RAPID COMMUNICATIONS

PHYSICAL REVIEW A, VOLUME 61, 010302共R兲

Unconditional teleportation of continuous-variable entanglement

P. van Loock1 and Samuel L. Braunstein1,2
1
SEECS, University of Wales, Bangor LL57 1UT, United Kingdom
2
Hewlett-Packard Labs, Mail Box M48, Bristol BS34 8QZ, United Kingdom
共Received 7 September 1999; published 8 December 1999兲
We give a protocol and criteria for demonstrating unconditional teleportation of continuous-variable en-
tanglement 共i.e., entanglement swapping兲. The initial entangled states are produced with squeezed light and
linear optics. We show that any nonzero entanglement 共any nonzero squeezing兲 in both of two entanglement
sources is sufficient for entanglement swapping to occur. In fact, realization of continuous-variable entangle-
ment swapping is possible using only two single-mode squeezed states.

PACS number共s兲: 03.67.⫺a, 03.65.Bz, 42.50.Dv

Quantum teleportation enables reliable transportation of tanglement gives us also a protocol for the experimental veri-
quantum information encoded in nonorthogonal quantum fication of entanglement swapping.
states. It is only possible with entanglement. Teleportation For our entanglement swapping scheme 共Fig. 1兲, we need
was originally proposed for discrete variables 关1兴 and later two entangled states of the electromagnetic field: a two-
also for continuous variables 关2,3兴. Discrete-variable telepor- mode squeezed state of mode 1 and mode 2 and a two-mode
tation has been performed experimentally for single-photon squeezed state of mode 3 and mode 4. This can be described
polarization states 关4,5兴. Continuous-variable teleportation in the Heisenberg representation by
has been realized for coherent states of electromagnetic field ⫺r 2 (0) 冑
modes 关6兴. But coherent states, although nonorthogonal, are x̂ 1 ⫽ 共 e ⫹r 1 x̂ (0)
1 ⫹e x̂ 2 兲 / 2,
very close to classical states. A real challenge for quantum
⫹r 2 (0) 冑
teleportation is the teleportation of truly nonclassical states p̂ 1 ⫽ 共 e ⫺r 1 p̂ (0)
1 ⫹e p̂ 2 兲 / 2,
like entangled states. This ‘‘entanglement swapping’’ was
⫺r 2 (0) 冑
first introduced in the context of single-photon polarization x̂ 2 ⫽ 共 e ⫹r 1 x̂ (0)
1 ⫺e x̂ 2 兲 / 2,
states 关7兴. It means to entangle two quantum systems that
have never directly interacted with each other. With single ⫹r 2 (0) 冑
p̂ 2 ⫽ 共 e ⫺r 1 p̂ (0)
1 ⫺e p̂ 2 兲 / 2,
photons, it has already been demonstrated experimentally
共1兲
关8兴. Practical uses of entanglement swapping have been sug- ⫺s 2 (0) 冑
x̂ 3 ⫽ 共 e ⫹s 1 x̂ (0)
3 ⫹e x̂ 4 兲 / 2,
gested 关9–12兴 and it has also been generalized for multipar-
ticle systems 关9兴. All these investigations have only referred ⫹s 2 (0) 冑
to discrete-variable systems, namely, two-level systems. We p̂ 3 ⫽ 共 e ⫺s 1 p̂ (0)
3 ⫹e p̂ 4 兲 / 2,
will demonstrate that entanglement swapping can also be
⫺s 2 (0) 冑
realized in continuous-variable systems where the source of x̂ 4 ⫽ 共 e ⫹s 1 x̂ (0)
3 ⫺e x̂ 4 兲 / 2,
entanglement is two-mode squeezed light. In contrast to the
scheme of Polkinghorne and Ralph 关13兴 where polarization- ⫹s 2 (0) 冑
p̂ 4 ⫽ 共 e ⫺s 1 p̂ (0)
3 ⫺e p̂ 4 兲 / 2,
entangled states of single photons are teleported using
squeezed-state entanglement, in our scheme both entangled where a superscript 共0兲 denotes initial vacuum modes. The
states are produced with squeezed light. This enables uncon- operators x̂ and p̂ represent the electric quadrature ampli-
ditional teleportation of entanglement without postselection tudes 共the real and imaginary parts of the mode’s annihila-
of ‘‘successful’’ events by photon detections. Unconditional tion operator兲. These two-mode squeezed vacuum states can
teleportation of continuous-variable entanglement has been be generated either directly as the output of a nondegenerate
independently investigated by Tan 关14兴. We will compare
Tan’s results with ours at the end.
Due to the finite degree of entanglement arising from
squeezed states, the entanglement that emerges from en-
tanglement swapping is never as good as the entanglement of
the two initial entanglement sources. However, entanglement
swapping as here proposed occurs every inverse bandwidth
time and is very efficient 共near unit efficiency兲. The fidelity
criterion for coherent-state teleportation 关15兴 will enable us
to recognize the entanglement produced from entanglement
swapping. The maximum average fidelity achievable using
the output of entanglement swapping for teleportation indi- FIG. 1. Entanglement swapping using four squeezed vacuum
cates the quality of the entanglement from entanglement states. Before the detections, mode 1 is entangled with mode 2 and
swapping. Applying this operational quality criterion for en- mode 3 is entangled with mode 4.

1050-2947/99/61共1兲/010302共4兲/$15.00 61 010302-1 ©1999 The American Physical Society

 RAPID COMMUNICATIONS

P. van LOOCK AND SAMUEL L. BRAUNSTEIN PHYSICAL REVIEW A 61 010302共R兲

optical parametric amplifier 关16兴 or by combining two inseparable states 关20兴. However, we will see that Alice and
squeezed vacuum modes at a beam splitter 共see Fig. 1兲. But Bob cannot use mode 1 and 4 for subsequent quantum tele-
note that the two independently squeezed single-mode states portation without information about Claire’s measurement
combined at a beam splitter to create entanglement need not results. At least one of them, Alice or Bob, has to receive
be equally squeezed. In fact, even one single-mode squeezed from Claire the information that the detection of mode 2 and
state superimposed with vacuum yields an entangled two- 3 has been performed and its results. Finally, the entangle-
mode state 关17,18兴, enabling quantum teleportation 关19兴. ment of the single projections is ‘‘unwrapped’’ via appropri-
Thus, we use four different squeezing parameters r 1 , r 2 , s 1 , ate displacements. Let us assume Bob obtains the classical
and s 2 . results from Claire. Now Bob can displace mode 4 as x̂ 4
Let us introduce ‘‘Alice,’’ ‘‘Bob,’’ and ‘‘Claire’’ to illus- →x̂ ⬘4 ⫽x̂ 4 ⫹g swap冑2x u , p̂ 4 →p̂ ⬘4 ⫽ p̂ 4 ⫹g swap冑2 p v . The pa-
trate the whole protocol with entanglement swapping and rameter g swap describes a normalized gain. Bob’s mode then
subsequent teleportation of a coherent state. Alice and Claire becomes
shall share the entangled state of mode 1 and 2 while Claire
and Bob are sharing the other entangled state of mode 3 and g swap g swap g swap⫺1
4 共Fig. 1兲. Thus, initially Alice and Bob do not share an x̂ ⬘4 ⫽ e ⫹r 1 x̂ (0)
1 ⫺ e ⫺r 2 x̂ (0)
2 ⫺ e ⫹s 1 x̂ (0)
entangled state. Now Alice wants to teleport an unknown 冑2 冑2 冑2 3

coherent state to Bob and asks Claire for her assistance.

Claire combines mode 2 and mode 3 at a beam splitter and g swap⫹1
⫺ e ⫺s 2 x̂ (0)
4 ,
detects the quadratures x̂ u ⫽(x̂ 2 ⫺x̂ 3 )/ 冑2, p̂ v ⫽(p̂ 2 冑2
⫹p̂ 3 )/ 冑2. Let us write Bob’s mode 4 as
g swap g swap g swap⫹1
x̂ 4 ⫽x̂ 2 ⫺ 共 x̂ 3 ⫺x̂ 4 兲 ⫺ 冑2x̂ u , p̂ 4⬘ ⫽ e ⫺r 1 p̂ (0)
1 ⫺ e ⫹r 2 p̂ (0)
2 ⫹ e ⫺s 1 p̂ (0)
冑2 冑2 冑2 3

共2兲
p̂ 4 ⫽p̂ 2 ⫹ 共 p̂ 3 ⫹p̂ 4 兲 ⫺ 冑2p̂ v , g swap⫺1
⫹ e ⫹s 2 p̂ (0)
4 . 共5兲
and Alice’s mode 1 as
冑2
As in ‘‘usual’’ teleportation, Alice now couples the unknown
x̂ 1 ⫽x̂ 3 ⫹ 共 x̂ 1 ⫺x̂ 2 兲 ⫹ 冑2x̂ u ,
input state she wants to teleport to Bob 共described by x̂ in ,
共3兲
p̂ 1 ⫽p̂ 3 ⫹ 共 p̂ 1 ⫹p̂ 2 兲 ⫺ 冑2p̂ v .
p̂ in) with her mode 1 at a beam splitter and measures the
superpositions x̂ u⬘ ⫽(x̂ in⫺x̂ 1 )/ 冑2, p̂ ⬘v ⫽(p̂ in⫹ p̂ 1 )/ 冑2. Based
Claire’s detection yields classical results x u and p v . Bob’s on the classical results sent to him from Alice, Bob displaces
mode 4 in Eqs. 共2兲 and Alice’s mode 1 in Eqs. 共3兲 then his ‘‘new’’ mode 4 ⬘ , x̂ 4⬘ →x̂ tel⫽x̂ 4⬘ ⫹g 冑2x u⬘ , p̂ 4⬘ →p̂ tel⫽p̂ 4⬘
become ⫹g 冑2 p ⬘v , with the gain g. For g⫽1 and nonunit detector
efficiencies, Bob’s outgoing mode then becomes
x̂ 4 ⫽x̂ 2 ⫺ 冑2e ⫺s 2 x̂ (0)
4 ⫺ 冑2x u ,
g swap⫺1 g swap⫹1
p̂ 4 ⫽ p̂ 2 ⫹ 冑2e ⫺s 1 p̂ (0) x̂ tel⫽x̂ in⫹ e ⫹r 1 x̂ (0)
1 ⫺ e ⫺r 2 x̂ (0)
3 ⫺ 冑2p v , 冑2 冑2 2

共4兲
x̂ 1 ⫽x̂ 3 ⫹ 冑2e ⫺r 2 x̂ (0)
2 ⫹ 冑2x u , g swap⫺1 g swap⫹1
⫺ e ⫹s 1 x̂ (0)
3 ⫺ e ⫺s 2 x̂ (0)
冑2 冑2 4

p̂ 1 ⫽ p̂ 3 ⫹ 冑2e ⫺r 1 p̂ (0)
1 ⫺ 冑2p v .
⫹g swap冑␩ ⫺2
c ⫺1 共 x̂ d ⫹x̂ e 兲
(0) (0)

For s 1 ⫽s 2 ⫽s→⬁, the quadrature operators of mode 4 be-

come those of mode 2 up to a 共random兲 classical phase-space ⫹ 冑␩ ⫺2
a ⫺1 共 x̂ f ⫹x̂ g 兲 ,
(0) (0)

displacement. In every single projection, mode 4 gets en-

tangled with mode 1 as mode 2 has been before. For r 1 g swap⫹1 g swap⫺1
⫽r 2 ⫽r→⬁, the quadrature operators of mode 1 become p̂ tel⫽ p̂ in⫹ e ⫺r 1 p̂ (0)
1 ⫺ e ⫹r 2 p̂ (0)
those of mode 3 up to a 共random兲 classical phase-space dis- 冑2 冑2 2

placement. In every single projection, mode 1 gets entangled

g swap⫹1 g swap⫺1
with mode 4 as mode 3 has been before. Mode 2 is perfectly ⫹ e ⫺s 1 p̂ (0)
3 ⫹ e ⫹s 2 p̂ (0)
teleported to mode 4 (s→⬁) or mode 3 is perfectly tele- 冑2 冑2 4

ported to mode 1 (r→⬁) apart from local classical displace-

ments. The entanglement of one of the initial two-mode ⫹g swap冑␩ ⫺2
c ⫺1 共 p̂ k ⫹ p̂ l 兲 ⫹ 冑␩ a ⫺1 共 p̂ m ⫹p̂ n 兲 .
(0) (0) ⫺2 (0) (0)

squeezed states is completely preserved for infinite squeez- 共6兲

ing in the other two-mode squeezed state. But also for any
nonzero squeezing in both initial entanglement sources, The parameters ␩ c and ␩ a describe detector efficiencies in
Claire’s detection of mode 2 and 3 projects mode 1 and 4 on Claire’s and Alice’s detections, respectively. Note that, for
010302-2
 RAPID COMMUNICATIONS

UNCONDITIONAL TELEPORTATION OF CONTINUOUS- . . . PHYSICAL REVIEW A 61 010302共R兲

g swap⫽1, Bob’s teleported mode from Eqs. 共6兲 is the same as ping took place. The reduced states of mode 1 and 4 after
if Alice teleports her input state to Claire with unit gain using detecting mode 2 and 3 are inseparable for any r⬎0 and s
the entangled state of mode 1 and 2, and Claire teleports the ⬎0 关20兴.
resulting output state to Bob with unit gain using the en- Let us now consider the case where each of the two initial
tangled state of mode 3 and 4. entangled states is generated with only one single-mode
The teleportation fidelity for a coherent state input ␣ in , squeezed state. We set r 1 ⫽s 1 ⫽r and r 2 ⫽s 2 ⫽0. With unit
defined as F⬅ 具 ␣ in兩 ␳ˆ tel兩 ␣ in典 ⫽ ␲ Q tel( ␣ in) 关15兴, is efficiency, we find the optimum gain g swap⫽tanh r. The op-
timum fidelity then becomes

F⫽
1
2 冑␴ x ␴ p
冋
exp ⫺ 共 1⫺g 兲 2 冉 x 2in
2␴x
⫹
p 2in
2␴p
冊册 , 共7兲 F opt⫽ 兵 关 1⫹2e ⫹2r / 共 e ⫹2r ⫹1 兲 ⫹ 共 tanh r 兲 2 共 ␩ ⫺2
c ⫺1 兲

⫹ ␩ ⫺2
a ⫺1 兴关 1⫹2e
⫺2r
/ 共 e ⫺2r ⫹1 兲
where ␴ x and ␴ p are the variances of the Q function of the
teleported mode for the corresponding quadratures. Using ⫹ 共 tanh r 兲 2 共 ␩ ⫺2 ⫺2
c ⫺1 兲 ⫹ ␩ a ⫺1 兴 其
⫺1/2
. 共11兲
Eqs. 共6兲, the fidelity becomes for g⫽1,

F⫽ 关 1⫹ 共 g swap⫺1 兲 2 共 e ⫹2r 1 ⫹e ⫹2s 1 兲 /4 Note that this fidelity can be optimized further for nonunit
efficiency, as we have only used the optimum gain for unit
⫹ 共 g swap⫹1 兲 2 共 e ⫺2r 2 ⫹e ⫺2s 2 兲 /4 efficiency. With unit efficiency ( ␩ c ⫽ ␩ a ⫽1) this fidelity ex-
ceeds the classical limit F opt⬎ 21 for any nonzero squeezing
⫹g swap
2
共 ␩ ⫺2 ⫺2
c ⫺1 兲 ⫹ ␩ a ⫺1 兴
⫺1/2
r 1 ⫽s 1 ⫽r⬎0. Thus, entanglement swapping can be realized
⫻ 关 1⫹ 共 g swap⫺1 兲 2 共 e ⫹2r 2 ⫹e ⫹2s 2 兲 /4 even with only two single-mode squeezed states, provided
that two initial entangled states are produced. Indeed, the
⫹ 共 g swap⫹1 兲 2 共 e ⫺2r 1 ⫹e ⫺2s 1 兲 /4 creation of entanglement is possible using only one single-
mode squeezed state for any nonzero squeezing 关17–19兴.
⫹g swap
2
共 ␩ ⫺2 ⫺2
c ⫺1 兲 ⫹ ␩ a ⫺1 兴
⫺1/2
. 共8兲 Therefore we can generally say that any nonzero entangle-
ment in both of the two initial entanglement sources is suf-
For unknown coherent input states, an 共average兲 fidelity ficient for entanglement swapping to occur. In order to
F⬎ 12 is only achievable using entanglement 关15兴. Thus, if achieve perfect teleportation of arbitrary coherent states with
for some g swap 共for some local operation on mode 4 by Bob fidelity F⫽1, four infinitely squeezed states r 1 ⫽r 2 ⫽s 1
based on Claire’s results兲 F⬎ 21 , entanglement swapping ⫽s 2 ⫽r→⬁ are necessary and Bob has to perform a dis-
must have taken place. Otherwise Alice and Bob, who ini- placement with g swap⫽1 using Claire’s results. It is impos-
tially did not share entanglement, were not able to beat the sible for Alice and Bob to achieve quantum teleportation of
classical fidelity limit using mode 1 and 4. The assumption unknown coherent states with F⬎ 21 for g swap⫽0, i.e., with-
g⫽1 is the optimal choice for Bob’s local operation based out a local operation based on the results of Claire’s detec-
on Alice’s results. tion. The optimum fidelity using mode 1 and 4 after en-
Let us first consider four equally squeezed states r 1 ⫽r 2 tanglement swapping is worse than the optimum fidelity
⫽s 1 ⫽s 2 ⫽r. In this case with unit efficiency ( ␩ c ⫽ ␩ a ⫽1), using the entanglement of the initial modes 1 and 2 or 3 and
the fidelity is optimized for g swap⫽tanh 2r (g⫽1) and be- 4. This indicates that the degree of entanglement after en-
comes F opt⫽(1⫹1/cosh 2r)⫺1. For any r⬎0, we obtain tanglement swapping deteriorates compared to the initial en-
F opt⬎ 12 . For ␩ c ⫽1 and ␩ a ⫽1, the optimum gain is g swap tangled states. In Fig. 2 is shown the comparison between the
⫽sinh 2r/(cosh 2r⫹␩⫺2 c ⫺1). For the more general case r 1 optimum fidelities of coherent-state teleportation using en-
⫽r 2 ⫽r and s 1 ⫽s 2 ⫽s, we find the optimum gain tangled states produced from entanglement swapping and
without swapping.
sinh 2r⫹sinh 2s The maximum fidelity achievable using entanglement
g swap⫽ . 共9兲
cosh 2r⫹cosh 2s⫹2 ␩ ⫺2
c ⫺2
produced with one single-mode squeezed state is F⫽1/冑2
关19兴. The maximum fidelity achievable using the output of
Using this gain we obtain the optimum fidelity with unit entanglement swapping with two equally squeezed single-
efficiency mode states is F⫽1/冑3. For 6-dB squeezing and detectors
with efficiency ␩ 2 ⫽0.99, the optimum fidelity using the out-
再
F opt⫽ 1⫹
cosh关 2 共 r⫺s 兲兴 ⫹1
cosh 2r⫹cosh 2s 冎 ⫺1
. 共10兲
put of entanglement swapping with two equally squeezed
single-mode states becomes F⫽0.5201. Squeezing of 10 dB
and the same efficiency yields F⫽0.5425. Here, the gain
This fidelity is equal to 21 and never exceeds the classical g swap⫽tanh r has been chosen, that is, the optimum gain with
limit if either r⫽0 or s⫽0. The reduced states of mode 1 two equally squeezed single-mode states for unit efficiency.
and 4 after the detection of mode 2 and 3 are separable if Tan proposes continuous-variable entanglement swapping
either r⫽0 or s⫽0 关20兴. Both initial two-mode states need as the teleportation of the signal beam of a parametric am-
to be squeezed and hence entangled for entanglement swap- plifier using the entanglement between signal and idler beam
ping to occur. Then, for any nonzero squeezing r⬎0 and s of another parametric amplifier 关14兴. The entanglement of
⬎0, we obtain F opt⬎ 21 , indicating that entanglement swap- the teleported signal beam with the idler beam in Tan’s pro-
010302-3
 RAPID COMMUNICATIONS

P. van LOOCK AND SAMUEL L. BRAUNSTEIN PHYSICAL REVIEW A 61 010302共R兲

rable for any nonzero entanglement in both initial sources,

they should enable quantum teleportation, as has been
shown. We have demonstrated, that in this case, the reduced
state ␳ˆ 14 can always be transformed by some local displace-
⬘ that provides F⬎ 21 in coherent-state tele-
ment to a state ␳ˆ 14
portation and has quadratures violating 具 (x̂ 1 ⫺x̂ 4⬘ ) 2 典 ⫹ 具 (p̂ 1
⫹ p̂ 4⬘ ) 2 典 ⭓1. We have given the optimum gain g swap of this
displacement for any verification, either by further teleporta-
tion or by simple detection 关22兴.
In Tan’s scheme 关14兴, unit gain g swap has been chosen.
Therefore entanglement is only confirmed if the initial states
exceed a certain degree of squeezing 共e.g., for equally
squeezed states 3 dB兲. By choosing the right gain, entangle-
ment can be verified for any squeezing, in principle even if
the initial entangled states are built from the minimal re-
FIG. 2. Optimum fidelity for the teleportation of an arbitrary
coherent state (g⫽1) using 共a兲 entanglement produced with two source of two single-mode squeezed states.
equally squeezed single-mode states, 共b兲 the output of entanglement Detecting the output and applying Tan’s 共or Duan et al.’s兲
swapping with four equally squeezed single-mode states, 共c兲 en- inequality for verification is easy, but it requires bringing the
tanglement produced with one single-mode squeezed state, 共d兲 the entangled subsystems together and measuring states that are
output of entanglement swapping with two equally squeezed single- now local. This provides only an indirect confirmation that
mode states, and 共e兲 the state as in 共d兲 with detector efficiencies the entanglement is preserved through teleportation. Our
␩ 2c ⫽ ␩ 2a ⫽0.95. verification leaves the subsystems separate and directly dem-
onstrates the entanglement by exploiting it in a second round
tocol is confirmed by combining them at a beam splitter and of teleportation.
detecting the x̂ quadrature at one output and the p̂ quadrature In summary, we have given a protocol for continuous-
at the other output: x̂⫽(x̂ 1 ⫺x̂ 4 )/ 冑2, p̂⫽( p̂ 1 ⫹p̂ 4 )/ 冑2. A variable entanglement swapping, i.e., for the unconditional
violation of the uncertainty relation 具 ⌬x̂ 2 典具 ⌬p̂ 2 典 ⭓1/16 teleportation of entanglement, using squeezed light and lin-
proves the entanglement of mode 1 and 4 关14兴. A similar, ear optics. Entanglement swapping occurs for any nonzero
sufficient inseparability criterion for continuous-variable sys- entanglement 共any nonzero squeezing兲 in both of the two
tems has been given very recently 关21兴 that in the context of initial entanglement sources. This can be realized even with
our entanglement swapping scheme would require the viola- only two single-mode squeezed states. Our verification
scheme would provide a compelling demonstration of uncon-
tion of 具 (x̂ 1 ⫺x̂ 4 ) 2 典 ⫹ 具 (p̂ 1 ⫹p̂ 4 ) 2 典 ⭓1.
ditional teleportation of entanglement.
It is obvious that the violation of these inequalities corre-
ponds to a fidelity F⬎ 21 in our subsequent coherent-state This work was funded by a ‘‘DAAD Doktorandenstipen-
teleportation. However, we have directly looked at the sepa- dium im Rahmen des gemeinsamen HSP III von Bund und
rability of the projected states of mode 1 and 4 after the Ländern’’ and EPSRC Grant No. GR/L91344. P.v.L. thanks
detection of mode 2 and 3 关20兴. Since these states are insepa- Sze M.Tan for sending his manuscript.

关1兴 C. H. Bennett et al., Phys. Rev. Lett. 70, 1895 共1993兲. 关13兴 R. E. S. Polkinghorne and T. C. Ralph, Phys. Rev. Lett. 83,
关2兴 L. Vaidman, Phys. Rev. A 49, 1473 共1994兲. 2095 共1999兲; e-print quant-ph/9906066.
关3兴 S. L. Braunstein and H. J. Kimble, Phys. Rev. Lett. 80, 869 关14兴 S. M. Tan, Phys. Rev. A 60, 2752 共1999兲.
共1998兲. 关15兴 S. L. Braunstein, C. A. Fuchs, and H. J. Kimble, e-print
关4兴 D. Bouwmeester et al., Nature 共London兲 390, 575 共1997兲. quant-ph/9910030.
关16兴 M. D. Reid and P. D. Drummond, Phys. Rev. Lett. 60, 2731
关5兴 D. Boschi et al., Phys. Rev. Lett. 80, 1121 共1998兲.
共1988兲.
关6兴 A. Furusawa et al., Science 282, 706 共1998兲.
关17兴 Y. Aharonov et al., Ann. Phys. 共N.Y.兲 39, 498 共1966兲.
关7兴 M. Zukowski et al., Phys. Rev. Lett. 71, 4287 共1993兲. 关18兴 M. G. A. Paris, Phys. Rev. A 59, 1615 共1999兲.
关8兴 J.-W. Pan et al., Phys. Rev. Lett. 80, 3891 共1998兲. 关19兴 P. van Loock and S. L. Braunstein, e-print quant-ph/9906021.
关9兴 S. Bose, V. Vedral, and P. L. Knight, Phys. Rev. A 57, 822 关20兴 P. van Loock and S. L. Braunstein 共unpublished兲.
共1998兲. 关21兴 L.-M. Duan et al., e-print quant-ph/9908056.
关10兴 S. Bose, V. Vedral, and P. L. Knight, Phys. Rev. A 60, 194 关22兴 Polkinghorne and Ralph 关13兴 have found a similar gain condi-
共1999兲. tion that ensures that the correlations of single polarization-
关11兴 H.-J. Briegel et al., Phys. Rev. Lett. 81, 5932 共1998兲. entangled photons are verified via Bell inequalities after tele-
关12兴 W. Dür et al., Phys. Rev. A 59, 169 共1999兲. porting them.

010302-4