Quantum Information and Computation, Vol. 1, No.

0 (2001) 000000
c Rinton Press

Photonic Entanglement for Fundamental Tests and Quantum Communication

Wolfgang Tittel
Group of Applied Physics, University of Geneva (GAP)
arXiv:quant-ph/0107156v1 31 Jul 2001

20, Rue de lEcole-de-Medecine, 1211 Geneve 4, Switzerland

Gregor Weihs
Institut of Experimental Physics, University of Vienna (UNIVIE)
Boltzmanngasse 5, 1090 Wien, Austria

Received (February 1, 2008)

Revised (February 1, 2008)

Entanglement is at the heart of fundamental tests of quantum mechanics like tests of

Bell-inequalities and, as discovered lately, of quantum computation and communication.
Their technological advance made entangled photons play an outstanding role in entan-
glement physics. We give a generalized concept of qubit entanglement and review the
state of the art of photonic experiments.

Keywords: entanglement, non-locality, quantum communication

1
2 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Contents
1 Introduction 3

2 The superposition principle: photonic qubits and entanglement 5

2.1 The qubit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 A theoretical approach . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Preparing and measuring a qubit . . . . . . . . . . . . . . . . . . . . 6
2.2 Two-particle entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Photon pair sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Types of entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Measuring entanglement . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 N-particle entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Fundamental tests of nonlocality 19

3.1 Bell inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.1 Closing loopholes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.2 Relativistic configurations . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 GHZ states and nonlocality . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4 Quantum communication 33
4.1 Quantum cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.1 Quantum cryptography based on faint laser pulses . . . . . . . . . . 33
4.1.2 Quantum cryptography based on photon-pairs . . . . . . . . . . . . 37
4.2 Quantum dense coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Quantum teleportation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Entanglement swapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.5 Purification and distillation . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Conclusion 50
 Wolfgang Tittel and Gregor Weihs . . . 3

1 Introduction
Due to its importance for understanding the properties of the quantum world and its
role in applications in the new domain of quantum computation and communication,
entanglement got more and more attention within the physics community throughout the
last 70 years, and, lately, even in the general public.aThe interest in entanglement, a term
invented by Erwin Schrodinger in 19351 ,bwas triggered by a paper by Einstein, Podolsky
and Rosen (EPR) that was published also in 1935.2 In this famous paper, often referred
to as EPR paradox, the authors analyze the predictions for a two-particle state where
neither particle can be considered in a state independent from the other. In contrast, both
subsystems, even if at arbitrarily large distance, form a single entangled system. Based
on the assumption of locality, i.e. that the choice of the type of measurement performed
on one particle can not influence the properties of the other particle, they argued that the
description of reality as given by the wavefunction is not complete.
The question whether or not this is true, or, in other words, whether or not entangle-
ment (and hence non-locality) exists became a very important fundamental issue. It was
Bells discovery of the so-called Bell inequalities in 19643 and their extension to experi-
mental conditions by Clauser et al. in 1969 and 19744,5 that transferred the former purely
philosophical debate to the realms of laboratory experiments. Beginning with the first test
of Bell inequalities in 1972,6 an increasing number of more and more refined experiments
has been performed.7,8 Although the type of particles entangled is of no importance to
demonstrate the existence of non-locality, by far most experiments relied on entangled
photons. Nowadays, although not all experimental loopholes have been closed simultane-
ously in a single experiment (but all of them have already been closed), it is commonly
believed that quantum non-locality is indeed real. Nevertheless, there is still interest in
performing more Bell-type tests. A first motivation is to examine the boundary between
the quantum and the classical world,9,10 a second one are experiments extending the tra-
ditional set-up for Bell-type tests to relativistic configurations and investigating so-called
relativistic non-locality.11 While massive particles prone to decoherence are used for
experiments of the first kind,12 it was again photons that served for experiments of the
second kind.13
Apart from these fundamental motivations, the recent discovery that processing and
exchange of information based on quantum systems enable new forms of computation
and communication, more powerful than its classical analogs, engendered further interest
in entangled systems (see i.e. Refs. [14,15,16 ]). Best known applications in the domain
of quantum communication hence in the domain where photons are most likely best
suited for are quantum cryptography (for a recent review see Ref. [17 ]) and quantum
teleportation.18
In this article, we review experiments based on photonic entanglement, addressing
both fundamental as well as applied aspects. However, although there has been consid-
erable progress in experiments based on continuous quantum variables as well (see e.g.
asearching the internet for entanglement, we found 45.000 pages!
b Entanglement is translated from the original German word Verschrankung.
4 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Refs. [19,20,21 ]), we will focus only on entanglement between discrete two-level quantum
systems (now called quantum bits or qubits). This enables us to pursue a rather formal
approach, summarizing all experiments under the aspect of experiments with entangled
qubits.
The article is structured along the following lines: In Section 2 we introduce the quan-
tum toolbox in form of sources and analyzers for qubits and entangled qubits, respectively,
and we explain various experimental approaches. The fact that all different realizations
of a qubit are formally equivalent (a qubit is a qubit) then renders the task of presenting
the variety of experiments quite simple: With different arrangements of these few basic
building blocks various issues can be addressed experimentally. This concerns tests of
non-locality (Section 3) as well as applications of entanglement in the domain of quantum
communication (section 4). As we will see, it is sometimes enough to change only minor
things like analyzer settings in order to continuously pass from one to the other side
like from tests of Bell inequalities to quantum tomography and quantum cryptography.
Finally, a short conclusion is given in Section 5.
 Wolfgang Tittel and Gregor Weihs . . . 5

2 The superposition principle: photonic qubits and

entanglement
One of the most important features of quantum theory is certainly the superposition
principle. While we find this property already in classical wave theories, e.g. in Youngs
famous double-slit experiment in classical optics, quantum theory allows for instance to
describe objects that have traditionally been considered well localized in space as being in
a superposition of different positions, well separated in space. The superposition-principle
can be applied to any physical property, and it is at the origin of the quantum-paradoxes
as well as of quantum information theory.

2.1 The qubit

2.1.1 A theoretical approach
The most important entity of classical information theory is the bit. A bit can either
have the value 0 or 1 with both values separated by a large energy gap so that the
unwanted spontaneous transition from one to the other value is extremely unlikely.c
The quantum mechanical analog of the bit is the quantum bit or qubit. It is a two-
state quantum system with the basic states |0i and |1i forming an orthogonal basis in the
qubit space. In contrast to the classical bit, it is possible to create qubits in a coherent
superposition of |0i and |1i, the general state being

|iqubit = |0i + ei |1i (2 + 2 = 1). (1)

Qubits can be represented graphically on the qubit-spheredpictured in Fig. 1. The states |0i
and |1i are localized on the poles of the sphere, any superposition of |0i and |1i with equal
coefficients and are represented on the equator, and qubits with different coefficients
lie on a circle with polar angle tan() = /. Note that any two states represented on
opposite sides of the sphere form a orthonormal basis in the two-dimensional Hilbert space
describing the qubit.
In contrast to classical bits, the outcome of a measurement of a qubit is not always
deterministic. For the general qubit state given in Eq. 1, one finds the value 0 with
probability 2 and the value 1 with probability 2 . Note that this could still be achieved
with a classical bit in a mixture between 0 and 1. However, the unique feature of a
quantum bit is that the basic states |0i and |1i are superposed coherently. Let us consider
the state
1 
| i = |0i + |1i .

(2)
2
Measuring this state in a rotated basis with eigenvectors |0 i = |0i+|1i and |1 i = |0i|1i,
we always find the result 0 . This contrasts with an incoherent mixture between |0i and
|1i that stays a mixture in any basis and leads to either result with equal probabilities.
c The bit error rate in standard telecommunication is 109 1012 .
dDepending on the physical property represented also known as Bloch- or Poincare-sphere.
6 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

The transition from a coherent superposition to an incoherent one can be caused by

decoherence. In contrast to the first case that is represented on the shell of the qubit
sphere, an incoherent superposition can be found closer to the origin of the sphere with a
completely incoherent one represented in the origin.

Fig. 1. The general qubit sphere. Coherent superpositions of |0i and |1i lie on the shell of the
sphere, incoherent ones closer to the origin. All states represented on opposite sides on the shell
of the sphere form a orthonormal basis in the two-dimensional qubit space.

2.1.2 Preparing and measuring a qubit

Although, from a theoretical point of view, a qubit is just a qubit independent of its
implementation, one must identify the abstract qubit space with a physical property when
planning an experiment. There are various ways in which qubits can be realized using
single photons.e Every degree of freedom that is available can in principle be exploited.
The available properties are the photons polarization, spatial mode, emission time, or
their frequency.
In addition, depending on the specific goal, there are initial considerations concerning
the wavelength of the photons used: If the goal is to demonstrate the existence of a certain
quantum effect, it is a good idea to work at a wavelength where high efficiency and low noise
single photon detectors (based on silicon avalanche photo diodes (APD)) are commercially
available, hence at around 700800 nm. If the wavelength has to be compatible with optical
fibers as often requested for quantum cryptography or other long distance applications of
e Notethat the generation of a single photon is far from being obvious. In quantum cryptography, single
photons are often mimicked by faint laser pulses with a mean photon number of 0.1.
 Wolfgang Tittel and Gregor Weihs . . . 7

quantum communication, the absorption of the fibers require to work in the second or
third telecommunication window (at 1310 and 1550 nm, respectively). Here, only home
made detectors based on Germanium or InGaAs APDs are available. Obviously, the same
reflections hold for the creation of entangled qubits (Section 2.2.1) as well. For a more
detailed discussion of technological issues, see Gisin et al.17

Polarization qubits The most well known realization of a qubit is probably the one
using orthogonal states of polarization. In this case, the qubit-sphere is identical with the
well-known Poincare sphere. We identify left |li and right |ri circular polarized photons
as our basis states |0i and |1i; they are represented on the poles of the sphere. Linear
polarization of any orientation as an equally weighted coherent superposition of |li and
|ri can be found on the equator, and elliptically polarized light elsewhere. Completely
depolarized light as an incoherent superposition of right and left circular polarized photons
is represented by a point located at the origin. Polarization qubits can be created and
measured using polarizers and waveplates oriented at various angles.

Spatial-mode qubits Another possibility to realize a qubit is shown in Fig. 2. Using

a variable coupler and a phase shifter, Alice can create any desired superposition of a
photon being in mode |0i and in mode |1i. A similar set-up serves to analyze the qubit.
If Bob uses for instance a symmetrical coupler and
a phase =0, a click in detector 0
collapses the photon state to |0 i = 12 |0i + i |1i and in detector 1 to the orthogonal
state |1 i = 12 |0i i |1i . Using a completely asymmetrical coupler, the photon state

is projected onto the basis spanned by the states |0 i = |0i and |1 i = |1i. Spatial-mode
qubits would not be very practical for transmitting quantum information since the phase
between |0i and |1i is easily randomized by different environments acting on the different
modes.

Fig. 2. Creation and measurement of a spatial-mode qubit.

Time-bin qubits A much more robust realization of |0i and |1i in so-called time-bin
qubits is shown in Fig. 3. The switch at Alices is used to transfer the amplitudes of
8 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

both spatial modes arriving with time-difference t large compared to the photons
coherence time (localization) without losses into one mode. The net effect is to create
a superposition of amplitudes describing a photon in two different time-bins. To undo this
transformation, Bob uses a second switch, delaying now the amplitude of the first time-
bin with respect to the amplitude of the second one so that both arrive simultaneously
at the variable coupler identically to the measurement of the mode qubit. This set-up
corresponds to systems developed for faint laser-pulse based quantum cryptography (see
Section 4.1) by British Telecom22 , Los Alamos National Laboratory23 and, in a modified
and even more robust plug&play form, by one of our groups (GAP).24,25 The good
performance of these systems underlines the robustness of time-bin qubits with respect to
decoherence effects as encountered while transiting down an optical fiber.

Fig. 3. Creation and measurement of a time-bin qubit.

Frequency qubits Finally, qubits can in principle be realized using a superposition of

basic states at frequencies |1 i and |2 i. This resembles much schemes in atomic physics
where different energy levels are used to realize a qubit. However, the superposition of
the two basic states is probably difficult to achieve with photons, and to our knowledge,
no experiment has been reported to date. Note nevertheless that there is related work
concerning cryptography with frequency states26 or based on phase-modulated light 27 .

Superposition in higher dimensions: qu-nits All we said so far was based on super-
position of two orthogonal states. Although this is general for polarization, two dimensions
are only one possibility for superpositions of different modes, emission times, frequencies,
or orbital angular momenta28 which are not restricted to two dimensional Hilbert space.
Fig. 4 shows the straight-forward generalization of a time-bin qubit to a 4-dimensional
qu-quart.29

2.2 Two-particle entanglement

Entanglement can be seen as a generalization of the superposition principle to multi-
particle systems, a principle which is already at the heart of the qubit. Entangled qubits
can be described by
 Wolfgang Tittel and Gregor Weihs . . . 9

Fig. 4. Generation and detection of qu-quarts. Depending on the coupling ratios and the phases
1 to 3 , Alice can create any four dimensional time-bin state. The analyzing device at Bobs is
identical to Alices preparation device. A click in one of his detectors corresponds to the projection
on one of the four eigenstates.

|i = |0iA |0iB + ei |1iA |1iB (3)

or
|i = |0iA |1iB + ei |1iA |0iB (4)
and again 2 + 2 = 1. The indices label the entangled photons. For equal amplitudes
and , and = 0, , Eqs. 3 and 4 reduce to the well-known Bell states

 = 1 |0i |0i |1i |1i

 

A B A B (5)
2
and
 = 1 |0i |1i |1i |0i
 

A B A B (6)
2
Entangled states states are states where each of the entangled particles has no property
of its own but where the property of the global state is well defined. This becomes clear
when we look at the density matrix representations for the global and the reduced (one
particle) state. Here we give the example of the -Bell state:

|00i |01i |10i |11i

0 0 0 0

1
 
 0
2 21 0
global =  =
0 1 1
(7)
2 2 0
0 0 0 0
and
reduced = TrA () = TrB () = 1 (8)
All properties that can be used to realize photonic qubits can be used to create en-
tangled qubits as well. Before focusing on the various realizations, we will briefly present
10 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

the two main types of photon pair sources sources that always create photons in pairs,
however, not necessarily in an entangled state.

2.2.1 Photon pair sources

Atomic cascades The first sources for entangled photons were constructed using two-
photon transitions in various elements with either very short-lived or even virtual inter-
mediate states.6,30,31 The most notable elements used were Ca and Hg. All these sources
suffered from the common drawback that the atomic decay with two emitted photons is a
three-body process. Therefore, the relative direction of one emitted photon with respect
to the other is completely uncertain. This reduces the achievable collection efficiency to
an extremely low value leading to numerous experimental problems.

Parametric down-conversion When Burnham and Weinberg32 discovered the pro-

duction of photon pairs by spontaneous parametric down-conversion (SPDC) in 1970,
they did not foresee the enormous wealth and precision of experiments that this technique
would allow. Ou and Mandel33,34 then triggered the extensive work on entanglement from
SPDC.
Spontaneous parametric down-conversion is part of a (2) nonlinear optical effect also
known as three-wave mixing. In the spontaneous case only one of the three interacting
fields usually called pump is initially excited. The two others are in the vacuum
state at first. Classically they would remain unexcited but quantum mechanically there
exists a small chance that a pump photon decays into two photons which emerge within
their coherence-time, 100 fs, simultaneously from the crystal.35
The rate of this process scales linearly with the pump intensity and the magnitude
of the nonlinear coefficient. Non-vanishing (2) -nonlinearities can only appear in non-
centrosymmetric materials. KDP (KD2 PO4 ), LiIO3 , KNbO3 , LiNbO3 , and BBO (-
BaB2 O4 ) are a few of the most widely used crystals.
Naturally, the process of SPDC is subject to conservation of energy and momentum.
The latter one is also called phase matching condition. They determine the properties
of the created photons, namely their polarization, the wavelength and the direction of
propagation (the mode). Phase matching can be achieved in birefringent crystals if either
both, or one of the down-converted photons are polarized orthogonally with respect to the
pump photons. The first type is referred to as type I phase matching, the second type
as type II phase matching. If the birefringence of the crystal can not be exploited, it is
possible to achieve so-called quasi phase matching in crystals where the sign of the (2)
non-linear coefficient is periodically reversed. SPDC is a very inefficient process; it needs
around 1010 pump photons to create one photon-pair in a given mode. Lately, two groups
demonstrated down-conversion in periodically poled lithium niobate waveguides,36,37 re-
porting unprecedented efficiencies as high as 106 four orders of magnitude more than
what has been achieved with bulk crystals .36
 Wolfgang Tittel and Gregor Weihs . . . 11

Spontaneous parametric down-conversion is possible using a continuous as well as a

pulsed pump (for problems associated to very short pump pulses, see Ref. [38 ] and ref-
erences therein). The latter is necessary if the creation time of a photon pair must be
determined exactly (see Sections 3.2, 4.3 and 4.4).

2.2.2 Types of entanglement

In order to produce photon pairs in entangled states (here entangled qubits), there must
be two possible ways of creating such a pair for example both photons in state |0i or
both photons in state |1i. First sources where based on the already mentioned cascaded
transitions in atoms (Section 2.2.1), going from a well defined state of total angular mo-
mentum of the atom to another such state of lower energy. The final state could be reached
via two possible ways producing polarization entangled photon pairs in its net effect. In
the following, we will omit these first realizations and focus only on recent sources of en-
tangled photons based on spontaneous parametric down-conversion. However, as already
mentioned, we will neither comment on squeezed states entanglement19,20 nor on entan-
glement of external angular momentum states.28 Finally, problems arising from the fact
that the number of photon-pairs is (almost) thermally distributed will be addressed only
in Section 4.1 (quantum cryptography). Here we assume photon-pairs in a n=1 Fock-state.

Polarization entanglement Most experiments to date have taken advantage of polar-

ization entanglement. The first down-conversion sources only generated pairs of photons
in product states and the entanglement was created using some additional optics. The
set-ups were rather simple, but they suffered from necessary postselection by coincidence
measurements. Nowadays more sophisticated configurations for down-conversion can pro-
duce polarization entanglement directly.

Entanglement with post-selection

The first sources were so-called type-I sources, which means that the two down-
converted photons carry identical polarization. Momentum conservation rules the
emission directions such that two photons out of an individual down-conversion pro-
cess emerge on cones centered around the pump beam. At the degenerate wavelength
(twice the pump wavelength) both photons will always be on the same cone opposite
of each other with respect to the pump beam.
The type-I down-conversion source can be used to produce polarization entanglement
if the polarization of one of the beams is rotated by 90 before it is superposed with
the other beam on a beam-splitter34 (see Fig. 5). Post-selecting the cases where the
photons exit in different spatial modes by means of a coincidence measurement yields
polarization entanglement. An appropriate description of the post-selected state is
given by
1 
|i = |0iA |1iB + ei |1iA |0iB

(9)
2
where |0i and |1i stand for horizontal and vertical polarized photons. This state can
easily be transformed into one of the Bell states (Eqs. 5 and 6).
12 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 5. Schematic of a non-collinear type-I SPDC source creating polarization entanglement after
coincidence post-selection behind a beam-splitter (BS).

A more direct way to achieve polarization entanglement is type-II down-conversion

where the down-converted photons are polarized orthogonally with respect to each
other. In the simplest way the crystal is cut such that all beams are collinear. The
pump is then separated by prisms or filters and the down-converted photons forming
a pair are probabilistically separated on a beam-splitter.39 Similar to the type-I
source, this procedure again necessitates post-selection by coincidence techniques.
Entanglement without post-selection
The non-collinear extension of type-II down-conversion helps to get rid of post-
selection since the two down-converted photons are emitted into different spatial
modes already from the beginning on. A schematic is shown in Fig. 6. This kind
of source has been realized for the first time by Kwiat et al. in 1995.40 The high
degree of entanglement and brightness (up to 400.000 coincidence counts per second
as counted with Silicon APDs) of this source made quantum teleportation and other
more complicated quantum communication protocols feasible (see Section 4).
The quest for better sources and stimulated by the success of polarization entangle-
ment from type-II sources led Kwiat et al.41 to invent a new scheme in which they
stack two thin type-I crystals with their optic axes at 90 of each other (Fig. 7). A
pump photon that has linear polarization at 45 with respect to the two optic axes
will equally likely down-convert in either crystal. The photons created in the first
crystal will be polarized orthogonally to the ones created in the second crystal. If the
crystals are thin enough this leads to polarization entanglement. Furthermore, by
varying the pump polarization one can also realize non-maximally entangled states.

Momentum or mode entanglement Rarity and Tapster have first realized momen-
tum entanglement in 1990.42 A schematic is shown in Figure 8, left-hand picture.
From the emission of a non-linear crystal, two pairs of spatial (momentum, direction)
modes are extracted by pinholes. Due to the phase matching conditions, photon pairs are
created such that whenever a photon at frequency 21 pump + is emitted into one of
 Wolfgang Tittel and Gregor Weihs . . . 13

Fig. 6. Schematic of non-collinear type-II parametric down-conversion. Extraordinary (vertical,

V) photons of of a certain wavelength emerge on the upper cone, ordinary (horizontal, H) on the
lower cone. The intersections are unpolarized and display polarization entanglement after proper
compensation of the birefringent delay incurred in the conversion crystal.

Fig. 7. Polarization entanglement using two stacked type-I down-conversion crystals with optic
axis oriented at 90 with respect to each other. Depending on the polarization of the pump,
maximally as well as non-maximally entangled states can be created.
14 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 8. Schematics of two different mode-entangled source. The phase shifter can act either locally
on one of the modes as shown in the left-hand picture (Rarity et al.42 ), or on both modes |1iA
and |1iB as shown in the right-hand picture (Ribeiro et al.43 ).

the inner two modes its partner at frequency 21 pump will be found in the opposite
outer mode. The momentum entanglement very much resembles the original EPR idea of
a state of two particles whose momenta are correlated in continuous space.2
Another realization has recently been published by Ribeiro et al.43 (see right-hand side
of Fig. 8). Two subsequent crystals are pumped by a laser having a coherence length
larger than the distance between the crystals. The superposition of the two amplitudes
describing a photon pair created either in crystal 1 or in crystal 2 leads to a mode- or
momentum-entangled state.

Time-bin entanglement Using a set-up similar to the one shown in Fig. 3, Brendel
et al.44 proposed and realized the first source for time-bin entangled qubits in 2000 (see
Fig. 9).f A classical light-pulse is split into two subsequent pulses by means of an interfer-
ometer with a large path-length difference. Pumping a nonlinear crystal, a photon pair is
created either by pulse 1 (in time-bin 1) or by pulse 2 (in time-bin 2). Depending on the
coupling ratios of the couplers in the interferometer and the phase , any maximally as
well as non-maximally entangled (pure) state can be realized similar to the polarization
entangled source mentioned above. Furthermore, this set-up can easily be extended to
create time-bin entangled qu-nits (see also Section 5).

Energy-Time entanglement Energy-time entanglement can be considered the con-

tinuous version of time-bin entanglement, hence does not belong to the class of entangled
qubits. It has been proposed already in 1989 (long before time-bin entanglement) by
Franson47 in connection with a novel test of Bell inequalities (see Section 3.1). Initially,
Franson considered a three-level atomic system with a relatively long lifetime for the initial
state 1 and for the ground state 3 , and a much shorter lifetime for the intermediate
f For related work, see also Refs. [45,46 ].
 Wolfgang Tittel and Gregor Weihs . . . 15

Fig. 9. Schematics of a source creating time-bin entangled qubits. Here we show a fiber-optical
realization of the interferometer.

state 2 (see Fig. 10). Therefore, the sum energy of both photons is very well defined
although the energy of each of the two emitted photons is very uncertain. Or, in the time
domain, although the precise emission time of a pair can be predicted only to within the
long lifetime of the initial atomic state, both photons are emitted almost simultaneously
only depending on the short lifetime of the intermediate state.
The first to realize and employ energy-time entanglement in terms of Franson-type
tests of Bell inequalities were Brendel et al.48 in 1992 and, almost simultaneously, Kwiat
et al.49 In contrast to the initial proposal which is based on cascaded atomic transitions,
both took advantage of SPDC in a non-linear crystal pumped by a coherent laser. The
long coherence time of the pump-laser now bounds the emission time of a photon pair
equivalent to the lifetime of the atomic state 1 and the coherence time of the down-
converted photons, which can be as small as 100 fs, determines the degree of simultaneity
of the emission of the photons.

Fig. 10. Schematics for the creation of energy-time entangled photon pairs based on electronic
transitions in a three-level atom. The lifetimes 1 and 3 of the initial and the final state are
supposed to be large compared to the lifetime 2 of the intermediate state.

2.2.3 Measuring entanglement

16 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Projection on Bell states Common to many protocols of quantum communication is

the necessity to determine the state of a two particle system. For qubits, this means to
project on a basis in 4-dimensional Hilbert space, spanned for instance by the four Bell
states (Eq. 5 and 6). Such a measurement is known as a Bell-, or Bell-state measurement.
Figure 11 shows a possible set-up for such a measurement in the cases of polarization
(left hand figure) and time-bin (right hand figure) qubits: The two particles enter a beam-
splitter (BS) from modes a and b. Behind the beamsplitter, each particle is subjected
to a single qubit projection measurement. In case of polarization qubits, this is often a
projection on the horizontal and vertical axis using a polarizing beam-splitter (PBS),50,51
in case of time-bin qubits, the simplest way is to chose a projection on time-bins by looking
at the detection time.44 If now both photons exit the BS via different output modes (c
and d, respectively) and are found to be projected on different eigenstates of the analyzers
one horizontal, one vertical, or both with one time-bin difference, respectively, and
there is no possibility to know which photon entered the BS from which mode, then the
two-photon state is projected on the -state. If both photons exit the BS in the same
mode but are still detected in orthogonal states, then the two-photon state is projected on
to the + state.

Fig. 11. Set-up using linear optics for projecting two polarization (right-hand picture) and two
time-bin (left-hand picture) qubits on a basis spanned by the four Bell states. The two photons
enter a beam-splitter (BS) via modes a and b and are then subjected to a polarization-, or arrival-
time measurement, respectively. For polarization qubits, a coincidence between detectors DV and
(or D state, a coincidence between D and
DH H and DV ) corresponds to a projection on the V
+
DH (or DH and DV ) on . For time-bin qubits, detection of both photons in different time bins
yields a projection on a state: if they are found in different spatial modes on , if they leave
the beam-splitter in same spatial mode on the + state similar to the polarization case. Note
that in both schemes only two of the four Bell states can be distinguished.

Surprisingly, as shown by Lutkenhaus et al.,52,53 there is no experimental possibility

to differentiate between all of the states, at least as far as linear optics is concerned. The
best one can do is to identify two of the four Bell states with the other two states leading
to the same, third, result. However, as theoretically shown by Kwiat and Weinfurter in
 Wolfgang Tittel and Gregor Weihs . . . 17

1998, complete Bell-state analysis is possible even with linear optics if the two particles
are entangled in another degree of freedom as well.54 However, this condition can not be
fulfilled by photons that come from independent sources as required e.g. for entanglement
swapping (Section 4.4).
Recently, Kim et al.55 performed an experiment within the frame of quantum tele-
portation (see Section 4.3) which could lead the way to a complete Bell measurement.
The authors were taking advantage of non-linear interactions, however, even though they
used a classical input, the efficiency was extremely small. Whereas in principle it would
be possible to extend this method to a single photon input, which could carry any qubit
realization, it does not seem feasible with current technology.
A major problem for every Bell state measurement is to erase the identity of the
originally incoherent wave-functions. It must not be possible to infer by any degree of
freedom whether a detected photon originates from a specific mode. One condition that
emerges from this criterion is that the coincidence window should be significantly smaller
than the coherence time of the photons (ultracoincident). It turns out that this situ-
ation is rather tricky to achieve in an experiment, especially if both photons come from
different sources. As can be seen in Ref. [56 ], it involves an elaborate application of a quan-
tum erasure technique using mode-locked pulsed lasers to make the coincidence window
independent of any electronic limitations while still maintaining reasonable count rates.

Quantum state reconstruction The most general approach to measure a quantum

state is the reconstruction of its density matrix (also called quantum tomography). In
the case of two-qubit states, this has first been investigated by White et al. in 199957
for polarization qubits. Obviously, this method can be generalized to any realization of
a qubit. The measuring scheme is depicted in Fig. 12. In contrast to the Bell-state
measurements mentioned before, not only one projection has to be measured, but the
density matrix is reconstructed from the statistical outcomes of different joint projection
measurements.
In general there is some freedom in the choice of the measurements that will ultimately
be performed. A density matrix for a partially mixed state in a d-dimensional Hilbert space
contains d2 1 independent parameters. Therefore one needs at least as many independent
measurements to be able to reconstruct such a matrix. For a two-qubit state this would
be 15 measurements.
Still, in general it is difficult to calculate the matrix elements given only some marginal
probabilities. Of course, it is possible to just directly invert the set of measurement re-
sults but taking into account the ubiquitous measurement errors it turns out that direct
inversion is not always suitable because it can sometimes even lead to unphysical density
matrices.58 A different approach this was also applied to experimental data in Ref. [59 ]
is the maximum likelihood method, in which a suitably chosen likelihood functional is
maximized by the physical state that is most likely to having produced the given measure-
ment results.
18 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 12. Schematics for reconstruction of the density matrix of a two-qubit state. Each of the two
particles is analyzed using a qubit-analyzer. The vectors ~a and ~b specify the bases to be projected
on. Note that the eigenvalues of the different bases lie on a two-dimensional subspace within the
shell of the respective qubit-spheres.

2.3 N-particle entanglement

With the pioneering work by Greenberger et al. 60 it became clear that entangled states of
three or more particles are very important for fundamental tests of quantum theory. Also,
applications in quantum communication like quantum secret sharing have been found that
require multi-particle entanglement.61,62
Although the interest in GHZ states is quite high, no efficient practically usable sources
have been discovered to date.g As long as deterministic quantum gates are not available
we depend on combining two-particle entangled states and projective measurements to
construct entangled states of three or more particles by postselection.h Examples of this
technique will be presented in Section 3.2.
Just as in the two particle case these higher entanglements could in principle be pre-
pared for any degree of freedom (for time-bin entanglement see Ref. [44 ]), and methods
for GHZ state analysis have been suggested.65 Still, as long as there is no efficient way
to produce these states, the corresponding applications will probably remain in the aca-
demic interest only. An exception is the quantum secret sharing scheme explained in
Section 4.1.2, where GHZ state correlations are mimicked by the correlations observed
between the pump photon and the two downconverted photons.66

g One might speculate about a generalization of parametric down-conversion to processes, where more than
two photons are generated. However, the corresponding non-linear coefficients are generally many orders
of magnitude lower than in the parametric down-conversion case. As we restricted ourselves to photonic
entanglement, it would be beyond the scope of this article to discuss many-particle entanglements that
have been produced with Rydberg atoms and ions in a trap.63,64
hAs always, postselection means that a state is not actually prepared at any stage in the experiment. Still,

it is possible to observe the correlations that quantum theory predicts for the specific state.
 Wolfgang Tittel and Gregor Weihs . . . 19

3 Fundamental tests of nonlocality

3.1 Bell inequalities

Since the early days of quantum physics the intrinsic randomness of measurement
outcomes puzzled many physicists. Even very prominent proponents believed that there
should be a more complete, realistic theory complementing quantum physics with some
extra information in order to be able to describe the observed randomness in classical
terms of statistical ensembles. This extra information has become widely known as
hidden variables.67,68 A state of a system with a certain set of hidden values would then
be called dispersion-free or super-pure.
The debate whether or not hidden variable theories correctly describe nature was trig-
gered in 1935 by the now famous paper by Einstein, Podolsky and Rosen (EPR).2 However,
the debate remained a purely philosophical one until 1964, when Bell derived a statement
which is in principle experimentally testable.3,69 He started from EPRs example in the
version given 1957 by Bohm.70 The latter considered a gedankenexperiment where a source
emits pairs of spin-1/2 particles we would now call them qubits into opposite di-
rections (see Fig. 13). The particles are analyzed by two independent qubit-analyzers:
Stern-Gerlach apparata in separate regions of space. From the idea of spatial separation,
the assumption of locality, namely that the setting of the analyzer on one side can not
influence the properties of the particle on the other side, is inferred as a very natural
restriction for any otherwise most general hidden-variable model. Bell found that the cor-
relation between the two measurements as predicted by any such model must necessarily
comply with a set of inequalities nowadays known as Bell inequalities. The most widely
used form reads

S(a, b, a , b ) := |E(a, b) E(a, b )| + |E(a , b ) + E(a , b)| 2, (10)

where E(a, b) is the correlation coefficient of measurements along a, a and b, b . S is

sometimes called Bell-parameter and has the meaning of a second-order correlation.
It is easily seen that a singletstate of two spin-1/2 particles (the Bell state) will
violate this inequality with S = 2 2 for a specific set of analyzer directions. We conclude
that a system governed by local hidden variables (LHV) a system that can be described
by a local theory cannot mimic the behaviour of entangled states and hence that
quantum theory must be a non-local theory. It is interesting to note that the set-up for
testing Bell inequalities is identical to the one needed to measure the density matrix of a
two-qubit state as explained in Section 2.2.3. Only, in order to get the full information
about a quantum system, there are more analyzer directions to measure then when testing
whether the two-particle system is correctly described by a local or by a non-local theory.i
i Anotherapproach to see whether hidden variable theories can mimic the quantum behaviour is based
on the concept of non-contextuality. This means that the measured value of an observable should not
depend on the context, i. e. other commuting observables, that are measured simultaneously. Kochen and
Specker71 and independently Bell72 showed that given an at least three dimensional Hilbert space and
20 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Shortly after Bells discovery it became clear that no existing experimental data could
be used to find out whether or not non-local behaviour can indeed be observed. Further-
more it turned out that there were still some problems in applying the inequality to real
experiments because of shortcomings of sources, analyzers, and detectors. These short-
comings are usually called loopholes (see Section 3.1.1) because they leave an escape route
open for local realistic theories. To arrive at a logically correct argument against LHV
one has to supplement the strong original Bell inequalities with additional assumptions
which allow to calculate the necessary correlation coefficients from the measured coinci-
dence count rates (see paragraph 3.1.1). Such modified weak inequalities as derived e.g.
1969 by Clauser, Horne, Shimony and Holt (CHSH)4 (Eq. 10), or 1974 by Clauser and
Horne,5 were then used in most experiments to interpret the data.
By today all these shortcomings have been eliminated, though not all of them simul-
taneously in one experiment. The overwhelming evidence suggests that indeed quantum
physics accurately describes nature and that LHV theories have been ruled out.

Fig. 13. General set-up for Bell experiments. A source emits correlated particles, each of which
can be described in a two-dimensional Hilbert space. The qubits then fly back to back towards
two qubit analyzers making projection measurements in bases defined by parameters a and b, re-
spectively. The outcomes of the measurements are correlated, enabling to test via Bell inequalities
whether the two-particle system can be described by a local or by a non-local theory. Note that it
is sufficient to project on two different bases with eigenvalues located on a one-dimensional sub-
space within the shell of the respective qubit-spheres, in contrast to quantum tomography (Section
2.2.3).

In the following we will present a brief history of Bell experiments, give an account of
the current state-of-the-art and discuss the experiments that were instrumental in closing
the loopholes.

starting from the assumption of non-contextuality, it is indeed possible to prove a theorem against hidden
variables. For many years it was an open question how to experimentally test the Bell-Kochen-Specker
theorem. Last year, Simon et al.73 proposed an experiment based on measurements on a single particle
that is in an entangled state of two different degrees of freedom.
 Wolfgang Tittel and Gregor Weihs . . . 21

History The following table is a necessarily incomplete account of events in the history of
tests of Bell inequalities. More experiments can be found in Refs. [74,30,31,75,39,76,40,77,78,79,80 ].
1972 Freedman and Clauser perform the first experimental test based on po-
larization entangled photons generated via cascaded atomic transitions,
demonstrating that indeed a Bell inequality is violated for an entangled
system and thus ruling out a local realistic description.6
1982 Aspect et al. measure polarization correlations with two-channel
analyzers.81 Later they carry out an experiment in which they vary the
analyzers during the flight of the particles under test.82
1988 Ou and Mandel and independently Shih and Alley do the first non-
locality experiments using parametric down-conversion sources to create
polarization entanglement.34,83
1990 Rarity and Tapster observe momentum entanglement in their
experiment.84
1992 Brendel et al.48 and a bit later Kwiat et al.49 realize Fransons idea of
a test based on energy-time entanglement.47 Although the source itself
does not produce entangled qubits directly (see Section 2.2.2), the use
of qubit-analyzers post-projects on such states.
1997 Tittel et al. show that the quantum correlations between energy-time
entangled photons are preserved even over distances of more than 10
km.85,86
1998 Weihs et al. close the spacelike separation (Einstein locality) loophole
using randomly switched analyzers. This experiment was based on po-
larization entanglement.87
2001 Rowe et al. perform the first test violating a strong Bell inequality. In
contrast to all other tests that where based on photons, this experiment
took advantage of entangled ions.88

Current status As can be seen from this table, various types of entanglement have
been used for tests of Bell inequalities. In addition, Tittel et al. recently employed time-
bin entangled photons for quantum cryptography (see Section 4.1), an experiment that
can be interpreted as a test of Bell inequalities as well.89 This supports the notion of a
completely abstract formulation of Bells gedankenexperiment in terms of two apparata
each with a variable parameter that produce output correlated results. The generalization
of the original formulation based on Bohms example of two spin 1/2 particles and Stern
Gerlach apparata can for instance be found in a paper by Mermin.90
Presently parametric down-conversion sources (s. Section 2.2.1) in various configura-
tions deliver the highest quality entangled states. The entanglement contrast can be as
high as 99.5% while maintaining reasonable count rates. In terms of standard deviations
and coincidence count rates, some of the most impressive violations of Bell inequalities
were published in Refs. [40,91 ].

3.1.1 Closing loopholes

22 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

As already stated above, there are certain loopholes in most of the realized experiments.
Here we discuss their current status from the experimentalists perspective

Detector efficiency Pearle92 first noticed already in 1970 that in real tests of Bells
inequality, which fall short of detecting all particles that are emitted by the source, one
can still construct a hidden variable model that accurately predicts the observed data
(see also recent work by Santos93 and by Gisin94 ). Afterwards this argument has been
named efficiency loophole because of the fact that various inefficiencies reduce the ratio
of counted to emitted particles sometimes to less than a few percent. These inefficiencies
include incomplete collection of particles from the source, imperfect transmission of optical
elements and analyzer devices, and most important, the far-from ideal detectors.
Violation of a Bell inequality without any supplementary assumption requires an overall
efficiency of more than 82.8%. For photon based experiments this is very difficult to
achieve with present photon counting technology. The threshold can be reduced to 67%
by the help of non-maximally entangled states.95 In both cases the experimental visibility
must be perfect. Although in principle there exist detectors which have high enough
efficiency, it is practically very unlikely that a photonic Bell experiment could achieve
these efficiency levels. To date, all these experiments invoked so-called fair sampling4
and no-enhancement assumptions5 which allow to derive inequalities into which only
coincidence count rates enter. However, very recently a beautiful experiment based on
two entangled ions in a microscopic trap succeeded in violating a strong Bell inequality at
nearly 100% efficiency.88 Many physicists consider this a closure of the discussed loophole.8

Separation Spatial separation is an important ingredient in the derivation of Bells

inequality. The lack of large spatial separation does not by itself constitute a loophole
but it was said that with increasing distance the quantum correlations could diminish.96
Therefore it is natural to try to extend the range of proven quantum phenomena to as
large a distance as possible. In addition, the fact that, and the question how entanglement
can be maintained over large distances is very important for quantum communication
protocols like quantum cryptography (see also Section 4.5).
The largest spatial separation has been achieved in tests that were carried out by
one of our groups (GAP) in 1997 and 1998.85,86,97 The experiments utilized energy-time
entanglement created by parametric down-conversion at a wavelength of 1310 nm, suitable
for transmission in standard telecommunication optical fibers. Clear violations of Bell
inequalities of up to 16 standard deviation were achieved for measurements that were
more than 10 km apart (see Figs. 14 and 15).

Einstein locality The other prominent loophole in experiments on Bells inequality is

called the spacelike separation or Einstein locality loophole.98 It is constituted by the
fact that for most experiments it was possible to explain the observed correlations by a
hypothetic subluminal (slower than or equal to the vacuum speed of light) link between
the two particles or apparata and particles. Given such a link it would in principle be
 Wolfgang Tittel and Gregor Weihs . . . 23

Fig. 14. Experimental arrangement for a test of Bells inequality with measurements made more
than 10 km apart. Source (in Geneva) and observer stations (interferometers in Bellevue and
Bernex, respectively) were connected by a fiber optic telecommunications network.

Fig. 15. Data from the long distance Bell test. The two curves show the correlation coefficients
for two different analyzer settings at Alices while varying the setting at Bobs. A clear violation
of the CHSH inequality is observed.
24 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

possible that the analyzer direction or even the measurement outcome is communicated
to the other side.

Fig. 16. Experimental set-up for a test of Bells inequality with independent observers. Data
are collected locally at the observer stations only and can be compared after the measurement is
completed.

Bell himself considered this fact as being important, calling it the vital time factor.99
However, operationally and theoretically it is hard to define this time factor. There is no
criterion that is generally agreed upon. Bohm talked already in 1957 about variation of the
analyzers while the particles are in flight.70 Over the years the idea emerged that it would
be necessary to vary the analyzers in a random way, where the randomness would have to
be drawn from local sources or from distant stars in opposite directions of the universe.
Including the generation of randomness and other delays the measurements should then be
completed within a time that is short compared to the time it takes to signal to the other
observer station. This prescription amounts to performing the measurement in spacelike
separated regions of spacetime.
Obviously these are extremely vague concepts and therefore it is not astonishing that
only two experiments have tried to pin down and answer these questions. The first one was
performed by Aspect et al.82 in 1982 and included a periodic variation of the analyzers.
Because periodic functions are in principle predictable it has been said that this experiment
was not definitive in closing the spacelike separation loophole.
In 1998, 16 years later, one of our groups (UNIVIE) was able to include the randomness
factor and, at the same time, to extend the spatial separation to 360 m yielding a large
safety margin for the spacelike separation issue.87 The experimental set-up is shown in
Fig. 16. It yielded a violation of Bells inequality by 30 standard deviations.
Another approach to attack the locality loophole has been demonstrated by Tittel et
 Wolfgang Tittel and Gregor Weihs . . . 25

al. in 1999.97 In this experiment, two analyzers with different parameter settings were
attached to each side of the source, and the random choice was done by a passive optical
coupler, sending the photons to one or the other analyzer. Hence, in contrast to the
experiment mentioned before, the randomness does not come from an external random
number generator but is engendered using the photons themselves and might therefore
seem less good.j

Analyzer R otation Angle

-0,50 -0,25 0,00 0,25 0,50

800 A +0/B 0
A +1/B 0
600
C oincidences in 5s

400

200

600

400

200 A +0/B +0
A +1/B +0
0
-100 -50 0 50 100
Bias Voltage (Alice) [V]

Fig. 17. Correlation curves taken for spacelike separated measurements on a polarization entangled
photon pair at a distance of 360 m. The measurements yield a violation of Bells inequality by 30
standard deviations.

3.1.2 Relativistic configurations

To find the quantum mechanical predictions for the results of Bell-type measurements,
one can think of the first measurement (traditionally at Alices) as a non-local state
preparation for the photon traveling to the second analyzer (at Bobs) like in entan-
glement based quantum cryptography (Section 4.1.2): In a first step one calculates the
probabilities for the different outcomes of Alices measurement that depend only on the
setting of her analyzer and the local quantum state.kKnowing the global two-particle state
enables in a second step to calculate Bobs local state. In a third step, equivalent to the
j However, note that the borderline between good and bad randomness is very vague.100
kThe local state is obtained by tracing over the distant system. In case of a maximally entangled global
state, it is completely mixed and the outcome of Alices measurement is completely random.
26 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

first one, one calculates the probabilities of the different outcomes of the measurement at
Bobs (which are now joint probabilities), determined by his local state hence by the
setting of the first analyzer and the specific two-particle (global) state used and by the
setting of the second analyzer.
Many people believe that the state vector is not endowed with reality but that it is
only a mathematical tool that helps to calculate the statistical outcomes of experiments.
Consequently, the reduction of the state of Bobs subsystem by Alices measurement must
be understood as an instantaneous modification of the knowledge of an observer of the first
measurement concerning the quantum state of the second particle. Indeed, possessing only
the second system, it is impossible to see any change as a result of the first measurement:
The density matrix describing the local state remains unchanged and thus there is no
possibility for superluminal signaling and hence no contradiction with special relativity.10
However, it has never been proven experimentally whether the state vector does indeed
not represent reality and that the collapse only changes the knowledge of the observer,
or whether the state vector is real and its collapse has to be considered a real physical
phenomenon as assumed e.g. by Ghirardi, Rimini and Weber.101

The speed of quantum information. If one assumes the collapse to be real, it is

natural to ask how fast it propagates from Alices to Bobs subsystem. The lower bound
of this speed of quantum information (or, following Einsteins words, the speed of the
spooky action at a distance) can easily be calculated from the distance between both
analyzers (the distance that has to be traveled) and the time left for the second particle
until reaching its analyzer. However, before doing so, one still has to define which parts
of the analyzers are considered crucial for the alignment. The most natural choice seems
to be the detectors where the transition from quantum to classical takes place, although
there are other possibilities as well.102,11 It is assumed in the following that the important
parts are indeed the detectors. From standard Bell experiments, one already knows that
quantum information propagates faster than with the speed of light c. However, these
experiments have not been devised in order to investigate the lower bound, and the pre-
cision of the alignment is not discussed in most works. In 2001, Zbinden et al.13 reported
on an experiment performed again with analyzers separated by more than 10 km (similar
to Fig. 14) where the fibers connecting the source with the detectors, this time each of 10
km length, were aligned such that the arrival time difference was smaller than 5 ps. This
allowed to set a lower bound of the speed of quantum information of 2/3 107 c as seen
from the laboratory (Geneva) reference frame. Obviously, this is not the only possible
choice of a reference frame but one can argue that this is a very natural one.
The speed of quantum information is very important for a class of problems that
can be labeled the search for a covariant description of the measurement process with
the example of the impossibility of a causal description of the instantaneous collapse in
an EPR experiment that would be valid in all frames. The introduction of a preferred
frame PF that would allow a realistic (obviously non-local) description of the quantum
measurement is a way out of these problems.103 However, the introduction of a PF is
still an intellectual tool (or trick) and there is no experimental evidence to support this
 Wolfgang Tittel and Gregor Weihs . . . 27

hypothesis. A good candidate for such a preferred frame is the frame from which the
cosmic microwave background radiation (CMB) is seen to be isotropic. Analyzing the
same data and taking into account the relative motion between the Geneva and the CMB
frame, Scarani found a lower bound of the speed of quantum information of 2 104 c as
seen from the CMB frame.104 Repeating this experiment with different alignments, each
corresponding to simultaneous measurements in a different frame, it would be possible
to test whether there is one (the preferred) frame from which the speed of quantum
information is seen to be limited.105

Moving detectors in different reference frames. The experiment establishing the

lower bound of the speed of quantum information has been extended13 to perform a first
test of relativistic non-locality (RNL)(or multisimultaneity) a theory unifying non-
locality and relativity of simultaneity that was proposed in 1997 by Suarez and Scarani.11
RNL predicts that the quantum correlation should disappear in a setting where both an-
alyzers are in relative motion such that each one in its own inertial frame is considered to
cause the collapse. In the initial proposal, the crucial part of an analyzer was assumed to
be the last beamsplitter. In the experiment, it was supposed again that the alignment has
to take into account the positions of the detectors. Although many assumptions concern-
ing the nature of a detector had to be made in order to make this experiment feasiblel, it
was possible for the first time to test an interpretation of quantum mechanics. The data
always reproduced the quantum correlations regardless the motion of the detectors, yield-
ing thus experimental evidence that the tested version of RNL does not correctly describe
nature, and making it more difficult to consider the collapse of the wave function as a real
phenomenon.

3.2 GHZ states and nonlocality

In 1989 (G)reenberger, (H)orne, and (Z)eilinger60 found that entangled states of at least
three quantum systems can exhibit contradictions with local realistic models that are much
more striking than those found for two particles in violations of Bell inequalities: Not even
the perfect correlation predicted by quantum mechanics for coincidence measurements
between more than two particles can be described by realistic models. The basic set-up
for such a GHZ-type test is shown in Fig. 18. While it became clear later, that, from
a fundamental point of view, there is no difference between GHZ- and Bell-type tests of
non-locality, at the conceptual level the so-called GHZ correlations remain a remarkable
feature of quantum physics.
Naturally, people have tried to construct such states and investigate their properties
but as there are no efficient and controllable natural sources for three or more photon
states the researchers had to resort to techniques allowing to construct higher dimensional
l Forexample, it was assumed that a black absorbing surface is sufficient to engender the collapse of the
wave function: A detected photon can in principle be observed, e.g. via increase of the temperature of
the black surface. This is similar to the fact that it is sufficient to observe distinguishability in principle
to make the fringes in an interference experiment disappear.
28 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 18. Schematics for a test of 3-particle GHZ-type non-locality using momentum entangled
states. The three qubits are each send to an analyzer. Coincidence measurements in identical or
orthogonal bases enable to test whether the three-particle state is described by a local, or by a
non-local model. Note that, in contrast to Bell-type tests of non-locality, only settings leading to
deterministic outcomes either perfect coincidences or perfect anti-coincidences are required.

entanglement from two-particle entangled states (see Section 2.3).

Generation of three-photon entangled states. In 1999 Bouwmeester et al.106 re-

ported the first observation of three-particle entanglement. The team employed two photon
pairs produced in non-colinear parametric downconversion from a pulsed UV pump beam
and combined them via beam-splitters to a conditional three particle-entangled state. The
schematic (Fig. 19) shows that whenever a trigger photon is received and simultaneuosly
photons are observed in modes 13, in subsequent correlation measurements (not shown)
one will observe results which can only be described by the state
1
|HiT [|HHVi123 + |VVHi123 ] . (11)
2
The fact that such a state exists must be considered a manifestation of quantum non-
locality.
While Bouwmeester demonstrated the existence of a three-photon GHZ state, Pan
et al.107 reported last year the first and still only experiment in which the test against
local realism was carried out in full. Although GHZ-type tests theoretically permit a
clear cut contradiction to local realism, in a real experiment one has to resort to an
inequality just like in Bells theorem. For this task, Pan measured in total 32 (4 groups
of 8) different combinations of polarizer settings, and the reasoning followed Mermins
 Wolfgang Tittel and Gregor Weihs . . . 29

Fig. 19. Schematic of the set-up used by Bouwmeester et al. to produce three-photon GHZ
correlations. Two independent pairs are generated simultaneously and one photon from each pair
is send right, one left. By action of the left polarizing beam-splitter (PBS), only horizontally
polarized particles can reach the trigger detector T. To register three-fold coincidences between
detectors 1, 2, and 3, the other photon must be reflected from that PBS, i.e. it must be vertically
polarized. This photon is subsequently rotated by 45 and can end up as V in detector 1 or as H
in detector 2. The only two possible ways that a triple coincidence event arises in detectors 1, 2,
and 3 are therefore when the right two photons split at the right beam-splitter and when the two
photons meeting at the upper PBS have are found to have identical polarization either both
horizontal or both vertical.
30 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 20. The graphs show the relative frequencies of three-fold coincidence events between outputs
D1 , D2 , and D3 as predicted and as measured (bottom). All analyzers project onto a L=linear
45 basis. P and M refer to plus and minus 45 linear polarization, respectively. MPM for
example marks the probability to measure a three-fold coincident event if polarizer settings are M
for particle 1, P for particle 2, and M for particle 3. The data violate the local realistic prediction
by 10 standard deviations.
 Wolfgang Tittel and Gregor Weihs . . . 31

argument for an entangled system of three spin-1/2 particles,108 adapted to an imperfect

(noisy) experimental set-up. Fig. 20 shows the results of the measurements of one of the
four groups and compares them to the local realistic as well as to the quantum predictions.
The data clearly violates the local realistic prediction, and agrees within experimental
errors with quantum physics.

Fig. 21. Experimental set-up to create and confirm the existence of a four-photon GHZ states.

Four-photon GHZ Extending the teleportation and entanglement swapping experi-

ments discussed in sections 4.3 and 4.4, one of our groups (UNIVIE) recently succeeded in
demonstrating four-photon entanglement.109 In this experiment again two down-conversion
pairs are used as an initial resource, but this time from processes occurring in different
spatial modes. One photon of each pair is directed towards a polarizing beam splitter.
(s. Fig. 21) Coincidence detection after this beam-splitter projects the incoming product
state of two entangled photon pairs onto a four-photon GHZ state.m

1
|i1234 = (|Hi1 |Vi2 |Vi3 |Hi4 + |Vi1 |Hi2 |Hi3 |Vi4 ) . (12)
2
A comparison of the measured 4-photon coincidence probabilities for various combi-
nations of H and V projections confirms that indeed only the desired |Hi1 |Vi2 |Vi3 |Hi4
and |Vi1 |Hi2 |Hi3 |Vi4 components have been created. The contrast in this measurement
was of more than 100:1. In addition, measurements in the conjugate 45 basis (s. Fig. 22)
demonstrate the coherent superposition of the two components, hence show the existence
of a four photon GHZ state and confirm the existence of non-locality.

mNote that the state is never realized as a freely propagating system.

32 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 22. Experimental data showing four-photon polarization correlation in the 45 linear basis.
At zero delay between the two photons that are superposed on the polarizing beam-splitter, in-
distinguishability is granted and interference occurs, demonstrating the coherent superposition of
|Hi1 |Vi2 |Vi3 |Hi4 and |Vi1 |Hi2 |Hi3 |Vi4 .
 Wolfgang Tittel and Gregor Weihs . . . 33

4 Quantum communication

4.1 Quantum cryptography

Quantum cryptography (QC) is certainly the most mature application of quantum
communication. Based on the non-classical feature of the quantum world, it provides two
parties, a sender Alice and a receiver Bob, with a means to distribute a secret key in a
way that guarantees the detection of any eavesdropping (Eve): Any information obtained
by an unauthorized third party about the exchanged key goes along with an increase of
the quantum bit error rate (QBER) of the transmitted data which can be checked using a
suitable subset of the data. It has been shown that, as long as the QBER of the sifted key
(the key after bases reconciliation) is below a certain threshold (11 or 15 %, respectively,
depending on the eavesdropping strategy assumed110,111 ), Alice and Bob can still distill
a secure key by means of classical error correction and privacy amplification protocols.112
This secret key can then be used together with the one-time pad to exchange a confidential
message in complete privacy.
There is vast literature covering theoretical as well as experimental aspects of quantum
cryptography. Since there is an extensive and very recent article17 reviewing both sides,
we will keep this section short and give only the necessary information to understand
entanglement-based quantum cryptography.

4.1.1 Quantum cryptography based on faint laser pulses

The first QC protocol has been published by Bennett and Brassard in 1984n; it is known
today as BB84 or four-state protocol.114 Fig. 23 illustrates the protocol with the example
of four polarization qubits of linear polarization. However, any property can in principle
be used to realize a qubit. Furthermore, any four states that fulfill the requirement that
they form two bases in the qubit space and that any two states belonging to a different
basis have an overlap of 21 will do as well.
The first experimental demonstration of quantum cryptography took place in 1989
at IBM when Bennett et al. realized the so-called B92 (or two-state) protocol115 based
on polarization coding with single photons over a distance of 30 cm in air.112 Since
then, a lot of experimental progress has been made, and from 1995 on, several groups
demonstrated that quantum cryptography is possible outside the laboratory over distances
of tens of kilometers as well.17 All experiments relied on faint laser pulses strongly
attenuated pulses that contain less than 1 photon on average to mimic single photons.
These realizations could, even today, provide secure communication in case cryptographic
protocols that are based on mathematical complexity turn out to be unsafe. However,
they still suffer from low bit-rateso, and the maximum span with todays technology is only
of around 100 km.
nThe work was inspired by an unpublished article by Wiesner from 1970 (published only in 1983113 )
o Togive an example, the secret key rate (after error correction and privacy amplification) reported by
Ribordy et al.116 over a distance of 23 km was of 210 Hz.
34 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 23. For each photon she sends to Bob, Alice chooses randomly a bit value (row 1) and a basis
(h/v, or 45o ), and prepares the photon in the corresponding state. Every time Bob expects a
photon to arrive, he activates his detectors and chooses randomly to analyze in the h/v basis, or
in the 45o basis. He records which basis he used (row 2) and, in case of a successful detection,
which result (in terms of bits) he got (row 3). After exchange of a sufficient large number of
photons, he publicly announces the cases where he detected a photon and the basis used for the
measurements. However, he does not reveal which results he got. Alice compares event by event
whether or not Bobs analyzer was compatible to her choice of bases. If they are incompatible
or if Bob failed to detect the photon, the bit is discarded. For the remaining bits (row 4), Alice
and Bob know for sure that they have the same value. These bits form the so-called sifted key.
The security of the key distribution is, roughly speaking, based on the fact that a measurement of
an unknown quantum system will, in most cases, disturb the system: If Alices and Bobs sifted
keys are perfectly correlated, no eavesdropper tried to eavesdrop the transmission and the key can
be used for encoding a confidential message using the one-time pad. If the sifted keys are not
100% correlated, then, depending on the QBER, Alice and Bob can either distill a secret key via
error correction and privacy amplification, or the key is discarded and a new distribution has to
be started.
 Wolfgang Tittel and Gregor Weihs . . . 35

Let us briefly elaborate on the maximum distance. In theory, i.e. using perfect experi-
mental equipment, the sifted key rate decreases exponentially with increasing transmission
losses but never drops to zero. It is given by the product of Alices pulse rate frep, the
number of photons per pulse , the probability tlink that a photon arrives at Bobs, and
the quantum efficiency that it is detected.
1
Rsifted = frep tlink (13)
2
The factor 1/2 is due to key sifting the fact that Alice and Bob use compatible bases in
only 50 % of the cases. In practice, there are always experimental imperfections, and there
will always be some errors in the sifted key even in the absence of an eavesdropper. For
the sake of the presentation, we assume here that the errors are only due to detector dark
counts (a signal generated by a detector without the presence of a photon), arising with
probability pdark :
1 1
Rwrong Rdark = frep pdark n (14)
2 2
The first factor 1/2 is again due to key sifting, the second one to the fact that a detector
dark count leads only in half of the cases to a wrong result, and n is the number of
detectors. Finally, the QBER is given by

Rwrong n pdark
QBER = (15)
Rsifted 2 tlink

Since Alice and Bob can never know for sure whether the observed QBER is due to
the imperfections of their equipment, or whether it is engendered by the presence of an
eavesdropper, they always have to assume to worst case, i.e. that there is an eavesdropper
that has the maximum information compatible with the observed QBER. Therefore, they
have to apply classical error correction and privacy amplification protocols to the sifted
key in order to distill a secret key.
Fig. 24 shows the secret bit rate after error correction and privacy amplification as a
function of distance. Here we assume that the photons are transmitted using optical fibers
with losses of 0.2 dB/km. The bit rate decreases exponentially for small distances. Then,
with larger distance (i.e. with decreasing transmission probability tlink , hence increasing
QBER), the bit reduction due to error correction and privacy amplification gets more and
more important, and at a QBER (hence distance) where the Alice-Bob mutual Shannon
information is equal to Eves Shannon Information, there are no bits left and the curve
representing the bit rate drops vertically to zero. Note that the maximum distance does
not depend on Alices pulse rate.
To achieve a better performance of a cryptographic system concerning bit rate or
maximum transmission distance, there are basically two things to improve: the detectors,
and the sources.

Todays single photon detectors, capable of counting photons at telecommunication

wavelength of 1.3 and 1.5 m where fiber losses are low, feature quantum efficiencies
36 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

Fig. 24. Secret bit rate after error correction and privacy amplification. The maximum transmis-
sion span is given by the distance where the QBER equals 15% (assuming symmetric individual
eavesdropping attacks110 ). Here we assume a initial pulse rate fpulse of 1 and 10 MHz, respec-
tively, losses of 0.2 dB/km of optical fibers at 1550 nm wavelength (close to the fundamental limit),
a quantum efficiency of 10%, dark count probability of 105 and a mean photon number per pulse
of 0.1 and 1, respectively. The use of a higher mean photon number leads to a higher secret
bit rate for a given distance and pulse rate, as well as to a larger maximum transmission span,
i.e. 130 instead of 90 km. A higher pulse rate engenders a higher secret key rate, however, does
not change the maximum span. In practice, if we take into account non-ideal error correction and
privacy amplification algorithm, multi-photon pulses and other optical losses not considered here,
the maximum distance is likely to be reduced by a factor of around two.
 Wolfgang Tittel and Gregor Weihs . . . 37

of only around 10%, paired with a high dark count probability of 105 per 1 ns
time window. These figures determine the secret key rate for a given distance and
pulse rate, and the maximum transmission span as shown in Fig. 24. In addition,
the detectors performance limits the pulse rate via the effect of so-called afterpulses:
These are avalanches that are not caused by the detection of a photon but by the
release of charges from trapping levels populated while a current transits through the
diode. Since the probability for observing an afterpulse after a detection of a photon
decreases exponentially with time, they can be suppressed using suitable dead times
with the drawback of limiting the maximum pulse rate. It is thus obvious that the
use of better detectors would have an important impact on experimental quantum
cryptography.
Mimicking single photons by faint pulses has a very important advantage: it is
extremely simple. Unfortunately this advantage is paired with two drawbacks. First,
a mean photon number smaller than 1 (the upper limit for quantum cryptography)
leads to a reduction of the bit rate (see Eq. 13 and Fig. 24). Second, since the
photon-number statistics for faint pulses is given by a Poissonian distribution, there
is always a possibility to find more than one photon in a weak pulse. This opens
the possibility of an eavesdropper attack based on multi-photon splitting.117,118 The
smaller the mean number of photons per pulse, the smaller this threat, however, the
smaller the bit rate as well.

4.1.2 Quantum cryptography based on photon-pairs

Single-photon based realizations In order to get around the problem of faint pulses
where the probability of having zero photons in a pulse is rather high, a good idea is to
replace the faint pulse source by a photon-pair source (see Section 2.2.1 and Fig. 25b)
where one photon serves as a trigger to indicate the presence of the other one.119 In this
case, Alice can remove the vacuum component of her source, and Bobs detectors are only
triggered whenever she sends at least one photon.p This leads to a higher sifted key rate
(assuming the same trigger rate than in the faint-pulse case) and a lower QBER for a
given distance (for given losses) and therefore to a larger maximum span (see Fig. 24). It
is important to note that photon-pairs can not be created in Fock states, similar to single
photons mimicked by faint pulses. Therefore, depending on the probability to create more
than one photon pair, the danger of multi-photon splitting eavesdropping attacks exists
as well.

Entanglement based realizations Finally, the potential of a source creating entangled

pairs is not restricted to create two photons at the same time one serving as a trigger
for the other one. It is possible to use the full quantum correlation to generate identical
keys at Alices and Bobs, and to test the presence of an eavesdropper via a test of a Bell
inequality (see Fig. 25c). This beautiful application of tests of Bell inequalities has been
pHere we assume that the collection efficiency for the photon traveling towards Bob is 1. In practice, a
more realistic value is 0.70.
38 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

pointed out by A. Ekert in 1991120 without knowing about the discovery of quantum
cryptography by Bennett and Brassard 7 years earlier. The set-up is similar to the one
used to test Bell inequalities with the exception that Alice and Bob each have to chose
from three different bases. Depending on the bases chosen for each specific photon pair,
the measured data is either used to establish the sifted key, to test a Bell inequality, or it
is discarded.
The security of the Ekert protocol is very intuitive to understand: If an eavesdropper
gets some knowledge about the state of the photon traveling to Bob, she adds some hidden
variables (hidden in the sense that only the eavesdropper knows about their value). If she
gets full knowledge about all states, i.e. the whole set of photons analyzed by Bob can
be described by hidden variables, a Bell inequality can not be violated any more. If Eve
has only partial knowledge, the violation is less than maximal, and if no information has
leaked out at all, Alice and Bob observe a maximal violation.
However, the Ekert protocol is not very efficient concerning the ratio of transmitted
bits to the sifted key length. As pointed out in 1992 by Bennett et al.121 as well as by
Ekert et al.,122 protocols originally devised for single photon schemes can also be used
for entanglement based realizations. This is not surprising if one considers Alices action
as a non-local state preparation for the photon traveling to Bob (see also Section 3.1.2).
Interestingly, it turns out that, if the perturbation of the quantum channel (the QBER)
is such that the Alice-Bob mutual Shannon information equals Eves maximum Shannon
information, then the CHSH Bell inequality (Eq. 10) can not be violated any more.123,110
Although this seems very natural in this case and a similar connection has recently been
found for n-party quantum cryptography and some n-particle Bell inequalities124 , it is not
clear yet to what extend the connection between security of quantum cryptography and
the violation of a Bell inequality can be generalized.
Compared to the faint pulse schemes, entanglement based QC features two advantages.
First, similarly to photon-pair based realizations, Alice removes the vacuum component
of her source. Actually, the entanglement based case is even more efficient since even the
optical losses in Alices preparation device are now eliminated as can be seen from Fig. 25c.
Second, even if two pairs are created within the same detection window hence two
photons travel towards Bob within the same pulse they do not carry the same bit,
although they are prepared in states belonging to the same basis. Beyond this passive
state preparation, it is even possible to achieve a passive preparation of bases using a
set-up similar to the one depicted in Fig. 25d. There is no external switch that forces
all photons in a pulse to be measured in the same basis but each photon independently
chooses its basis and bit value. Therefore, eavesdropping attacks based on multi-photon
pulses do not apply in entanglement based QC. However multi-photon pulses lead to errors
at Bobs who detects from time to time a photon that is not correlated to Alices.
Although all Bell experiments intrinsically contain the possibility for entanglement
based QC, we list here only experiments that have been devised in order to allow a fast
change of measurement bases.
 Wolfgang Tittel and Gregor Weihs . . . 39

Fig. 25. Single photon based quantum cryptography using a) a faint-pulse source, b) a two-photon
source, c) entanglement based quantum cryptography with active, and d) with passive choice of
bases. 2h denotes the photon-pair source, and the parameters and characterize the settings
of the qubit-analyzers.
40 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

1982 Interestingly enough, the first experiment that fulfills the above defini-
tion is the test of Bell inequalities using time-varying analyzers, per-
formed by Aspect et al.82 with polarization entangled qubits in order
to close the locality loophole (see also Section 3.1.1) at a time where
quantum cryptography was not yet known, not even the single photon
based version.
1998 Weihs et al. demonstrate a violation of Bell inequalities with polar-
ization entangled qubits at 700 nm wavelength and randomly switched
analyzers, separated by 360 km of optical fiber.87 This experiment has
been devised to close the locality loophole.
1999 Tittel et al. perform a Bell experiment, again to be seen in the context
of the locality loophole, incorporating a passive choice of bases.97 Two
fiber-optical interferometers are attached to each side of a source cre-
ating energy-time entangled photons at 1.3 m wavelength. However,
similar to both before-mentioned experiments, the bases chosen for the
measurements are chosen in order to allow a test of Bell inequalities and
not to establish a secret key.
2000 Three publications on entanglement based cryptography appear in the
same issue of Phys. Rev. Lett.:
1.) Using a set-up similar to the one mentioned already in the second
entry of this table, Jennewein et al. realize a quantum cryptography sys-
tem including error correction over a distance of 360 m.125 Two different
protocols are implemented, one based on Wigners inequality (a special
form of Bell inequality), the other one following BB84. Sifted key rates
of around 400 and 800 bits/s, respectively, are obtained, and QBERs of
around 3% observed. Using the same assumptions that lead to Fig. 24,
this amounts to a secret key rate of 300 and 600 bits/s, respectively.
2.) Naik et al. demonstrate the Ekert protocol in a free space experiment
over a short (laboratory) distance.126 The experiment takes advantage of
polarization entangled qubits at a wavelength of around 800 nm. Sifted
key rates of around 10 bits/s paired with a QBER of 3% are reported,
leading to a secret key rate of 6 bits/s after implementation of error cor-
rection and privacy amplification. In addition to the key exchange, the
authors simulate different eavesdropping strategies and find an increase
of the QBER with increasing information of the eavesdropper, according
to theory. The experiment has recently been extended127 to realize the
so-called six state protocol.128,129
3.) Tittel et al.89 report on a fiber-optical realization of quantum cryp-
tography in a laboratory experiment using the BB84 protocol. This
experiment is based on time-bin entangled qubits at telecommunication
wavelength of 1.3 m and takes advantage of phase-time coding and a
passive choice of bases. Sifted key rates of 33 Hz and a QBER of 4% are
obtained, leading to a calculated secret key rate of 21 bits/s.
2001 Ribordy et al.130 realize a QC system based on energy-time entangle-
ment. In contrast to the schemes mentioned before, this realization takes
advantage of an asymmetric set-up, optimized for QC, instead of a set-
up designed for tests of Bell inequalities where the source is generally
located roughly in the middle between Alice and Bob. Here, one photon
(at 810 nm wavelength) is send to a bulk-optical interferometer, located
directly next to the source, the other one (at 1550 nm wavelength) is
transmitted through 8.5 km of fiber on a spool to a fiber optical interfer-
ometer. Implementing the BB84 protocol and a passive choice of bases,
a sifted key rate of 134 bits/s and a mean QBER of 8.6% (over 1 hour)
is observed. From these values, one can calculate a secret key rate of 45
bits/s.
 Wolfgang Tittel and Gregor Weihs . . . 41

Three party quantum cryptography In addition to the mentioned two-party QC

schemes, Tittel et al. reported in 2001 a proof-of principle demonstration of quantum
secret sharing (three party quantum cryptography) in a laboratory experiment.66 This
rather new protocol enables Alice to send key material to Bob and Charlie in a way that
neither Bob nor Charlie alone have any information about Alices key, however, when
comparing their data, they have full information. The goal of this protocol is to force both
of them to collaborate.
In contrast to proposed implementations using three-particle GHZ states,61,62 pairs of
time-bin entangled qubits were used to mimic the necessary quantum correlation of three
entangled qubits, albeit only two photons exist at the same time (see Fig. 26). This is
possible thanks to the symmetry between the preparation interferometer acting on the
pump pulse and the interferometers analyzing the down-converted photons. Indeed, the
data describing the emission of a bright pump pulse at Alices is equivalent to the data
characterizing the detection of a photon at Bobs and Charlies: all specify a phase value
and an output, or input port, respectively. Therefore, the emission of a pump pulse can
be considered as a detection of a photon with 100% efficiency, and the scheme features
a much higher coincidence count rate compared to the initially proposed GHZ-state type
schemes.

Fig. 26. Basic set-up for three-party quantum secret sharing using pseudo-GHZ states. Compare
with true GHZ states as shown in Fig. 18.

4.2 Quantum dense coding

Whenever two parties A (Alice) and B (Bob) wish to communicate, they first have to
42 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

agree on a coding procedure, that is, they have to associate symbols with physical states.
In classical communication, one usually uses a two letter alphabet where the different
symbols (bit-values) are represented by (classical) optical pulses with different individual
properties. In quantum physics we can encode information in a novel way into joint
properties of elementary systems in entangled states, leading in principle to the possibility
to transmit two bits of information by sending only one qubit. This striking application
of quantum communication is known as quantum dense coding.

Fig. 27. Experimental set-up to demonstrate quantum dense coding based on Bell state analysis
of entangled polarization qubits.51 After locally preparing the joint state of the entangled particles
by means of wave-plates, Alice sends her particle to Bob. Performing a Bell measurement
 on the
two-particle state, Bob can distinguish between two different Bell states ( ), with the two
 
other ones ( ) leading to the same, third, result. Therefore, Alice can encode 1.58 bit of
information sending only one qubit.

The maximally entangled Bell basis (Eqs. 5 and 6) has a very important and interesting
property which was exploited by Bennett and Wiesner115 in their proposal for quantum
dense coding: In order to switch from any one of the four Bell states to all others, it is
sufficient to manipulate only one of the two qubits locally. Thus, the sender, Alice, can
actually encode two bits of information into the whole entangled system by just acting on
one of the two qubits.
In order to read out this information, the receiver, Bob, needs to be able to identify
the four Bell states, that is, he needs to perform a Bell measurement as explained in
Section 2.2.3. However, using linear optics, only two out of the four Bell states can be
distinguished unambiguously52,53 whereas the other two states lead to identical signatures.
Still, as has been experimentally demonstrated in 1996 by Mattle et al.,51 this is enough
to encode three-valued information (corresponding to 1.58 bit of information) into each
transmission event (see Fig. 27).
 Wolfgang Tittel and Gregor Weihs . . . 43

4.3 Quantum teleportation

Fig. 28. Schematic of quantum teleportation of an unknown state. Particle 1 is given to Alice
who subjects it to a Bell-state measurement (BSM) together with particle 2, the latter one being
entangled to particle 3 (at Bobs). Depending on the result of this measurement, Bob applies
a unitary transformation to particle 3 which then ends up in precisely the same state in which
particle 1 was originally.

If, in some sense, the aim of quantum key distribution is the communication of classical
bits, quantum teleportation, discovered in 1993 by Bennett et al.,131 can be thought of
as being the exchange of quantum bits. One might define quantum teleportation as the
art of transferring the state of an unknown qubit located at Alices to a second quantum
system, located at Bobs, the motivation being that it might be impossible to send the
physical system itself. In this application of quantum communication, the qubit to be sent
is unknown to the parties involved in the transfer, but it might be known to a third party,
Charlie.
In a world of classical physics, teleportation is nothing remarkable. It suffices to mea-
sure the properties of the (classical) bit and then communicate the information about
its composition to Bob, who then reconstructs the bit. This strategy must fail in the
quantum case where the measurement of an unknown qubit without disturbing it is im-
possible and cloning is forbidden.132,133 Surprisingly, quantum communication provides a
way out of this problem (see Fig. 28). Before Charlie hands over the particle to Alice
who then teleports it to Bob, the latter have to share a pair of entangled particles. Alice
now makes a Bell measurement on Charlies particle and her part of the entangled pair
(see Section 2.2.3). She thus projects the two-particle state randomly onto one of the four
Bell states. Note that this measurement only reveals the joint state of both particles, but
not the individual states. The outcome of this measurement projects Bobs particle onto
one of four different states as well. Using two classical bits, Alice now tells Bob about the
outcome of her measurement and depending on her message, Bob performs one of four
44 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

unitary operations on his particle: the identity operation, a bit flip, a phase flip, or a bit
and a phase flip. This finally leaves it in the state of the particle Charlie had initially
handed over to Alice, albeit neither Alice nor Bob know about this state. It is important
to note that Charlies particle is left in an arbitrary state after the Bell measurement, and
that no cloning has taken place. Moreover, since Alices classical information is needed to
reconstruct the state of Charlies particle, faster than light communication is not possible.
The experimental realization of quantum teleportation has invoked a strong reaction
in the public. Whereas one can clearly say that quantum teleportation in its current form
has no relation to disembodied transport of objects or even humans, it is also true that
the idea to transmit a quantum state without sending a particle is an intriguing concept.
Three different experiments based on polarization qubits have been reported.
1997 Bouwmeester et al.134 are the first to demonstrate quantum teleportation
based on a Bell measurement using linear optics. Although this allows
in principle to teleport in 50% of all cases, only the projection onto the
state is used in the experiment. The result of the measurement is
shown in Fig. 29.
1998 Boschi et al.135 demonstrate a teleportation set-up in which all four Bell
states can be identified even using only linear optics. The entangled
state is realized using k-vector (mode) entanglement, and the polariza-
tion degree of freedom of one of the entangled photons is employed to
prepare the unknown state. However, this scheme can not be imple-
mented for photons that come from independent sources as required for
instance for entanglement swapping (see Section 4.4).
2001 Kim et al.55 demonstrated quantum teleportation based on a Bell mea-
surement implementing non-linear interaction. This enables a projection
onto all four Bell states, however, with very small efficiency of around
one out of 1010 . In order to compensate for the efficiency, the input
state (send by Charlie) is a classical pulse from a fs laser. Nevertheless,
this experiment shows that a complete Bell measurement is in principle
possible, even when using single-photons and without having to take
advantage of additional degrees of freedom of the entangled pair.

In addition, Furusawa et al.20 demonstrated 1998 quantum teleportation based on contin-

uous quantum variables.

4.4 Entanglement swapping

If we think of quantum teleportation as the transfer of an unknown (but still well
defined) state we usually require that all states of the space we select are teleported
perfectly. A natural extension is that any relation that the original particle has with
respect to other systems should be transferred as well. Specifically, if our particle was
entangled to another system we would require from a faithful teleportation machine that
this entanglement is transferred to the particle that inherits the state at Bobs location.
This generalized concept, mentioned for the first time in 1993 by Zukowski et al.136 ,
has become known as entanglement swapping (or teleportation of entanglement). It sym-
 Wolfgang Tittel and Gregor Weihs . . . 45

Teleportation of a

+45 polarized photon +90 polarized photon

100 120

100
4-fold coincidences per 4000 seconds

80

80
60
60

40
40

20 -45 0
20
(a) (c)
0 0
120
100

100
80
80

60
60

40
40

20 20
(b) +45 +90
(d)

0 0
-150-100 -50 0 50 100 150 -150-100 -50 0 50 100 150

Delay (m) Delay (m)

Fig. 29. Experimental data showing faithful teleportation of an independently created polarization
qubit. Two linear polarization states are tested: 45 and 90 . The fidelity is roughly the same for
both cases.

Fig. 30. Entanglement swapping between two EPR-pairs by a Bell-state measurement.

46 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

metrizes the teleportation scheme to a procedure that can be applied to two or more
entangled systems.137 The lowest order protocol joining two entangled two-qubit systems
is depicted in Fig. 30. If, say, both entangled systems are produced in the | i12 and
| i34 Bell-states, and we project onto a | i23 state we find particles 1 and 4 in the
state


23 [| i12 | i34 ] = (16)
1
= [h01| h10|]23 [|01i |10i]12 [|01i |10i]34
8
1
= [h01| h10|]23 [|0101i |1001i |0110i + |1010i]1234
8
1
= [|01i |10i]14 =  14 .
 
8
Therefore, particles 1 and 4 end up in an entangled state although they never interacted
locally.
Any experimental verification of this protocol is a clear demonstration of the quantum
physical projection postulate: the joint measurement of particles 2 and 3 results in a
preparation of the joint state of particles 1 and 4, regardless of whether one decides to
project on a product, or on an entangled state. In the specific case considered here, the
projection postulate predicts the change of the joint state of particle 1 and 4 from a product
state to an entangled state. This can easily be verified by subjecting particle 1 and 4 to a
test of a Bell inequality.
A first attempt to demonstrate that the entanglement is indeed swapped has been
reported by Pan et al.138 in 1998. The experiment was based on polarization entanglement.
However, although the observed degree of entanglement surpasses the limit of a classical
wave theory, it was not high enough to manifest itself in a violation of a Bell inequality. Yet,
after some refinements, Jennewein et al. could recently demonstrate a violation of Bells
inequality with swapped entanglement.139 The measurements yielded S exp = 2.42 0.09
which exceeds the limit of 2 for local realistic theories by 4 standard deviations.

4.5 Purification and distillation

In the context of extending quantum information protocols to larger distances, con-
siderable interest has grown for measures against decoherence as encountered while trans-
mitting quantum states through a noisy environment. Various schemes of entanglement
distillation, purification and concentration, and of quantum error correction have been
proposed (see references in 16,15 ), however, to date, only one experimental demonstration
of distillation of photonic entanglement has been reported.
As predicted by Gisin in 1996,140 entanglement purification can be achieved by local
filtering. This scheme has been demonstrated experimentally 2001 by Kwiat et al.59 (Fig.
31). To demonstrate the phenomenon of hidden non-locality, certain partially entan-
gled, partially mixed states that do not violate Bell inequality were prepared utilizing the
polarization entanglement source based on two stacked thin type-I crystals with their optic
 Wolfgang Tittel and Gregor Weihs . . . 47

Tomographic/Bell
inequality analyzers
Quartz Distillation
decoherers filters 1
Ar + UV HWP
laser PBS BBO
QWP HWP PBS
HWPUV
HWP
(sets )
2

Fig. 31. Experimental setup to demonstrate entanglement distillation and hidden non-locality.
Half-wave plates and two 1cm thick quartz elements (decoherers) allow to generate partially en-
tangled, partially mixed states. After filtering they are analyzed by means of two qubit analyzers
either in order to reconstitute the density matrix via quantum tomography, or to test the CHSH
Bell inequality.
48 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

axes at 90 as discussed in Section 2.2.2. It has been shown that the resulting states after
a suitable local filtering operation violate Bell-CHSH inequalities (Fig. 32).
Actually, much more than from decoherence, applications of quantum communication
like quantum cryptography suffer from transmission losses paired with detector noise.q As
argued in section 4.1, the fact that the QBER increases with losses (Eq. 15) limits the max-
imum transmission distance. A way out would be to use a quantum non-demolition (QND)
measurement141 at Bobs and to switch on the detectors only if a photon is known to arrive
(hence = 1 and remaining tlink =1) (see Fig. 33). Unfortunately, QND measurements for
detection of visible or telecommunication photons do not exist yet.r Another possibility
that is in reach with current technology is to use a concatenation of two-particle sources
and Bell state measurements and to establish entangled photons at Alices and Bobs via
entanglement swapping. In this scheme, Bobs detectors are only switched on if Alice
detected a photon and if the last Bell state measurement (at Bobs) was conclusive. It
therefore much resembles a QND measurement. Interestingly, this idea is closely related
to the original motivation for entanglement swapping:136 testing Bell inequalities with
Event-Ready-Detectors. However, although this scheme allows to extend the maximum
transmission span, it will at the same time significantly reduce the bit rate.

Fig. 32. Measured density matrices before (left) and after (right) distillation. A violation of CHSH
Bell inequality is observed for the state after distillation (Sfiltered =2.22) while the initial state
does not manifest non-local behavior (Sinitial =1.82)

q This contrasts with proposals for quantum computing using massive particles where the particles hardly
get lost and good detectors exist, but where decoherence is the major problem.
r QND measurements for microwave photons have recently been demonstrated by Nogues et al.142
 Wolfgang Tittel and Gregor Weihs . . . 49

Fig. 33. Comparison of QND measurement based (a), and entanglement swapping based (b)
realization of event ready detectors at Bobs. 2h denote entangled two-photon sources, BSM
a Bell-state measurement, and and are the settings at Alices and Bobs qubit analyzers.
50 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

5 Conclusion
In this article, we tried to review the major developments in experiments based on
entanglement of photonic qubits, both in the traditional domain of fundamental tests
of quantum non-locality, as well as in the new approach of using entanglement as a re-
source for quantum communication. In traditional tests of spin 1/2 Bell-inequalities no
major surprises are expected any more at least concerning experiments with photons.
But we are only at the very beginning of tests of non-locality using higher dimensional
systems,143 systems employing more than two particles (Section 3.2), or experiments test-
ing other interpretations of the quantum world (Section 3.1.2). Furthermore, the fact that
entanglement can be used as a resource for applications in quantum communication came
like a great surprise and stimulated much interest in the physics as well as in the general
community. Quantum cryptography is a good candidate of becoming the first industrial
application, a development that would have a major impact on the whole field.
It is very interesting that tests of Bell-inequalities, traditionally considered part of
fundamental research, recently became an application with the Ekert scheme for quantum
cryptography, and are nowadays routinely employed in the laboratory to characterize
the quality of the entanglement of photon pairs before passing on to more complicated
experiments. However, not only the change from fundamental to applied aspects can be
observed, the opposite is possible as well: It is for instance still an open fundamental
question if the connection between security of quantum cryptography and violation of a
Bell inequality can be generalized to all sorts of protocols.
The enormous progress obtained in laser and optical fiber technology, single photon
detectors, and the availability of non-linear crystals enable nowadays experiments that
were only gedankenexperiments not even long ago. However, new ideas and technology
are needed for further steps. For instance, the secret bit rate and maximum transmission
distance in quantum cryptography experiments are limited by the performance of detectors
and single photon sources, and there is big need for efficient sources creating more than
two entangled particles. Maybe photon sources based on individual nitrogen-vacancy
color centers in diamond144,145 , quantum dots146,147,148 and parametric down-conversion
in PPLN waveguides36,37 will turn out to enable more refined experiments in the future.
Although quantum non-locality has recently been observed with atoms, photons still
play an outstanding role whenever it comes to experiments employing entanglement. They
are best suited as a carrier of quantum information since decoherence effects due to inter-
action with the environment are very small. But unfortunately, photons do not interact
with other photons, a major problem when it comes to Bell state measurements or pro-
cessing of quantum information in general. A solution might be to map photonic quantum
states onto massive particles like atoms or ions that are in principle well suited for all
applications where two-particle interaction is needed. A first experiment in this context
has been performed 149 very recently.
 Wolfgang Tittel and Gregor Weihs . . . 51

Acknowledgements
The authors acknowledge financial support by the Swiss FNRS and the Austrian Science
Fund (FWF) project no. S1506, as well as by the IST-FET QuComm project of the
European Commission, partly financed by the Swiss OFES. W. T. would like to thank
his colleagues from GAP, especially N. Gisin and H. Zbinden for theoretical and practical
support during the mentioned experiments. G. W. would like to thank T. Jennewein and
A. Zeilinger for helpful discussions and continued support.

References

1. E. Schrodinger, Naturwissenschaften 23, 807 (1935).

2. A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev. 47, 777 (1935).
3. J. Bell, Physics 1, 195 (1964).
4. J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Phys. Rev. Lett. 23, 880 (1969).
5. J. F. Clauser and M. A. Horne, Phys. Rev. D 10, 526 (1974).
6. S. J. Freedman and J. F. Clauser, Phys. Rev. Lett. 28, 938 (1972).
7. A. Aspect, Nature 398, 189 (1999).
8. P. Grangier, Nature 409, 774 (2001).
9. W. H. Zurek, Physics Today 44, 36 (1991).
10. G. C. Ghirardi, A. Rimini, and T. Weber, Lett. Nouv. Cim. 27, 293 (1980).
11. A. Suarez and V. Scarani, Phys. Lett. A 232, 9 (1997).
12. E. Hagley et al., Phys. Rev. Lett. 79, 1 (1997).
13. H. Zbinden, J. Brendel, N. Gisin, and W. Tittel, Phys. Rev. A 63, 022111 (2001).
14. Physics World 11, (1998).
15. The Physics of Quantum Information, edited by D. Bouwmeester, A. Ekert, and A. Zeilinger
(Springer, Berlin, 2000).
16. H.-K. Lo, S. Popescu, and T. Spiller, Introduction to quantum computation and informa-
tion (World Scientific, Singapore, 1998).
17. N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, Rev. Mod. Phys. (2001).
18. A. Zeilinger, Sci. Am. 282, 50 (2000).
19. Z. Y. Ou, S. F. Pereira, H. J. Kimble, and K. C. Peng, Phys. Rev. Lett. 68, 3663 (1992).
20. A. Furusawa et al., Science 282, 706 (1998).
21. C. Silberhorn et al., Phys. Rev. Lett. 86, 4267 (2001).
22. P. D. Townsend, Opt. Fiber Tech. 4, 345 (1998).
23. R. Hughes, G. Morgan, and C. Peterson, J. Mod. Opt. 47, 533 (2000).
24. H. Zbinden et al., Electron. Lett. 33, 586 (1997).
25. G. Ribordy et al., J. Mod. Opt. 47, 517 (2000).
26. S. N. Molotkov, Journal of Experimental and Theoretical Physics 87, 288 (1998), quant-
ph/9811038.
27. J.-M. Merolla, Y. Mazurenko, J.-P. Goedgebuer, and W. Rhodes, Phys. Rev. Lett. 82, 1656
(1999).
28. A. Mair, A. Vaziri, G. Weihs, and A. Zeilinger, Nature (2001), (in press).
29. H. Bechmann-Pasquinucci and W. Tittel, Phys. Rev. A 61, 062308 (2000).
30. E. S. Fry and R. C. Thompson, Phys. Rev. Lett. 37, 465 (1976).
31. A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett. 47, 460 (1981).
32. D. C. Burnham and D. L. Weinberg, Phys. Rev. Lett. 25, 84 (1970).
33. Z. Y. Ou, C. K. Hong, and L. Mandel, Phys. Lett. A 122, 11 (1987).
34. Z. Y. Ou and L. Mandel, Phys. Rev. Lett. 61, 50 (1988).
 52 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

35. C. K. Hong, Z. Y. Ou, and L. Mandel, Phys. Rev. Lett. 59, 2044 (1987).
36. S. Tanzilli et al., Electronics Letters 37, 26 (2001), quant-ph/0012053.
37. K. Sanaka, K. Kawahara, and T. Kuga, Phys. Rev. Lett. 86, 5620 (2001), quant-ph/0012028.
38. Y.-H. Kim et al., Phys. Rev. A 63, 062301 (2001).
39. T. Kiess, Y. Shih, A. Sergienko, and C. Alley, Phys. Rev. Lett. 71, 3893 (1993).
40. P. G. Kwiat et al., Phys. Rev. Lett. 75, 4337 (1995).
41. P. G. Kwiat et al., Phys. Rev. A 60, R773 (1999), quant-ph/9810003.
42. J. G. Rarity and P. R. Tapster, Phys. Rev. A 41, 5139 (1990).
43. P. H. S. Ribeiro, quant-ph/0009112 (unpublished).
44. J. Brendel, N. Gisin, W. Tittel, and H. Zbinden, Phys. Rev. Lett. 82, 2594 (1999).
45. T. Keller, M. Rubin, and Y. Shih, Phys. Lett. A 244, 507 (1998).
46. Y.-H. Kim, M. Chekhova, S. Kulik, and Y. Shih, Phys. Rev. A 54, R37 (1999).
47. J. D. Franson, Phys. Rev. Lett. 62, 2205 (1989).
48. J. Brendel, E. Mohler, and W. Martienssen, Europhys. Lett. 20, 575 (1992).
49. P. G. Kwiat, A. M. Steinberg, and R. Y. Chiao, Phys. Rev. A 47, 2472 (1993).
50. M. Michler, K. Mattle, H. Weinfurter, and A. Zeilinger, Phys. Rev. A 53, R1209 (1996).
51. K. Mattle, H. Weinfurter, P. G. Kwiat, and A. Zeilinger, Phys. Rev. Lett. 76, 4656 (1996).
52. N. Lutkenhaus, J. Calsamiglia, and K.-A. Suominen, Phys. Rev. A 59, 3295 (1999).
53. J. Calsamiglia and N. Lutkenhaus, Appl. Phys. B 72, 67 (2001).
54. P. G. Kwiat and H. Weinfurter, Phys. Rev. A 58, R2623 (1998).
55. Y.-H. Kim, S. P. Kulik, and Y. Shih, Phys. Rev. Lett. 86, 1370 (2001).
56. M. Zukowski, A. Zeilinger, and H. Weinfurter, Annals of the N.Y. Acad. of Sciences 755, 91
(1995).
57. A. G. White, D. F. V. James, P. H. Eberhard, and P. G. Kwiat, Phys. Rev. Lett. 83, 3103
(1999).
58. J. Rehacek, Z. Hradil, and M. Jezek, Phys. Rev. A 63, 040303(R)1 (2001).
59. P. G. Kwiat, S. Barazza-Lopes, A. Stefanov, and N. Gisin, Nature 409, 1014 (2001).
60. D. M. Greenberger, M. Horne, and A. Zeilinger, in Bells Theorem, Quantum Theory, and
Conceptions of the Universe, edited by M. Kafatos (Kluwer Academic, Dordrecht, The
Netherlands, 1989), pp. 6972.
61. M. Zukowski, A. Zeilinger, M. Horne, and H. Weinfurter, Acta Phys. Pol. A. 93, 187 (1998).
62. M. Hillery, V. Buzek, and A. Berthiaume, Phys. Rev. A 59, 1829 (1999).
63. A. Rauschenbeutel et al., Science 288, 2024 (2000).
64. C. A. Sackett et al., Nature 404, 256 (2000).
65. J.-W. Pan and A. Zeilinger, Phys. Rev. A 57, 2208 (1998).
66. W. Tittel, H. Zbinden, and N. Gisin, Phys. Rev. A 63, 042301 (2001).
67. D. Bohm, Phys. Rev. 85, 166 (1952).
68. D. Bohm, Phys. Rev. 85, 180 (1952).
69. J. S. Bell, Speakable and Unspeakable in Quantum Mechanics (Cambridge UP, Cambridge,
1987).
70. D. Bohm, Phys. Rev. 108, 1070 (1957).
71. S. Kochen and E. Specker, J. Math. Mech. 17, 59 (1967).
72. J. S. Bell, Rev. Mod. Phys. 38, 447 (1966).
73. C. Simon, M. Zukowski, H. Weinfurter, and A. Zeilinger, Phys. Rev. Lett. 85, 1783 (2000).
74. J. F. Clauser, Phys. Rev. Lett. 36, 1223 (1976).
75. W. Perrie, A. J. Duncan, H. J. Beyer, and H. Kleinpoppen, Phys. Rev. Lett. 54, 1790 (1985).
76. T. Pittman, Y. Shih, A. Sergienco, and M. Rubin, Phys. Rev. A 51, 3495 (1995).
77. P. Kwiat et al., Phys. Rev. A 60, R773 (1999).
78. P. Tapster, J. Rarity, and P. Owens, Phys. Rev. Lett. 73, 1923 (1994).
79. D. Strekalov et al., Phys. Rev. A 54, R1 (1996).
 Wolfgang Tittel and Gregor Weihs . . . 53

80. A. A. et al., Phys. Lett. B 422, 339 (1998).

81. A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett. 49, 91 (1982).
82. A. Aspect, J. Dalibard, and G. Roger, Phys. Rev. Lett. 49, 1804 (1982).
83. Y. H. Shih and C. O. Alley, Phys. Rev. Lett. 61, 2921 (1988).
84. J. G. Rarity and P. R. Tapster, Phys. Rev. Lett. 64, 2495 (1990).
85. W. Tittel et al., Phys. Rev. A 57, 3229 (1998).
86. W. Tittel, J. Brendel, H. Zbinden, and N. Gisin, Phys. Rev. Lett. 81, 3563 (1998).
87. G. Weihs et al., Phys. Rev. Lett. 81, 5039 (1998).
88. M. A. Rowe et al., Nature 409, 791 (2001).
89. W. Tittel, J. Brendel, H. Zbinden, and N. Gisin, Phys. Rev. Lett. 84, 4737 (2000).
90. D. Mermin, Physics Today 38, 38 (1985).
91. C. Kurtsiefer, M. Oberparleiter, and H. Weinfurter, quant-ph/0101074 (unpublished).
92. P. Pearle, Phys. Rev. D 2, 1418 (1970).
93. E. Santos, Phys. Lett. A 212, 10 (1996).
94. N. Gisin and B. Gisin, Phys. Lett. A 260, 323 (1999).
95. P. H. Eberhard, Phys. Rev. A 47, R747 (1993).
96. W. H. Furry, Phys. Rev. 49, 393 (1936).
97. W. Tittel, J. Brendel, N. Gisin, and H. Zbinden, Phys. Rev. A 59, 4150 (1999).
98. E. Santos, Phys. Lett. A 200, 1 (1995).
99. J. S. Bell, Journal de Physique 42, C2 41 (1981).
100. N. Gisin and H. Zbinden, Phys. Lett. A 264, 103 (1999).
101. G. Ghirardi, A. Rimini, and T. Weber, Phys. Rev. D 34, 470 (1986).
102. D. Bohm, Quantum Theory (Prentice-Hall, Englewood Cliffs, 1951).
103. L. Hardy, Phys. Rev. Lett. 68, 2981 (1992).
104. V. Scarani, W. Tittel, H. Zbinden, and N. Gisin, Phys. Lett. A 276, 1 (2000).
105. P. Eberhard, private communication (unpublished).
106. D. Bouwmeester et al., Phys. Rev. Lett. 82, 1345 (1999).
107. J.-W. Pan et al., Nature 403, 515 (2000).
108. N. D. Mermin, Physics Today 43, 9 (1990).
109. J.-W. Pan et al., Phys. Rev. Lett. 86, 4435 (2001).
110. C. A. Fuchs et al., Phys. Rev. A 56, 1163 (1997).
111. D. Mayers, Journal of the Association of Computer Machinery , (in press) quant-ph/9802025.
112. C. H. Bennett et al., J. Cryptology 5, 3 (1992).
113. S. Wiesner, Sigact news 15, 78 (1983).
114. C. H. Bennett and G. Brassard, in Proceedings of the International Conference on Com-
puter Systems and Signal Processing (PUBLISHER, Bangalore, India, 1984), pp. 175179.
115. C. H. Bennett, Phys. Rev. Lett. 68, 3121 (1992).
116. G. Ribordy et al., Electron. Lett. 34, 2116 (1998).
117. N. Lutkenhaus, Phys. Rev. A 61, 052304 (2000).
118. G. Brassard, N. Lutkenhaus, T. Mor, and B. C. Sanders, Phys. Rev. Lett. 85, 1330 (2000).
119. C. K. Hong and L. Mandel, Phys. Rev. Lett. 56, 58 (1986).
120. A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991).
121. C. Bennett, G. Brassard, and N. D. Mermin, Phys. Rev. Lett. 68, 557 (1992).
122. A. Ekert, J. G. Rarity, P. R. Tapster, and G. M. Palma, Phys. Rev. Lett. 69, 1293 (1992).
123. N. Gisin and B. Huttner, Phys. Lett. A 228, 13 (1997).
124. V. Scarani and N. Gisin, quant-ph/0104016 (unpublished).
125. T. Jennewein et al., Phys. Rev. Lett. 84, 4729 (2000).
126. D. S. Naik et al., Phys. Rev. Lett. 84, 4732 (2000).
127. D. Enzer, P. Hadley, G. Peterson, and P. G. Kwiat, private communication (unpublished).
128. D. Bruss, Phys. Rev. Lett. 81, 3018 (1998).
 54 Photonic Entanglement for Fundamental Tests and Quantum Communication . . .

129. H. Bechmann-Pasquinucci and N. Gisin, Phys. Rev. A 59, 4238 (1999).

130. G. Ribordy et al., Phys. Rev. A 63, 012309 (2001).
131. C. H. Bennett et al., Phys. Rev. Lett. 70, 1895 (1993).
132. W. Wootters and W. Zurek, Nature 299, 802 (1982).
133. D. Dieks, Phys. Lett. A 92, 271 (1982).
134. D. Bouwmeester et al., Nature 390, 575 (1997).
135. D. Boschi et al., Phys. Rev. Lett. 80, 1121 (1998).
136. M. Zukowski, A. Zeilinger, M. A. Horne, and A. K. Ekert, Phys. Rev. Lett. 71, 4287 (1993).
137. S. Bose, V. Vedral, and P. L. Knight, Phys. Rev. A 57, 822 (1998).
138. J.-W. Pan, D. Bouwmeester, H. Weinfurter, and A. Zeilinger, Phys. Rev. Lett. 80, 3891 (1998).
139. T. Jennewein, J.-W. Pan, G. Weihs, and A. Zeilinger, to be published (unpublished).
140. N. Gisin, Phys. Lett. A 210, 151 (1996).
141. P. Grangier, J. A. Levenson, and J.-P. Poizat, Nature 396, 537 (1998).
142. G. Nogues et al., Nature 400, 239 (1999).
143. J. C. Howell, A. Lamas-Linares, and D. Bouwmeester, quant-ph/0105132 (unpublished).
144. C. Kurtsiefer, S. Mayer, P. Zarda, and H. Weinfurter, Phys. Rev. Lett. 85, 290 (2000).
145. R. Brouri, A. Beveratos, J.-P. Poizat, and P. Grangier, Opt. Lett. 25, 1294 (2000).
146. J.-M. Gerard et al., Phys. Rev. Lett. 81, 1110 (1998).
147. J. Kim, O. Benson, H. Kan, and Y. Yamamoto, Nature 397, 500 (1999).
148. A. Kiraz et al., Appl. Phys. Lett. 78, 3932 (2001).
149. B. Julsgaard, A. Kozhekin, and E. Polzik, quant-ph/0106057 (unpublished).