used to calculate the relative noises
shown in Fig. 3. Each histogram is generated from
a set of 7200 raw slice areas that were divided by
the mean slice area so that the mean of the scaled
data was one. This is the reason the horizontal axis
is labeled ‘‘normalized slice area.’’ Fig. 4(a) showsthe histogram for the lowest input pump power
and a 5 ns slice width. This histogram is highly
skewed, with a long tail extending to slice energies
that are many times the mean energy. (The nega-
tive values on the graph are due to the effects of
0 2 4 6 8
Cou
nts
0
200
400
600
800
1000
1200(a) (
0.0 0.5 1.0 1.5 2.0
Cou
nts
0
200
400
600
800(c)
Normalized Slice Area
Normalized Slice Area
(
Fig. 4. Sample slice energy histograms for narrow and wide slices at t
6.14 W, 180 ns slice; (c) 18.4 W, 5 ns slice; (d) 18.4 W, 90 ns slice. Not
the same input pump power.
the integrated electronic noise, which fluctuated
about an average value of zero, pushing a few of
the very weak SBS slice areas below zero. Less
than 1% of the data points were negative for this
histogram, and these points were much smaller inmagnitude than the mean so that the effect on
the relative noise was negligible.) By way of con-
trast, Fig. 4(b) shows the histogram for the same
input pump power but a slice width of 180 ns. This
histogram is much closer to being symmetric, like a
Gaussian distribution, with energies that are much
more tightly confined around the mean. Frames (c)
and (d) of Fig. 4 show histograms for the highestinput pump power for slice widths of 5 and 90
ns, respectively. Both histograms for the highest
pump power are symmetric, but the increase in
the sharpness of the distribution is clearly evident
for the longer duration slice. These changes in
0 2 4 6 8
Cou
nts
0
200
400
600
800
1000
1200b)
Normalized Slice Area
Normalized Slice Area
0.0 0.5 1.0 1.5 2.0
Cou
nts
0
200
400
600
800d)
wo different input pump peak powers: (a) 6.14 W, 5 ns slice; (b)
e the contrast between the narrow and wide slice histograms for
![Page 9: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/9.jpg)
0 10 20 30 40 50 60
Cor
rela
tion
Coe
ffic
ient
Cor
rela
tion
Coe
ffic
ient
Cor
rela
tion
Coe
ffic
ient
0.0
0.2
0.4
0.6
0.8
1.0(a)
0 10 20 30 40 50-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0(b)
Delay (ns)
Delay (ns)
Delay (ns)
0 5 10 15 20 25 30-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0(c)
Fig. 5. Intensity correlation functions for three values of input
pump peak power: (a) 5.99 W, (b) 12.0 W, and (c) 18.0 W.
J. Correa et al. / Optics Communications 242 (2004) 267–278 275
shape and/or sharpness of the histograms as slice
width is increased are consistent with the idea that
the intensity fluctuations in different sections of the
SBS waveform are independent or only weakly
correlated.To provide experimental evidence that the
trends seen in the slice energy statistics can be ex-
plained in terms of the Central Limit Theorem, we
turn to measurements of the intensity correlation
function for the SBS waveforms. The results of
these measurements for three different input pump
powers are shown in Fig. 5. As mentioned earlier,
in calculating the intensity correlation function theinitial point had a fixed position near the leading
edge of the SBS waveform while the delayed point
was systematically swept to later times to calculate
the correlation coefficient for varying delays be-
tween the two points. The structure of the intensity
correlation function does not seem to be sensitive
to the particular choice of the initial point�s loca-tion in the leading section of the SBS waveform.We chose an initial point that excludes most of
the shot-to-shot jitter in the leading edge of the
SBS waveform just as we did for the leading edge
of the gate pulse in the slice energy noise measure-
ments described earlier. In all of the graphs, near
neighbors have intensity fluctuations that are
highly correlated. The strong positive correlation
between near neighbors tends to fade after the de-lay reaches the 5–10 ns range. For the lowest input
pump power, the correlation function has an ex-
tended shoulder beyond the initial period of strong
positive correlation where the correlation coeffi-
cient is small and positive. The extended duration
of positive correlations between intensity fluctua-
tions in different parts of the SBS waveform may
explain why the measured slice energy noises aresystematically larger than the noises calculated
using the simple statistical model since positive
correlations between different slices will enhance
the fluctuations over the case when the slices are
uncorrelated. For the intermediate input pump
power, there is a long interval of negative intensity
correlations beyond the initial strong positive cor-
relations. These negative correlations may explainwhy the measured slice energy noises are systemat-
ically lower than the calculated values since nega-
tive correlations between different slices will
suppress the fluctuations below the value for
uncorrelated slices. For the highest input pump
power the strong positive correlations between
nearest neighbors is followed by relatively strong
negative correlations for a narrow range of delays
followed by very weak correlations for larger de-
lays that fluctuate between positive and negative
![Page 10: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/10.jpg)
Time (ns)
0 20 40 60 80 100
Nor
mal
ized
Int
ensi
ty
0.0
0.2
0.4
0.6
0.8
1.0
Fig. 6. Transmitted pump temporal profile for three values of
input pump peak power: solid curve, 6.10 W; dashed curve, 12.2
W; dash-dot curve, 18.3 W.
276 J. Correa et al. / Optics Communications 242 (2004) 267–278
values. The comparison between the measured and
calculated slice energy noises is arguably the best
for this power. The measured value dips below
the calculated value for intermediate slice dura-
tions, which could be the consequence of the shortinterval of negative correlations discussed above
[37,38]. We can summarize these results by noting
that for all three input pump powers the regime of
strong correlation, either positive or negative, is
confined to points that are within roughly 10–20
ns of one another, which is comparable to the pho-
non decay time typical for silica fibers. So for long
integration intervals we are indeed adding up sec-tions of the waveform that are independently fluc-
tuating or only weakly correlated.
There is one final feature of the intensity corre-
lation functions that merit comment: the weak
oscillations seen for time delays falling after the
initial region of strong correlations. These oscilla-
tions are most prominent in Fig. 5(a), which is the
intensity correlation function in the absence ofpump depletion. This feature could be the result
of the roughly regular spacing of intensity peaks
in the SBS waveforms [26]. The correlation func-
tion represents the ensemble average of the overlap
of an SBS waveform with a delayed version of it-
self. When the delay corresponds to an integer
multiple of the average spacing between adjacent
intensity spikes, one expects to observe peaks inthe correlation function. However, the amplitude
of the correlation peaks will be diminished by shot
to shot jitter in the SBS waveforms.
We also made measurements to determine the
degree of depletion of the transmitted pump by
SBS. The normalized average temporal profile of
the transmitted pump is shown in Fig. 6 for several
different input pump powers that match the pumppowers for the SBS data presented in this article.
Each of the transmitted pump waveforms was
averaged over 500 shots of the laser. Looking at
the figure, we note that the trailing edge of the
pump pulse is eaten away by the SBS process as
the input pump power is increased. This is consist-
ent with the idea that Brillouin scattering tends to
be initiated near the peak of the pump pulse, sinceSBS is an intensity dependent process, and then
counter-propagates through the trailing edge of
the pump pulse growing by stimulated scattering
as it sweeps through the pump. For the highest
pump power, the peak of the transmitted pump
pulse is just beginning to flatten due to the onset
of significant stimulated Raman scattering in the
forward direction. Any further increase in pump
power would result in the preferential depletion
of the peak of the pump pulse by stimulated Ra-man scattering. It should be noted that all of our
SBS data was taken for input pump powers below
the threshold for significant Raman induced pump
depletion. Finally, it should be pointed out that
there is negligible pump depletion for the lowest
input pump power. We verified this by overlaying
the normalized input and transmitted pump
pulses. The shapes of the two pulses were essen-tially identical for the lowest pump power.
The information about pump depletion can be
used to explain features of the SBS intensity statis-
tics that are quantified in Figs. 3 and 5. Starting
with Fig. 3, note that there is a systematic shift
to lower relative noises as the input pump power
is increased, which can be seen by comparing data
points for the same slice widths from frames (a)–(c). In the absence of pump depletion in the weak
scattering limit, spontaneous scattering dominates
the statistics of the resulting intensity spikes. The
fluctuating amplitude of the thermally generated
acoustic waves that drive the spontaneous scatter-
ing can be modeled by a zero-mean Gaussian ran-
dom process; consequently, for sufficiently weak
scattering the distribution of energies contained
![Page 11: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/11.jpg)
J. Correa et al. / Optics Communications 242 (2004) 267–278 277
in individual intensity spikes should be negative
exponential. The result is nearly full-scale energy
fluctuations when the slice width is comparable
to the width of an individual intensity spike
[13,14]. For strong scattering the intensity of indi-vidual spikes begins to saturate as the pump is de-
pleted, and stimulated scattering is now the
dominant contributor to the growth of intensity
spikes. The resulting intensity spikes will have sub-
stantially larger energies with less spread in energy
about the mean, giving rise to a symmetric proba-
bility distribution with a narrow width. For longer
duration slices the relative energy noise drops asthe pump power increases because we are summing
energies of individual slices that are more stable.
Finally, let us consider the effects of pump
depletion on the intensity correlation functions
shown in Fig. 5. At high input pump powers,
strong scattering in the early sections of the fiber
tends to deplete the pump so that only weak
intensity spikes are generated later in the fiber.The region of strong negative correlation seen in
the intensity correlation function for the higher
input pump powers is a consequence of the effects
of pump depletion. Intensity spikes in the early
part of the SBS waveform are effectively compet-
ing for the energy supplied by the pump, and this
competition demands that they be negatively cor-
related. For weak SBS the pump remains strongthroughout the fiber so that different intensity
spikes are not competing for the energy supplied
by the pump, and one would expect the correla-
tion function to drop to zero, or perhaps display
weak oscillations about zero, after the narrow re-
gion of strong positive correlation between near-
est neighbors. However, the data in Fig. 5(a)
shows an extended region of weak positive corre-lations rather than an abrupt drop to zero, the
prominent ‘‘shoulder’’ beyond the main peak of
the correlation function. We believe that this pos-
itive shoulder may reflect the impact of small
shot-to-shot fluctuations in the pump pulse peak
amplitude on the SBS waveform fluctuations.
Pump pulses with larger peak amplitudes will
lead, on average, to trains of larger intensityspikes within an SBS waveform. The reverse will
be true for pump pulses with smaller peak ampli-
tudes. Thus, the successive spikes in an SBS wave-
form will tend to rise and fall in concert with the
rise and fall of the pump pulse peak amplitude.
Once pump depletion comes into play, this fea-
ture disappears because of the strong competition
between individual spikes within an SBSwaveform.
4. Conclusions
In this article we have presented a detailed
experimental study of the statistical properties of
pulsed SBS in optical fiber for pulse widths com-parable to the phonon lifetime but much shorter
than the fiber transit time. Under these conditions
the temporal dynamics of the scattered signal
should be purely stochastic, with fluctuations
determined by the thermally generated sound
waves that initiate the scattering and by any resid-
ual amplitude fluctuations of the single mode
pump laser. We have provided experimental evi-dence and simple theoretical arguments to assess
whether or not different sections of a SBS wave-
form represent statistically independent scattering
episodes from the same input pump pulse. The
temporal structure of the SBS waveforms suggest
this interpretation, especially for weak scattering,
since the scattered light from a smooth single-
peaked pump pulse consists of a long series ofirregular intensity spikes with durations less than
the pump pulse width. Our analysis indicates that
this interpretation is a reasonable approximation
that can explain the general trends of the statistics
of the integrated SBS waveform energy, but that
there are intensity correlations that cause modest
deviations of the statistics from the idealistic case
of completely independent fluctuations. We be-lieve these correlations between the intensity fluc-
tuations within an SBS waveform are the
consequence of the finite damping time for the
thermally excited acoustic waves, the influence of
residual pump pulse amplitude fluctuations, and
the competition that is induced by pump
depletion.
In the future we plan to extend this work in sev-eral ways. First, we are working on a theoretical
model for pulsed SBS that will incorporate ther-
mal and pump fluctuations. We hope to perform
![Page 12: Correlations between intensity fluctuations within stimulated Brillouin waveforms generated by scattering of Q-switched pulses in optical fiber](https://reader031.fdocuments.us/reader031/viewer/2022020511/57501e1a1a28ab877e8ef4bc/html5/thumbnails/12.jpg)
278 J. Correa et al. / Optics Communications 242 (2004) 267–278
simulations and experiments that thoroughly
investigate the interplay of these two sources of
noise in pulsed SBS. It might also be interesting
to investigate the interplay of Brillouin and Ra-
man scattering to determine how the statisticalproperties of the scattered light, in the forward
and backward directions, are impacted by compet-
ing nonlinear processes. We have recently com-
pleted an experimental and theoretical study of
noise shaping by stimulated Raman scattering in
optical fiber that should help with this investiga-
tion [39,40].
Acknowledgements
This work is supported by the National Science
Foundation through grant PHY-0140305. J.R.
Thompson was also supported by a faculty re-
search leave granted by the University Research
Council at DePaul University.
References
[1] H. Haken, Synergetics an Introduction: Nonequilibrium
Phase Transitions and Self-Organization in Physics, Chem-
istry, and Biology, third ed., Springer-Verlag, New York,
1983 (Chapter 2).
[2] R.Y. Chiao, C.H. Townes, B.P. Stoicheff, Phys. Rev. Lett.
12 (1964) 592.
[3] E.P. Ippen, R.H. Stolen, Appl. Phys. Lett. 21 (1972) 539.
[4] D.T. Hon, Opt. Lett. 5 (1980) 516.
[5] M.J. Damzen, H. Hutchinson, IEEE J. Quant. Electron. 19
(1983) 7.
[6] S. Schiemann, W. Ubachs, W. Hogervorst, IEEE J. Quant.
Electron. 33 (1997) 358.
[7] A. Heuer, R. Menzel, Opt. Lett. 23 (1998) 834.
[8] H.J. Eichler, A. Mocofanescu, Th. Riesbeck, E. Risse, D.
Bedau, Opt. Commun. 208 (2002) 427.
[9] T. Tanemura, Y. Takushima, K. Kikuchi, Opt. Lett. 27
(2002) 1552.
[10] A. Loayssa, D. Benito, M.J. Garde, Opt. Lett. 25 (2000)
1234.
[11] S. Norcia, S. Tonda-Goldstein, D. Dolfi, J.-P. Huignard,
R. Frey, Opt. Lett. 28 (2003) 1888.
[12] R. Feced, T.R. Parker, M. Farhadiroushan, V.A. Han-
derek, A.J. Rogers, Opt. Lett. 23 (1998) 79.
[13] R.W. Boyd, K. Rzazewski, P. Narum, Phys. Rev. A. 42
(1990) 5514.
[14] A.L. Gaeta, R.W. Boyd, Phys. Rev. A. 44 (1991) 3205.
[15] A.A. Fotiadi, R. Kiyan, O. Deparis, P. Megret, M.
Blondel, Opt. Lett. 27 (2002) 83.
[16] I. Bar-Joseph, A.A. Friesem, E. Lichtman, R.G. Waarts,
J. Opt. Soc. Am. B. 2 (1985) 1606.
[17] R.G. Harrison, J.S. Uppal, A. Johnstone, J.V. Moloney,
Phys. Rev. Lett. 65 (1990) 167.
[18] M. Dammig, G. Zinner, F. Mitschke, H. Welling, Phys.
Rev. A. 48 (1993) 3301.
[19] R.G. Harrison, P.M. Ripley, W. Lu, Phys. Rev. A. 49
(1994) R24.
[20] D. Yu, W. Lu, R.G. Harrison, Phys. Rev. A. 51 (1995)
669.
[21] G.P. Agrawal, Nonlinear Fiber Optics, second ed., Aca-
demic Press, Boston, 1995 (Chapter 9).
[22] A.A. Fotiadi, R.V. Kiyan, Opt. Lett. 23 (1998) 1805.
[23] K.-D. Park, B. Min, P. Kim, N. Park, J.-H. Lee, J.-S.
Chang, Opt. Lett. 27 (2002) 155.
[24] E.A. Kuzin, M.P. Petrov, A.E. Sitnikov, A.A. Fotiadi,
Sov. Phys. Tech. Phys. 33 (1988) 1420.
[25] A.A. Fotiadi, E.A. Kuzin, M.P. Petrov, A.A. Ganichev,
Sov. Tech. Phys. Lett. 15 (1989) 434.
[26] A.A. Fotiadi, E.A. Kuzin, Tech. Phys. 40 (1995) 740.
[27] Z.M. Benenson, F.V. Bunkin, D.V. Vlasov, E.M. Dianov,
A.Ya. Karasik, A.V. Luchnikov, E.P. Shchebnev, T.V.
Yakovleva, JETP Lett. 42 (1985) 202.
[28] E.M. Dianov, A.Ya. Karasik, A.V. Luchnikov, Sov. Phys.
Tech. Phys. 32 (1987) 928.
[29] E.M. Dianov, A.Ya. Karasik, A.V. Lutchnikov, A.N.
Pilipetskii, Optical and Quantum Electronics, vol. 21, p.
381.
[30] H. Li, K. Ogusu, Jpn. J. Appl. Phys. 38 (1999) 6309.
[31] K. Ogusu, J. Opt. Soc. Am. B. 17 (2000) 769.
[32] H. Li, K. Ogusu, J. Opt. Soc. Am. B. 18 (2001) 93.
[33] V.I. Bespalov, A.A. Betin, G.A. Pasmanik, A.A. Shilov,
JETP Lett. 31 (1980) 630.
[34] M.V. Vasil�ev, A.L. Gyulameryan, A.V. Mamaev, V.V.
Ragul�skii, P.M. Semenov, V.G. Sidorovich, JETP Lett. 31
(1980) 634.
[35] N.G. Basov, I.G. Zubarev, A.B. Mironov, S.I. Mikhallov,
A.Yu. Okulov, JETP Lett. 31 (1980) 645.
[36] T. Crawford, C. Lowrie, J.R. Thompson, Appl. Opt. 35
(1996) 5861.
[37] D.C. Montgomery, G.C. Runger, Applied Statistics and
Probability for Engineers, John Wiley and Sons, New
York, 1994 (Chapter 5).
[38] P.R. Bevington, D.K. Robinson, Data Reduction and
Error Analysis for the Physical Sciences, second ed.,
McGraw Hill, Boston, 1992 (Chapter 3).
[39] L. Garcia, A. Jalili, Y. Lee, N. Poole, K. Salit, P. Sidereas,
C.G. Goedde, J.R. Thompson, Opt. Commun. 193 (2001)
289.
[40] L. Garcia, J. Jenkins, Y. Lee, N. Poole, K. Salit, P.
Sidereas, C.G. Goedde, J.R. Thompson, J. Opt. Soc. Am.
B. 19 (2002) 2727.