21Fiber-optic quantum informationtechnologies
Prem Kumar*, Jun Chen†, Paul L. Voss†, Xiaoying Li†,
Kim Fook Lee†, and Jay E. Sharping‡
*Technological Institute, Northwestern University, Evanston,IL, USA†Center for Photonic Communication and Computing,EECS Department, Northwestern University, Evanston, IL, USA‡University of California, Merced, CA
21.1 INTRODUCTION
Quantum mechanics (QM), born almost a century ago, is one of the most aston-
ishing pieces of knowledge that human beings have ever discovered about
Nature. Its rules are surprisingly simple: linear algebra and first-order partial
differential equations, and yet its predictions are so unimaginably precise and
unbelievably accurate when compared with experimental data. Such a successful
theory, however, is not without its own imperfections (or mysteries). For
example, the orthodox interpretation of QM—the Copenhagen interpretation—
does not give a satisfactory explanation about how and why the wavefunction
of a particle (e.g., an electron) would suddenly collapse once some measure-
ment has been made on it (e.g., an electron has been registered by a particle
detector). This so-called measurement problem has not been properly understood
since the very early days of QM, until recently when the process of quantum
measurement is thoroughly studied, and the concept of “decoherence” is pro-
posed [1].
Optical Fiber Telecommunications V A: Components and Subsystems
Copyright � 2008, Elsevier Inc. All rights reserved.
ISBN: 978-0-12-374171-4 829
Another “mysterious” feature of QM, which we will explain in further detail, is
the superposition principle and the ensuing quantum entanglement. The former
allows a quantum mechanical system to be in any state that is spanned by the basis
vectors of its Hilbert space. For example, if the system can be in two orthogonal
states j0i and j1i (i.e., h0j1i = 0), it can also be in a linear combination of these
two states—�j0i þ �j1i, where � and � are complex numbers satisfying j�j2 þj�j2 = 1. This innocuous-looking principle is, in fact, the origin of a lot of
“quantum weirdness” not observed in our classical everyday experiences.
Entanglement is such a counterintuitive example. Consider the following bipartite
state for particles A and B:
j�i= 1ffiffiffi2p ðj0iAj0iB þ j1iAj1iBÞ; (21.1)
which states that whenever particle A is detected to be in state j0i, particle B must
also be found in j0i (in its own subspace), and vice versa. The same rule applies if
particle A is found in state j1i. That is, we are assured to find particle B in its own
j1i state with unit probability. This may not sound so surprising at first sight; after
all, classical objects sometimes exhibit this kind of correlation too. For example,
we can take a coin and split it into half. Then we put the two half-coins into two
separate envelopes, which are sealed afterward. Suppose we do it in such a way
that nobody, not even we, know exactly which half of the coin ends up in which
envelope. It is obvious that if we open one of the envelopes and find that we get the
“head” portion of the coin, we can infer with 100% certainty that the other
envelope contains the “tail” portion of the coin, and vice versa. The two halves
of the same coin, just like the two particles in Eqn (21.1), can be spatially
separated. The deterministic correlation between the two parties in both cases
remain the same, no matter how far they are from each other. An obvious question
naturally arises: as the coin game is something we can play everyday, what is so
shocking about the correlation that we find in the entanglement example?
The fundamental difference between quantum entanglement and classical corre-
lation lies in the fact that particles are quantum-mechanical objects which can exist
not only in states j0i and j1i but also in states described by �j0i þ �j1i (allowed by
the superposition principle!), while half-coins, being classical objects, can only live
in one of the two deterministic states (“head” or “tail”), and not something in
between. To put it more bluntly, in the case of coins, even though we do not
know a priori which portion of the coin ends up in which envelope, we are confident
that one of the envelopes must contain the “head” portion of the coin, the other
envelope the “tail” portion of the coin. This “confident ignorance” about the results
of classical correlation cannot be safely extended into the regime of quantum
entanglement. In fact, before we decide which basis to use for measuring the states
of the particles in Eqn (21.1), we do not know anything about what results we will
eventually get. We cannot even say, before measurement, each particle is in some
deterministic state, we just do not know which is which. It is not that we do not have
830 Prem Kumar et al.
the knowledge about the states of individual particles; they, in fact, do not come into
being until we make the measurement. In other words, the individual particle in
quantum entanglement does not have a well-defined pure state before measurement.
Each of them are in a mixed state; the joint state of both particles constitutes the pure
state in Eqn (21.1), which we call “entanglement.”
It was Schrodinger who first realized the strangeness of entanglement, or
“Verschrankung” as it was originally coined in German. He pondered the quantum-
to-classical-transition problem at the same time, and extended the concept of
entanglement to the ill-defined boundary between quantum and classical worlds,
where the contrast is most extreme. He imagined a macroscopic object, e.g., a cat
(which later becomes the notorious Schrodinger’s cat), is somehow entangled with a
microscopic object, e.g., an atom. The poor cat”s fate depends solely on the decay-
ing property of the atom. If the atom decays, the cat dies; if the atom does not decay,
the cat lives. As discussed before, the atom can live in a superposition state “decay–
not decay,” and as the two are entangled, the cat is then forced to be living in a state
“dead–alive.” This is very counterintuitive, since normally we do not observe a half-
dead, half-alive cat in our daily lives.
This problem was brought into focus by Einstein, Podolsky, and Rosen (EPR) [2]
in a famous paper in 1935, in which they pointed out the incompatibility between
QM and local realism. The latter notion consists of two parts: locality is a very
reasonable assumption that directly follows our everyday physical intuition, which
postulates the nonexistence of “action-at-a-distance”; realism demands the existence
of “elements of physical reality” in every physical system, which should take
definite values prior to any conceivable measurements. In their example, EPR
considered a quantum system composed of two particles such that neither one of
them has well-defined position or momentum, but the sum of their positions (their
center-of-mass) and the difference of their momenta (their individual momenta in
the center-of-mass system) are precisely defined. It then follows that measurement
of either particle’s momentum (position) would immediately determine the measure-
ment outcome for the other particle’s momentum (position), without even interact-
ing with that particle. Since the two particles can be separated by arbitrary distances,
and properties like position and momentum of a particle are “elements of reality”
according to EPR that must assume definite values before any measurement, EPR
then suggest this “spooky action-at-a-distance” must imply that QM is at least
incomplete, if not incorrect; and that there should be a deeper theory, possibly
with some hidden degrees of freedom (later known as “hidden variables”), which
can faithfully reproduce every result that QM has achieved, and hopefully retain our
familiar deterministic classical world view—local realism. Niels Bohr [3] replied by
arguing that the two particles in the EPR case are always parts of one quantum
system, and thus measurement on one particle changes the possible predictions that
can be made for the entire system and consequently on the other particle; QM is
indeed complete and there is no need for a “more complete” theory.
While for a long time, the famous Einstein–Bohr debate has been widely regarded
as merely philosophical, David Bohm [4] in 1951 introduced spin-entangled
21. Fiber-Optic Quantum Information Technologies 831
systems, as a discrete version of the original continuous EPR-entangled systems. In
1964, John Bell [5] pointed out that for such spin-entangled systems, classical
hidden-variable theories would make different predictions from QM on measure-
ments of correlated quantities. The theorem he published, later known as Bell’s
theorem, quantified just how strongly quantum particles were correlated than would
be classically allowed. This effectively opened up the possibility of experimentally
testing quantum mechanical predictions against those of classical hidden variable
theory. By now a number of experiments have been performed, and the results are
almost universally accepted to be fully in favor of QM [6–10]. However, from a
strictly logical point of view the problem is not completely closed yet, because some
loopholes in these existing experiments still make it at least logically possible to
uphold a local realist world view [11, 12].
More recently, since the beginning of 1990s, the field of quantum information
and quantum communication has opened up and expanded rapidly [13, 14].
Quantum entanglement, once the core concept and the sole mystery of the decade-
long Einstein–Bohr debate, has begun to take on a new look. It is still an unresolved
mystery, philosophically speaking, as it forces us to abandon either one of the two
familiar notions that we hold dear since the beginning of modern-day science:
locality and realism. But this sacrifice we have to make, much in the spirit of
Niels Bohr’s comment that we have no right to tell God what to do; we are only
entitled to discover what God’s plans are and accept them, or better yet, utilize them
to our full advantage. After taking this more humble and more practical point of
view, a whole new world of “quantum ideas” have been ignited and are actively
being pursued. Quantum teleportation [15] and quantum cryptography [16] are two
prominent examples. Quantum entanglement plays a central role in the former and
can lead to many advantages in the latter.
21.2 FIBER NONLINEARITY AS A SOURCEFOR CORRELATED PHOTONS
Efficient generation and transmission of quantum-correlated photon pairs, espe-
cially in the 1550-nm fiber-optic communication band, is of paramount impor-
tance for practical realization of the quantum communication and cryptography
protocols [17]. The workhorse source employed in all implementations thus far
[18] has been based on the process of spontaneous parametric down-conversion
(SPDC) in second-order (�(2)) nonlinear crystals. Such a source, however, is not
compatible with optical fibers as large coupling losses occur when the pairs are
launched into the fiber. This severely degrades the correlated photon-pair rate
coupled into the fiber, because the rate depends quadratically on the coupling
efficiency. From a practical standpoint it would be advantageous if a photon-
pair source could be developed that not only produces photons in the commu-
nication band but also can be spliced to standard telecommunication fibers with
high efficiency. Over the past few years various attempts have been made to
832 Prem Kumar et al.
develop more efficient photon-pair sources, but all have relied on the �(2)
down-conversion process [19–25]. Of particular note is Ref. [26], in which the
effective �(2) of periodically poled silica fibers was used. In this chapter, we report
the first, to the best of our knowledge, photon-pair source that is based on the Kerr
nonlinearity (�(3)) of standard fiber. Quantum-correlated photon pairs are observed
and characterized in the parametric fluorescence of four-wave mixing (FWM) in
dispersion-shifted fiber (DSF).
The FWM process takes place in a nonlinear-fiber Sagnac interferometer
(NFSI), shown schematically in Figure 21.1. Previously, we have used this NFSI
to generate quantum-correlated twin beams in the fiber [27]. The NFSI consists of
a fused-silica 50/50 fiber coupler spliced to 300 m of DSF having zero-dispersion
wavelength �0 = 1537 nm. It can be set as a reflector with proper adjustment of
the intraloop fiber polarization controller (FPC) to yield a transmission coefficient
<�30 dB. When the injected pump wavelength is slightly greater than �0, FWM in
the DSF is phase-matched [19]. Two-pump photons of frequency !p scatter into a
signal photon and an idler photon of frequencies !s and !i, respectively, where
!s þ !i = 2!p. Signal/Idler separations of ’20 nm can be easily obtained with
use of commercial DSF [27]. The pump is a mode-locked train of ’3 ps long
pulses that arrive at a 75.3-MHz repetition rate. The pulsed operation serves two
important purposes: (i) the NFSI amplifier can be operated at low-average powers
(typical values are � 2 mW, corresponding to � 9-W peak powers) and (ii) the
production of the fluorescence photons is confined in well-defined temporal
windows, allowing a gated detection scheme to be used to increase the signal-to-
noise ratio. A 10% (90/10) coupler is employed to inject a weak signal, which is
parametrically amplified, and the output signal and the generated idler are used for
alignment purposes. For the photon-counting measurements described in this
chapter, the input signal is blocked.
Pumpin
FPC
300 mDSF loop
Signalin
Primarydiffraction
grating
Secondarydiffractiongratings
APD2
APD1
Coincidencecounter
Signal andidler out
FPC
50/50coupler
90/10
Rejectedpump
Figure 21.1 Diagram of the experimental setup; FPC, fiber polarization controller (this figure may be
seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 833
After passing through the 90/10 coupler, the fluorescence photons are directed
toward free-space filters that separate the signal and the idler photons from each
other and from the pump photons. To measure the nonclassical (i.e., quantum)
correlations between the signal and the idler photons, one must effectively sup-
press the pump photons from reaching the detectors. Because a typical pump pulse
contains ’108 photons and we are interested in detecting ’0.01 photons/pulse, a
pump-to-signal (idler) rejection ratio in excess of 100 dB is required. To meet this
specification, we constructed a dual-band spectral filter based on a double-grating
spectrometer. A primary grating (holographic, 1200 lines/mm) is first employed
to spatially separate the signal, the pump, and the idler photons. Two secondary
gratings (ruled, 600 lines/mm) are then used to prevent the pump photons that are
randomly scattered by the primary grating (owing to its nonideal nature) from
going toward the signal and idler directions. The doubly diffracted signal and idler
photons are then recoupled into fibers, which function as the output slits of the
spectrometer.
Transmission spectrum of the dual-band filter, measured with a tunable source
and an optical spectrum analyzer (OSA), is shown in Figure 21.2. The shape is
Gaussian in the regions near the maxima of the two transmission bands, which are
centered at 1546 nm (signal) and 1528 nm (idler), respectively, and the full-width
at half-maximum (FWHM) is ’0.46 nm. For pulse trains separated by 9 nm, which
is the wavelength difference between the pump and the signal (or idler), this filter
is able to provide an isolation ‡75 dB; the measurement being limited by the
intrinsic noise of the OSA. The combined effect of the Sagnac loop and the
double-grating filter thus provides an isolation ‡105 dB from the pump photons
in the signal and idler channels. The maximum transmission efficiency in the
–90
–80
–70
–60
–50
–40
–30
–20
–10
0
1523 1528 1533 1538 1543 1548
Wavelength (nm)
Filt
er tr
ansm
issi
on (
dB)
Instrumentnoise
Idler Signal
Figure 21.2 Transmission curves of the signal and idler channels in the dual-band filter (this figure
may be seen in color on the included CD-ROM).
834 Prem Kumar et al.
signal channel is 45% and that in the idler channel is 47%. The total collection
efficiency for the signal (idler) photons is thus 33% (35%), with inclusion of the
losses in the Sagnac loop (18%) and at the 90/10 coupler (10%).
The separated and filtered signal and idler photons are directed toward fiber-
pigtailed InGaAS/InP avalanche photodiodes (APDs, Epitaxx EPM239BA). In
recent years, the performance of InGaAs APDs as single-photon detectors for
use in the fiber communication window around 1550 nm has been extensively
studied by several groups [28–31]. The pulsed nature of the photon pairs allows
us to use the APDs in a gated Geiger mode. In addition, the quality of our
APDs permits room-temperature operation with results comparable to those
obtained by other groups at cryogenic temperatures. A schematic of the electro-
nic circuit used with the APDs is shown in the inset in Figure 21.3. A bias
voltage VB (’�60 V), slightly below the avalanche breakdown voltage, is
applied to each diode and a short-gate pulse (�8 V, 1 ns FWHM) brings the
diodes into the breakdown region. The gate pulse is synchronized with the
arrival of the signal and idler photons on the photodiodes. Due to limitations
of our gate-pulse generator, the detectors are gated once every 128 pump pulses,
giving a photon-pair detection rate of 75.3 MHz/128 = 588 kHz. We expect this
rate to increase by more than an order of magnitude with use of a better pulse
generator. The electrical signals produced by the APDs in response to the
incoming photons (and dark events) are reshaped into 500-ns-wide transistor-
transistor logic (TTL) pulses that can be individually counted or sent to a TTL
AND gate for coincidence counting.
In Figure 21.3 we show a plot of the quantum efficiency vs the dark-count
probability for the two APDs used in our experiments. A figure of merit for the
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 1 2 3 4 5 6 7
Dark-count probability (×10–3 counts/pulse)
Qua
ntum
effi
cien
cy
APD 1
APD 2
VBTo counter
or AND gate
Invertingamplifier
Pulseshaper
Figure 21.3 Quantum efficiency vs dark-count probability for the two APDs used in the experiments.
The inset shows a schematic of the electronic circuit used with the APDs.
21. Fiber-Optic Quantum Information Technologies 835
APDs can be introduced through the noise-equivalent power NEP = (h�/�)(2RD)1/2
[20], where h is the Planck constant, � is the frequency of light, � is the detector
quantum efficiency, and RD is the dark-count rate measured during the gate time.
The best values of NEP obtained by optimizing VB are 1.0 � 10�15 W/Hz�1/2 for
APD1 and 1.6 � 10�15 W/Hz�1/2 for APD2. These values are comparable to those
reported in Refs [28–31] for cryogenically cooled APDs. Under the optimized
conditions, the efficiency of APD1 (APD2) is 25% (20%) and the corresponding
dark-count probability is 2.2 � 10�3/pulse (2.7 � 10�3/pulse).
As a first test of our photon-pair source, and of the filtering process, we measure
the number of scattered photons detected in the signal (idler) channel, NS (NI), as a
function of the number of pump photons, NP, injected into the NFSI. The results
for the idler channel are shown in the inset in Figure 21.4. We fit the experimental
data with NS =ND þ s1NP þ s2N2P, where ND is the number of dark counts during
the gate interval, and s1 and s2 are the linear and quadratic scattering coefficients,
respectively. The fit clearly shows that the quadratic scattering owing to FWM in
the fiber can dominate over the residual linear scattering of the pump due to
imperfect filtering.
In Figure 21.4 we present the coincidence counting results. The diamonds
represent the rate of coincidence counts as a function of the rate of the signal
and idler photons generated during the same pump pulse. For convenience, we
0
0.0004
0.0008
0.0012
0.0016
0.002
0 0.005 0.01 0.015 0.02 0.025 0.03
Signal/idler photon rate (counts/pulse)
Coi
ncid
ence
rat
e (c
ount
s/pu
lse)
0.10.1 1 10 100 1000
1
10
100
ND +
s1NP
+ s2NP
2
ND
s2NP2
s1NP
Pump photons (×106/pulse)
Sig
nal p
hoto
ns (
×10– 3
/pul
se)
Figure 21.4 Coincidence rates as a function in the single-photon rates in two different cases: signal/
idler fluorescence produced by a pump pulse (diamonds) and signal/idler fluorescence produced by two
consecutive pump pulses (triangles). The line represents the calculated “accidental” counts. The inset
shows a plot of the detected idler photons as a function of the injected pump photons (hollow circles).
A second-order polynomial is shown to fit the experimental data. The contributions of the dark counts,
linear scattering, and quadratic scattering are plotted separately as well (this figure may be seen in color
on the included CD-ROM).
836 Prem Kumar et al.
have plotted the coincidence rate as a function of the geometric mean of the signal
and idler count rates; in fact, because the efficiency of the two detectors is
different, we measure different single-photon count rates in the two channels.
Dark counts have been subtracted from the plotted count rates. For the coincidence
rates, both dark–dark and photon–dark coincidences have to be taken in account,
but for the counting rates in our experiment the former are negligible. Thus far,
we have achieved a maximum coincidence rate of 103 counts/s (= coincidence
rate/pulse� gate-pulse rate), which is expected to go up by at least a factor of
10 with use of a higher repetition-rate gate-pulse generator.
We have performed two independent experiments to demonstrate the nonclassi-
cal nature of the coincidences. Results of the first experiment are shown by the
triangles in Figure 21.4, which represent the measured coincidence rate as a
function of the signal-photon count rate when the signal is delayed with respect to
the idler by one pulse period. The delay was achieved by inserting a fiber patch-cord
of appropriate length in the signal path from the output of the filter to APD1. For
two independent photon sources, each with a count rate RS� 1, the “accidental”
coincidence rate RC is given by RC =R2S, regardless of the photon statistics of the
sources. This quadratic relation is plotted as the solid line in Figure 21.4, which fits
the delayed-coincidence data (triangles) very well. These measurements then show
that while the fluorescence photons produced by the adjacent pump pulses are
independent, those coming from the same pump pulse show a strong correlation,
which is a signature of their nonclassical behavior.
In the second experiment, measurements were performed to demonstrate the
nonclassicality test described in Ref. [24]. It can be shown that the inequality
RC � RðaÞC � 2 RS=2 � R
ðaÞS=2þ RI=2 � R
ðaÞI=2
� �� 0 (21.2)
is valid for two classical light sources, where RC is the coincidence-count rate for
the two sources, RðaÞC is the calculated “accidental” coincidence-count rate corre-
sponding to the same photon-count rate for the two sources, RS/2 and RI/2 are the
coincidence-count rates measured by passing the light from each of the two
sources through a 50/50 splitter and detecting the two halves independently, and
RðaÞS=2 and R
ðaÞI=2 are the calculated “accidental” coincidence-count rates in the 50/50
splitting measurements. When we substitute the experimental data, Eqn (21.2)
yields (64 – 9) 10�6� 0, where the error is statistical. The inequality for classical
sources is thus violated by over seven standard deviations.
In conclusion, we have demonstrated, for the first time to our knowledge, a
source of quantum-correlated photon pairs that is based on FWM in a fiber near
1550 nm. We have also developed and tested a room-temperature coincidence
detector for the photons in that window. The photon-pair detection rate (’103
coincidence counts/s) at present is limited by the electronics employed in our
setup. In addition, we believe that the spectral filter used for rejecting the pump
photons can be implemented with fiber Bragg gratings, making this source integr-
able with the existing fiber-optic infrastructure.
21. Fiber-Optic Quantum Information Technologies 837
21.3 QUANTUM THEORY OF FOUR-WAVE MIXINGIN OPTICAL FIBER
Four-wave mixing has long been studied, especially in the context of isotropic
materials, e.g., optical fibers [25, 32]. Generally speaking, it is a photon–photon
scattering process, during which two photons from a relatively high-intensity beam,
called pump, scatter through the third-order nonlinearity (�(3)) of the material (silica
glass in the case of optical fibers) to generate two daughter photons, called signal
and idler photons, respectively. The frequencies of the daughter photons are
symmetrically displaced from the pump frequency, satisfying the energy conserva-
tion relation !s þ !i = 2!p, where !j ( j = p, s, i) denotes the pump/signal/idler
frequency, respectively. They are predominantly copolarized with the pump beam,
owing to the isotropic nature of the optical Kerr nonlinearity:
�ð3Þxxxx =�
ð3Þxxyy þ �ð3Þxyxy þ �ð3Þxyyx = 3�
ð3Þxxyy. The daughter photons also form a time–
energy-entangled state, in the sense that the two-particle wavefunction cannot be
factorized into products of single-particle wavefunctions: �ð!s; !iÞ 6¼ �ð!sÞ� ’ð!iÞ.This four-photon scattering (FPS) process is intrinsically interesting and particularly
useful when applied to the field of quantum information processing (QIP), in which
generation of entangled states and test of Bell’s inequalities play an important role.
A great amount of original work, both theoretical and experimental, has been done
in the rapidly expanding field of QIP (see, e.g., Ref. [33] for a general review). The
workhorse process for generating entangled states is the process of SPDC in second-
order (�(2)) nonlinear crystals, which has been studied exhaustively during the past
decades. However, unlike its �(2) counterpart, the �(3) process of FWM has received
relatively lesser theoretical attention in the quantum mechanical framework, despite
its apparent benefits in the applications of QIP. To name a few, the ubiquitous readily
available fiber plant serves as a perfect transmission channel for the FWM-generated
entangled qubits, whereas it remains a technical challenge to efficiently couple �(2)-
generated entangled photons into optical fibers due to mode mismatch. Besides, the
excellent single-mode purity of the former makes it suitable for applications that
require multiple quantum interactions. Furthermore, it is also possible to wavelength
multiplex several different entangled channels from the broadband parametric spec-
trum of FWM by utilizing the advanced multiplexing/demultimplexing devices
developed in connection with the modern fiber-optic communications infrastructure.
The only drawback of this scheme that has been identified is the process of sponta-
neous Raman scattering (SRS), which inevitably occurs in any �(3) medium and
generates uncorrelated photons into the detection bands, leading to a degradation in
the quality of the generated entanglement [28]. Various efforts have been made to
minimize the negative effect that SRS imposes [29, 30].
In this section, we present a quantum theory that models the FWM process in an
optical fiber, without inclusion of the Raman effect. The pump is treated as a classical
narrow (picosecond-duration) pulse due to its experimental relevance. The signal and
idler fields form a quantum mechanical two-photon (or “biphoton” [31, 34]) state at
838 Prem Kumar et al.
the output of the fiber. From the experimental point of view, what we are mostly
interested in is the nonclassicality that the two-photon state exhibits. It is this
unique quantum feature that makes the two-photon state a valid candidate for various
quantum-entanglement-related experiments, including quantum cryptography [35],
quantum teleportation [15], etc. Coincidence-photon counting, or second-order coher-
ence measurement of the optical field [36], serves as a measurement technique that
distinguishes a quantum mechanically entangled state from a classically correlated
state, which will form a central part of our investigation.
A sample coincidence-counting result from Ref. [37] is shown in Figure 21.5.
The top (bottom) series of data points represents the total (accidental) coincidence-
count rate as a function of the single-channel count rate. SRS and dark counts from
the detectors account for the major part of the accidental coincidence counts. Our to-
be-developed theory, however, only takes into account the photon counts generated
by the FWM process. To reconcile the theory with experiments, the contributions
from SRS and dark counts from the detectors are independently measured [29, 30],
and subsequently subtracted from both the single counts and the total coincidence
counts. Overall quantum efficiencies of detection in both the signal and idler
channels are also separately measured. The single-count rates are divided by the
respective quantum efficiencies and the coincidence-count rate by the product of the
efficiencies in the signal and idler channels to arrive at rates at the output of the fiber
for comparison with the prediction of our theory. The dependence of the photon-
counting results on various system parameters, for instance, the pump power, pump
bandwidth, filter bandwidth, etc., can be studied.
Polarization entanglement has also been generated by time and polarization
multiplexing two such FWM processes [38, 39]. However, the theory for that
0
0.0004
0.0008
0.0012
0.0016
0.002
0 0.005 0.01 0.015 0.02 0.025 0.03
Signal/idler photon rate (counts/pulse)
Coi
ncid
ence
rat
es (
coun
ts/p
ulse
)
Figure 21.5 (Color online) Experimental results. Diamonds, total coincidences; triangles, accidental
coincidences; curve, theoretical fit y = x2 for statistically independent photon sources (this figure may
be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 839
particular experiment is a straightforward extension of our current theory, and
therefore will not be included in the analysis to follow.
Having described the experiment in the previous section, we are ready to start
building up the theoretical model for that experiment. We take the standard
approach of modern quantum optics, i.e., finding out the interaction Hamiltonian
and calculating the evolution of the state vector using the Schrodinger picture. We
accomplish the first task by seeking connections with the well-known classical
FWM theory in optical fibers [32]. The coupled classical-wave equations for the
pump, signal, and idler fields are
qAp
qz¼ ijApj2Ap;
qAs
qz¼ i 2jApj2As þ A2
pA�i e�iDkz
h i;
qAi
qz¼ i 2jApj2Ai þ A2
pA�s e�iDkz
h i;
(21.3)
where the usual undepleted-pump approximation has been made, and we only keep
terms that are significant, i.e., to OðA2pÞ. Fiber loss is neglected from the above
equations. The Aj ( j = p, s, i) denote electric-field amplitudes for the pump,
signal, and idler, respectively, and all of them have been normalized such that
their unit isffiffiffiffiffiWp
Dk= ks þ ki � 2kp is the magnitude of the wave-vector mis-
match. = 2n2/�Aeff is the nonlinear parameter of interaction, wherein
n2 = ð3=4n2�0cÞReð�ð3ÞxxxxÞ is the nonlinear-index coefficient, �0 is the vacuum
permittivity, Aeff is the effective mode area of the optical fiber, and ���p,s,i is
the wavelength involved in the FWM interaction.
Due to the highly nonresonant nature of FWM in optical fibers, we expect the
quantum equations of motion, which describe the interplay between and evolution
of the fields at the photon level, to fully correspond with their classical counterparts.
In light of this correspondence principle, we write the quantum equations of motion
by replacing the classical amplitudes in Eqn (21.3) with electric-field operators:
qEðþÞp
qz¼ i� Eð�Þp EðþÞp EðþÞp ;
qEðþÞs
qz¼ i� ½2Eð�Þp EðþÞp EðþÞs þ E
ð�Þi EðþÞp EðþÞp ;
qEðþÞi
qz¼ i� ½2Eð�Þp EðþÞp E
ðþÞi þ Eð�Þs EðþÞp EðþÞp ;
(21.4)
where EðþÞj =
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið�h!j=2�0VQÞ
paj ð j= p; s; iÞ are the positive-frequency electric-
field operators, corresponding to photon annihilation operators, and VQ is the
840 Prem Kumar et al.
quantization volume. Here we omit the Hermitian-conjugate equations corre-
sponding to Eqn (21.4) for simplicity. We have assumed that the photon fields
phase match, i.e., Dk = 0. In Eqn (21.4), �= � ð�ð3ÞAeff nL!=2VQcÞ is a const-
ant similar to in the classical Eqn (21.3); the exact form of this constant differs
from its classical cousin to compensate for the unit discrepancy between the two
sets of equations (note that the operator EðþÞj is of unit V/m, and the amplitude Aj is
of unitffiffiffiffiffiWp
). The correct form of the interaction Hamiltonian that we are
seeking should lead to Eqn (21.4) through the Heisenberg equation of motion
for the field operators, namely, i�hðqE=qtÞ= ½E;HI, where E stands for any
electric-field operator. Utilizing the mathematical facts q/qt (c/n)(q/qz) and
½EðþÞj ðzÞ;Eð�Þk ðz0Þ ¼ ð�h!=2�0VQÞ � ðz� z0Þ �jk, we arrive at the following form
for our interaction Hamiltonian:
HI ¼ � �0 �ð3ÞZ
V
dV Eð�Þp Eð�Þp EðþÞp EðþÞp
hþ 2Eð�Þs E
ð�Þi EðþÞp EðþÞp
þ 2Eð�Þp Eð�Þp EðþÞs EðþÞi þ 4Eð�Þp EðþÞp Eð�Þs EðþÞs þ 4Eð�Þp EðþÞp E
ð�Þi E
ðþÞi
i; (21.5)
where � is an overall unknown constant related to the specific experimental
details, which will be determined later when we compare our theory with the
experiment; �(3) is the nonlinear electric susceptibility whose tensorial nature is
ignored as all the optical fields are assumed to be linearly copolarized. The integral
is taken over the entire volume of interaction, namely, the effective volume of the
optical fiber. We label the first term in the integrand of Eqn (21.5) as the self-phase
modulation (SPM) of the pump field, the next two terms as the FPS among the
optical fields, and the last two terms as the cross-phase modulation (XPM)
between pump and signal (idler) fields.
After obtaining the Hamiltonian responsible for the quantum FWM process,
we are ready to tackle our next task: calculate the state vector evolution. It is
worthwhile, at this point, to define the various electric field operators appearing
in the Hamiltonian, in accordance with the experiment we are trying to model.
The pump field is taken to be a classical narrow pulse, which is linearly
polarized, propagating in the z direction (parallel with the fiber axis), with a
central frequency �p and an envelope of arbitrary shape eEp. Mathematically, it
can be written as
EðþÞp ¼ e�i�pteEpðz; tÞ
¼ e�i�pt
Zd�p Epð�pÞ eikpz�i�pt; (21.6)
wherein the bandwidth of the pump field is much smaller than �p, satisfying the
quasi-monochromatic approximation. The signal and idler fields are quantized
21. Fiber-Optic Quantum Information Technologies 841
electromagnetic fields, copolarized and copropagating with the pump, as given by
the following multimode expansion:
Eð�Þs =X!s
ffiffiffiffiffiffiffiffiffiffiffiffi�h!s
2�0VQ
sayks
nð!sÞe�i½ksð!sÞz�!st; (21.7)
Eð�Þi =
X!i
ffiffiffiffiffiffiffiffiffiffiffiffi�h!i
2�0VQ
sayki
nð!iÞe�i½kið!iÞz�!it; (21.8)
where ayks
is the creation operator for the signal mode with frequency !s, ks (!s) =n(!s)!s/c is its wave-vector magnitude. The idler field is defined in an analogous
fashion. The central frequencies of the signal and idler fields are individually
denoted by �s and �i, which are symmetrically distanced from the central fre-
quency of the pump field �p, satisfying the energy conservation relation
�s þ �i = 2�p.
To simplify our calculation and to compare our results with the experiments,
two assumptions are further made about the pump field: it has a Gaussian spectral
envelope and its SPM is included in a straightforward manner, i.e.,
EðþÞp = e�i�pte�iPpzEp0
Zd�p e
�ð�2p=2 2
pÞ eikpz�i�pt; (21.9)
where Pp 2ffiffiffip
Aeff�0 c n 2p E
2p0 is the peak power of the pump pulse, which is
treated as a constant under the undepleted pump approximation, and p is the
optical bandwidth of the pump. The first assumption is justified by the fact that
our experimental optical filter for the pump can be well approximated by a
Gaussian function in the frequency domain. The validity of the second assumption
can be seen when we solve the classical equation of motion for the pump
field, namely, the complex conjugate form of the first equation in Eqn (21.3),
which reads
qA�pqz
= � ijApj2A�p: (21.10)
We choose to study the complex conjugate form of the equation because it is A�pthat corresponds to E
ðþÞp . Straightforward calculations show that the solution to
Eqn (21.10) is A�pðzÞ=A�pð0Þ e�iPpz, where Pp = jApj2 is the same undepleted
peak power of the pump pulse. The SPM term of the pump, e�iPpz, which is the
nonlinear phase factor in the classical FWM theory that determines the phase-
matching condition [32], now manifests itself in our quantum mechanical calcula-
tion as a “phase tag” for the pump field through its propagation along the optical
fiber. Finally, the undepleted-pump approximation holds because the loss in the
fiber is negligible and only a few photons are scattered (�1 out of 108) through the
nonlinear interaction.
842 Prem Kumar et al.
The two-photon state at the output of the fiber is calculated by means of first-
order perturbation theory, i.e.,
j�i= j0i þ 1
i�h
Z 1�1
HIðtÞ dtj0i: (21.11)
Retaining of higher order terms in the perturbation series involves generation of
multiphoton states, which will be ignored in our calculation owing to their
smallness. We can see that only the FPS terms in the interaction Hamiltonian
contribute to the formation of the signal/idler two-photon state. This is because all
terms vanish when acting on the vacuum state j0i with the exception of
Eð�Þs E
ð�Þi E
ðþÞp E
ðþÞp þ h:c:, which we denote as
HFPS ��0�ð3ÞZ
V
dVðEð�Þs Eð�Þi EðþÞp EðþÞp þ h:c:Þ; (21.12)
where � = 2�, and h.c. stands for Hermitian conjugate.
The state vector is then given by
j�i= j0i þ 1
i�h
Z 1�1
HFPS dtj0i; (21.13)
which is a superposition of the vacuum and the two-photon state. Substituting
Eqns (21.7–21.9) and (21.12), into Eqn (21.13), after some algebra, leads to the
following form of the state vector:
j�i= j0i þXks;ki
Fðks; kiÞ ayks aykij0i; (21.14)
Fðks; kiÞ ¼g
Z 0
�L
dz1ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1� ik00ð�pÞ 2pz
q exp � ik00ð�pÞz4
ð�s � �i þ DÞ2�
�2iPpz� ð�s þ �iÞ2
4 2p
); (21.15)
g =�2�ð3Þ Pp
i �0VQ n3�p p: (21.16)
The function F(ks, ki) is called the two-photon spectral function [34]. Here
k00ð�pÞ= d2k=d!2��!¼�p
is the second-order dispersion at the pump central
frequency [also known as the group-velocity dispersion (GVD)], which can be
obtained from k00ð�pÞ¼ � ð�2p=2 cÞDslopeð�p � �0Þ, where �0 is the zero-
dispersion wavelength of the DSF, Dslope = 0.06 ps/nm2 km is the experi-
mental value of the dispersion slope in the vicinity of �0. D�s��i is the
21. Fiber-Optic Quantum Information Technologies 843
central-frequency difference between signal and idler fields. �s and �i are related to
!s and !i through the following relation: �s = !s��s, �i = !i��i.
In lieu of giving the detailed derivation of the two-photon state (which is
lengthy), we highlight several noteworthy mathematical maneuvers along the
way. The following identification of the Dirac �-function is useful in handling
the time integral:
Z 1�1
eið!þ!0�2�p��p��0pÞtdt = 2�ð!þ !0 � 2�p � �p � �0pÞ; (21.17)
which reinforces the energy conservation requirement in the FPS process. The
volume integralRdV is reduced to a length integral
Rdz by usingR R
dxdy! Aeff, which is a valid approximation for single spatial-mode
propagation and interaction in optical fibers. Taylor expansion of the various
wave-vector magnitudes kp, ks, ki around the pump central frequency �p has
been used to simplify their relationship. In terms of the mathematical structure
of the two-photon spectral function, we note that the GVD term k00(�p) as well
as the pump SPM term Ppz play important roles in shaping the two-photon
state, in contrast with the observation that the pump SPM term is virtually
nonexistent in the �(2)-generated two-photon states. The appearance of the
pump SPM is therefore a unique signature of the �(3) two-photon state, when
comparing with its �(2) counterparts.
In this and the next section, we will make use of the previously derived
formulas for the two-photon state [Eqns (21.14)–(21.16)] to obtain the photon-
counting formulas for the single channels as well as for the coincidences. The
mathematics involved for the two cases are similar to each other, so it suffices to
present a detailed version for the former. The signal-band single-photon counting
rate can be calculated using the following formula [36]:
Sc =
Z 10
h�jEð�Þs EðþÞs j�i dT: (21.18)
It is obvious that an analogous approach can be applied to the idler band as well.
As Sc denotes single-photon counting probability for one pump pulse, it is by
definition a dimensionless quantity. It is customary, in this case, to use the photon-
number unit for the electric field operator [40]. In this unit, the electric field
operator has dimensionality 1=ffiffiffiffiffiffiffisecp
, as shown below:
EðþÞs =X
ks
ffiffiffiffiffiffiffiffiffiffic Aeff
4VQ
saks e
�i!s te�½ð!s��sÞ2=2 20; (21.19)
844 Prem Kumar et al.
where the Gaussian filter in front of the detector has been included. The integrand
in Eqn (21.18) can be written as
h�jEð�Þs EðþÞs j�i ¼c Aeff
4VQ
Xki;k
0i
h0jakiayk0ij0i X
k1;k2;ks;k0s
h0jaksayk1
ak2ayk0sj0iei!s t
� e�½ð!s��sÞ2=2 20e�i!
0s te�½ð!
0s��sÞ2=2 2
0 F�ðks; kiÞFðk0s; k0iÞ:
(21.20)
Nonvanishing results emerge only when the wave vectors observe the following
restrictions:
ki = k0i; ks = k1; k0s = k2: (21.21)
The integrand may be further simplified into
h�jEð�Þs EðþÞs j�i ¼c Aeff
4VQ
Xki
Xks
ei!s te�½ð!s��sÞ2=2 20F�ðks; kiÞ
" #
�½Xk0s
e�i!0s te�½ð!
0s��sÞ2=2 2
0Fðk0s; kiÞ
=c Aeff
323uð!iÞu2ð!sÞ
Zd!i
Zd!s e
�i!ste�½ð!s��sÞ2=2 20Fð!s; !iÞ
���� ����2, (21.22)
where in the last step we have invoked the following identity to transform wave-
vector summations into angular frequency integrals:
Xkj
!V
1=3Q
2
Zdkj =
V1=3Q
2
Zd!j
uð!jÞ: (21.23)
Here u(!j) = d!j/dkj, j = s, i, is the group velocity of the jth mode, and is to be
taken as a constant c/n in our simplified calculation.
Equation (21.18) can be written in the following form after all the above steps
have been absorbed:
Sc ¼2�2½�ð3Þ2 Aeff P2
p
16 �20 V2
Q n3 c2 �2p
2p
Z 0
�L
dz1
Z 0
�L
dz2
e�2iPpðz1�z2Þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið1� ik00 2
pz1Þð1þ ik00 2pz2Þ
q Zd�s
Zd�i
� exp �ð�s þ �iÞ2
2 2p
� �2s
20
� ik00
4ðz1 � z2Þð�s � �i þ DÞ2
( ): (21.24)
21. Fiber-Optic Quantum Information Technologies 845
The frequency double integral can be analytically integrated through a change of
variables and completion of squares, namely, let
�þ=�s þ �i
2; (21.25)
��= �s � �i: (21.26)
The frequency double integral can be rewritten in terms of the new variables asZd�s
Zd�i e
�½ð�sþ�iÞ2=2 2p � �
2s
20
� ik00
4ðz1 � z2Þð�s � �i þ DÞ2
¼Zd�þ exp �
2 20 þ 2
p
20
2p
�þ þ 2p��
2ð2 20 þ 2
pÞ
" #28<:
9=;Zd�� exp �
�2�
2ð2 20 þ 2
pÞ� ik00ðz1 � z2Þð�� þ DÞ2
4
( ): (21.27)
The first part of the integral, concerning only Gaussian functions with real variable
as arguments, is easily integrated as
Zd� þ exp �
2 20 þ 2
p
20
2p
�þ þ 2p��
2ð2 20 þ 2
pÞ
" #28<:
9=; ¼ffiffiffip
p 0ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 2
0 þ 2p
q : (21.28)
The second part of the integral, having a Gaussian function with complex argu-
ment as integrand, has a closed analytical form by using the integral formula from
Ref. [41], i.e.,Zd� � exp � u2
�2ð2 2
0 þ 2pÞ� ik00ðz1 � z2Þð�� þ DÞ2
4
( )
¼ffiffiffipffiffiffi
ap ffiffiffiffiffiffiffiffiffiffiffiffiffi
1þ b24p exp � cb2
1þ b2þ i
2arctanðbÞ þ ir
1þ b2
� �; (21.29)
where
a =1
2ð2 20 þ 2
pÞ; b ¼ �
k00ðz1 � z2Þð2 20 þ 2
pÞ2
;
c =D2
2ð2 20 þ 2
pÞ; and r = � k00ðz1 � z2ÞD2
4:
846 Prem Kumar et al.
We therefore obtain the following final form of the single-photon counting
formula:
Sc = A1ðPpLÞ2 0
pIsc; (21.30)
A1 =�2 n A3
eff
18ffiffiffi2p
V2Q
; (21.31)
Isc¼1
L2
Z 0
�L
dz1
Z 0
�L
dz2
exp �2iPpðz1�z2Þ�cb2
1þb2þ i
2arctanðbÞþ ir
1þb2
� �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið1�ik00 2
pz1Þð1þik00 2pz2Þ
q ffiffiffiffiffiffiffiffiffiffiffiffi1þb24p ; (21.32)
where A1 is an unknown constant with � and VQ as fitting parameters, and Isc is a
double-length integral that has to be investigated numerically. Despite the seemingly
complicated form of the single-counts formula, the physics behind it is quite clear.
Apart from a small contribution from the double integral, the single counts scale
quadratically with pump power, which coincides with the intuitive FPS picture that
requires two pump photons to scatter into the signal/idler modes. It also scales
linearly with the ratio of the filter bandwidth to pump bandwidth. This makes
sense in that if one broadens the filter bandwidth, more photons will be collected;
and conversely if the filter bandwidth is narrowed, one would expect to count less
photons. The dependence on pump bandwidth is more clearly seen in the time
domain. As the pulse width becomes wider (thus the pump bandwidth narrower)
while maintaining the peak power to be the same, the probability of FPS increases
linearly with pulse width (thus decreases linearly with pump bandwidth) simply
because there is more time for the pump photons to interact; the reverse is also true.
The more intricate dependence on pump power, pump bandwidth, and filter band-
width is described by the double integral Isc, which takes into account phase match-
ing, SPM of the pump, and the Gaussian shapes of pump and filter spectrum.
Calculations of the coincidence-counting rate with Gaussian filters can be
performed in a similar way to those of the single counting rate. We start with
the probability of getting a coincidence count for each pulse [36]:
Cc =
Z 10
dT1
Z 10
dT2h�jEð�Þ1 Eð�Þ2 E
ðþÞ2 E
ðþÞ1 j�i: (21.33)
The electric fields are free fields propagating through Gaussian filters evaluated at
detectors 1 and 2, defined in the photon-number unit:
EðþÞ1 ¼
Xk1
ffiffiffiffiffiffiffiffiffiffic Aeff
4 VQ
sak1
e�i!st1 e�½ð!s��sÞ2=2 20; (21.34)
21. Fiber-Optic Quantum Information Technologies 847
EðþÞ2 ¼
Xk2
ffiffiffiffiffiffiffiffiffiffic Aeff
4 VQ
sak2
e�i!it2e�½ð!i��iÞ2=2 20; (21.35)
where the Gaussian filters take the form fð!j � �jÞ= fje�½ð!j��jÞ2=2 2j , fj = 1 and
j = 0 for j = s, i are assumed to simplify the calculation. ti = Ti� li/c is the
time at which the biphoton wavepacket leaves the output tip of the fiber, which in
our case is almost the same for the signal and idler as there is negligible group-
velocity difference between the two closely spaced (in wavelength), copolarized
fields. li denotes the optical path length from the output tip of the fiber to the
detector i, i = 1, 2, and can be carefully path matched to be the same.
The integrand in Eqn (21.33) can be written in the following form:
h�jEð�Þ1 Eð�Þ2 E
ðþÞ2 E
ðþÞ1 j�i ¼ jh0jE
ðþÞ2 E
ðþÞ1 j�ij
2
¼ jAðt1; t2Þj2; (21.36)
where A(t1, t2) is the biphoton amplitude introduced in Refs [31, 34]. While the
concept of a biphoton amplitude plays an important role in the study of frequency
and wave-number entanglement inherent in the two-photon state, it serves merely
as a calculational shorthand for our purpose in determining the coincidence
counting rate. It is straightforward to show that
Aðt1; t2Þ=cAeff
4 VQ
Xks;ki
Fðks; kiÞe�ið!st1þ!it2Þe�½�2sþ�2
i=2 2
0: (21.37)
The fact that the biphoton cannot be written as a function of t1 times a function of
t2 may be readily observed from the form that Eqn (21.37) takes. It is also
nonfactorable in the wave numbers ks and ki, displaying its entangled nature in
those degrees of freedom. However, it is not entangled in polarization, due to the
fact that all the fields involved are collinear with respect to one another and the
polarization states can be factored out.
When everything is taken into account, after some similar steps shown in
Section 21.4, we arrive at the following form for the coincidence counting
formula:
Cc = A2ð PpLÞ2 20
p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2p þ 2
0
q Icc; (21.38)
A2 =�2 n2 A4
eff
144 V8=3Q
; (21.39)
848 Prem Kumar et al.
Icc¼1
L2
Z 0
�L
dz1
Z 0
�L
dz2
exp �2iPpðz1� z2Þ�ðc0b02=1þb02Þþði=2Þarctanðb0Þþðir0=1þb02Þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1� ik00 2pz1Þð1þ ik00 2
pz2Þq ffiffiffiffiffiffiffiffiffiffiffiffi
1þb24p ; (21.40)
where
b0= � ½k00ðz1 � z2Þ 20=2; c0= ðD2=2 2
0Þ; and r0= � ½k00ðz1 � z2ÞD2=4:
From Eqns (21.30) and (21.38), we can see that the single counts and the
coincidence counts both scale quadratically with the pump peak power. This is a
distinct feature of the �(3) interaction, in contrast to the linear dependence on pump
power in �(2) SPDC. Whereas one might expect to see an exact linear relation
between the single counts and the coincidence counts under ideal detection con-
ditions (unity quantum efficiency of the detectors, no loss, no dark counts), the
linearity is absent due to the broadband nature of the pump field and the presence
of the filters. Some of the correlated twin photons are lost during the filtering
process, and some uncorrelated photons are detected instead. The explicit depen-
dence of Cc on the quantity 20= p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2p þ 20
qcan be understood from its limiting
cases. When the pump bandwidth is wide compared with the filter bandwidth, i.e.,
p= 0, every individual frequency component of the pump spectrum will
generate its own energy-conserving signal/idler pairs. The filters, being narrow,
are only effective at collecting a small portion of the correlated photons. There-
fore, the coincidence counts should be proportional to 20= 2p. However, if the
pump bandwidth is sufficiently narrow, i.e., p� 0, the photons being filtered
(and subsequently collected by the detectors) are more likely to be correlated with
each other, in which case the coincidence counts should scale with 0= p. Both
cases are verified when we look at the asymptotic limits:
lim p� 0
20
p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2p þ 2
0
q = 2
0
2p
; (21.41)
lim p� 0
20
p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2p þ 2
0
q = 0
p: (21.42)
To pin down the unknown parameters � and VQ in Eqns (21.31) and (21.39), we
fit our theory to two sets of experimental data, where the ratio of pump bandwidth
to filter bandwidth is varied. The commonly used least-squares fitting technique
has been employed. The results are shown in Figure 21.6, where the central solid
curve corresponds to the optimum fitting parameters, which are determined to be
21. Fiber-Optic Quantum Information Technologies 849
� = 0.237 and VQ = 1.6 � 10�16 m3. k00 has also been found to be �0.116 ps2/km,
corresponding to the wavelength difference �p��0 = 1.52 nm, which agrees well
with the measured experimental value. We also show the robustness of the fit by
perturbing either one of the fitting parameters around its optimum value by as
small as 5%. For example, the dotted curve corresponds to the case where we
set VQ = 1.5 � 10�16 m3 while keeping � optimum, and the dot-dashed curve
corresponds to the case where we set VQ = 1.7 � 10�16 m3 while keeping �optimum. The remaining two curves are generated when we keep VQ optimum
and set � = 0.250 (short-dashed curve), or � = 0.220 (long-dashed curve), respec-
tively. The large discrepancies between the experiment and the theory induced by
this operation are shown in the same figure, which boosts our confidence in the
correctness of the theory.
We have provided a detailed discussion of the two-photon state originating
from the third-order nonlinearity in optical fibers. This �(3) two-photon state shares
Pump bandwidth = 0.8 nm
Filter bandwidth = 0.8 nm
Pump bandwidth = 0.45 nm
Filter bandwidth = 0.8 nm Pump bandwidth = 0.45 nm
Filter bandwidth = 0.8 nm
Pump bandwidth = 0.8 nm
Filter bandwidth = 0.8 nm
00 0.5 1 1.5
Pp (W)2 2.5
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
Sc
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Sc
Cc
00.0050.01
0.0150.02
0.0250.03
0.0350.04
0.045
Cc
(a)
0 0.5 1
Pp (W)1.5
(c)
0 0.5 1
Pp (W)1.5
(d)
0 0.5 1 1.5
Pp (W)2 2.5
(b)
Figure 21.6 Experiment vs theory: squares correspond to experimental data and the curves correspond to
theoretical predictions. (a) Single counts with p = 0.8 nm and 0 = 0.8 nm; (b) Coincidence counts with
p = 0.8 nm and 0 = 0.8 nm; (c) Single counts with p = 0.45 nm and 0 = 0.8 nm; (d) Coincidence
counts with p = 0.45 nm and 0 = 0.8 nm. The central solid curve represents the theoretical fit with the
optimum fitting parameters (� = 0.237, VQ = 1.6 � 10�16 m3), whereas the other curves correspond to
fits with nonoptimum fitting parameters: dotted, � = 0.237, VQ = 1.5 � 10�16 m3; dot-dashed,
� = 0.237, VQ = 1.7 � 10�16 m3; short-dashed, � = 0.250, VQ = 1.6 � 10�16 m3; long-dashed,
� = 0.220, VQ = 1.6 � 10�16 m3 (this figure may be seen in color on the included CD-ROM).
850 Prem Kumar et al.
some similar features with the �(2) two-photon state generated from SPDC, yet it
also has some distinct characteristics. Coincidence photon-counting rate, which is
a significant nonclassical figure of merit of the two-photon state, has been shown
to depend heavily upon various experimental parameters. The dependence on the
ratio of the pump bandwidth to filter bandwidth is of practical importance, because
it serves as a guideline for optimizing the measurement of coincidence counts.
Single-photon counting rate has also been studied, and both fit to the experimental
data reasonably well. While in this chapter we are only concerned with parametric
fluorescence from a single pump pulse, the current theory can be readily extended
to include multiphoton-state generation from one pulse [42], and multiple two-
photon-states generation from adjacent pulses [38, 39] to study polarization entan-
glement. The effect of SRS can also be included in our model by taking into
account the noninstantaneous nature of the third-order nonlinearity in optical fiber.
21.4 FIBER NONLINEARITY AS A SOURCE FORENTANGLED PHOTONS
Quantum entanglement refers to the nonclassical interdependency of physically
separable quantum subsystems. In addition to being at the heart of the most
fundamental tests of QM [43–46], it is an essential resource that must be freely
available for implementing many of the novel functions of QIP [47, 48]. In
photonic systems, the ongoing developments in lasers, optical-fiber technology,
single-photon detectors, and nonlinear optical materials have led to enormous
experimental progress in both the fundamental [49–53] and applied domains
[54–56]. A popular approach to generating entangled pairs of photons is based
on the nonlinear process of parametric down-conversion in �(2) crystals [57–59].
Although much progress has been made using this approach, formidable engineer-
ing problems remain in coupling the entangled photons into standard optical fibers
[60] for transmission, storage, and manipulation over long distances.
The coupling problem can be obviated if the entangled photons can be gener-
ated in the fiber itself and, desirably, in the fiber’s low-loss propagation window
near 1.5 mm, since that would minimize losses during transmission as well. Apart
from the inherent compatibility with the transmission medium, a fiber-based
source of entangled photons would have other advantages over its crystal counter-
parts [57, 58, 59, 61–63]. Particularly, the spatial mode of the photon pair would be
the guided transverse mode of the fiber, which is a very pure Gaussian-like single
spatial mode in modern fibers. A well-defined mode is highly desirable for
realizing complex networks involving several entangling operations. In this chap-
ter, we describe the first, to the best of our knowledge, optical fiber source of
polarization-entangled photon pairs in the 1550-nm telecom band. A variety of
biphoton interference experiments are presented that show the nature of the
entanglement generated with this source. All four Bell states can be prepared
21. Fiber-Optic Quantum Information Technologies 851
with our setup and the CHSH form of Bell’s inequality is violated by up to 10
standard deviations of measurement uncertainty.
Recently, our group has demonstrated that parametric fluorescence accompa-
nying nondegenerate FWM in standard optical fibers is an excellent source of
quantum-correlated photon pairs [64, 65]. The quantum correlation arises from
FPS events, wherein two-pump photons at frequency !p scatter through the Kerr
nonlinearity of the fiber to simultaneously create a signal photon and an idler
photon at frequencies !s and !i, respectively, such that !s þ !i = 2!p. For a
linearly polarized pump with wavelength close to the zero-dispersion wavelength
of the fiber, the FWM process is phase-matched and the accompanying parametric
fluorescence is predominantly copolarized with the pump. Two such parametric
scattering processes can be time and polarization multiplexed to create the desired
polarization entanglement. For example [see Figure 21.7(a)], when the fiber is
pumped with two orthogonally polarized, relatively delayed pulses, the signal/idler
photon pairs scattered from each pulse are copolarized with that pump pulse and
relatively delayed by the same amount. The distinguishing time delay between the
orthogonally polarized photon pairs, however, can be removed by passing the pairs
through a piece of birefringent fiber of appropriate length, wherein the photon-pair
traveling along the fast axis of the fiber catches up with the other pair traveling
along the slow axis. When the emerging signal and idler photons are separated
based on their wavelength, each stream of photons is completely unpolarized
because any polarizer/detector combination is unable to determine which pump
pulse a detected photon originated from. When the signal and idler photons are
detected in coincidence, it is still impossible to determine which pump pulse
created the detected pair. This indistinguishability gives rise to polarization entan-
glement in our experiment.
A schematic of the experimental setup is shown in Figure 21.7(b). Signal and
idler photon pairs at wavelengths of 1547.1 and 1525.1 nm, respectively,
are produced in an NFSI [65, 66]. The NFSI consists of a fused-silica 50/50
fiber coupler spliced to 300 m of DSF that has a zero-dispersion wavelength at
�0 = 1535 – 2 nm. Because the Kerr nonlinearity is weak, for this length of fiber
only about 0.1 photon-pair is produced with a typical 5-ps duration pump pulse
containing �107 photons. Thus, to reliably detect the correlated photon pairs, a
pump-to-signal rejection ratio in excess of 100 dB is required. We achieve this by
first exploiting the mirror-like property of the Sagnac loop, which provides a pump
rejection of >30 dB, and then sending the transmitted fluorescence photons along
with the leaked pump photons through a free-space double-grating spectral filter
(DGSF) that provides a pump rejection ratio in excess of 75 dB [65]. The doubly
diffracted signal and idler photons are then recoupled into fibers, whose numerical
apertures along with the geometrical settings of the gratings determine the pass
bands for the signal and idler channels. The FWHM bandwidth for both the
channels is 0.6 nm.
During the experiment, for alignment and phase-control purposes, input signal
and reference pulses are also needed that are temporally synchronized with the
852 Prem Kumar et al.
Signal
0 20 40 60 80 100 120 140 160 180 200
Time (s)
Ref.
(a)
(b)
Delay
PM fiber
Hs Hi
Vi
Vs
f
sPH
PV
Temporally delayedorthogonally polarized
pump pulsesTemporally delayed
orthogonally polarizedsignal/idler pairs Polarization-entangled
signal/idler pairs
is
iφis
HHeHHψ
χ(3) FWM
OP
O
APD1
G1
M4M3
Delay
QWP1
PZT
QWP2
Pump
300mDSF Loop
Out
FPC1
EDFA 90/1050/50
FPC2
4% BS
HWP2 &QWP3
F
APD2
Countingsystem
Sig. Det.
Ref. Det.
PM fiber
Signal in
Ref. light
50/50
FPC4
FPC3
P1
P3
P4 &HWP3
G2
P5 &HWP4
M5
M2
M1 G3
G4 P2
HWP1
(c)
Figure 21.7 (a) Conceptual representation of the multiplexing scheme used to create polarization
entanglement. (b) Schematic of the experimental setup. P1–P5, polarization beam splitters; G1–G4,
diffraction gratings; M1–M5, mirrors; FPC1–FPC4, fiber polarization controllers; QWP, quarter-wave plate;
HWP, half-wave plate; F, flipper mirror. (c) Sinusoidal variations (or constancy at the peaks and troughs) of
the photocurrents obtained from the signal (top traces) and the reference detectors (bottom traces) upon
linearly sweeping the voltage (or maintaining a fixed voltage) on the PZT, piezoelectric transducer. The
clarity of the traces demonstrates minutes-long stability of the polarization interferometer formed between P1
and P3 (P2) for signal (reference) light (this figure may be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 853
pump pulses. The main purpose of the signal pulses is to ensure that the time
distinguishability between the orthogonally polarized photon pairs is effectively
removed. By spectrally carving [66] the �150-fs pulse train from an optical
parametric oscillator (OPO) [Coherent Inc., model Mira-OPO], we obtain trains
of 5-ps pump pulses, 2.8-ps signal pulses, and 4-ps reference pulses at central
wavelength of 1536, 1547, and 1539 nm, respectively. The pump pulses are then
amplified by an erbium-doped fiber amplifier (EDFA) to achieve the required
average pump power. Light at the signal and idler wavelengths from the OPO that
leaks through the spectral-carving optics and the amplified spontaneous emission
(ASE) from the EDFA are suppressed by passing the pump pulses through a 1-nm
bandwidth tunable optical filter (Newport, TBF-1550-1.0).
A 30-ps relative delay between the two orthogonally polarized pump pulses is
introduced by adding separate free-space propagation paths for the two pulses with
use of a polarization beam splitter (PBS) P1, quarter-wave plates (QWP) QWP1
and QWP2, and mirrors M1 and M2. Mirror M2 is mounted on a piezoelectric
transducer (PZT)-driven translation stage, which allows precise adjustment of the
relative delay and phase difference between the orthogonally polarized pump-
pulse pairs. After the NFSI, the delay is compensated by propagating the scattered
photon pairs along the fast- and slow-polarization axes of a 20-m-long polariza-
tion-maintaining (PM) fiber. A careful alignment procedure is implemented to
properly orient the axes of the PM fiber, taking into consideration the change of
polarization state incurred by an input signal-pulse pair upon maximally amplified
reflection from the NFSI [67]. Alignment is performed prior to the actual experi-
ment by injecting weak path-matched signal-pulse pairs, having identical temporal
and polarization structure as the pump pulses, into the NFSI through the 50/50 and
90/10 couplers. First the signal amplification is maximized by adjusting FPC2,
while monitoring the signal gain on a detector (ETX500) placed after P3. Then the
fringe visibility of the polarization interferometer formed between P1 and P3 is
maximized by adjusting FPC3, HW2, and QWP3, while observing the fringes in
real time upon periodic scanning of M2. Once the alignment is completed, the
injected signal is blocked and further measurements are made only on the para-
metric fluorescence.
After compensation of the time delay, the following polarization-entangled
state is generated at the output of the PM fiber: j�i = jHisjHii þ e2i�pjVisjVii,where �p is the relative phase difference between the two delayed, orthogonally
polarized pump pulses. This source can produce all four polarization-entangled
Bell states. When �p = 0, /2, the states j�–i = jHisjHii – jVisjVii are created.
The other two Bell states j�–i = jHisjVii – jVisjHii can be prepared by insert-
ing a properly oriented HWP in the idler channel. Nonmaximally entangled
pure states with an arbitrary degree of polarization entanglement can also be
created with our setup by choosing the two-pump pulses to have unequal
powers.
To actively monitor and control the relative phase �p during the course of data-
taking, weak reference-pulse pairs of about 50 mW average power are injected into
854 Prem Kumar et al.
the NFSI through the 50/50 and 90/10 couplers. The reference-pulse pairs have
identical temporal and polarization structure as the pump pulses, except the
temporal location of the reference-pulse pairs is mismatched with respect to the
pump-pulse pairs and their wavelength is slightly detuned, so that they neither
interact with the pump pulses nor are seen by the single-photon detectors used in
the signal and idler channels. During the course of measurements on the polariza-
tion-entangled states, the relative phase between the reference-pulse pairs, �ref, is
monitored by measuring the photocurrent from a low-bandwidth reference detector
placed after P2 to make observations on one output port of the polarization
interferometer [see Figure 21.7(b)]. The voltage created by this photocurrent is
compared to a reference voltage and the difference is used to stabilize �ref by
feeding back on the PZT through an electronic circuit. The excellent overall
stability of the system is shown by the near-perfect classical interference fringes
displayed in the inset in Figure 21.7(c), which were simultaneously obtained with
injected signal light and with reference light while scanning �ref by ramping the
voltage on the PZT. The relative phase between the reference-pulse pairs, �ref, is
related to the relative phase between the pump-pulse pairs through �p = �ref þ �,where � results from dispersion in the DSF owing to slightly different wavelengths
of the pulse pairs.
The photon-counting modules used for detecting the signal and idler photons
consist of InGaAs/InP APDs (Epitaxx, EPM 239BA) operated in a gated Geiger
mode [65]. The measured quantum efficiencies for the two detectors are 25% and
20%, respectively. The overall detection efficiencies for the signal and idler
photons are about 9% and 7%, respectively, when the transmittance of the
Sagnac loop (82%), 90/10 coupler, DGSF (57%), and other optical components
(90%) are included. Given a parametric scattering probability of ’0.1 pairs/pulse
in the DSF, corresponding to 0.39 mW of average pump power in each direction
around the Sagnac loop, and the gate rate of 588 kHz, we typically observe
’4000 counts/s in the signal and idler channels when detecting the parametric
fluorescence.
The polarization-entanglement generation scheme described here uses the fact
that the FPS efficiency does not depend on the pump-polarization direction. We
verify this by monitoring the parametric fluorescence while varying the polariza-
tion direction of the injected pump pulses with use of a half-wave plate (HWP1).
The individual counts for the signal and idler photons, and their coincidence
counts, vs the HWP1 angle are shown in Figure 21.8(a). The slight variation
observed in the count rates is due to polarization-dependent transmission of the
DGSF. Note that for the measurements shown in Figure 21.8(a) the input pump
delay, the PM-fiber delay compensation, and the detection analyzers were
removed.
Polarization correlations are measured by inserting adjustable analyzers in the
paths of signal and idler photons, each consisting of a PBS (P4, P5) preceded by an
adjustable HWP (HWP3, HWP4). For the state j�i = jHisjVii þ ei2�pjVisjHii,when the polarization analyzers in the signal and idler channels are set to �1 and
21. Fiber-Optic Quantum Information Technologies 855
–4.71 –3.14 –1.57 0 1.57 3.14
Relative phase φ ref (rad)
Ref
. det
. out
put (
a.u.
)
0
4
8
12
–4.71 3.14
0
40
80
120
0
1
2
3
4
5
0 50 100 150 200HWP1 orientation (°)
Sin
gle
coun
ts(×
103 /
s)
Single counts(×
103/20
s)
Single counts(×
103/30
s)
0
50
100
150 Coincidences (s
–1)
–0.2
1.8
3.8
5.8
7.8
–50 150 350
Relative angle θ1 – θ2 (°)
Coi
ncid
ence
s(×
102 /2
0 s)
Coi
ncid
ence
s(×
102 /
30 s
)
0
20
40
60
80
(b)
δ
(c)
(a)
(d)
Figure 21.8 (a) Observed polarization (in)dependence of parametric fluorescence in the DSF. (b) Coincidence counts and single counts detected over 30 s when the
relative phase fref is varied. The solid curve is a fit to Eqn (21.43). (c) Output from the reference detector vs fref showing the ordinary one-photon interference with
twice the fringe spacing as in (b). (d) Measurement of polarization entanglement: Coincidence counts and single counts detected over 20 s as the analyzer angle in
the idler channel is varied while keeping the signal-channel analyzer fixed at 45 relative to vertical (this figure may be seen in color on the included CD-ROM).
85
6P
remK
um
aret
al.
�2, respectively, the single-count probability for the signal and idler photons is
Ri = �i/2 (i = 1, 2) and the coincidence-count probability R12 is given by
R12 ¼ 2�1�1�2½sin2 �1 cos2 �2 þ cos2 �1 sin
2 �2
þ 2 cosð2�pÞ sin �1 cos �1 sin �2 cos �2; (21.43)
where �i is the total detection efficiency in each channel.
We performed three sets of experiments to evaluate the degree of polarization
entanglement of our source. The first measurement consisted of setting both
analyzers at 45 and slowly scanning �ref by applying a voltage ramp on the
PZT. As shown in Figure 21.8(b), the coincidence counts reveal sinusoidal varia-
tion with a fringe visibility of 93% (dark counts and accidental-coincidence counts
have been subtracted), while the single counts remain unchanged. The output
from the reference detector is also recorded simultaneously, which is shown in
Figure 21.8(c). The relative shift of the sinusoidal variation of two-photon inter-
ference in Figure 21.8(b) from that of reference-light interference in Figure 21.8(c)
is a direct measure of the phase shift �, which is used below to properly set �p for
measurements of the violation of Bell’s inequality.
In the second set of measurements on polarization entanglement, we locked the
generated state to j��i = jHisjVii� jVisjHii by applying an appropriate feedback
on the PZT, fixed the angle of the polarization analyzer in the signal channel to
45 , and varied the analyzer angle in the idler channel by rotating HWP4. The
result is shown in Figure 21.8(d). As expected, the coincidence-count rate displays
sinusoidal interference fringes with a visibility of 92%, whereas the variation in
the single-count rate is only 4% (once again, dark counts and accidental coin-
cidences have been subtracted).
In the third set of experiments, we characterized the quality of polarization
entanglement produced with our source through measurements of Bell’s
inequality violation. By recording coincidence counts for 16 different combina-
tions of analyzer settings with �1 = 0 , 90 , �45 , 45 and �2 = � 22.5 , 67.5 ,22.5 , 112.5 , we measured the quantity S in the CHSH form of Bell’s inequal-
ity [46], which satisfies jSj � 2 for any local realistic description of our experi-
ment. The results, which are presented in Table 21.1, show that (a) the CHSH
Table 21.1
Measured values of S for the four Bell states.
Bell state S Violation ( )
jHisjHii þ jVisjVii 2.75 – 0.077 10
jHisjHii� jVisjVii 2.55 – 0.070 8
jHisjVii þ jVisjHii 2.48 – 0.078 6
jHisjVii� jVisjHii 2.64 – 0.076 8
21. Fiber-Optic Quantum Information Technologies 857
inequality is violated, i.e., jSj > 2, for all four Bell states produced with our
setup and (b) the violation occurs by up to 10 standard deviations ( ) of
measurement uncertainty.
To ascertain the degree of entanglement produced by the true FPS events in our
setup, the accidental coincidences resulting from the uncorrelated background
photons and the dark counts in the detectors were measured for each set of data
acquired in the three polarization-entanglement experiments described above. The
rate of accidental coincidences was as large as the rate of “true” coincidences,
plotted in Figure 21.8 by subtracting the accidental coincidences, and the raw
visibility of TPI was only ’30%. We believe, the majority of background photons
in our setup arise from SRS as verified by our recent measurement of the noise
figure of fiber-optic parametric amplifiers [68–70]. Our recent measurement with a
modified DGSF has shown that the contribution of accidental coincidences can be
made <10% of the total measured coincidences [64]. With these improvements, a
raw TPI visibility of >85% would be obtained, i.e., without any postmeasurement
corrections.
In conclusion, we have developed and characterized a fiber-based source of
polarization-entangled photon pairs. The pair-production rate can be dramati-
cally increased by using state-of-the-art pulsed lasers that have been developed
for fiber-optic communications. These lasers operate at 10–40-GHz repetition
rates and can have the requisite peak-pulse powers with use of medium-power
EDFAs. Bulk-optic implementations of the pump delay apparatus and the
detection filters were used in these proof-of-principle experiments for purposes
of tunability and control. All-fiber versions of these subsystems can be readily
realized with use of PM fibers, wavelength-division-multiplexing filters, and
fiber polarizers. Finally, we have understood the origin of the large number of
accidental coincidences in the experiment and subsequent system improvements
are expected to significantly improve the degree of entanglement produced with
our system. Therefore, we believe that such fiber-based entangled-photon pairs
will prove to be an efficient source for developing quantum communication
technologies.
21.5 HIGH-FIDELITY ENTANGLEMENT WITHCOOLED FIBER
For many QIP applications it is desirable to produce entangled photon-pairs at
telecom wavelengths directly in the fiber by use of the fiber’s Kerr nonlinearity.
Previously, our group has developed the use of an NFSI to generate quantum-
correlated twin beams [66], correlated twin-photon pairs [71], and polarization
entanglement [72–74]. We have pointed out that SRS, which gives rise to the
majority of background photons, prevents us from observing TPI with unit visibi-
lity [74–76]. In these experiments, we have reduced the contribution of Raman
858 Prem Kumar et al.
photons by lowering pump power, selecting signal/idler photon-pairs at small
detuning from the pump (’5 nm), and using polarizers to remove the cross-
polarized Raman photons. In this chapter, we present our measurements of the
ratio between coincidence counts (coincidences arising from the same pulse) and
accidental-coincidence counts (coincidences arising from adjacent pulses) as a
function of pump power, and with the DSF at three different temperatures: ambient
(’300 K), dry ice (195 K), and liquid nitrogen (77 K). Then, we create polarization
entangled photon-pairs with the fiber at the above three temperatures using a
compact counterpropagating scheme (CPS) [77, 78]. We observe TPI with visibi-
lity >98% and Bell’s inequality violation by >8 standard deviations of measure-
ment uncertainty at 77 K.
As shown in Figure 21.9, the pump is a mode-locked pulse train with pulse
duration ’5 ps and repetition rate of 75.3 MHz, which is obtained by using a
diffraction grating to spatially disperse the output of an optical parametric oscil-
lator (Coherent Inc., model Mira-OPO). The pump central wavelength is set at
1538.7 nm. To achieve the required power, the pump pulses are further amplified
by an EDFA. The ASE from the EDFA is suppressed by passing the pump through
a tunable filter (Newport, model TBF-1550-1.0) with 1-nm FWHM passband.
A fused-silica 90/10 fiber coupler is used to split 10% of the total pump power
to a power meter for monitoring the power and stability of the pump pulses. The
remaining 90% of the pump power goes through " fiber polarization controller
(FPC1) and a linear polarizer (LP), whose purpose is to adjust the input pump field
90/10coupler
FPC2
FP
FP
FPC1
FP
FP
LP
HWP1at 22.5°
PBS1
Signal Idler
PV
PH
Fiber spool
Filter>100 dB
L4
L2 L3
L5
Polarization analyzer
PBS
PBS
HWP
QWP
HWP
QWP
D1
D2
Idler = 1543.5 nm
Signal = 1533.9 nm
Figure 21.9 A schematic of the experimental setup. FP, fiber-port; LP, linear polarizer; L2, L3, L4, L5,
fiber-to-free space collimators; PBS, polarization beam splitter; HWP, QWP, half- and quarter-wave
plates; FPC, fiber polarization controller (this figure may be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 859
to be horizontally polarized. A half-wave plate (HWP1) can be inserted in the
pump path to split the pump pulse into horizontally and vertically polarized
components for creating polarization entanglement. Initially, we use a single
pump with horizontal polarization to pump a straight 300 m of DSF whose zero-
dispersion wavelength is at �0 = 1538.7 nm. The fiber is wound on a spool of
9.32 cm diameter. A fiber polarization controller (FPC2) is used to compensate the
bend-birefringence induced changes in the polarization states in the fiber. To
reliably detect the scattered photon-pairs, an isolation between the pump and
signal/idler photons in excess of 100 dB is required. We achieve this by using
two cascaded WDM filters with FWHM of about 1 nm in the signal and idler
channels, which provide total pump isolation greater than 110 dB [74]. The
selected signal and idler wavelengths are 1543.5 and 1533.9 nm, respectively,
corresponding to ’4.8 nm detuning from the pump’s central wavelength. Two
sets of polarization analyzers, each composed of a quarter-wave plate (QWP), a
half-wave plate (HWP), and a PBS are constructed and are individually inserted
into the signal and idler channels. With proper settings of the QWP and HWP in
each channel, the signal/idler photons with horizontal polarization can be made to
arrive at the detectors with negligible loss. Raman photons that are copolarized
with the pump also reach the detectors, while cross-polarized Raman photons are
blocked by the PBSs. The copolarized Raman photons are inevitably detected,
and hence contribute to the background photons in coincidence detection. It is
known that Stokes and anti-Stokes Raman-scattering noise photons are emitted
at a rate proportional to nth þ 1 and nth [79, 86], respectively, where nth =1/[exp(h�/kT)� 1] is the Bose population factor, � is the frequency shift of
Stokes and anti-Stokes from the pump frequency, T is the temperature of the
fiber, h is Planck’s constant, and k is Boltzmann’s constant.
The photon-counting modules consist of InGaAs/InP APDs (Epitaxx, Model
EPM239BA) operated in the gated Geiger mode at room temperature [71].
The 1-ns-wide gate pulses with the FHWM window of about 300 ps arrive at a
rate of 588 kHz, which is downcounted by 1/128 from the original pump
pulses. The gate pulses are adjusted by an electronic delay generator to
coincide with the arrival of the signal and idler photons at the APDs. The
quantum efficiency of APD1 (APD2) is about 25% (20%), with a correspond-
ing dark-count probability of 2.2 � 10�3 (2.7 � 10�3) per pulse. The total
detection efficiencies for the signal and idler photons are about 7% and 9%,
respectively. Because the size of the fiber spool used in our experiment is
smaller than the regular spool provided by Corning, it is convenient to cool
the DSF by dry ice in a small homemade box or by liquid nitrogen in a small
dewar. The loss due to fiber bending and handling, even in the low-tempera-
ture environment, is negligible. The loss due to cooling the fiber is less than
1% (4%) at 195 K (77 K). Cooling the fiber also causes contraction of the
fiber length, which in turn shortens the propagation time for the photon-pairs
to arrive at the detectors. The advancement of the arrival time is 1.6 ns (3.0 ns)
at temperature of 195 K (77 K).
860 Prem Kumar et al.
With the fiber at each temperature, we record coincidence and accidental-
coincidence counts with an integration time of 60 s as the pump power. Note
that accidental coincidence is the measured coincidence counts as the signal
channel is delayed with respect to the idler channel by one pulse period. After
subtracting the detector dark counts, we plot the ratio between the coincidence
and accidental coincidence vs signal (or idler) counts per pulse. The plot at each
temperature has a similar shape as those observed by other groups [81, 82]. The
reason for such shape is not yet clearly understood. At high pump powers, the
ratios are similar for the cooled and uncooled fibers. This may be due to
multiphoton effects or leakage of the pump photons through the filters leading
to increasing accidental-coincidence counts. As the pump power is decreased,
we observe a ratio as high as 111:1 at 77 K and 60:1 at 195 K compared to a
ratio of 28:1 at room temperature (300 K). The reason is that as the fiber is
cooled, the Raman scattering is reduced as given by nth, and leads to a decrease
in the accidental-coincidence counts. The relative increase of the ratio is
commensurate with the temperature dependence of nth, which is about a factor
of 4.5 (1.6) for the fiber at 77 K (195 K) compared to its room temperature
value. At 300 and 195 K, the peaks in the ratio occur at an average pump power
of about 50 mW (133 mW peak power). At 77 K, the peak occurs at a higher
average pump power about 75 mW (200 mW peak power) because of cooling-
induced loss in the fiber. At these peaks, the photon-pair production rate
is ’0.01/pulse.
To further test the fidelity of the generated photon pairs, we create polariza-
tion entanglement with the cooled fiber by using a CPS [77, 78]. The single
horizontally polarized pump pulse is now split into two equally powered, ortho-
gonally polarized components PH and PV by inserting and properly setting the
half-wave plate (HWP1) in front of the polarization beam splitter (PBS1). For
low-FPS efficiencies, where the probability for each pump pulse to scatter more
than one pair is low, the clockwise (CW) and counterclockwise (CCW) pump
pulses scatter signal/idler photon pairs with probability amplitudes jHiijHsi and
jViijVsi, respectively. After propagating through the DSF, these two amplitudes
of the photon pair are then coherently superimposed through the same PBS1.
This common-path polarization interferometer has good stability for keeping
zero-relative phase between horizontally and vertically polarized pumps, and
hence is capable of creating polarization entanglement of the form jHiijHsi þei2�pjViijVsi at the output of PBS1, where the relative pump phase �p is set to 0
by setting HWP1 to 22.5 . The polarization analyzers in signal and idler channels
are used to set the detection polarization angles �1 and �2 of the entangled two-
photon state. In our experiment, we set �1 = 45 , vary �2, and record single
counts for both signal and idler channels as well as coincidence counts between
the two channels for each value of �2. We repeat these measurements with the
fiber spool at 300, 195, and 77 K with corresponding integration times of 10, 50,
and 60 s, respectively. At room temperature, we observe TPI with 91% visibility.
As the fiber is cooled to 195 and 77 K, the TPI visibility increases to 95% and 98%,
21. Fiber-Optic Quantum Information Technologies 861
respectively, as shown in Figure 21.10. All these TPI fringes are obtained when the
pump power is adjusted to match the peak values of the ratio and are recorded
without subtraction of accidental-coincidences. The observed higher visibility at low
temperatures can be attributed to the suppression of Raman photons. The fitting
function used in the above figures is cos2(�1� �2). We believe that the imperfection
in spatial-mode matching of the correlated nondegenerate photon pairs at the PBS1
and the remaining copolarized Raman photons prevent this scheme from achieving
unit-visibility TPI fringes.
0
100
200
300
400
500
0
1 × 104
2 × 104
3 × 104
4 × 104
0
5 × 104
1 × 104
1.5 × 104
2 × 104
2.5 × 104
3 × 104
3.5 × 104
0 50 100 150 200
Coi
ncid
ence
cou
nts
in 5
0 s
0
100
200
300
400
500
600
700
800
Coi
ncid
ence
cou
nts
in 6
0 s
Sig
nal/i
dler
cou
nts
in 6
0 s
Sig
nal/i
dler
cou
nts
in 5
0 s
(a)
Signal
Idler
Two-photoninterference
Signal
Idler
Two-photoninterference
Relative angle θ1 − θ2 (°)
0 50 100 150 200
(b)
Relative angle θ1 − θ2 (°)
Figure 21.10 Two-photon interference with the fiber at (a) 195 K and (b) 77 K. The observed
visibility is about 95% and 98.3%, respectively (this figure may be seen in color on the included
CD-ROM).
862 Prem Kumar et al.
We further confirm the nonlocal behavior of the polarization entangled photon
pairs generated from this source by making Bell’s inequality measurements. For
this purpose, we prepare the singlet state jHiijVsi� jViijHsi (where �p = /2) by
inserting a QWP at 0 after HWP1 and by adding a HWP at 45 (HÐV) in the
signal channel. By recording coincidence counts for 16 different combinations of
analyzer settings with �1 = 0 , 90 , �45 , 45 and �2 = � 22.5 , 67.5 , 22.5 ,112.5 , we measure the quantity jSj in the CHSH form of Bell’s inequality
[46], wherein jSj � 2 holds for any local realistic system. The results are shown
in Table 21.2. At 77 K we measure jSj = 2.76 – 0.09, which amounts to Bell’s
inequality violation by over eight standard deviations of measurement uncertainty.
All these measurements are made without subtraction of the background Raman
photons. Given these achievements, we believe we can reliably implement Ekert’s
QKD protocol with our polarization-entangled photon-pairs source.
21.6 DEGENERATE PHOTON PAIRS FOR QUANTUMLOGIC IN THE TELECOM BAND
The seminal paper by Knill et al. [83] has rekindled considerable amount of interest
in the field of linear optical quantum computing (LOQC). For real applications of
LOQC, single-photon as well as entangled-photon sources are indispensable [84].
Quantum interference arising from two indistinguishable photons, such as the well-
known Hong-Ou-Mandel (HOM) interference [85], lies at the heart of LOQC.
Therefore, it is of great significance to develop photon sources that generate
identical photons in well-defined spatiotemporal modes. Additionally, for distributed
LOQC it is desirable to produce such photons in the 1550-nm telecom band.
21.6.1 Polarization-Entangled Degenerate Photon-PairGeneration in Optical Fiber
Entangled photons in well-defined time slots are usually obtained using pulsed
SPDC in �(2) crystals [86], wherein a high-frequency pump photon (!p) fissions
into two identical daughter photons (!1 = !2 = !p/2). Recently, Fan et al. [87]
Table 21.2
Violation of Bell’s inequality for the state jHiijVsi� jViijHsi.s is the standard deviation in the measurement of jSj.
Temperature (K) jSj Violation ( )
300 2.22 – 0.06 4
77 2.76 – 0.09 >8
21. Fiber-Optic Quantum Information Technologies 863
have also produced identical correlated photons at the mean frequency (!c) of two
pump frequencies (!p1 þ !p2 = 2!c) by using a reverse degenerate FWM process
in a piece of �(3) microstructure fiber (MF). In that experiment, the generated
photons reside in the visible/near-infrared wavelength region (�< 800 nm). The
spatial-mode profile of the identical photons is thus incompatible with standard
single-mode fiber, making them unsuitable for distributed LOQC. Here, we
describe a source useful for distributed LOQC by producing identical photons at
a telecom-band wavelength using standard DSF. As a further endeavor, we use a
novel dual-pump, CPS [78, 88] to make the otherwise identical photons entangled
in polarization. We thus present the first, to the best of our knowledge, telecom-
band, degenerate, correlated/entangled photon source based on a spool of standard
DSF, which constitutes a promising step toward practical implementation of
LOQC using fiber-based devices.
Figure 21.11 depicts our experimental setup. Out of the broadband spectrum of
a femtosecond laser (repetition rate ’50 MHz), we spectrally carve out our desired
pump central wavelengths (�p1 = 1545.95 nm, and �p2 = 1555.92 nm, pulse
width ’ 5 ps) by cascading two free-space double-grating filters (DGF1 and
DGF2, FWHM ’0.8 nm for each passband, see inset in Figure 21.12). An EDFA
is sandwiched in between the two DGFs to provide pump power variability. The
out-of-band ASE photons from the EDFA are suppressed by DGF2. Here we
utilize the same nonlinear process as in Ref. [87], namely, reverse degenerate
(a)
Laser DGF1 EDFA
50/50BS1
DGF2
50/50BS2
Pump
FPC1
FPC2
λp1
λp1
λp2
λp2
(b)
Pump
λp1 λp2
DSF 300 m
OBPF
Coincidencecounter
50/50BS3
λc
FPC3
HWP1
QWP1
LP
HWP2PBS1
HWP3
HWP4
QWP2
QWP3
PBS2
PBS3
APD1APD2
H
V
Figure 21.11 Experimental layout: (a) Pump preparation; and (b) CPS scheme. See text for details (this
figure may be seen in color on the included CD-ROM).
864 Prem Kumar et al.
FWM. The pump central wavelengths are selected such that their mean wave-
length (�c = 1550.92 nm) is located near the zero-dispersion wavelength of the
300-m-long DSF. The two-pump pulses emerging from the second 50/50 beam
splitter (BS2), before being launched into the CPS, have to satisfy the following
criteria to maximize the FWM efficiency and the resulting entanglement in the
two-photon state: (i) overlapped in time, (ii) parallel polarization, and (iii) equal
power. The first criterion is met by careful path matching, and the resulting
overlapped pulses from the unused port of BS2 can be monitored on a high-
speed oscilloscope. The second criterion is satisfied by adjusting the fiber polar-
ization controllers (FPC1 and FPC2) and the half-wave plate/quarter-wave plate
combination (HWP1 and QWP1) such that the power exiting the LP is maximized.
The last criterion is fulfilled by balancing the transmission efficiencies in the two
arms of the cascaded DGFs. SRS in the DSF is suppressed by cooling the DSF to
77 K [89, 90]. Compared with the cascade-MF approach in Ref. [87], this setup is
straightforward and simpler in design.
We first characterize our degenerate correlated-photon source by measuring
temporal coincidences between the two identical photons. We call a “coincidence”
count when the two detectors fire in the same triggered time slot, and an “accidental-
coincidence” count when they fire in the adjacently triggered time slots. In the
literature [87, 89, 90], a figure of merit for such a source has been established, i.e.,
the coincidence-to-accidental ratio (CAR). We abbreviate it as CAR hereafter.
Intuitively, CAR is a measure of the purity of a correlated photon source, and a
high CAR value indicates a relatively high purity of the source in
coincidence basis, i.e., coincidence events due to uncorrelated noise photons are
0
20
40
60
80
100
120
0 0.001 0.002 0.003 0.004 0.005
77 K
300 K
Single counts/pulse
CA
R
–80
–70
–60
–50
–40
–30
–20
–10
0
–150
–100
–50
0
1540 1545 1550 1555 1560Wavelength (nm)
DGFs
OBPF
Rel
ativ
e tr
ansm
issi
onsp
ectr
um (
dB)
Figure 21.12 CAR plotted as a function of single counts/pulse at two different temperatures [77 K
(dots) and 300 K (squares)]. Inset: spectral shapes of the cascaded DGFs and OBPF (this figure may be
seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 865
very rare. To measure the CAR, we set HWP2 so that the two-frequency pump field
maintains its horizontal polarization upon entering the CPS. After propagating
through the DSF, the pump and its accompanying degenerate FWM photons are
made to keep the same polarization by proper adjustment of FPC3. We use an optical
bandpass filter (OBPF) composed of two cascaded 100-GHz-spacing wavelength-
division-multiplexing filters (FWHM ’0.3 nm, see Figure 21.12, inset) with a
central wavelength of �c at the output of the CPS to collect the correlated FWM
photon pairs, and to provide the >100 dB pump isolation needed to effectively
detect those photons. The OBPF is followed by a 50/50 beam splitter (BS3) to
probabilistically split the two daughter photons. We are only interested in the cases
where the two daughter photons split up (which happen with 50% probability) to
give coincidence counts; the other equally probable cases where the two photons
bunch together are not recorded in coincidence detection. Nevertheless, the latter
contribute to the single-count measurement results. For this part of the experiment,
the polarization analyzers (HWP3/QWP2/PBS2 and HWP4/QWP3/PBS3) shown in
Figure 21.11 are taken out, as the polarization properties of the detected photons are
not of concern in this part of the experiment. Each photon is directed to a fiber-
coupled APD1 and APD2 (Epitaxx EPM 239BA) operating in the gated Geiger
mode, whose detection results are recorded and analyzed by a “coincidence
counter” software. For the measurement reported in this chapter, the detection
rate is 1/64 of the laser’s repetition rate (’780 kHz), which is limited by the
electronics used to arm and trigger the detectors. The overall efficiencies of the
two detectors, including propagation losses and detector quantum efficiencies, are
7% and 9%, respectively. We subtract detector dark-count contributions from all of
our measurement results.
A sample experimental result is shown in Figure 21.12, in which we plot the
CAR as a function of the single-channel photon-detection rate. A preliminary
version of this result has been reported previously [91]. The CAR is measured at
the ambient temperature (300 K) and when the DSF is immersed in liquid nitrogen
(77 K). A CAR value as high as 116 is obtained at 77 K, whereas the peak of the
CAR at 300 K is around 25. This is consistent with the expectation that SRS is
more severe at a higher temperature, resulting in degraded purity of the source.
The shape of the CAR function is similar to those reported previously [87, 89, 90]
and can be qualitatively explained by a quantum model [92], which correctly
predicts the CAR’s temperature dependence. However, further work is required
for a quantitative comparison between experiment and theory.
After determining the single-count rate for the optimum CAR value, we adjust
HWP2 to split the dual-frequency pump field into two orthogonally polarized dual-
frequency components with equal power (more specifically, P1H = P1V = P2H =P2V). Each pump component probabilistically produces its own degenerate signal/
idler photon pair, and the two probability amplitudes superpose upon each other at
the output of the CPS, as there is no distinguishability between them. As a result of
reverse degenerate FWM and the CPS, polarization entanglement of the form
jHsHii þ jVsVii is generated, wherein the signal/idler photons are of identical
866 Prem Kumar et al.
wavelength �c. After passing through the 50/50 beam splitter BS3, the photons are
then detected in coincidence as before, except we add in the polarization analyzers
shown in Figure 21.11(b), because it is the polarization entanglement properties of
the photon pairs that are to be examined here. In the experiment, the polarization
analyzer in one photon’s path is set to a fixed linear polarization angle �0, and we
stepwise rotate the HWP in the other photon’s path to go from an initially parallel
polarization angle �0 to a final polarization angle �0 þ , which corresponds to a
HWP rotation angle of D� ˛ [0, /2]. Both single counts and coincidence counts
are recorded as a function of D�.A sample experimental result is shown in Figure 21.13. We demonstrate the high
purity of our entanglement source by exhibiting TPI with visibility >97% together
with polarization-independent single counts. These results are obtained when we pump
the CPS with relatively low power [P1H = P1V = P2H = P2V ’ 90 mW (peak power
’ 0.36 W)] and the DSF is cooled to 77 K. The high-TPI visibility means that we are
indeed generating the maximally entangled state jHs Hii þ jVsVii. Nonmaximally
entangled states can also be generated if we change the relative pump-power ratio
between the CW and the CCW paths. The relatively large error bars associated with the
TPI data points can be attributed to the low-data-collection rate in our current imple-
mentation, which can be improved by either using an OBPF with a wider passband, or
using faster detection electronics [93]. As is the case with other CPS entanglement
sources [78, 88], the dual-pump CPS source presented here can be easily configured to
produce all four Bell states (jHs Hii – jVs Vii, jHs Vii – jVsHii) and the Bell’s inequality
violation can be inferred from the observed >71% TPI visibility.
We further investigate the scenario when only a single-frequency pump (at
either �p1 or �p2) is injected into the CPS. As shown in Figure 21.14, we
0
50
100
150
200
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
–20 0 20 40 60 80 100
Δθ (°)
Coi
ncid
ence
/120
s
Sin
gle
coun
ts/1
20 s
Signal
Idler
TPI
Figure 21.13 Two-photon interference (TPI) with visibility >97% is shown, while single counts in
both channels exhibit no polarization dependence. Solid curve is a cosine fit to the data (this figure may
be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 867
observe polarization-independent behavior for both the single counts and the
coincidence counts. In fact, the coincidence counts after subtraction of the dark-
count contributions in both cases are zero within their error-bar ranges. This is
expected because the FWM photons in this case are produced not by the reverse
degenerate FWM process, but by the conventional FWM process involving a
degenerate-frequency pump. As the filters in the experiment are configured
to look at coincidences between photons in a single band (i.e., without its
conjugate), the pairwise nature of the conventional FWM photon production is
not manifested. Hence all the recorded coincidence counts are light–dark
0
–20 0 20 40
Δθ (°)60 80 100
200
400
600
800
1000
0
5,000
10,000
15,000
20,000
signal
idler
CC
(a)
–20 0 20 40
Δθ (°)60 80 100 120
(b)
0
200
400
600
800
1000
0
5,000
10,000
15,000
20,000
25,000
signal
idler
CC
Sin
gle
coun
ts/6
0 s
Sin
gle
coun
ts/6
0 s
Coi
ncid
ence
/60
sC
oinc
iden
ce/6
0 s
Figure 21.14 Single counts and coincidence counts (CC) show no dependence on the HWP angle D�when only a single-frequency pump at �p1 = 1545.95 nm (a), or at �p2 = 1555.92 nm (b), is used (this
figure may be seen in color on the included CD-ROM).
868 Prem Kumar et al.
coincidences, which after subtraction of the dark counts lead to zero light–light
coincidence counts. This serves as a further evidence that the degenerate-
frequency polarization-entangled photon pairs are indeed produced by the
dual-frequency pumps, and not by the single-frequency pumps. The polariza-
tion-independent single counts in Figure 21.14 are a result of the CPS-induced
polarization indistinguishability.
To conclude this section, we have presented an efficient, high-purity source of
correlated/entangled photon pairs in a single spatiotemporal mode in the telecom
band. It is foreseeable that further development of such fiber-based photon sources
will lead to practical applications of various quantum-communication protocols
and LOQC.
21.6.2 Hong-Ou-Mandel Interference with Fiber-GeneratedIndistinguishable Photons
The nascent field of quantum information science, motivated by the extraordinary
computing power of a full-fledged quantum computer, naturally selects photons—
the fundamental energy packets of the electromagnetic field—as the carrier of
quantum information from one computing node to another, a task typically
associated with quantum communication. The omnipresent telecommunication
fibers constitute the quantum channels for quantum cryptography [16], quantum
gambling [94], and quantum games [95], which utilize remote sharing of quantum
entanglement as a resource. In addition, indistinguishable photons—photons hav-
ing identical wave packets—play a major role in the arena of LOQC [96]. There-
fore, a telecom-band indistinguishable photon-pair source is particularly useful for
the above QIP tasks. The traditional method of producing such photons—SPDC
in second-order (�(2)) nonlinear crystals [97]—faces formidable engineering
challenges when these photons are launched into a single-mode optical fiber for
long-distance transmission or mode cleansing. Large coupling losses inevitably
occur due to mode mismatch [98], limiting the usefulness of such a source.
Recently, we demonstrated a fiber-based source of indistinguishable photon pairs
at a telecom-band wavelength near 1550 nm that utilizes the third-order (�(3))
nonlinear process of FWM [99] in the fiber itself. This approach automatically
takes care of the aforementioned mode-matching issue, because the photonic
spatial mode of the generated photon pair is the same as that of standard optical
fiber.
Four-wave mixing is a third-order process mediated by the Kerr nonlinearity of
optical fiber, wherein two-pump photons annihilate to give birth to a pair of time-
energy entangled daughter photons, usually denoted as signal and idler. Energy
conservation as well as momentum conservation are obeyed during the FWM
process: !p1 þ !p2 = !s þ !i, and ~kp1 þ ~kp2 =~ks þ ~ki, where !j and ~kj stand
for the frequency and wavevector of the jth photon, and subscripts p1, p2, s,
21. Fiber-Optic Quantum Information Technologies 869
and i denote the two pump photons, the signal photon, and the idler photon,
respectively. Our group has previously demonstrated generation [29, 30], distribu-
tion [100, 101], and storage [38, 39] of nondegenerate photon pairs (s 6¼ i) with a
single-frequency pump pulse (p1 = p2). More recently, we have also achieved
success in generating degenerate photon pairs (s = i) with dual-frequency pump
pulses (p1 6¼ p2) [99]. While it is straightforward to separate the nondegenerate
photon pairs by means of their different wavelengths through a wavelength-
division (de)multiplexer, it remains unclear how to deterministically separate the
fully degenerate photon pairs, as the two photons share the same properties in all
degrees of freedom: spatial, temporal (frequency), and polarization. This situation
is analogous to a type-I down-conversion with collinear output photons [102]. Up
to now, researchers have used a regular 50/50 beam splitter (BS) to probabilisti-
cally split the identical photons [99, 102, 103], which in actuality produces the
following state [schematically shown in Figure 21.15(a)]:
j�iin ¼ j2iaj0ib;
j�iout ¼1ffiffiffi2p j2idj0ic � j0idj2icffiffiffi
2p þ iffiffiffi
2p j1icj1id
1ffiffiffi2p �2002 þ
iffiffiffi2p �11: (21.44)
The presence of the �2002 component limits the usefulness of such a probabilistic
identical photon source, which has a maximum attainable HOM dip visibility of
only 50% [102].
However, if the output wavefunction only consists of the �11 component, the
HOM dip visibility can in principle reach 100%.
(a) (b)
d c
FPC
CWCCW
Input
50/50 coupler
P/2 P/2
P
02 ba20 b
iδae
ab
50/50BS
a
b
c
d
ψin ψout
Figure 21.15 (Color online) Quantum interference at a beam splitter. (a) Schematic of input/output
state transformation for a 50/50 beam splitter (BS). (b) Illustration of quantum interference in a Sagnac
loop between photon pairs with different phase shift �. CW, clockwise; CCW, counterclockwise; FPC,
fiber polarization controller (this figure may be seen in color on the included CD-ROM).
870 Prem Kumar et al.
One way to produce a “clean” �11 output is to utilize a cross-polarized dual-
frequency pump to excite the �xyxy component of the Kerr nonlinear susceptibility
tensor [104, 105]. This is analogous to a type-II collinear down-conversion [106],
where the output orthogonally polarized degenerate-frequency photon pairs are
split by a PBS in a deterministic fashion. However, the intrinsically weaker nature
of �xyxy, which is 1/3 of the collinear tensor component �xxxx in a Kerr nonlinear
medium like fused-silica glass, makes this approach much less efficient than the
copolarized FWM process. Here we introduce a new type (or topology) of copo-
larized identical-photon source, called the “50/50 Sagnac-loop” source, which is
capable of producing a clean �11 state. Due to its intended use to deterministically
split up the identical photons, it is also given the name “quantum splitter,” or QS
for short.
The idea behind the QS source comes from the realization that, not only can a
fiber Sagnac loop be configured as a total reflector [107] (TR) or a total transmitter
(TT), it also can be set in an equally transmissive and reflective state (50/50). The
difference between the above operational modes lies in the different settings of the
intraloop fiber polarization controller (FPC), which results in different relative
phase shifts between the CW and the CCW paths. As shown in Figure 21.15(b), the
pump is injected from port d into the Sagnac loop, which is composed of a 50/50
fiber coupler, a piece of DSF of suitable length, and an FPC. The pump peak power
is assumed to be P, which is split into two equally powered pulses (P/2) by the
coupler. The two-pump pulses traverse the DSF in a counterpropagating manner,
each of which probabilistically scatters copolarized FWM photon pairs. Here we
neglect the case where both pumps undergo FWM scattering, as it corresponds to a
higher order process of multiphoton generation, whose probability is vanishingly
small when the pump power is low. The two identical-photon probability ampli-
tudes, with a differential phase � controlled by the setting of the FPC, are then
recombined at the coupler before coming out of the Sagnac loop. The input state in
this case is written as
j�iin =j2iaj0ib þ ei�j0iaj2ibffiffiffi
2p : (21.45)
The corresponding output state is obtained from the standard BS input/output
relationship, and is given by
j�iout1� ei�
2�2002 þ
ið1þ ei�Þ2
�11: (21.46)
We can readily see, from Eqn (21.46), that a pure �11 state is obtained when we
set the differential phase � to be 0, which corresponds classically to the case of a
50/50 Sagnac loop. This can be physically interpreted as time-reversed HOM
interference. One can also verify that when � is set to be (-), the Sagnac loop
is totally reflective (transmissive), and one obtains the pure state �2002.
21. Fiber-Optic Quantum Information Technologies 871
We now demonstrate the identical nature of the photons produced by the QS
source. The observation of a HOM dip of high visibility has been established as a
figure of merit for identical-photon sources [108], and has been demonstrated
using a variety of devices, including identical photon pairs from the same SPDC
source [109, 110], independent indistinguishable photons from separate SPDC
sources [111–113], and indistinguishable photons from a quantum-dot single-
photon source [114]. Our scheme bears some resemblance to the first category,
but yet is different enough in that it is fiber-based and thus easily integrable into
fiber-optic networks [115]. The experimental setup is depicted in Figure 21.16
with considerable amount of detail. Figure 21.16(a) shows how the dual-frequency
copolarized pump is prepared [99]. Out of the broadband spectrum of a
femtosecond laser (repetition rate ’50 MHz), we spectrally carve out our desired
pump central wavelengths (�p1 = 1545.95 nm and �p2 = 1555.92 nm, pulse width
’ 5 ps) by cascading two free-space double-grating filters [DGF1 and DGF2,
FWHM ’ 0.8 nm for each passband; see Figure 21.17(a)]. An EDFA is
Pump1
Pump2
50/50 Sagnacinterferometer
OBPF2
OB
PF
1
λ m
λ p1
λ p2
FPC3
Circ
50/50BS3
Coincidence
counter
Translationstage
HWP1
HWP2
QWP2
QWP3
QWP4QWP1 PBS1 PBS2
PBS3
BS
APD1
e
APD2
f
g
h
M1
M2
(b)
Laser DGF1λ p1
λ p2
50/50BS1
EDFADGF2
50/50BS2
FPC1
FPC2
FPBS
Pump1
Pump2
λ p1
λ p2
(a)
λ m
Figure 21.16 (Color online) Schematic experimental setup to observe the Hong-Ou-Mandel
interference between identical photons generated from the QS source. (a) Preparation of the dual-
frequency copolarized pump. BS, beam splitter; DGF, double-grating filter; FPBS, fiber polarization
beam splitter; FPC, fiber polarization controller; EDFA, erbium-doped fiber amplifier. (b) Hong-Ou-
Mandel experiment. PBS, polarization beam splitter; HWP, half-wave plate; QWP, quarter-wave plate;
OBPF, optical band-pass filter; Circ, circulator; APD, avalanche photodiode (this figure may be seen in
color on the included CD-ROM).
872 Prem Kumar et al.
sandwiched in between the two DGFs to provide pump power variability. The out-
of-band ASE photons from the EDFA are suppressed by DGF2. The pump central
wavelengths are selected such that their mean wavelength (�m = 1550.92 nm) is
located near the zero-dispersion wavelength of the 300-m-long DSF in the Sagnac
loop to maximize the FWM efficiency [116]. The two-pump pulses emerging from
the second 50/50 beam splitter (BS2) are in turn passed through a fiber polarization
beam splitter (FPBS) to ensure their copolarized property. The other necessary
properties of the two-pump pulses, namely, temporal overlapping and equal power,
are individually addressed by careful path-matching and transmission-efficiency
balancing for the two pulses. Figure 21.16(b) shows the QS source and its intended
use in a HOM experiment. The Sagnac loop is preceded by a circulator (Circ),
which redirects the Sagnac-loop reflected photons to a separate spatial mode. The
output degenerate FWM photons from the Sagnac loop are selected by two optical
bandpass filters (OBPF1 and OBPF2), whose transmission spectrum is shown in
Figure 21.17(a) (center wavelength = 1550.92 nm, passband ’ 0.8 nm). The
OBPFs also provide the necessary >100-dB isolation from the pump to effectively
detect those filtered photons. The alignment of the Sagnac loop to its 50/50 state
consists of using a continuous-wave laser source with its wavelength set to �m as
the input to the Sagnac loop, and adjusting FPC3 so that the transmitted and
reflected powers are equal (i.e., Pe = Pf). In practice, we measure the powers at
the output of the OBPFs (Pg and Ph), and demand that Pg/�eg = Ph/�fh, where �eg
(�fh) is the transmission efficiency from point e (f) to point g (h). The experimental
values of �eg and �fh are 0.749 and 0.768, respectively. SRS in the DSF is
suppressed by cooling the DSF to 77 K using liquid nitrogen [117, 118].
Before the actual HOM experiment, we first characterize the QS source
by measuring its CAR [99], which has been established to be a figure of merit of
Single counts/pulse
CA
R
0
20
40
60
80
100
0
1000
2000
3000
4000
0 0.00125 0.0025 0.00375 0.005
Coi
ncid
ence
/60
s
Single counts/pulse
Coincidence
Accidentalcoincidence
–801540 1550 15551545 1560 0 0.001 0.002 0.003 0.004 0.005
–70
–60
–50
–40
–30
–20
–10
Wavelength (nm)
(a) (b)
Tra
nsm
issi
on s
pect
rum
(dB
) OBPFsDGFs
Figure 21.17 (Color online) Experimental parameters. (a) Transmission spectra of the OBPFs and
DGFs. (b) Experimentally measured CAR function of the QS source with the fiber at 77 K. Inset shows
the measured coincidence and accidental-coincidence counts as a function of single counts/pulse (this
figure may be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 873
such correlated-photon sources [103, 104, 117, 118]. To do that, the path-matched
outputs of the QS source [points g and h in Figure 21.16(b)] are directly connected to
two APD1 and APD2 (Epitaxx EPM 239BA) for coincidence detection, bypassing all
the intermediate free-space optics. The APDs are operated in a gated Geiger mode,
with detection rate ’780 kHz and overall detection efficiencies of 7% and 9%,
respectively. Single-channel counts as well as coincidence counts are recorded by a
“coincidence counter” software. Detector dark-count contributions are subtracted from
all of our measurement results. We call a “coincidence” count when the two detectors
fire in the same triggered time slot, and an “accidental- coincidence” count when they
fire in the adjacently triggered time slots. We then vary the pump power and record
the corresponding coincidence and accidental- coincidence counts. Their ratio,
CAR, is plotted as a function of the single-channel count rate in Figure 21.17(b).
It has a similar shape to those reported previously [99, 118]. The CAR peak value
of about 100 occurs at a single-count rate of around 4 � 10�4/pulse (photon-pair
production rate ’2.2 � 10�3/pulse), and decreases rather rapidly on both sides.
We then proceed with our HOM experiment. After the identical photons come
out from g, and h, they are collimated into free space through lenses, and two sets
of polarization compensators composed of half-wave plates and quarter-wave
plates (HWP1/QWP1 and HWP2/QWP2) are used to restore each photon’s polar-
ization to horizontal before they enter PBS1. Each photon then traverses one arm
of the Mach–Zehnder-like interferometer before being combined at the 50/50
beamsplitter (BS) cube. Quarter-wave plates QWP3 and QWP4 are each set at
45 , so that when combined with a mirror behind, they function as 45 -oriented
half-wave plates, rotating the horizontal polarization of the incident light by 90 .In principle, the combinations PBS3/QWP3/M1 and PBS2/QWP4/M2 can be
replaced with just two mirrors to direct the photons to the input ports of the BS.
In practice, we choose to implement the more complicated version, because it is
less susceptible to translation-induced misalignment of the photon wavepackets at
the BS. Careful path-matching is done to ensure that the two photons reach the BS
at approximately the same time (i.e., their arrival-time difference is within the
tuning range of the translation stage placed under M1). The identical photons,
before hitting the BS, are both of linear vertical polarization. The output photons
from the BS are each coupled into single-mode fibers, which are connected to the
two APDs for coincidence detection. The HOM experiment is performed by
recording the coincidence counts at each setting of the translation stage, which is
equivalent to recording the coincidence counts as a function of the overlap
between the two identical photon wavepackets. The two scales are related by
�� = 2Ds/c, where �� is the temporal difference between the photon wavepackets,
Ds is the difference in readings of the translation stage, and c = 3 � 108 m/s is the
speed of light in vacuum.
We first pump the QS source with relatively high pump power, and obtain a
sample experimental result as shown in Figure 21.18(a). The single-count rate is
’3.4 � 10�3/pulse, corresponding to a low-CAR value of around 2 according to
Figure 21.17(b). A HOM dip of visibility ’50% is observed. This 50% dip
874 Prem Kumar et al.
visibility is the upper bound for the case of two classically random photon sources
as inputs to the BS, and can be explained by using a simple classical electro-
magnetic theory [112]. It is an expected result, because in the high-pump condi-
tion, the photon source emits mainly spontaneously scattered Raman photons,
along with multiple FWM photon pairs, whose random behavior masks the true
correlated nature of a single pair of identical FWM photons. We then lower down
the pump power to observe a much higher HOM dip visibility of 94 – 1%, as
shown in Figure 21.18(b). This result is obtained at a single-count rate of around
4 � 10�4/pulse, which corresponds to the CAR peak in Figure 21.17(b). The near-
unity visibility of the HOM dip is well beyond the classical limit of 50%, and
clearly demonstrates the high indistinguishability of the QS-generated identical
photons. Combined with its telecom-band operation and fiber-optic networkabil-
ity, we expect to be able to achieve more complicated QIP applications with this
fiber-based QS source.
Theoretical modeling of the QS source and its HOM experiment will be
provided elsewhere [119]. Here we show theoretical simulation results of the
HOM dip in Figure 21.18(b), along with the experimental data. The two theory
curves, generated for different OBPF spectral shapes (see Figure 21.18 caption for
details), appear to fit the experimental data remarkably well. Both fits agree on the
ideally attainable HOM dip visibility of 100%, which corresponds to the fact that
the two-photon probability distribution function is symmetric with respect to its
two frequency arguments [106]. The missing 6(�1)% visibility can be explained
by taking into account the following real-life imperfections: (i) The BS used in
the experiment (R = 0.474, T = 0.526, where R and T are the BS intensity
0
100
200
300
400
500
600
700
–15 –10 –5 0 5 10 15 –12 –8 –4 0 4 8 12
Coi
ncid
ence
cou
nts/
30 s
Coi
ncid
ence
cou
nts/
60 s
δπ (ps) δπ (ps)
0
20
40
60
80
100
120
(a) (b)
Figure 21.18 Hong-Ou-Mandel experimental results. (a) Hong-Ou-Mandel dip visibility of 50% is
observed when the pump power is high. The solid curve is a least-squares Gaussian fit to the data.
(b) Hong-Ou-Mandel dip visibility of 94.3% is observed when the pump power is low. The solid curve
is a least-squares Gaussian fit to the data. The dotted curve is a theoretical fitting for Gaussian OBPFs,
while the dot-dashed curve is that for super-Gaussian OBPFs (see Ref. [119] for details) (this figure may
be seen in color on the included CD-ROM).
21. Fiber-Optic Quantum Information Technologies 875
coefficients of reflection and transmission, respectively) deviates from its ideal
performance, but this gives rise to a negligible correction factor [110] 2RT/
(R2 þ T2)’ 0.994. (ii) The spatial-mode mismatch of the two photon wavepackets
at the BS results in some distinguishing information between the two coincidence-
generating probability amplitudes. A simple calculation [119] shows that a small
angular mismatch of around 30 mrad can bring the HOM dip visibility down
to 94%. (iii) Some remaining �2002 component due to nonideal alignment of the
50/50 Sagnac loop may lead to a degradation of the dip visibility [102]. (iv)
Existence of unsuppressed noise photons, such as Raman photons and FWM
photons induced by a single-frequency pump, may also degrade the attainable
HOM dip visibility.
21.7 CONCLUDING REMARKS
Because this volume deals with advances in Optical Fiber Telecommunications,
our focus in this chapter has been on the generation of correlated and entangled
photons in the telecom band with use the Kerr nonlinearity in DSF. Over the past
decade, microstructure or holey fibers (MFs) have come on the scene, which
because of their tailorable dispersion properties allow phase-matching to be
obtained over a wide range of wavelengths. Generation of correlated and entangled
photons in MFs has been demonstrated and rapid progress is taking place.
REFERENCES
[1] D. Heiss, Fundamentals of Quantum Information: Quantum Computation, Communication,
Decoherence, and all that, Berlin Heidelberg: Springer-Verlag, 2002.
[2] A. Einstein, B. Podolsky, and N. Rosen, “Can quantum-mechanical description of physical
reality be considered complete?” Phys. Rev., 47, 777–780, 1935.
[3] N. Bohr, “Can quantum-mechanical description of physical reality be considered complete?”
Phys. Rev., 48, 696–702, 1935.
[4] D. Bohm, Quantum Theory, Prentice-Hall, 614, 1951.
[5] J. S. Bell, 1964, Physics (Long Island City, N.Y.) 1, reprinted in J. S. Bell, Speakable and
Unspeakable in Quantum Mechanics (Cambridge: Cambridge University, 1987).
[6] S. J. Freedman and J. S. Clauser, Phys. Rev. Lett., 28, 938–941, 1972.
[7] A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett., 47, 460, 1981.
[8] A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett., 49, 91, 1982.
[9] A. Aspect, J. Dalibard, and G. Roger, Phys. Rev. Lett., 49, 1804, 1982.
[10] P. G. Kwiat, K. Mattle, H. Weinfurter et al., Phys. Rev. Lett., 75, 4337, 1995.
[11] P. M. Pearle, Phys. Rev. D, 2, 1418, 1970.
[12] J. F. Clauser and A. Shimony, Rep. Prog. Phys., 41, 1881, 1978.
[13] M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, Cambridge:
Cambridge University Press, 2000.
[14] M. Nielsen and I. Chuang, Quantum Computation and Quantum Information, Cambridge:
Cambridge University Press, 2002.
876 Prem Kumar et al.
[15] C. H. Bennett, G. Brassard, C. Crepeau et al., “Teleporting an unknown quantum state via dual
classical and einstein-poldosky-rosen channels,” Phys. Rev. Lett., 70, 1895, 1993.
[16] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, “Quantum cryptography,” Rev. Mod. Phys., 70,
145–195, 2002.
[17] C. H. Bennett and P. W. Shor, “Quantum information theory,” IEEE Trans. Inf. Theory, 44,
2724–2742, October 1998.
[18] J. Mod. Opt., 48(13), November 2001.
[19] R. H. Stolen and J. E. Bjorkholm, “Parametric amplification and frequency conversion in optical
fibers,” IEEE J. Quantum Electron., 18, 1062–1072, July 1982.
[20] A. Lacaita, P. A. Francese, F. Zappa, and S. Cova, “Single-photon detection beyond 1 mm:
performance of commercially available InGaAs/InP detectors,” Appl. Opt., 35, 2986–2996,
June 1996.
[21] G. Ribordy, J.-D. Gautier, H. Zbinden, and N. Gisin, “Performance InGaAs/InP avalanche
photodiodes as gated-mode photon counters,” Appl. Opt., 37, 2272–2277, April 1998.
[22] J. G. Rarity, T. E. Wall, K. D. Ridley et al., “Single-photon counting for the 1300–1600-nm range
by use of Peltier-cooled and passively quenced InGaAs avalanche photodiodes,” Appl. Opt., 39,
6746–6753, December 2000.
[23] P. A. Hiskett, G. S. Buller, A. Y. Loudon et al., “Performance and design of InGaAs/InP photo-
diodes for single-photon counting at 1.55 mm,” Appl. Opt., 39, 6818–6829, December 2000.
[24] X. Y. Zou, L. J. Wang, and L. Mandel, “Violation of classical probability in parametric down-
conversion,” Opt. Commun., 84, 351–354, August 1991.
[25] Y. R. Shen, IEEE J. Quantum Electron., QE-22(8), 1196, 1986.
[26] G. Bonfrate, V. Pruneri, P. G. Kazansky et al., “Parametric fluorescence in periodically poled
silica fibers,” Appl. Phys. Lett., 75, 2356–2358, October 1999.
[27] J. E. Sharping, M. Fiorentino, and P. Kumar, “Observation of twin-beam-type quantum correla-
tion in optical fiber,” Opt. Lett., 26, 367–369, March 2001.
[28] P. L. Voss and P. Kumar, Opt. Lett., 29, 445, 2004.
[29] X. Li, J. Chen, P. L. Voss et al., Opt. Express, 12, 3737, 2004.
[30] X. Li et al., Phys. Rev. Lett., 94, 053601, 2005.
[31] M. H. Rubin, D. N. Klyshko, Y. H. Shih, and A. V. Sergienko, Phys. Rev. A, 50, 5122, 1994.
[32] G. P. Agrawal, Nonlinear Fiber Optics, San Diego: Academic Press, 2001, chapter 10.
[33] D. Bouwmeester, A. Ekert, and A. Zeilinger, The Physics of Quantum Information: Quantum
Cryptography, Quantum Teleportation, Quantum Computation, 1st edn, Springer Verlag, 2000.
[34] T. E. Keller and M. H. Rubin, Phys. Rev. A, 56, 1534, 1997.
[35] A. K. Ekert, Phys. Rev. Lett., 67, 661, 1991.
[36] R. J. Glauber, Phys. Rev., 130, 2529, 1963; 131, 2766, 1963.
[37] M. Fiorentino, P. L. Voss, J. E. Sharping, and P. Kumar, IEEE Photon. Technol. Lett., 14, 983, 2002.
[38] X. Li et al., Opt. Lett., 30, 1201, 2005.
[39] X. Li, P. L. Voss, J. E. Sharping, and P. Kumar, Phys. Rev. Lett., 94, 053601, 2005.
[40] J. H. Shapiro, IEEE J. Quantum Electron., QE-21(3), 237, 1985.
[41] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products.
[42] M. Eibl, S. Gaertner, M. Bourennane et al., Phys. Rev. Lett., 90, 200403, 2003.
[43] A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev., 47, 777, 1935.
[44] J. S. Bell, Physics, 1, 195, 1964.
[45] D. M. Greenberger et al., Am. J. Phys., 58, 1131, 1990.
[46] J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Phys. Rev. Lett., 23, 880, 1969.
[47] C. H. Bennett and S. J. Wiesner, Phys. Rev. Lett., 69, 2881, 1992.
[48] C. H. Bennett et al., Phys. Rev. Lett., 70, 1895, 1993.
[49] A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett., 49, 91, 1982.
[50] Z. Y. Ou and L. Mandel, Phys. Rev. Lett., 61, 50, 1988.
[51] Z. Y. Ou et al., Phys. Rev. A, 41, 566, 1990.
[52] Y. H. Shih and C. O. Alley, Phys. Rev. Lett., 61, 2921, 1988.
21. Fiber-Optic Quantum Information Technologies 877
[53] J. W. Pan et al., Nature, 403, 515, 2000.
[54] K. Mattle et al., Phys. Rev. Lett., 76, 4656, 1996.
[55] D. Bouwmeester et al., Nature, 390, 575, 1997.
[56] N. Gisin et al., Rev. Mod. Phys., 74, 145, 2002.
[57] J. G. Rarity and P. R. Tapster, Phys. Rev. Lett., 64, 2495, 1990.
[58] J. Brendel, E. Mohler, and W. Martiennsen, Europhys. Lett., 20, 575, 1992.
[59] P. G. Kwiat, K. Mattle, H. Weinfurter et al., “New high-intensity source of polarization-
entangled photon pairs,” Phys. Rev. Lett., 75, 4337–4341, 1995.
[60] F. A. Bovino et al., Opt. Commun., 227, 343, 2003.
[61] K. Sanaka, K. Kawahara, and T. Kuga, Phys. Rev. Lett., 86, 5620, 2001.
[62] J. Kim et al., Nature, 397, 500, 1999.
[63] C. Kurtsiefer et al., Phys. Rev. Lett., 85, 290, 2000.
[64] X. Li et al., Opt. Express, 12, 3337, 2004.
[65] M. Fiorentino, P. L. Voss, J. E. Sharping, and P. Kumar, “All-fiber photon-pair source for
quantum communications,” IEEE Photon. Technol. Lett., 14, 983–985, 2002.
[66] J. E. Sharping, M. Fiorentino, and P. Kumar, “Observation of twin-beam-type quantum correla-
tion in optical fiber,” Opt. Lett., 26, 367–369, 2001.
[67] D. B. Mortimore, “Fiber loop reflectors,” J. Lightw. Technol., 6, 1217–1224, 1988.
[68] P. L. Voss, R. Tang, and P. Kumar, Opt. Lett., 28, 549, 2003.
[69] P. L. Voss and P. Kumar, Opt. Lett., 29, 445, 2004.
[70] P. L. Voss and P. Kumar, J. Opt. B: Quantum Semiclass. Opt., 6, S762, 2004.
[71] M. Fiorentino, P. L. Voss, J. E. Sharping, and P. Kumar, IEEE Photon. Technol. Lett., 14, 983, 2002.
[72] X. Li, P. L. Voss, J. E. Sharping, and P. Kumar, Phys. Rev. Lett., 94, 053601, 2005.
[73] X. Li, P. L. Voss, J. Chen et al., Opt. Lett., 30, 1201, 2005.
[74] X. Li, C. Liang, K. F. Lee et al., Phys. Rev. A, 2006; to appear. See arXiv:quant-ph/0601087
[75] X. Li, J. Chen, P. L. Voss et al., Opt. Express, 12, 3737, 2004.
[76] X. Li, P. L. Voss, J. Chen et al., Opt. Express, 13, 2236, 2005.
[77] P. Kumar, M. Fiorentino, P. L. Voss, and J. E. Sharping, U.S. Patent No. 6897434, 2005.
[78] H. Takesue and K. Inoue, Phys. Rev. A, 70, 031802, 2004.
[79] S. A. E. Lewis, S. V. Chernikov, and J. R. Taylor, Opt. Lett., 24, 1823, 1999.
[80] R. H. Stolen and G. D. Bjorkholm, “Parametric amplification and frequency conversion in optical
fibers,” IEEE J. Quantum Electron., 18, 1062–1072, 1982.
[81] H. Takesue and K. Inoue, Opt. Express, 13, 7832, 2005.
[82] J. Fan, A. Dogariu, and L. J. Wang, Opt. Lett., 30, 1530, 2005.
[83] E. Knill, R. Laflamme, and G. J. Milburn, Nature, 409, 46, 2001.
[84] P. Kok, W. J. Munro, K. Nemoto et al., arXiv: quant-ph/0512071
[85] C. K. Hong, Z. Y. Ou, and L. Mandel, Phys. Rev. Lett., 59, 2044, 1987.
[86] D. Branning, W. Grice, R. Erdmann, and I. A. Wamsley, Phys. Rev. A, 62, 013814, 2000.
[87] J. Fan, A. Dogariu, and L. J. Wang, Opt. Lett., 30, 1530, 2005.
[88] P. Kumar, M. Fiorentino, P. L. Voss, and J. E. Sharping, U.S. Patent No. 6897434.
[89] K. F. Lee, J. Chen, C. Liang et al., to appear in Opt. Lett. (2006). See also post deadline paper
presented at the Frontiers in Optics 2005, paper PDP-A4.
[90] H. Takesue and K. Inoue, Opt. Express, 13, 7832, 2005.
[91] J. Chen, K. F. Lee, C. Liang, and P. Kumar, in Proc. IEEE/LEOS Summer Topical Meetings,
paper TuB2.3, 2006.
[92] J. Chen, P. L. Voss, M. Medic, and P. Kumar, “On the coincidence-to-accidental-coincidence
ratio of fiber-generated photon pairs,” to be submitted.
[93] C. Liang, K. F. Lee, P. L. Voss et al., in Proc. SPIE, 5893, 2005, p. 282.
[94] L. Goldenberg, L. Vaidman, and S. Wiesner, Phys. Rev. Lett., 82, 3356, 1999.
[95] J. Eisert, M. Wilkens, and M. Lewenstein, Phys. Rev. Lett., 83, 3077, 1999.
[96] E. Knill, R. Laflamme, and G. J. Milburn, Nature, 409, 46–52, 2001.
[97] Articles in J. Mod. Opt., 48(13), November 2001.
878 Prem Kumar et al.
[98] S. Castelletto et al., New J. Phys., 6, 87, 2004.
[99] J. Chen et al., Opt. Lett., 31, 2798, 2006.
[100] C. Liang et al., presented at OFC/NFOEC 2005, post-deadline paper, paper PDP35, 2005.
[101] C. Liang et al., Opt. Express, 15, 1322, 2007.
[102] M. Halder et al., Phys. Rev. A, 71, 042335, 2005.
[103] J. Fan, A. Dogariu, and L. J. Wang, Opt. Lett., 30, 1530, 2005.
[104] J. Fan and A. Migdall, Opt. Express, 13, 5777, 2005.
[105] J. Chen et al., presented at Frontiers in Optics 2006/Laser Science XXII, paper FTuR4.
[106] W. P. Grice and I. A. Walmsley, Phys. Rev. A, 56, 1627–1634, 1997.
[107] D. B. Mortimore, J. Lightw. Technol., 6, 1217, 1988.
[108] P. Kok et al., ArXiv: quant-ph/05 12071, 2006.
[109] W. P. Grice et al., Phys. Rev. A, 57, R2289, 1998.
[110] C. K. Hong, Z. Y. Ou, and L. Mandel, Phys. Rev. Lett., 59, 2044–2046, 1987.
[111] H. de Riedmatten et al., Phys. Rev. A, 67, 022301, 2003.
[112] J. G. Rarity, P. R. Tapster, and R. Loudon, ArXiv: quant-ph/9702032, 1997.
[113] R. Kaltenbaek et al., Phys. Rev. Lett., 96, 240502, 2006.
[114] C. Santori et al., Nature, 419, 594, 2002.
[115] X. Li et al., Phys. Rev. A, 73, 052301, 2006.
[116] J. Fan, A. Dogariu, and L. J. Wang, Opt. Lett., 30, 1530, 2005.
[117] H. Takesue, and K. Inoue, Opt. Express, 13, 7832, 2005.
[118] K. F. Lee et al., Opt. Lett., 31, 1905, 2006.
[119] J. Chen, K. F. Lee, and P. Kumar, ArXiv:quant-ph/0702176, 2007.
[120] P. G. Kwiat, E. Waks, A. G. White et al., “Ultrabright source of polarization-entangled photons,”
Phys. Rev. A, 60, R773–R776, August 1999.
[121] S. Tanzilli, H. De Riedmatten, W. Tittel et al., “Highly efficient photon-pair source using
periodically poled lithium niobate waveguide,” Electron. Lett., 37, January 2001.
[122] C. E. Kuklewicz, E. Keskiner, F. N. C. Wong, and J. H. Shapiro, “A high-flux entanglement
source based on a doubly-resonant optical parametric amplifier,” J. Opt. B: Quantum and
Semiclass. Opt., 4, June 2002.
[123] J. Chen, X. Li, and P. Kumar, Phys. Rev. A, 72, 033801, 2005.
[124] J. L. O’Brien et al., Nature, 426, 264–267, 2003.
[125] N. K. Langford et al., Phys. Rev. Lett., 95, 210504, 2005.
[126] N. Kiesel et al., Phys. Rev. Lett., 95, 210505, 2005.
[127] R. Okamato et al., Phys. Rev. Lett., 95, 210506, 2005.
[128] P. Walther and A. Zeilinger, Phys. Rev. A, 72, 010302, 2005.
[129] Z. Zhao et al., Phys. Rev. Lett., 94, 030501, 2005.
[130] J. Chen, K. F. Lee, and P. Kumar, submitted to Phys. Rev. Lett.
[131] J. Chen et al., Opt. Lett., 31, 2798, 2006.
[132] M. A. Jaspan et al., Appl. Phys. Lett., 89, 031112, 2006.
[133] R. H. Hadfield et al., Opt. Express, 13, 10846, 2005.
[134] G. N. Gol’tsman et al., Appl. Phys. Lett., 79, 705, 2001.
[135] Y. J. Lu, R. L. Campbell, and Z. Y. Ou, Phys. Rev. Lett., 91, 163602, 2003.
[136] D. F. V. James et al., Phys. Rev. A, 64, 052312, 2001.
[137] J. B. Altepeter, E. R. Jeffrey, and P. G. Kwiat, Advances in Atomic, Molecular, and Optical
Physics, Elsevier, 52, 2005, chapter 3.
[138] W. J. Munro, K. Nemoto, and A. G. White, J. Mod. Opt., 48, 1239, 2001.
[139] N. Patel, Nature, 445, 144–146, 2007.
[140] D. Bouwmeester, A. Ekert, and A. Zeilinger, The Physics of Quantum Information, Berlin:
Springer-Verlag, 2000.
[141] D. C. Burnham and D. L. Weinberg, “Observation of simultaneity in parametric production of
optical photon pairs,” Phys. Rev. Lett., 25, 84–87, 1970.
21. Fiber-Optic Quantum Information Technologies 879
[142] X. Li, P. Voss, J. E. Sharping, and P. Kumar, “Violation of Bell’s Inequality Near 1550 nm Using
an All-Fiber Source of Polarization-Entangled Photon Pairs,” in Proc. QELS’2003, Optical
Society of America, QTuB4, Washington, DC, 2003.
[143] X. Li, P. Voss, J. E. Sharping, and P. Kumar, “Optical-fiber source of polarization-entangled
photon pairs in the 1550 nm telecom band,” Quantum Physics Archive, 0402191 (2004), http://
xxx.lanl.gov/abs/quant-ph/0402191
[144] P. Russell, “Photonic crystal fibers,” Science, 299, 358–362, 2003.
[145] M. Fiorentino, J. E. Sharping, P. Kumar et al., “Soliton squeezing in microstructure fiber,” Opt.
Lett., 27, 649–651, 2002.
[146] A. Dogariu, J. Fan, L. J. Wang, and J. A. West, “Photon-Pairs Generation in Micro-Structured
Fiber,” in Proc. QELS’2003, Optical Society of America, QTuB6, Washington, DC, 2003.
[147] J. E. Sharping, M. Fiorentino, A. Coker et al., “Four-wave mixing in microstructure fiber,” Opt.
Lett., 26, 1048–1050, 2001.
[148] J. E. Sharping, M. Fiorentino, P. Kumar, and R. S. Windeler, “Optical-parametric oscillator based
on four-wave mixing in microstructure fiber,” Opt. Lett., 19, 1675–1677, 2002.
[149] S. Lloyd, M. S. Shahriar, J. H. Shapiro, and P. R. Hemmer, “Long distance unconditional
teleportation of atomic states via complete Bell state measurements,” Phys. Rev. Lett., 87,
167903, 2001.
[150] G. P. Agrawal, Nonlinear Fiber Optics, 3rd edn, San Diego, CA: Academic Press, 2000.
[151] D. Ouzounov, D. Homoelle, W. Zipfel et al., “Dispersion measurements of microstructured fibers
using femtosecond laser pulses,” Opt. Commun., 192, 219–223, 2001.
[152] P. L. Voss and P. Kumar, “Raman-noise-induced noise-figure limit for chi (3) parametric
amplifiers,” Opt. Lett., 29, 445–447, 2004.
[153] J. D. Harvey, R. Leonhardt, S. Coen et al., “Scalar modulation instability in the normal dispersion
regime by use of a photonic crystal fiber,” Opt. Lett., 28, 2225–2227, 2003.
880 Prem Kumar et al.