Long distance entanglement swapping with photons from separated sources.

H. de Riedmatten, I. Marcikic, J.A.W. van Houwelingen, W. Tittel, H. Zbinden, and N. Gisin

Group of Applied Physics, University of Geneva, Switzerland

(Dated: September 11, 2018)

We report the first experimental realization of entanglement swapping over large distances in
optical fibers. Two photons separated by more than two km of optical fibers are entangled, although
they never directly interacted. We use two pairs of time-bin entangled qubits created in spatially
separated sources and carried by photons at telecommunication wavelengths. A partial Bell state
measurement is performed with one photon from each pair which projects the two remaining photons,
formerly independent onto an entangled state. A visibility high enough to violate a Bell inequality
is reported, after both photons have each travelled through 1.1 km of optical fiber.
arXiv:quant-ph/0409093v1 15 Sep 2004

Quantum teleportation is a process that enables the involved only two photons instead of four [11, 12]. Fi-
quantum state of an object to be transferred from one nally, an experiment demonstrating the principle using
place to a distant one without ever existing anywhere in continuous variables has also been reported [13]. All the
between. The quantum teleportation channel is noth- experiments realized so far have demonstrated the prin-
ing like an ordinary channel: it follows no path in space, ciple of entanglement swapping over short distances (of
but consists of entangled particles. Entanglement is a the order of a meter).
property at the roots of quantum physics which leads to In this paper, we present the first experimental demon-
non-local correlations between distant particles that can-
not be explained by classical physics. Entangled particles entangled photons
that never interacted
behave as if they were a single object, non separable into
its constituents. Now, entanglement itself can be tele-
ported, if the state to be teleported is part of an entan-
gled state. This process, called entanglement swapping
[1], allows one thus to concatenate quantum teleporta-
tion channels. This protocol beautifully illustrates the Bell state
oddness of quantum physics since it enables one to en- measurement
tangle two particles that have never directly interacted.
Hence, two particles with no common past can act as a
A B C D
single object. The principle of entanglement swapping
is explained in Fig.1. Besides its fascinating aspect, en-
tanglement swapping also plays an essential role in the EPR source EPR source
context of quantum information science. It is for instance
the building block of protocols such as quantum repeaters FIG. 1: Scheme of principle of entanglement swapping. The
[2, 3] or quantum relays [4, 5, 6] proposed to increase idea is to start from two independent pairs of entangled par-
the maximal distance of quantum key distribution and ticles (EPR sources) and to subject one particle from each
pair to a joint measurement called Bell state measurement
quantum communication. It also allows the implemen-
(BSM). This BSM projects the two other particles, formerly
tation of an heralded source of entangled photon pairs independent onto an entangled state [1]
[1]. Finally, it is a key element for the implementation
of quantum networks [7] and of Linear Optics Quantum
Computing [8]. stration of entanglement swapping over large distances
The entanglement swapping protocol has been first pro- in optical fibers. We use two pairs of time-bin entangled
posed by Zukowski and colleagues in 1993 [1]. The qubits encoded in photons at telecommunication wave-
first experimental demonstration has been reported in lengths created by parametric down conversion (PDC).
1998, using polarization entangled qubits encoded in Contrary to previous swapping experiments involving
photons around 800nm [9]. In 2002, an improved ver- four photons, the two pairs are created in spatially sep-
sion of this experiment allowed a violation of a Bell in- arated sources although pumped by the same laser. A
equality with the teleported entanglement [10], thus con- partial Bell state measurement (BSM) is performed onto
firming the non-local character of this protocol. More one photon from each pair, entangling the two remaining
recently, two quantum teleportation experiments using photons which have each travelled over separated 1.1 km
mode-entangled qubits have been performed, that can be spools of optical fiber. A two photon interference visi-
interpreted as entanglement swapping experiments, al- bility high enough to violate a Bell inequality is demon-
though they differ from the original proposal since they strated, conditioned on a successful BSM. Hence, two
 2

photons separated by more than two km of optical fibers Note that when photons B and C are projected onto the
exhibit non-local correlations although they have been state |ψ + i or |ψ − i, the state of photons A and D is in-
created in spatially separated sources and have conse- dependent of the phase δ which appears only as a global
quently never interacted. The use of time-bin encoding at factor. This means that in this case, the creation process
telecommunication wavelength is an important extension is robust against phase fluctuations in the pump interfer-
compared to previous experiments, since it has proven to ometer [18]. If however photons B and C are projected
be well suited for long distance transmission in optical onto the state |φ+ i or |φ− i the state of photons A and
fibers [14]. In addition, time-bin entanglement can be D depends on twice the phase δ. In our experiment,
easily extended to high-dimensional Hilbert spaces in a we make a partial BSM, looking only at projections of
scalable way with only two photons [15]. Moreover, as ex- photons B and C onto the |ψ − i Bell state. Apart from
plained below, the present scheme is intrinsically robust robustness, another interesting feature to note is that all
against phase fluctuations and pump laser wavelength the four Bell states are involved in the experiment, since
drifts in the preparation stage, provided that we restrain we start from |φ+ i and |φ− i states, make a projection
ourself to a partial BSM. Hence, this experiment can also onto the |ψ − i state, which projects the two remaining
be considered as a (post-selected) heralded source of en- photons onto the |ψ + i state.
tangled photons pairs robust against phase fluctuation in A scheme of our experiment is presented in Fig.2. Fem-
the preparation stage [16]. tosecond pump pulses are sent to an unbalanced bulk
Time-bin entangled qubits can be seen as photon pairs Michelson interferometer with a travel time difference of
created in a coherent superposition of two emission times τ = 1.2ns. Thanks to the use of retroreflectors, we can
with a well defined relative phase [17]. They are created utilize both outputs of the interferometer, which are di-
first by splitting a laser pulse into two subsequent pulses rected to spatially separated Lithium triborate (LBO)
using an unbalanced interferometer called pump interfer- non linear crystals. Collinear non degenerate time-
ometer. One photon pair is then created by PDC. The bin entangled photons at telecommunication wavelength
down-converted photons originate from the two pulses (1310 and 1550 nm) are eventually created by paramet-
with a relative phase δ, hence the photon pair quantum ric down-conversion (PDC) in each crystal. Because of
state is |φ+ (δ)i = c0 |0, 0i + eiδ c1 |1, 1i, where |0, 0i corre- the phase acquired at the beam splitter in the pump in-
sponds to a photon pair created in the early time-bin and terferometer there is an additional relative phase of π
|1, 1i to a photon pair created in the delayed time-bin, between the terms |0, 0i and |1, 1i in the second output
with c20 + c21 = 1. of the interferometer. This explains why a state |φ+ (δ)i
In our experiment, we employ two spatially separated is created in one crystal while a state |φ− (δ)i is created
sources of entangled photons. In one of these sources, we in the other one.
create a state |φ+ (δ)iA,B while in the other one we create The created photons are coupled into single-mode opti-
a state |φ− (δ)iC,D = |φ+ (δ + π)iC,D . Initially, the two cal fibers and separated with a wavelength-division mul-
photon pairs are independent and the total state can be tiplexer (WDM). The two photons at 1310 nm (B and
written as the tensor product: C) are sent to a Bell state analyzer (BSA), where only
the Bell state |ψ (−) iBC is detected. We use an interfero-
|ΨABCD i = |φ+ (δ)iAB ⊗ |φ− (δ)iCD (1) metric BSA based on a standard 50-50 fiber beam-splitter
[19]. It can be shown that whenever photons B and C are
This state can be rewritten in the form: detected in different outputs modes and different time-
1 + bins, the desired projection is achieved [6]. For this kind
|ΨABCD i = [|φ iBC ⊗ |φ− (2δ)iAD of measurement, the two incoming photons must be com-
2
+|φ− iBC ⊗ |φ+ (2δ)iAD pletely indistinguishable in their spatial, temporal, spec-
tral and polarization mode. The indistinguishability is
+|ψ + iBC ⊗ eiδ |ψ − iAD verified by a Hong-Ou-Mandel experiment [20, 21]. The
+|ψ − iBC ⊗ eiδ |ψ + iAD ] (2) two photons at 1310 nm are filtered with 5 nm band-
width interference filters (IF) in order to increase their
where the four Bell states are: coherence length to 500 fs, larger than the pump pulses
1 duration (200 fs), which is necessary in order to make
|φ± (δ)i = √ |0, 0i ± eiδ |1, 1i


2 the photon temporally indistinguishable [22]. As a con-
1 sequence of the use of femtosecond pulses, the bandwidth
(±)
|ψ i = √ (|1, 0i ± |0, 1i) (3) of the down-converted photons is large, leading to severe
2
depolarization effects in long fibers. Time-bin encoding
When photons B and C are measured in the Bell basis is thus an advantage in this context, since it it not sensi-
(Eq. 3), i.e. projected onto one of the four Bell states tive to depolarization effects.
via a so-called Bell state measurement, photons A and The two photons at 1550 nm, filtered to 18 nm bandwith
D are projected onto the corresponding entangled state. (A and D) each travel over 1.1 km of dispersion shifted
 3

Alice Bob
FM
of the 1550 nm photons inside optical fibers. Hence, the
Charlie FM
BSM
swapping process is completed when the photons are al-
b
ready two km apart. A home made programme allows us
a
Ge InGaAs to register any desired combination of coincidence count
FM FM
C1 C2 rate between the four detectors, which is useful to char-
acterize the stability of the whole setup during the mea-
InGaAs
surement process. In our experiment, the average pump
Beam splitter 50:50 InGaAs
1550 nm
power for each source was about 80 mW, leading to a
1310 nm 1310 nm 1550 nm
probability of creating an entangled pair per laser pulse
A B C D
1.1 km DSF 1.1 km DSF of around 6%.
WDM WDM
SF SF If entanglement swapping is successful, the two photons
LBO LBO A and D at 1550 nm should be in the entangled state
|ψ (+) i, conditioned on a projection on the |ψ (−) i Bell
state. However, as real measurements are imperfect,
Pump there will be some noise, that we suppose equally dis-
interferometer
d tributed between all possible outcomes. Hence, the cre-
ated state can be written:
Femto-second laser
lp=710 nm 1 − F2
ρ = F2 |ψ (+) ihψ (+) | + (|ψ (−) ihψ (−) |
3
FIG. 2: Experimental setup. The pump laser is a mode- +|φ(+) ihφ(+) | + |φ(−) ihφ(−) |)
locked femtosecond Ti-Sapphire laser producing 200 fs pulses (1 − V )
at a wavelength of 710 nm with a repetition rate of 75 MHz. = V |ψ (+) ihψ (+) | + 11 (4)
4
After the crystals, the pump beam are blocked with silicon fil-
ters (SF). The Faraday mirrors (FM) are used to compensate where V is the visibility and F2 the two-qubit fidelity
polarization fluctuations in the fiber interferometers defined as: F2 = hψ (+) |ρ|ψ (+) i. V is related to F2 as:
4F2 − 1
V = (5)
3
fiber (DSF). Their entanglement is then analyzed with The Peres criteria [23] shows that the two photons are
two fibre Michelson interferometers with the same travel entangled (i.e. in a non separable state) if V > 31 , and
time difference as the pump one. The phase of each in- consequently if F2 > 21 . It can also be shown that Bell’s
terferometer can be varied with a piezo actuator (PZA). inequalities can be violated if V > √12 [24].
In order to control the phase and to obtain a long term To verify the entanglement swapping process we perform
stability, the fiber inteferometers are actively controlled a two-photon interference experiment with the two pho-
using a frequency stabilized laser (Dicos) and a feed-back tons at 1550 nm, conditioned on a successful BSM. This
loop on the PZA. The phase of each fiber interferometer is is done by sending the two photons A and D to two inter-
probed periodically and is locked to a user-defined value ferometers. The evolution of |ψ (+) i in the interferometers
[14]. This technique allowed us to obtain a excellent sta- is:
bility tested over up to 96 hours. Note that the phase of |ψ (+) i → |0A , 1D i + eiα |1A , 1D i + eiβ |0A , 2D i
the pump interferometer is not actively stabilized.
The photons are detected with avalanche photodiode +ei(α+β) |1A , 2D i + |1A , 0D i + eiα |2A , 0D i
(APD) single photon detectors. One of the 1310 nm +eiβ |1A , 1D i + ei(α+β )|2A , 1D i (6)
photon (detector C1 ) is detected with a liquid Nitrogen
where |iA , jD i corresponds to an event where the pho-
cooled Ge APD (NEC), with an efficiency of around 10%
ton A is in time-bin i and the photon D is in time-bin
for 40KHz of dark counts. The three other photons are
j. A photon travelling through the long arm of an in-
detected with InGaAs APDs (id-Quantique) with an effi-
terferometer passes from time-bin i to time-bin i + 1. If
ciency of 30% for a dark count probability of around 10−4
the arrival time difference between photon A and D are
per ns. The trigger signal for those detectors is given
recorded, Eq. (6) shows that there are 5 different time
by a coincidence between the Ge APD and the emission
windows, with ∆τ = tA − tB = {0, ±τ, ±2τ }. This is
time of the pump pulses. The coincidence events between
in contrast with previous experiments using time-bin en-
different detectors are recorded with a multistop time-to
tangled qubits in the state |φ(±) i, where only three time-
digital converter (TDC). The coincidence between the Ge
windows were present (see e.g.[14]). If only the event
APD and the emission time of the laser is used as START
with ∆τ = 0 are selected, there are two indistinguishable
while the other APDs are used as STOPs. Note that the
events leading to a coincident count rate:
classical information about the BSM is delayed electron-
ically by roughly 5 µs, corresponding to the travel time Rc ∼ 1 + V cos(α − β) (7)
 4

where V is the visibility of the interference which can in obtained after the swapping process was high enough to
principle attain the value of 1 but is in practice lower demonstrate a teleportation of entanglement and to in-
than 1 due to various experimental imperfections. Fig.3 fer a violation of Bell inequalities with photons separated
by more than 2 km of optical fibers that have never di-
rectly interacted. This constitutes a promising approach
100
5000 to push quantum teleportation and entanglement swap-
90
Four-photon coincidences [/6h]

ping experiments out of the lab, using the existing optical

3-photon coincidences [/6h]

80
70
4000 fiber network.
The authors would like to thank Claudio Barreiro and

without BSM
60
3000
50 Jean-Daniel Gautier for technical support. Financial
40 2000 support by the Swiss NCCR Quantum Photonics, and
30
by the European project RamboQ is acknowledged.
20 1000
10
0 0
100 200 300 400 500 600
Phase [degrees]
[1] M. Zukowski, A. Zeilinger, M. A. Horne, and A. K. Ekert
Phys. Rev. Lett. 71, 4287-4290 (1993)
FIG. 3: Two-photon interferences for swapped photons, as [2] H.-J. Briegel et al, Phys. Rev. Lett. 81, 59325935 (1998)
a function of the phase of one interferometer. Plain squares [3] L.-M. Duan, M.D. Lukin, J.I.Cirac and P.Zoller, Nature
represents the detection between photons A and D, without 414, 413 (2001)
conditioning on the BSM. Errors bars are too small to be [4] E. Waks et al, Phys. Rev. A 65, 052310 (2002), B. C.
represented. Open circles represent four-photon coincidences, Jacobs et al, Phys.Rev.A 66 052307 (2002) .
i.e two photon interference conditioned on a BSM [5] D.Collins, N.Gisin and H.de Riedmatten,
quant-ph 0311101, accepted for publication in
J.Mod.Opt.
[6] I. Marcikic et al, J.Mod.Opt. 51, 1011 (2004)
shows a measurement of two photon interference. The [7] S.Bose, V.Vedral and P.L.Knight, Phys.Rev. A 57, 822
plain squares represent coincidences between Alice’s and (1998)
Bob’s photons, without conditioning on a BSM as a func- [8] E. Knill, R.Laflamme and G. Milburn, Nature 409, 46
tion of the phase of one interferometer. The fact that the (2001)
coincidence count rate does not vary significantly with [9] J.W.Pan, D. Bouwmeester, H.Weinfurter and A.
the phase is a confirmation that the two photons are Zeilinger, Phys. Rev. Lett. 80, 3891 (1998)
[10] T. Jennewein, G. Weihs, J.-W. Pan, and A. Zeilinger,
completely independent in this case. However, if we now Phys.Rev.Lett. 88, 017903 (2002).
condition on a successful BSM (open circles), we see a [11] F.Sciarrino, E. Lombardi, G. Milani, and F. De Martini,
sinusoidal variation with a fitted raw (i.e. without noise Phys. Rev. A 66, 024309 (2002)
subtraction) visibility of (80 ± 4%), leading to a fidelity [12] D. Fattal et al Phys. Rev. Lett. 92, 037904 (2004)
F2 of (85 ± 3.25%) high enough to demonstrate a tele- [13] X. Jia et al, quant-ph/0407024
portation of entanglement and to violate a Bell inequality [14] I.Marcikic et al, quant-ph/0404124
[15] H. de Riedmatten et al, Phys.Rev.A 69, 050304(R)
with the teleported photons by more than two standard
(2004)
deviations. The whole measurement lasted 78 hours, [16] Because of the probabilistic nature of PDC, the probabil-
which demonstrates the robust character of our scheme. ity to create two photon pairs in one source is the same
The non perfect visibility of the interference fringe is at- as the probability of creating one pair in each source.
tributed mainly to the limited fidelity of the BSM. The Hence, the four photons need to be detected in order to
main limiting factor is the non-vanishing probability of post-select the cases where (at least) one pair is created
creating multiple photon pairs in one laser pulse, due in each source, which are the events we are interested in.
to the probabilistic nature of PDC [21, 25]. The visi- [17] J. Brendel, W. Tittel, H. Zbinden and N. Gisin, Phys.
Rev. Lett. 82, 2594 (1999).
bility could be improved by reducing the pump power [18] Note that this is valid only if the laser used is the same for
but this would reduce the four-photon coincidence count both sources. In case of distant sources pumped by inde-
rate. Note that the key parameters in order to increase pendent lasers, the pump interferometers phases should
the four photon coincidence count rate without degrading be stabilized.
the correlations are the quantum efficiencies of detectors [19] H.Weinfurter, Europhys.Lett 25, 559 (1994)
and the coupling efficiencies into the single mode fibers. [20] C. K. Hong, Z. Y. Ou, and L. Mandel, Phys. Rev. Lett.
59, 2044-2046 (1987)
In summary, we have reported the first demonstration
[21] H.de Riedmatten et al, Phys. Rev. A 67, 022301 (2003)
of entanglement swapping over long distance in optical [22] M.Zukowski et al, Ann. NY Acad. Sci. 775, 91-102 (1995)
fibers. We used two pairs of time-bin entangled qubits [23] A.Peres, Phys. Rev. Lett. 77, 1413(1996)
encoded into photons at telecommunications wavelengths [24] J.F. Clauser, M.A. Horne, A. Shimony, and R.A. Holt,
and created in spatially separated sources. The visibility Phys. Rev. Lett. 23, 880 (1969).
 5

[25] V.Scarani et al, quant-ph/0407189