You are on page 1of 21

This article was downloaded by: [University of Saskatchewan Library]

On: 31 August 2012, At: 12:43


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Journal of Modern Optics


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tmop20

Single-photon sources–an introduction


a
Stefan Scheel
a
Quantum Optics and Laser Science, Blackett Laboratory, Imperial College London, London,
UK

Version of record first published: 07 Oct 2010

To cite this article: Stefan Scheel (2009): Single-photon sources–an introduction, Journal of Modern Optics, 56:2-3, 141-160

To link to this article: http://dx.doi.org/10.1080/09500340802331849

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
Journal of Modern Optics
Vol. 56, Nos. 2–3, 20 January–10 February 2009, 141–160

TOPICAL REVIEW
Single-photon sources – an introduction
Stefan Scheel*
Quantum Optics and Laser Science, Blackett Laboratory, Imperial College
London, London, UK
(Received 31 January 2008; final version received 8 July 2008)

This review surveys the physical principles and recent developments in manufacturing single-photon sources.
Special emphasis is placed on important potential applications such as linear optical quantum computing
(LOQC), quantum key distribution (QKD) and quantum metrology that drive the development of these sources
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

of single photons. We discuss the quantum-mechanical properties of light prepared in a quantum state of definite
photon number and compare it with coherent light that shows a Poissonian distribution of photon numbers. We
examine how the single-photon fidelity directly influences the ability to transmit secure quantum bits over
a predefined distance. The theoretical description of modified spontaneous decay, the main principle behind
single-photon generation, provides the background for many experimental implementations such as those using
microresonators or pillar microcavities. The main alternative way to generate single photons using postselection
of entangled photon pairs from parametric down-conversion, will be discussed. We concentrate on describing the
underlying physical principles and we will point out limitations and open problems associated with single-photon
production.
Keywords: single photons; cluster states; linear optical quantum computing; cavity QED

1. Introduction article we will discuss the main uses of single photons


In recent years, rapid advances in quantum informa- on demand, their physical properties, and describe
tion theory and metrology were the main driving forces ways of generating them. We will focus on the physical
behind the development of a novel technological tool – principles of single-photon generation rather than
a controllable source of single photons on demand. technical details of specific implementations which
can be found in the relevant literature. In particular,
Single photons on demand are an important resource
we will concentrate on the two most widely used
in various areas of the emerging quantum technologies
methods, modifying spontaneous decay to achieve
such as quantum key distribution and all-optical
unidirectionality and near-determinism, and postselec-
quantum information processing. They are the basic
tion from entangled photon pairs generated in
prerequisite for unconditional security in quantum key parametric down-conversion.
distribution protocols and a key ingredient for fault- This article is organized as follows. We will
tolerant quantum computing schemes. In other areas introduce the three main applications that became
such as metrology, an accurate standard for the the driving forces behind the development for sources
luminous intensity, or candela, one of the seven of single photons on demand in Section 2. We then
primary SI units, can be built using single photons briefly review the differences between Fock states and
on demand in well-defined modes. attenuated coherent states in Section 3 and explain the
Generating one and only one photon at a well- intrinsic difficulties associated with single-photon
defined instance, however, proves to be a formidable generation. The main part of this review, Section 4, is
task. It amounts to producing a highly nonclassical devoted to reviewing the principal designs for determi-
state of light with strongly nonclassical properties. nistic single-photon sources and their basic functions.
Single photons on demand must therefore originate This list is necessarily far from complete, with new
from a source that operates deep in the quantum designs or combinations of mentioned designs
regime and that is capable of exerting a high degree of constantly being developed. The literature list which
quantum control to achieve sufficient purity and concludes this review should be taken as a guidance
quantum efficiency of photon production. In this and does not claim to be complete.

*Email: s.scheel@imperial.ac.uk

ISSN 0950–0340 print/ISSN 1362–3044 online


ß 2009 Taylor & Francis
DOI: 10.1080/09500340802331849
http://www.informaworld.com
142 S. Scheel

2. Applications for single photons Table 1. BB84 protocol for secret key distribution. Only in
those cases in which sender and receiver choose the same
Single photons on demand in well-defined spatio- polarization basis for preparation and measurement, a secure
temporal modes find a widespread use in modern bit is transmitted.
quantum technologies. Apart from fundamental appli-
cations in tests of quantum mechanics, single photons sender % " & ! % " ! " & % "
receiver "
! &
% &
% "
! &
% &
% "
! "
! &
% "
! "
!
play a vital part in quantum key distribution, all- key 1 1 0 1 0 1 0
optical quantum computing schemes, and radiometry.
We have already noted that light prepared in a photon-
number eigenstate possesses properties that classical
light, containing many photons that are Poissonian pulse (see Section 3), this protocol is susceptible to
distributed, cannot have. In this section, we will briefly eavesdropping. A particularly effective eavesdropping
describe those three main applications. attack is the photon-number splitting attack which
consists of performing a quantum non-demolition
measurement on each pulse. In case of more than
one photon available per pulse, the ‘surplus’ photons
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

2.1. Quantum key distribution


can be split off and can be used to perform projective
The most advanced application is certainly quantum
measurements in the basis publicly communicated
key distribution (QKD) [1–3] in which the exchange of
between the sender and receiver [4,5]. The obvious
quantum particles between two parties is used to
way to make sure that such an attack does not succeed
generate a string of secure bits of classical information.
is to send one and only one photon at any time, and
The secure classical bits then form a key with which
several experiments have been performed that imple-
a classical message can be encrypted by the sender,
ment the BB84-protocol with single-photon sources
transmitted, and deciphered by the receiver. It should
[6,7].
be stressed again that QKD does not involve encrypt-
There exist numerous extensions and modifications
ing the message into quantum states, only the exchange
to the original BB84 protocol that enhance the security
of the key involves quantum particles.
of the key distribution against various attacks despite
In the original version of the BB84 protocol [1],
the use of weak coherent pulses. Out of the many
information is encoded in the polarization of a stream of
photons. The sender switches randomly between two variants, we only mention the differential phase shift
orthogonal basis sets, for example between vertical (") protocol [8] which has resulted in the longest terrestrial
and horizontal (!) polarization and a basis consisting secure key distribution thus far [9]. In this experiment,
of 45 (%) and 135 (&) polarized photons. To each a secure bit rate of 17 kbit s1 at 105 km and 12.1 bit
individual basis state one assigns a classical bit value, for s1 at 200 km length of the optical fibre, respectively,
example ‘0’¼ (", %) and ‘1’¼ (!, &). The receiver also was achieved with coherent pulses with on average 0.2
chooses to measure randomly in one of the two basis sets photons per pulse produced at a repetition rate of 10
and collects the respective bit values. GHz. The transmission line of 200 km of optical fibre
The receiver has no information about the polar- showed 42.1 dB channel loss which is already beyond
ization basis the sender had actually chosen to send the the loss tolerance threshold of 33 dB needed to
photons in. If the receiver decided to measure in the implement QKD in satellite-based systems [10]. It
wrong basis, the measurement outcome will, with should be stressed, however, that for any given channel
probability 1/2, yield a wrong result. If, however, loss rate, an ideal single-photon source will always
sender and receiver by chance use the same polariza- yield higher secure key generation rates than any
tion basis to send and measure (which also happens in protocol based on weak coherent states. In Section 3.3
half of all cases on average), then ideally both share we show the maximal distance over which secure
information about the classical bit value encoded communication can be achieved as a function of the
within the two-dimensional space spanned by the second-order coherence function g(2)(0).
orthogonal basis states. Sender and receiver therefore
have to classically communicate their choice of basis
(note that this does not reveal any information about 2.2. All-optical quantum information processing
the bit value). This protocol is schematically depicted Single photons in well-defined quantum states are an
in Table 1. indispensable tool in most linear-optical quantum
In principle, the BB84 protocol could be realized computing (LOQC) schemes (for recent reviews on
with attenuated coherent laser pulses instead of single this topic, see [11,12]). The earliest notable proposal for
photons. However, because of the non-vanishing LOQC with single-photon Fock states in dual-rail
probability to find more than one photon per coherent polarization encoding is due to Knill, Laflamme and
Journal of Modern Optics 143

referring explicitly to a Mach–Zehnder interferometric


set-up has been proposed by Knill [18]. This network
operates with a slightly higher success rate of 2/27.
The major drawback of the circuit-based KLM
Figure 1. Left: the nonlinear sign (NS) gate forms the basis scheme is that, for it is an inherently probabilistic
of the KLM scheme. The beam splitter reflectivities protocol, it requires an unacceptably large amount of
R1 ¼ R3 ¼ (4  2(21/2))1 and R2 ¼ 1  21/2 ensure that with resources (beam splitters, single-photon sources,
a success probability of 1/4 the transformation photon-number-resolving detectors) to make it
j i ¼ c0j0i þ c1j1i þ c2j2i ° j 0 i ¼ c0j0i þ c1j1i  c2j2i is rea-
approximately deterministic [19]. Improvements to
lized. Right: two NS gates in the arms of a Mach–Zehnder
interferometer realize a probabilistic controlled-phase gate. this scheme were introduced later [20–22] and made
use of the concept of one-way computing [23] using
graph states. Within the framework of one-way
Milburn (KLM). In their seminal paper [13] they quantum computing, the computational process is
showed that, with the help of single photon sources not realized by a succession of unitary operations
and detectors and linear optical elements such as beam acting on an input state as in the gate model of
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

splitters and phase shifters, it would be in quantum computation, but rather by a sequence of
principle possible to create a universal quantum (projective) measurements on a highly entangled
computer. Their scheme combines off-line resources resource state, a cluster state or graph state [24].
of highly entangled states, quantum teleportation and Cluster states (a special subset of graph states) are
quantum error correction protocols. created by preparing all qubits (denoted by the symbol
The basic building block of the KLM scheme is the ) in an equal superposition of their two logical basis
nonlinear sign (NS) gate shown on the left in Figure 1. states j0i and j1i and performing a controlled-phase
It acts only on the two-photon component of an operation between neighboring pairs (denoted by the
arbitrary (unknown) quantum state of the form link —). For example, the two-qubit cluster state is
j i ¼ c0j0i þ c1j1i þ c2j2i and changes the phase of obtained by applying the controlled-phase gate to the
this component by  or, equivalently, performs the product state (j0i þ j1i)  (j0i þ j1i) (disregarding nor-
transformation c2 ° c2. The unitary operator that malization factors) to give
implements such a transformation is

U^ NS ¼ exp½ipnð
^ n^  1Þ=2 ð1Þ
which, after performing a Hadamard operation
with n^ denoting the photon number operator acting on j0i ° j0i þ j1i, j1i ° j0i  j1i on the second qubit, is
the signal mode containing the state j i. equivalent to the Bell state j00i þ j11i. Similarly,
The interaction Hamiltonian that would generate extending the two-qubit cluster yields (after
such a unitary operator is formally equivalent to a Hadamard operation on the third qubit) the GHZ
a (quartic) cross-Kerr interaction Hamiltonian, state
H^ / ð3Þ n^2 . The phase shift that this cross-Kerr
nonlinearity needs to generate is, however, far too
large to be experimentally accessible as the nonlinear Single-qubit measurements affect a cluster state in
crystal would have to be excessively long. different ways. A measurement in the logical basis
Hence, such type of nonlinear operation can only {j0i, j1i}, also known as a ^z measurement, removes
be implemented by means of projective measurements. a qubit from the cluster,
However, these measurement-induced nonlinearities
are only generated with less than unit probability,
and the maximally achievable success rate is just 1/4
[14–16]. Two of these NS gates sandwiched between
two symmetric beam splitters in a Mach–Zehnder
interferometer (see right figure in Figure 1) result in shown here for a simple two-dimensional cluster. In
a probabilistic (success rate 1/16) controlled-phase gate contrast, a measurement in the ^x direction not only
that introduces a  phase shift on the j1, 1i-component removes a qubit from the cluster but also joins the
adjacent qubits,
of a two-mode quantum state. Such a controlled-phase
gate, together with the ability to perform arbitrary
single-mode operations, would be sufficient for
universal quantum computing [17]. A beam splitter as shown here for the GHZ cluster. It then turns out
network realizing the controlled-phase gate without that any quantum-gate network can be rewritten in
144 S. Scheel

terms of single-qubit measurements on a specific single-photon production efficiency p (see Section


cluster state (or graph state). 7.2) and the single-photon detection efficiency  (see
Because each cluster (or graph) realizes a particular Section 7.1) satisfy the inequality [32]
quantum operation depending on its graph connectiv-
2
ity, producing clusters with the same number of qubits p 4 : ð2Þ
but different links between them are equivalent to 3
realizing different quantum networks. For example, With the currently highest single-photon detection
computation with purely one-dimensional clusters, efficiency of 70–80% using superconducting nanowire
i.e. linear chains of qubits, can be efficiently detectors (SSPDs) and single-photon generation effi-
classically simulated [25]. However, two-dimensional ciencies of at most 50–60%, this bound has not yet
clusters can provide genuine, classically non-simulable, been reached, but is not too far off experimental
computation. realizability.
In the experiment reported in [26], a four-qubit box It is worth noting that alternative schemes for
cluster cluster-state quantum computation have been pro-
posed, where the qubits are stored in extra internal
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

degrees of freedom in the single-photon sources and


entanglement is mediated by single-photon interference
has been generated which implements Grover’s search effects [33,34].
algorithm. Even bigger, six-qubit clusters in star and
ladder formations
2.3. The ‘quantum candela’
The potential to create single photons on demand in
well-defined spatio-temporal modes at high repetition
rates opens up the possibility for application of
deterministic single photon sources in the area of
metrology. Up until now, we have no accurate means
have been reported in [27]. The basic principle behind
of building a standard to measure the luminous
the experimental realizations is the production of
intensity of a light source, the candela, one of the
entangled photon pairs from parametric down-conver-
seven primary SI units. The current definition of the
sion which, as we have seen, yields two-qubit clusters
candela as it has been adopted since 1979 is [35]:
(or twin photons in old money). Using type-II fusion
gates described in [22] these two-qubit clusters can be The candela is the luminous intensity, in a given
merged to form larger clusters of almost arbitrary direction, of a source that emits monochromatic
radiation of frequency 5401012 hertz and that has
shapes.
a radiant intensity in that direction of 1/683 watt per
Despite the fact that the controlled-phase opera- steradian.
tions between two photonic qubits have to be realized
with probabilistic gates, the resource overhead drops The candela is currently realized by cryogenic radio-
dramatically compared to the KLM scheme. metry in which the optical power of a light source
The generation of an L-shaped two-dimensional heating an absorbing cavity is compared to the electric
cluster, regarded as the central building block for power heating the same sample during alternating
arbitrary clusters, requires on average 52 Bell pairs optical and electrical heating cycles. This technique is
[22], which are several orders of magnitude fewer than called Electric Substitution Radiometry (ESR)
in the original KLM proposal. With the additional use (see Figure 2).
of classical percolation methods, the creation of NN A far more elegant and potentially more accurate
cluster states could be possible using N2o(log3 N ) way of realizing the candela would be to design a tool
4-qubit clusters [28]. The resource overheads therefore that can (i) produce photons at a well-defined
grow only polynomially rather than exponentially. frequency, i.e. energy, in a well-defined spatio-
Cluster-state computation can in principle be made temporal mode and (ii) that is capable of releasing
fault-tolerant [29–31]. Fault-tolerant computing sets a predetermined number of photons in a certain time
stringent requirements on the performance and quality window. In this way, a well-defined energy flux per
of individual component elements, above which unit time and area can be produced. A deterministic
efficient error correction can be successfully applied. single-photon gun fulfils both of these requirements.
Within the framework of one-way quantum computing Photons released from some of the single-photon
it has been shown that there exist cluster-state schemes sources described below have very well-defined
that allow efficient LOQC provided that the spatio-temporal properties and, since they can be
Journal of Modern Optics 145

Section 4.2), twin photons find a wide range of


applications in quantum-optical coherence tomogra-
phy [37,38], quantum imaging [39–41] and potentially
loophole-free tests of Bell’s inequalities.
Bell’s inequality states that certain correlations
between measurements performed on spatially sepa-
rated and maximally entangled pairs of photons (EPR
pairs) exist that cannot be described by a classical local
hidden-variable model [42]. Tests of the validity of this
inequality are of fundamental importance for our
understanding of the quantum nature of our world.
However, all experimental verifications to date suf-
fered from a lack of sufficient evidence that would
support an interpretation of the obtained results in
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

favour of quantum mechanics rather than a local


hidden-variable model [43,44]. Two main loopholes
have been identified that make a quantum-mechanical
Figure 2. Schematic set-up of a cryogenic radiometer using interpretation of some experimental data difficult.
the Electric Substitution Radiometry technique. Picture The locality loophole is associated with the necessity
taken from NPL website http://www.npl.co.uk. For explana-
tion see text. (The color version of this figure is included in
to choose two independent, space-like separated
the online version of the journal.). measurement settings without the possibility of infor-
mation transfer between them. This loophole has
already been closed in photon experiments [45,46].
produced on demand, can be counted. As each of these The so-called detection loophole is related to measure-
photons carries a well-defined energy  h!, and the ments with insufficient detection efficiency whose
number of photons per unit time is known, photon statistics can be interpreted using a local hidden-
counting amounts to a very precise power measure- variable model [47]. To close this loophole in photon
ment. Such a device promises to revolutionize the experiments using nonmaximally entangled states,
existing luminous intensity standards (a ‘quantum photodetectors with at least 67% efficiency are
candela’) [36]. With the help of a single-photon needed [48]. In asymmetric atom–photon systems, in
source, there exists the realistic possibility to re-define which the atom is assumed to be detected with unit
the candela as [36]: efficiency, the minimum photodetection efficiency can
The candela is the luminous intensity, in a given be as low as 50% [49] or even 43% [50].
direction, of a source that emits photons of frequency
540  1012 hertz at a rate of 4.092  1015 photons per
second per steradian in that direction. 3. Characterization of quantum states of light
Given that the current radiometric realization In order to fully appreciate the technological achieve-
already has a relative precision of 104, single-photon ments that we are going to review in this article,
sources will have a long way to go before the new one needs to understand the fundamental difficulties
definition can be established. The reason is that using associated with the generation of single photons.
heralded single-photon sources, as is being envisaged Let us begin with examining the differences
for this particular realization, makes one dependent on between coherent states produced by a laser, and
the probabilistic nature of the postselection process Fock states.
(for details see Section 4.2).

3.1. Field quantization, Fock states versus coherent


2.4. Other applications states
So far, we have concentrated on single-photon sources. The electromagnetic field in free space (and in the
A host of new potential applications arise if one takes Coulomb gauge) is quantized by regarding the classical
twin photons (alternatively known as two-qubit cluster vector field A(r) and its canonically conjugate momen-
states or Bell states in the context of all-optical tum &(r) ¼ "0E?(r) as Hilbert-space operators Â(r)
quantum information processing) into consideration. and ^
&ðrÞ. The classical Poisson brackets
Besides their immediate use as a heralded source of {Ai(r), j(r0 )} ¼ ?
ij ðr  r0
Þ have to be replaced by
single photons by postselected measurement (see the corresponding equal-time commutators
146 S. Scheel

½A^ i ðrÞ, 
^ i ðr0 Þ ¼ i
h? 0
ij ðr  r Þ. The field operators can be electric field irrespective of the number of photons in
decomposed into monochromatic modes of frequencies the state, raise the question of whether there are other
! as quantum states of light with more familiar statistical
X properties. In particular, we are looking for states in
^ ¼
AðrÞ A ðrÞa^  þ H.c., ð3Þ which the mean value of the vector potential (and

hence of all electromagnetic field operators) could be
X obtained by replacing the photonic amplitude opera-
^
&ðrÞ ¼ i"0 ! A ðrÞa^ þ H.c., ð4Þ tors in Equation (6) by their complex amplitudes,

â °  and ây ° *. Clearly, from these arguments we
where the classical mode functions A(r) form immediately find that these states should be eigenstates
a complete set of orthonormal functions and are the of the non-Hermitian operator â with eigenvalue , i.e.
solution of the Helmholtz equation âji ¼ ji.
A proper definition of these states can be obtained
!2
A ðrÞ þ A ðrÞ ¼ 0: ð5Þ as follows. Since the Fock states form a complete set of
c2
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

orthonormal states, each unitarily transformed set of


The bosonic field commutation relations translate into states will also be sufficient to form a basis in the same
bosonic commutation relations for the annihilation Hilbert space. The coherent states ji are realized by a
and creation operators â and a^ y , ½a^  , a^y0  ¼ 0 , whose particular such unitary transformation. Transforming
elementary excitations are the photons of mode  [51]. the annihilation operator â with the displacement
For example, in cartesian coordinates the solutions operator ^
DðÞ ¼ expða^y   aÞ,
^ one obtains
to the Helmholtz equation are plane waves exp(ik  r) 0 ^ ^ y
a^ ¼ DðÞa^ D ðÞ ¼ a^  . The transformed ground
with k2 ¼ !2 =c2 which, for each mode frequency !, ^
state ji ¼ DðÞj0i is a right-hand eigenstate of the
support two orthogonal polarizations, un-transformed (non-Hermitian) annihilation operator
X2 ð  1=2 â with eigenvalue , âji ¼ ji. In contrast to
^ d3 k h

AðrÞ ¼ 3=2 2" kc
e expðik  rÞa^ ðkÞ þ h.c, Fock states, in a coherent state the mean value and
¼1 ð2pÞ 0
variance of the photon number operator are hjnji ^ ¼
ð6Þ hjðnÞ^ 2 ji ¼ jj2 . The mean value of the electric-field
and the bosonic commutation relations satisfied by the strength is hjE^ i ðrÞji ¼ i!½Ai ðrÞ  Ai ðrÞ  and thus
photonic amplitude operators read closely corresponds to classical light, with the variance
h i of the electric-field strength being independent of the
a^  ðkÞ, a^ y 0 ðk0 Þ ¼ 0 ðk  k0 Þ: ð7Þ coherent amplitude, hj[Êi(r)]2ji ¼!2jAi(r)j2, and
equal to its vacuum value.
The operator of the electromagnetic energy, i.e. the The number distribution of photons in a coherent
Hamiltonian, is then simply the infinite sum of state is Poissonian,
harmonic oscillator Hamiltonians, jj2n  
  pn ¼ jhnjij2 ¼ exp jj2 : ð10Þ
X X y 1 n!
^
H¼ ^
H ¼ ^ ^
h! a a þ :
 ð8Þ
 
2 An immediate consequence of the fact that the
photon-number variance is non-zero is that devices
Since the number operator, n^  ¼ a^y a^ , clearly commu- that produce coherent states can never be used as
tes with the Hamiltonian (8), its eigenstates (the deterministic single-photon sources. For example,
photon-number or Fock states), n^  jn i ¼ n jn i, are with jj ¼ 0.1 1, the probability of finding no
also the energy eigenstates of Ĥ with energy photon at all is p0
0.99, whereas the single-photon
E ¼ h! ðn þ 12Þ. Because of that, the photon- probability is p1
jj2 ¼ 0.01. However, there are
number variance vanishes in a Fock state, also non-negligible contributions from higher photon
^ 2 jni ¼ 0
hnjðnÞ ð9Þ numbers, e.g. p2
jj4/2 ¼ 5  105. The existence of
contributions from Fock states with photon number
[from now on, we drop the mode index ]. In contrast larger than one is detrimental for quantum key
to classical light, the mean value of the electric-field distribution and certain flavors of quantum informa-
strength vanishes, hnjÊ(r)jni ¼ 0, and its field fluctua- tion processing as they open the door for beam
tions are linear in the number of photons, splitting attacks (in case of quantum key distribu-
hnj[Êi(r)]2jni ¼ !2jAi(r)j2(2n þ 1). tion) and introduce irrecoverable errors (in quantum
These rather unexpected statistical properties of the information processing with photon-number
Fock state, in particular the vanishing of the mean encoding).
Journal of Modern Optics 147

On the other hand, we have seen that quantized


light prepared in a single-photon Fock state j1i shows
no fluctuation in photon number at all,
^ 2 j1i 0, because the Fock states are eigenstates
h1jðnÞ
of the number operator with the photon numbers being
its sharp eigenvalues. Hence, a deterministic source of
single photons must necessarily generate Fock states. Figure 3. Different pathways two indistinguishable photons
impinging on a beam splitter can take. For explanation see
Before we properly turn to the problem how to text. (The color version of this figure is included in the online
generate single photons, let us investigate ways of version of the journal.).
judging how close a given quantum state of light is
from a single-photon Fock state. In particular, we will
look at two types of correlation properties, the beam splitter. Following [52], we write the joint
Hong–Ou–Mandel quantum interference (or mutual probability for detecting one photon at times t and
indistinguishability) and photon anti-bunching (mea- t þ , respectively, at the detectors 1 and 2 as
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

sured by second-order intensity correlation).


P12 ð Þ ¼ Sh jE^ ðÞ ^ ðÞ ^ ðþÞ ^ ðþÞ
1 ðtÞE2 ðt þ ÞE2 ðt þ ÞE1 ðtÞj i ,
ð11Þ
3.2. Hong–Ou–Mandel quantum interference
where S is the spectral response function of the (broad-
The Hong–Ou–Mandel effect is a quantum interfer- band) detector. The electric-field operators appearing
ence effect that occurs when two indistiguishable in Equation (11) are the operators of the outgoing
photons impinge onto a symmetric beam splitter [52]. fields which have to be related to the electric-field
Based on the quantum theory of the beam splitter [53] operators of the incoming fields. For a symmetric
(we also refer the reader to standard textbooks, e.g. beam splitter one obtains [we set t ¼ 0 and remove
[51]) one can devise an intuitive argument based on constant prefactors]
adding probability amplitudes.
In Figure 3 we show the possible pathways two P12 ð Þ / h jE^ ðÞ ^ ðÞ ^ ðþÞ ^ ðþÞ
1 ð0ÞE2 ð ÞE2 ð ÞE1 ð0Þj i
identical single photons can take when they impinge on
þ h jE^ ðÞ ^ ðÞ ^ ðþÞ ^ ðþÞ
2 ð0ÞE1 ð ÞE1 ð ÞE2 ð0Þj i
a beam splitter. Although the photons are indistin-
guishable, we label them by different colors in order to  h jE^ ðÞ ð0ÞE^ ðÞ ð ÞE^ ðþÞ ð ÞE^ ðþÞ ð0Þj i
1 2 1 2
trace their paths. Paths (a) and (b) depict the two  h jE^ ðÞ ^ ðÞ ^ ðþÞ ^ ðþÞ
2 ð0ÞE1 ð ÞE2 ð ÞE1 ð0Þj i: ð12Þ
possibilities of having one photon reflected and the
other transmitted through the beam splitter. In these The electric-field operators in Equation (12) are now
cases both photons end up in the same output arm and the input field operators with respect to which the
leave the other port empty. quantum state j i has to be defined.
Reflection from an interface between two dielectric We assume the initial quantum state to be a pure
materials leads to a phase shift. For simplicity, we will product state of two single photons that may have
assume that the photons arriving from the bottom different spatiotemporal mode profiles. In order to
experience a  phase shift. In paths (c) and (d) the define these profiles, let us return to the definition of
photons leave the beam splitter in separate output the monochromatic mode expansion, Equation (4),
ports. However, the amplitudes of paths (c) and (d) and let us unitarily transform the annihilation
have opposite phases, and because of the indistinguish- operators
ability of the two photons, the amplitudes of these two X
paths cancel each other. Hence, two photons impinging a^ 0 ¼ U
a^
ð13Þ

on a symmetric beam splitter will always leave in the


same output port. so that the positive-frequency component of the
In contrast to Fock states, a tensor product of two electric-field strength in the Heisenberg picture can be
coherent states remains a tensor product of coherent written as
states, even if they are mixed at a beam splitter. In this !
sense, coherent states behave like classical light. X X
^EðþÞ ðr, tÞ ¼ i U ! A ðrÞ expði! tÞ a^ 0 : ð14Þ

The Hong–Ou–Mandel quantum interference effect 

can be used to define a measure for the indistinguish-
ability between two single-photon sources. For this Setting ¼ (1, 2), the functions Uð1,2Þ (i.e. the columns
purpose, we have to calculate the joint probability for of the unitary matrix U ) determine the spatiotemporal
detecting one photon in each of the output ports of the mode profiles of the (non-monochromatic) modes 1
148 S. Scheel

and 2 in terms of the monochromatic modes A(r).


Hence, we can write
X
E^ ðþÞ
1 ðr, tÞ ¼ i U1 ! A ðrÞ expði! tÞa^ 01 , ð15Þ


X Figure 4. Schematic set-up of a Hanbury-Brown–Twiss


E^ ðþÞ
2 ðr, tÞ ¼ i U2 ! A ðrÞ expði! tÞa^02 : ð16Þ experiment. The intensity correlation function with a delay

time is measured at the correlator C.
The input state can now be defined in terms of these
non-monochromatic modes as
3.3. Photon anti-bunching
0 0
j i¼ a^ 1y a^2y j0, 0i: ð17Þ Another feature of nonclassical states of light such as
Fock states that distinguishes them from classical (or
A straightforward calculation then reveals that the near-classical) states such as coherent states is the
joint probability (12) can be expressed as
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

effect of anti-bunching. In an experiment of the


Hanbury-Brown–Twiss type [55] (see Figure 4) one
P12 ð Þ / jG1 ð0Þj2 jG2 ð Þj2 þ jG1 ð Þj2 jG2 ð0Þj2 measures the intensity correlation function
 
 G1 ð0ÞG2 ð0ÞG1 ð ÞG2 ð Þ þ c.c. , ð18Þ Gð2Þ ð Þ ¼ lim hE^ ðÞ ðtÞE^ ðÞ ðt þ ÞE^ ðþÞ ðt þ ÞE^ ðþÞ ðtÞi:
t!1
where we have defined the (temporal Fourier) ð22Þ
transforms
For classical light, when the operator character of the
X
Gi ðtÞ ¼ Ui ! A ðrÞ expði! tÞ: electric field does not play a role, one can write
ð19Þ Equation (22) as


When integrating the joint probability over Gð2Þ


class ð Þ ¼ lim hIðt þ ÞIðtÞiclass , ð23Þ
t!1
a measurement time that is assumed to be longer
than the spread of Gi(t), we obtain an expression for where I ¼ E() E(þ) is the classical intensity.
the expected number of coincidences as The classical average is performed using the joint
ð ð probability distribution pclass(I1, t þ , I2, t) for two
Nc / jG1 ð0Þj2 d jG2 ð Þj2 þ jG2 ð0Þj2 d jG1 ð Þj2 intensities I1 and I2 as [51]
 ð  ðð
 
 G1 ð0ÞG2 ð0Þ d G1 ð ÞG2 ð Þ þ c.c. : ð20Þ Gð2Þ
class ð Þ ¼ lim dI1 dI2 I1 I2 pclass ðI1 , t þ , I2 , tÞ: ð24Þ
t!1

On examining Equation (20), we see that only if Application of the Cauchy–Schwarz inequality yields
G1(t) ¼ G2(t) the coincidence count drops to zero. ð 1=2
ð2Þ 2
Clearly, the photons are indistiguishable and interfer Gclass ð Þ lim dI1 I1 pclass ðI1 , t þ Þ
t!1
perfectly. On the other hand, if Ðthe spatiotemporal ð 1=2
mode profiles do not overlap, d G1 ð ÞG2 ð Þ ¼ 0,
 dI2 I22 pclass ðI2 , tÞ ¼ Gð2Þ
class ð0Þ , ð25Þ
the coincidence counts are maximized. Hence, we
can define a degree of mutual indistinguishability
where pclass(I, t) is the marginal probability distribu-
as [54]
tion. With this definition we find that for classical light
Ð
Re½G1 ð0ÞG2 ð0Þ d G1 ð ÞG2 ð Þ
M¼ Ð Ð , ð21Þ Gð2Þ ð2Þ
class ð Þ Gclass ð0Þ,
ð26Þ
jG1 ð0Þj2 d jG2 ð Þj2 þ jG2 ð0Þj2 d jG1 ð Þj2
meaning that the probability of detecting a photon
which takes its values in the interval M 2 [0, 1]. immediately after the arrival of another photon is
The value M ¼ 1 corresponds to two perfectly nonzero. Photons in classical light always arrive in
indistinguishable photons. By definition, the depth clumps (as do buses in London), this effect is called
of the Hong–Ou–Mandel dip is proportional to photon bunching. Equality holds for coherent states.
1  M. While mutual indistinguishability is a vital There are, on the other hand, nonclassical quantum
ingredient in linear-optical quantum computing states of light, that obey the reverse inequality
schemes, it is not necessary for a successful quantum
key distribution. Gð2Þ ð Þ 4 Gð2Þ ð0Þ: ð27Þ
Journal of Modern Optics 149

Figure 6. Theoretically allowed transmission loss 1  jTj2 in


Figure 5. Second-order correlation function g(2)( ) in arbi- dB as a function of the value of the intensity correlation at
trary units for classical light (red curve) with g(2)(0) ¼ 2, zero time delay, g(2)(0), to achieve a secure bit rate of 108
coherent light with g(2)( ) ¼ 1 (green curve), and perfectly bits per pulse. After [6].
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

anti-bunched (nonclassical) light with g(2)(0) ¼ 0 (blue curve).


The width of the curves reflects the coherence time of the
photons. (The color version of this figure is included in the
online version of the journal.). Table 2. Experimental values for the intensity correlation
function gð2Þ
exp ð0Þ at zero time delay.

This means that after the arrival of one photon, the Single-photon source gð2Þ
exp ð0Þ Refs.
probability of finding a second photon immediately
NV color center 0.08 [7]
after the first is greatly reduced. The photons tend to 87
Rb in optical cavity 0.054 [56]
arrive separately. This effect is called photon anti- 133
Cs in optical cavity 0.063 [57]
bunching. 40
Caþ in ion-trap cavity 0.015 [58]
In the literature, one finds the normalized intensity semiconductor polymer 0.08 [59]
correlation function (or simply second-order single molecule fluorescence 0.25 [60]
electrically driven InAs quantum dot 0.28 [61]
coherence) electrically driven InGaAs/GaAs quantum dot
0 [62]
GaN quantum dot 0.3 [63]
hE^ ðÞ ðtÞE^ ðÞ ðt þ ÞE^ ðþÞ ðt þ ÞE^ ðþÞ ðtÞi InAs quantum dot in micropillar 0.02 [64,65]
gð2Þ ð Þ ¼ lim :
t!1 hE^ ðÞ ðt þ ÞE^ ðþÞ ðt þ ÞihE^ ðÞ ðtÞE^ ðþÞ ðtÞi
ð28Þ
In Figure 6 we show the maximal distance over
Let us examine the possible values of g(2)(0) for which secure communication can be established with
different single-mode quantum states. Using the a number of secure bits per pulse of 108 as a function
definition (14) and restricting ourselves to one mode of g(2)(0). Additional parameters are the dark count
( ¼ 0), we immediately see that for a single-photon
0
probability of pdark ¼ 2.5  109 (corresponding to the
Fock state j i ¼ a^ y0 j0i we must have G(2)(0) ¼ 0 and experimental data presented in [9]; time jitter of the
thus g(2)(0) ¼ 0 because the action of two annihilation detectors j ¼ 50 ps, dark count rate d ¼ 50 Hz,
operators on a single excitation destroys the state. On pdark ¼ d j ¼ 2.5  109) and a baseline signal error
the other hand, for a single-mode coherent state rate of
¼ 4%.
obeying a^ 0 ji ¼ ji we find that g(2)( ) ¼ 1 which Typical experimental values lie in the range
marks the boundary between classical and nonclassical between gð2Þ exp ð0Þ ¼ 0:015 and gð2Þ exp ð0Þ ¼ 0:3 (see
states. A thermal P state with density operator Table 2), with one experiment reporting a value of
%^ ¼ ½1  expðxÞ n expðxnÞjnihnj (x ¼ h!/kBT ) gð2Þ
exp ð0Þ
0 [62]. However, the measured values are
shows g(2)(0) ¼ 2 (see Figure 5). generally higher than the actual values for the sources
Measurements of the intensity correlation function themselves. Dark counts in the detectors increase the
in a Hanbury-Brown–Twiss set-up can therefore reveal probability to register two photons at the same time,
the quantum nature of light and, as the Hong–Ou– one from the single-photon source and one from the
Mandel quantum interference effect, can be used to background light. Given the signal-to-background
determine the degree of nonclassicality of a quantum ratio x ¼ S/B, one can estimate the actual value of
state of light. For example, the value of g(2)(0) g(2)( ) from the measured value gð2Þ
exp ð Þ as [66]
determines the maximal channel loss, and thus the
maximal transmission distance, that can be tolerated gð2Þ
exp ð Þ  1 x
gð2Þ ð Þ ¼ 1 þ , ¼ : ð29Þ
for secure quantum key distribution [4,6]. 2 xþ1
150 S. Scheel

The dramatic effect of background noise in the 4.1. Spontaneous decay


detector on the measured value of the zero-delay The basic problem one faces when attempting to design
coincidence probability has been demonstrated in [67]. a single-photon source is to make sure that at any time
In this work, single-photon emission from the same one and only one photon is produced. The idea behind
semiconductor quantum dot has been studied with two virtually all designs is the release of a photon by
different types of detectors, two InGaAs avalanche a single atomic or molecular system with a wide variety
photodiodes (APD) and two NbN nanowire super- of emitters such as trapped neutral atoms or ions,
conducting single-photon detectors (SSPD). While the quantum dots, color centers or organic molecules being
APD measurement registered a coincidence probability investigated. Spontaneous emission, however, is an
g(2)(0) ¼ 0.38, the same experiment using the SSPDs gave isotropic and probabilistic process. But what is in fact
a much improved value of g(2)(0) ¼ 0.18 due to the much needed is a unidirectional, i.e. a highly anisotropic and
lower dark count rate of the SSPD. deterministic process. The common way to proceed is
to tailor the spontaneous decay to make it approxi-
mately unidirectional. How can this be achieved?
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

4. Single-photon sources Recall that the rate of spontaneous decay is given by


After this brief overview of the characterization and Fermi’s Golden Rule as
possible applications of single photons, we will now
turn our attention to the latest developments in 2p
designing single-photon sources. There already exist Gi!f ð!Þ ¼ ð!ÞjMi!f j2 , ð30Þ
h
excellent reviews on this topic [54,68] that give very
detailed accounts of a wide range of experiments. In where (!) is the density of states and Mi!f the
this review we will concentrate on the basic functional relevant dipole transition matrix element.
principles of some of the implementations as well as Dielectric structures or resonators can now be used
their limitations. to alter the density of modes in the frequency window
Single-photon sources can be classified in various occupied by the atomic transition (Purcell effect) [69].
ways. Oxborrow and Sinclair [54] distinguish various In a non-magnetic bulk material with (real) refractive
realizations by their control or trigger mechanism. index n(!), the spontaneous decay rate changes from
The so-called Class I sources implement a controlled its free-space value 0 to (!) ¼ n(!)0 [70]. Structured
pump mechanism where the release of the photon is materials can generate effective refractive indices with
random, the Class II sources, on the other hand, rather complicated frequency dependencies with the
continuously pump the emitter and control the release prospect of enhancing [69] or inhibiting [71,72]
of the photon. A combination of both, the Class III spontaneous decay. For example, Bragg stacks or
sources, exerts full control over both pump and trigger photonic band-gap structures are known to reduce the
mechanisms. We will give examples of these classes in density of modes in a certain frequency window by
the following sections. Another possible classification, multiple reflections that lead to destructive interfer-
introduced by Lounis and Orrit [68], is concerned with ence. It should, however, be noted that due to a sum
the photon emitter itself. As we will see, the photon rule valid for the spontaneous decay rate in lossless
emitters can be atoms or ions in resonant cavities, materials [73],
organic molecules, colour centres, semiconductor
ð1 ð1
nanocrystals and quantum dots. The single-photon Gð!Þ  G0
d! ¼ d!½nð!Þ  1 ¼ 0, ð31Þ
emitters themselves are embedded in resonant dielectric 0 G0 0
structures to ensure unidirectional emission.
The other main method to produce single photons, the density of modes has to be increased in other
albeit not necessarily on demand, is by postselection. frequency regions. The last equality in Equation (31)
Here, a parametric down-conversion source is used to follows from a sum rule for the (real part of the)
generate entangled, i.e. quantum-mechanically corre- refractive index and is a result of the superconvergence
lated, pairs of photons. Registering one of these theorem for Hilbert transform pairs [74]. The enhance-
photons in a photodetector ensures that a twin of ment effect near a band edge in photonic band-gap
this photon must exist in the mode that had been materials is used in micropillar structures (see Section
entangled with the measured photon. Although this 5.1). Similarly, resonator-like structures (see Section 6)
process is inherently probabilistic, the instance of are used to enhance the spontaneous decay in one
photon creation is known and ‘heralded’ by the particular spatial direction.
detection of its twin. This section describes the physical The discussion above is strictly correct only for
principles behind both methods. lossless materials. In case of absorbing materials,
Journal of Modern Optics 151

Fermi’s Golden Rule (30) has to be replaced by the local-field corrected modified spontaneous decay rate
generalized formula [75–77] that we encountered earlier for real refractive index.
The sum rule (31) can now be extended to
2!2A
Gð!A Þ ¼ d  Im GðrA , rA , !A Þ  d , ð32Þ absorbing materials. It can be shown that the
" 0 c2 
h regularized decay rate
where d is the matrix element of the atomic dipole

3ð"  1Þ 1 9ð"  1Þð4" þ 1Þ 1
moment and G(r, r0 , !) denotes the dyadic Green G0 ð!Þ ¼ G  G0 Im þ
2" þ 1 3 5ð2" þ 1Þ2
function associated with the classical scattering pro-
blem. More precisely, the Green tensor satisfies the ð37Þ
partial differential equation fulfils the generalized sum rule [80]
ð1
!2 G0 ð!Þ  G0
;  ;  Gðr, r0 , !Þ  "ðr, !ÞGðr, r0 , !Þ ¼ ðr  r0 ÞI , d! ¼ 0: ð38Þ
c2 0 G0
ð33Þ
This means that the rate 0 (!) can be associated with
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

where "(r, !) is the (complex) space- and frequency- decay processes into photonic final states, whereas the
dependent dielectric permittivity and I denotes the unit subtracted terms contain all non-radiative decay
dyad. Analytical and numerical methods for comput- processes. The strength of the latter, assuming
ing the dyadic Green function in various geometries a single-resonance Lorentz model, turns out to be
can be found, for example in [78]. proportional to the squared plasma frequency which
Equation (32) describes the spontaneous decay rate itself is proportional to the number of electrons per
of a two-level atomic system in a general absorbing unit volume of the dielectric material [80]. Sum rules
environment, including non-radiative decay processes for decay rates reflect the fact that inhibition and
(Förster transfer) and local-field corrections. Assuming enhancement of spontaneous decay are delicately
a real-cavity model in which the radiating atom is balanced over all frequencies, examples of which will
placed in an empty cavity of radius Rcavity, the latter be provided in Sections 5.1 and 6.
can be explicitly written as [79]
G ¼ G0 þ Gcavity þ Gbulk ð34Þ 4.2. Postselected single-photon sources
with the contribution to the decay rate associated with In certain cases, it is not necessary to generate single
the real cavity [ ¼ Rcavity!A/c, " ¼ "(rA, !A)] photons deterministically. Instead, it might be suffi-
cient just to know when a photon had been produced,
Gcavity
for example, when a trigger signals the arrival of a
G0 photon. If the trigger itself is a photodetection process,


3ð"  1Þ 1 9ð"  1Þð4" þ 1Þ 1 9"5=2 the triggering photon must have been entangled with
¼ Im þ þi 1
2" þ 1 3 5ð2" þ 1Þ2 ð2" þ 1Þ2 the photon that one wishes to make use of.
ð35Þ Entangled photons, or biphotons, are the result of
a (2) nonlinear process known as spontaneous para-
and the bulk contribution without the cavity, metric down-conversion (PDC) which was first reported
" 2 # in [81]. Energy and momentum conservation requires
2!2A 3" ðS Þ
Gbulk ¼ d  Im Gbulk ðrA , rA , !A Þ  d, the pump, signal and idler photons to obey the relations
" 0 c2 
h 2" þ 1
!pump ¼ !signal þ !idler and kpump ¼ ksignal þ kidler :
ð36Þ
ð39Þ
where GðS Þ
bulk is the scattering part of the Green function
of the medium without the cavity. The first two terms The Hamiltonian governing such a process must
in the rate contribution cavity are proportional to the therefore be of the form
imaginary part of the dielectric permittivity "(rA, !A) H^ int ¼ hð2Þ a^ ypump a^ signal a^idler þ h.c: ð40Þ
which is responsible for absorption. It thus follows that
these two terms describe non-radiative near-field In the parametric approximation, i.e. when depletion
processes, i.e. those that are not associated with of the pump is negligible, one replaces the operator
a photonic final state. In other words, a decay process âpump by its coherent amplitude and writes the
into one of those channels will not yield a photon and interaction Hamiltonian in the simpler form
is thus detrimental to the performance of a single-
photon source. The third term in cavity is the H^ int
hð2Þ pump a^signal a^ idler þ h.c. ð41Þ
152 S. Scheel

The associated unitary time evolution operator


h) corresponds to a squeeze operator.
Û ¼ exp(iĤintt/
Starting from the vacuum in signal and idler modes,
the two-mode quantum state of signal and idler Figure 7. Heralded single-photon source. A parametric
photons can be written in the Fock basis as down-conversion source based on a (2)-nonlinear material
X
1 produces correlated (entangled) photons. Upon detection of
j PDC i ¼ ð1  2 Þ1=2 n jni1 jni2 : ð42Þ the idler photon, a trigger T allows the signal photon to pass.
n¼0

Ideally, signal and idler photons would be perfectly


correlated (entangled) as can be seen from the (pure) probability of registering a photon in the idler arm by
quantum state (42). In reality, however, there are many p1 as above, one could in principle operate bp11 c PDC
more degrees of freedom to be taken into account than sources in parallel and multiplex the outputs to a single
just the photon number content. Spatial and temporal output.
pulse shapes as well as polarization play important
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

roles in determining the full quantum state of signal


and idler photons. A full characterization and control 5. Quantum dots in dielectric structures
of all degrees of freedom is necessary to produce an
entangled state of the form (42) [82]. Eliminating In this and the following sections we will describe
temporal and frequency correlations by employing examples of typical experimental implementations of
narrow-band filters, however, decreases the rate at single-photon sources that make use either of the
which these entangled photons can be generated. principle of modified spontaneous decay in solid-state
Alternative approaches such as engineering the structures (this section) or free-space microresonators
(2)-material and properties of the pump have been (Section 6), or use postselection from entangled
implemented that reduces these unwanted correlations biphotons (Section 7).
[83,84].
The effective squeeze parameter  that determines
the photon number content in the signal and idler 5.1. Micropillars
modes is usually much smaller than unity. Dielectric structures such as Bragg reflectors and
The probability p1 of finding exactly one photon in photonic band-gap (PBG) materials have been pro-
either of the two modes is thus p1 ¼ posed and are frequently used to tailor the mode
h PDC jn^ ð1 or 2Þ j PDC i ¼ (1  2)2 1. Although the structure of the electromagnetic field in selected spatial
probability p2 of finding two photons in each mode, directions. Recall that one requires both unidirection-
p2 ¼ 2(1  2)4, is even smaller than p1, their relative ality as well as a strong enhancement of spontaneous
weight, p2/p1 ¼ 22
2p1, is not negligible. These multi- decay to make a single-photon device as deterministic
photon contributions are detrimental for the security as possible. The band edge provided by a layered Bragg
of quantum key distribution as they allow photon- stack fulfils both requirements at once. Because it
splitting attacks, but they can be filtered out by using massively increases the decay in the direction perpen-
a photon-number resolving detector instead of dicular to the stack layers, and leaves the decay in the
a ‘bucket’ (photon/no photon) detector. transverse directions almost untouched, it creates
With such a source of entangled photons, a strong anisotropy for the otherwise isotropic
a heralded single photon source can be realized by spontaneous decay.
placing a photodetector in the path of the idler photon For simplicity, we will focus on a one-dimensional
(see Figure 7) [85]. Upon registering a photon in the periodic structure that consists of alternating layers of
detector, by Equation (42), there must also be a photon thickness d1 and d2 made of dielectric materials with
in the signal mode. This method has several draw- refractive indices n1 and n2 such that the relation
backs. The most obvious disadvantage is that it is 0 pc
a probabilistic way of generating photons, as opposed n1 d1 ¼ n2 d2 ¼ ¼ ð43Þ
4 2!0
to a deterministic photon gun. Although in some
applications such as in versions of LOQC there is no is fulfilled for some fixed wavelength 0 corresponding
need for deterministic photon generation, in single- to a midgap frequency !0. Let us further define the
photon radiometry with application to the quantum average (bulk) group velocity vbulk ¼ c/2(1/n1 þ 1/n2)
candela heralding is not good enough. Attempts have and the ratio
¼ n1/n2 of the refractive indices which
been made to make PDC sources quasi-deterministic we assume without loss of generality to be greater
by operating them in parallel [86]. If we denote the than one.
Journal of Modern Optics 153

Figure 8. Left figure: an isolated atom or a quantum dot located in a spacer between two quarter-wave Bragg stacks experiences
a strong change in the local mode density. Right figure: top view of photonic-crystal defect pillar microcavities with triangular or
square lattice structure. Figure taken from [90]. Reprinted with permission from Ho, Y.-L.D., Cao, T., Ivanov, P.S., Cryan, M.J.,
Craddock, I.J., Railton, C.R., Rarity, J.G. IEEE J. Quantum Electron. 2007, 43, 462–472. Copyright 2007, IEEE.

In an N-period quarter-wave stack, the density of Improvements to the performance of micropillar


modes at exactly the midgap frequency !0 is then [87] devices can be made by considering more sophisticated
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

   structures that go beyond circular pillar geometries


midgap 1
þ 1
2N  1 1 such as triangular or square lattice photonic crystals in
N ¼ , ð44Þ
2vbulk
 1
2N þ 1 N place of the spacer [90] (right figure in Figure 8). These
photonic crystal structures are there to suppress
which is seen to decrease with increasing number of
photon emission into transverse modes that diminish
periods N. On the other hand, at the band-edge
the performance of a single-photon device.
resonance, the density of modes for large stacks
increases quadratically with N,

1 ð
 1Þ2 2 5.2. Triggering the photon emission – shifting the
band -edge
N!1
N : ð45Þ
N
p2 vbulk
band gap
In the nomenclature of Oxborrow and Sinclair [54],
Note that in both Equations (44) and (45) the average single-photon sources based on quantum dots in
density of modes is equal to the inverse of the bulk micropillars fall into Class I, in which the excitation
group velocity vbulk so that vbulk N becomes process is controlled and the emission process is
a dimensionless quantity. The characteristic depen- spontaneous (albeit modified to make it as unidirec-
dence of the local density of modes on frequency is tional as possible). We have seen that unidirectionality
shown in Figure 9, Section 5.2. can be achieved by enhancing the spontaneous decay
The change in the density of modes is, as noted in rate in one selected direction by placing the atom in
Section 4, reflected in the modification of the a one-dimensional dielectric band-gap structure which
spontaneous decay rate. This effect can be enhanced has been designed in such a way that the atomic
by either increasing the number of periods N or by transition frequency falls into the band-edge
increasing the refractive index contrast
. Whereas resonance.
midgap
N [Equation (44)] stays approximately constant The location of the band gap is determined by the
for sufficiently large
, the density of modes at the refractive index ratio
¼ n1/n2. A change in one of the
band edge, band -edge [Equation (45)], increases linearly
N refractive indices will change the location of the stop
with
. gap [91]. Such a change can be induced externally if
A commonly used design is to use micropillars with one of the materials shows a substantial Kerr
a diameter of a few micrometers (see left figure in nonlinearity. The optical Kerr effect results in an
Figure 8). The quarter-wave layers in the propagation intensity-dependent refractive index of the form
direction consist of GaAs or AlAs layers alternating n(I) ¼ n0 þ (n)I. Thus, irradiating a Bragg stack with
with InAs layers with refractive indices nGaAs ¼ 3.55 light will shift the stop gap by an amount that depends
and nInAs ¼ 2.94 [88,89]. The one-wavelength GaAs on n and the number of layers N.
spacer contains either an atom that has been injected In Figure 9 we show the shift of the normalized
during the growing process of the stack, or a quantum mode density for a 40-layer stack with n1 ¼ 1, n2 ¼ 2
dot. The latter method commonly requires selecting (solid line) and n2 ¼ 2.006 (dashed line) [92]. We assume
those pillar microcavities containing quantum dots a structure whose midgap frequency is tuned such that
with the desired properties. Confinement in the !A ¼ 0.7814!0. Without the externally applied field,
direction perpendicular to the emission direction is !A is in the stop gap, and no de-excitation can occur
ensured by the high GaAs/InAs–air refractive index (other than non-radiative decay). Shifting the refractive
contrast. index from n2 ¼ 2 to n2 ¼ 2.006 brings !A right into the
154 S. Scheel
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

Figure 9. Left figure: normalized local mode density at the position of the atom or ion in a 40-layer stack with n1 ¼ 1 and n2 ¼ 2.
Right figure: the displaced bandgap for n2 ¼ 2 (solid line) and n2 ¼ 2.006 (dashed line). Figures taken from [92]. (The color version
of this figure is included in the online version of the journal.).

band-edge resonance and due to the strong Purcell 1/Q ¼ 2 /(2f ). Thus, Q ¼ 2c/(2 ). The finesse F, on
enhancement the atom almost instantaneously releases the other hand, is related to the energy loss per photon
a photon. In the nomenclature of Oxborrow and round trip through the cavity of length L, F ¼ 2c/
Sinclair [54], such a process would fall into Class III [2 (2L)]. Therefore, quality factor and finesse are
where both excitation and emission processes are related by Q ¼ (2L/)F.
controlled. The absorption-assisted part of the Q-factor can be
expressed in terms of the real (n0 ) and imaginary (n00 )
parts of the refractive index of the cavity mirrors as [94]
6. Single atoms in resonant microcavities
n0 ð!Þ
As in most other schemes, cavity QED relies on the Qabs ’ : ð46Þ
strong enhancement of spontaneous decay in a selected 2n00 ð!Þ
spatial direction. In the case of a cavity with high The total Q-factor is then determined by taking both
quality factor Q, this is achieved by confining the absorptive losses and radiative losses into account, 1/
electromagnetic field between highly reflecting Q ¼ 1/Qabs þ 1/Qrad. A quality factor of Q ¼ 4.2  1010
confocal mirrors. The effect of this confinement is to and a finesse of F ¼ 4.6  109 has been achieved in
select a discrete number of (equally spaced) supported a superconducting microwave cavity [95]. Typical
modes with a line width proportional to the cavity values for the finesse of high-quality optical cavities
decay rate . The spacing of the modes is given by the used for single-photon generation are F
60,000 [96].
condition for constructive interference; the allowed Note that in this case a very high finesse is not even
cavity modes form standing waves (for a review on desired as the photon has to be able to leave the cavity
optical microcavities, see e.g. [93]). through one of the mirrors. For example, the cavity
From the discussion on sum rules [see Equation used in [96] consisted of one (almost) perfect mirror
(31)] it is clear that the redistribution of modes goes and one with a 25-fold increased transmittivity thereby
hand in hand with a strong enhancement of the density reducing the potentially achievable finesse.
of states in the frequency regions defined by the cavity A different line of approach has been to combine
modes, and a decrease in mode density elsewhere. After techniques of trapping atoms near microstructured
all, Equation (31) can be thought of as a conservation surfaces (known as atom chips) and cavity QED.
law for the integrated density of states, i.e. the total Microcavities consisting of a spherical reflector etched
number of available modes. into the silicon wafer surface and a polished optical
There are several figures of merit that are used to fiber with a dielectric Bragg reflector attached to it
describe the quality of a cavity, the quality factor Q have already achieved a finesse of F 4 5000 [97]
and the finesse F. The quality factor Q is defined as the (Figure 10). In addition, because of the available
ratio of the centre frequency and the line width of the microstructuring techniques it is straightforward to
cavity resonance. Equivalently, Q1 is proportional build large regular arrays of these microcavities, each
to the fraction of energy lost per field oscillation, of which can serve as a single-photon source.
Journal of Modern Optics 155

For a successful deterministic photon production at


least an atomic three-level system is needed (see
Figure 11). The states je, 0i and jg, 1i are connected
by the strong cavity coupling g, whereas the states je, 0i
and jl, 0i are driven by a near-resonant laser with Rabi
frequency p. The process starts with an atom being
placed into the cavity in the state jl, 0i. A laser pulse is
then applied to drive the atom towards the state je, 0i
which itself is strongly coupled to the state jg, 1i. Thus,
Figure 10. Schematic diagram of a microcavity mirror
the pump laser plus the cavity transition couple the
etched into a silicon substrate. The second cavity mirror
consists of a dielectric multilayer glued onto the cleaved tip of three states such that a Raman transition can
an optical fibre. Figure taken from [97]. Reused with be driven.
permission from M. Trupke, Appl. Phys. Lett. 2005, 87, One of the three eigenstates of the atom–cavity
211106. Copyright 2005, American Institute of Physics. system is [96]
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

2gjl, 0i  Op jg, 1i
jd i ¼ 1=2 : ð48Þ
4g2 þ O2p

This particular eigenstate jd i is not coupled to the


atomic excited state je, 0i and is hence called a dark
state. The atom is never excited and therefore cannot
deposit a photon via spontaneous decay into any mode
other than the cavity mode. With a sufficiently slowly
varying (adiabatic) pump laser intensity the atom–
Figure 11. Schematic three-level system in which the states cavity system will stay in this eigenstate for all times.
je, 0i and jg, 1i are connected by the strong cavity coupling g.
The transition between the states je, 0i and jl, 0i is driven by This process is known as Stimulated Raman Adiabatic
a laser with Rabi frequency p. Passage (STIRAP) [100,101].
Starting with the pump laser turned off, the atom–
cavity system will be in the state jl, 0i. Increasing the
pump laser intensity slowly to a value p  2g, the
6.1. Atom–field dynamics in a high-Q cavity – the system will move towards the state jg, 1i thereby
Jaynes–Cummings model depositing a single photon into the cavity mode.
Let us now turn to the atom–field dynamics inside This photon can then escape the cavity through one
a cavity. For sufficiently large quality factor, that is, of the mirrors leaving the atom–cavity system in the
when the line width of the cavity response is much state jg, 0i. This state is not coupled to any other state.
smaller than the mode spacing, the dynamics of the A repumping laser is then used to re-excite the atom to
atom–cavity system is well described by the Jaynes– je, 0i so that it can decay spontaneously back to the
Cummings model [98,99]. In this model, a single initial state jl, 0i and the system is ready to produce
electromagnetic-field mode of frequency ! is coupled another photon.
to an atomic two-level system with frequency !A; the Several experiments have been reported using
Hamiltonian in the rotating-wave approximation is rubidium 85Rb [56,96] and 87Rb [102], cesium 133Cs
[57] in optical microcavities, and calcium 40Caþ in an
1 ion-trap cavity system [58].
H^ JC ¼ 
h!a^ y a^ þ  hgða^^ y þ a^ y Þ:
h!A ^ z   ^ ð47Þ
2
The operators  ¼ jgihej,  y ¼ jeihgj, and
^z ¼ jeihej  jgihgj connecting the ground (jgi) and 6.2. Quantum-state extraction from a cavity
excited (jei) states form the usual SU(2)-algebra. As we have described in the previous section,
The eigenstates of the Jaynes–Cummings the quality of a cavity is commonly measured by
Hamiltonian are superpositions of the unperturbed either the quality factor Q, the finesse F or, alterna-
states {je, ni, jg, nþ1i} with a splitting equal to the tively, by the cavity decay rate . The latter determines
n-photon Rabi frequency n ¼ 2g(nþ1)1/2. The set of the photon power loss per unit time from the cavity,
states that is interesting for single-photon sources is i.e. it relates to cavity leakage. However, a photon
{je, 0i, jg, 1i} which are split by the vacuum Rabi being lost from the cavity does not necessarily reappear
frequency 0 ¼ 2g. outside the cavity walls as a real and usable photon.
156 S. Scheel

It is wrong to assume that, only because the reflectivity swept into free space. This process repeats until, for
R of the cavity mirror is close to unity, the photon loss large extraction times, the steady-state solution (49) is
from the cavity is solely due to transmission through reached and the full photon wavefunction (minus the
the mirror. These radiative losses might be dominant portion absorbed by the mirrors) can be found
(and described by the quality factor Qrad), but traveling away from the cavity.
absorptive losses (described by Qabs) can never be It should be stressed again that the effect men-
excluded. We have mentioned this fact earlier in this tioned in this section only plays a role if it is necessary
section, but let us elaborate on this effect. to extract the full quantum state of the photon without
Generally, high-quality cavity mirrors have reflec- performing a measurement. Some heralded schemes
tivities R 1. Photon-number conservation implies that infer the presence of a photon and thus the success
that jRj2 þ jTj2 þ jAj2 ¼ 1, where T and A are the of a desired operation by destructive measurements are
transmission and absorption coefficients, respectively. not susceptible to a finite extraction efficiency.
Obviously, we have jTj, jAj 1. However, nothing can
be said about the ratio jTj/jAj which is clearly of
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

utmost importance for the extraction of a photon from 7. Heralded single-photon sources
a cavity. For high-Q cavities, one typically has jTj
jAj
Common realizations of heralded single-photon
which means that the probability of a photon being
sources use (2) nonlinear materials in the form of
absorbed in the cavity mirror is of the same order of
crystals made from beta barium borate (BBO), lithium
magnitude as the probability of actually recording
niobate, or potassium dihydrogen phosphate (KDP).
a photon outside the cavity. Since, in contrast to Depending on the orientation of the nonlinear crystal
heralded single-photon sources (see Section 4.2), there and the phase-matching relations, signal and idler
is no way of telling whether a photon has left the photons may propagate either collinearly or they may
cavity, we have to ‘trust’ the cavity to deliver a photon. diverge. In addition, the down conversion process is of
Let us associate a cavity decay rate  rad with the type-I or type-II depending on whether the polariza-
transmission of a photon through the cavity mirror, tions in signal and idler beams are oriented parallel or
and a decay rate  abs with the absorption of a photon. perpendicular to one another (see left figure in
Because we are ‘trusting’ the photon to leave the Figure 12).
cavity, but it might not actually do so, the quantum The right figure in Figure 12 depicts schematically
state will be a mixed state of the form the experimental set-up reported in [27] to generate
%^ ¼ pj1ih1j þ ð1  pÞj0ih0j, ð49Þ six-qubit cluster states from three parametric down-
conversion sources using BBO crystals, and postselec-
where the extraction efficiency p is given in the limit of tion by type-II fusion gates. Particular detection
large extraction time by [103,104] patterns in the photon counters D1T . . . D6T signals
rad the successful generation of one of the six-qubit
p¼ : ð50Þ clusters sketched above. In principle, this set-up
rad þ abs
could be enlarged by attaching yet another down-
This means that for the single-photon contribution to conversion step and adding more fusion gates. In the
dominate over the vacuum state, the extraction near future, we will see that progress in generating
efficiency must exceed 50%, that is, radiative decay cluster states with photons will be made at a rapid pace
must dominate absorption,  rad 4  abs. with more qubits being added all the time.
For finite extraction times, one observes transient However, increasing the number of successive
‘tidal wave’ behaviour [105] which can be understood down-conversion steps exponentially decreases the
as follows. The initially excited atom undergoes half success probability. In the experiment in [27], the
a Rabi cycle and deposits a photon inside the cavity. average two-fold coincidence counts of around
Due to the finite transmittivity of the semi-transparent 9.3  104 s1 dropped dramatically to on the order of
mirror the photon wavefunction starts to leak out into one six-fold coincidence event per minute.
free space (a small portion never leaves the resonator
as it is absorbed in the mirror). At the same time, the
coherent evolution inside the cavity continues and the 7.1. Imperfect detection and dark counts
atom reabsorbs the photon from the cavity field. Our arguments regarding the generation of a single
Hence, the already leaked portion of the photon photon by triggered detection of an idler photon was
wavepacket gets ‘sucked’ back into the cavity and based on the assumption of perfect photodetection
a new Rabi cycle begins. With each of the cycles which is normally not justified. In general, one can
a slightly larger portion of the photon wavepacket is only detect photons with an efficiency  5 1 which
Journal of Modern Optics 157
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

Figure 12. Left figure: diverging photons from a type-II nonlinear crystal used to generate entangled photon pairs at the
intersection points of signal and idler cones. Figure taken from http://www.quantum.at. Right figure: schematic experimental set-
up for the generation of six-qubit cluster states by parametric down-conversion and postselecting type-II fusion gates. Figure
taken from [27]. (The color version of this figure is included in the online version of the journal.) Reproduced with permission.
Copyright 2007, Nature Publishing Group.

means that generally the number of registered photons detection, Equation (51) can be formally solved by the
is smaller than the number of incident photons. inverse Bernoulli transformation,
The marginal probability P0m of recording m photons X
1
is given by the joint probability Pmjn()pn summed over pn ¼ Pnjm ð1 ÞPm , ð54Þ
the number n of available photons m¼0

which, however, does not converge for  5 0.5 as the


X
1
P0m ¼ Pmjn ðÞpn , ð51Þ conditional probabilities Pnjm(1) become unbounded
n¼m (see also the discussion in [106,107]). In more
complicated cases, maximum-likelihood inversion
where pn is the probability of finding n photons in the methods can be used [108].
incident light. The conditional probability Pmjn () for
finding m photons conditioned on n incoming photons
is given by the Bernoulli distribution [85] 7.2. Postprocessing with linear optics
 
n m Single-photon sources quite often have only a finite
Pmjn ðÞ ¼  ð1  Þnm , n  m: ð52Þ
m efficiency by which we mean that they produced at
If in addition to a detection efficiency  5 1 the most a mixed state of the form
photodetector produces dark counts, i.e. it registers %^ ¼ pj1ih1j þ ð1  pÞj0ih0j , p 5 1: ð55Þ
photons that do not originate from the source, the
probability distribution (51) has to be convoluted with For example, a photon might have been deposited into
the dark count probability distribution pdark,r as an optical cavity by a STIRAP process, but it might
not be transmitted through the semi-transparent
Xm
Pm ¼ pdark, r P0mr mirror due to a finite absorption probability. In the
r¼0 light of the mentioned applications in quantum key
X
m X1   distribution and metrology, an efficiency less than one
n
¼ pdark, r mr ð1  Þnmþr pn : ð53Þ translates into an increase of the bit error rate in QKD
r¼0 n¼mr m  r
and a drop in accuracy of a potential ‘quantum
In practise, one is interested in retrieving the candela’.
photon-number probabilities pn from the measured It would therefore be expedient to come up with
marginal probabilities Pm. For unit detection effi- a scheme that could improve upon the single-photon
ciency, Pmjn ¼ mn, and zero dark count probability, efficiency by postprocessing, for example using passive
pdark,r ¼ r0, we find that Pm ¼ pm. In case of imperfect linear optical elements and postselection. Such ideas
158 S. Scheel

are similar in spirit to the schemes for entanglement [9] Takesue, H.; Nam, S.W.; Zhang, Q.; Hadfield, R.H.;
purification which are used to concentrate quantum Honjo, T.; Tamaki, K.; Yamamoto, Y. Nature Photonics
correlation from several weakly entangled photon pairs 2007, 1, 343–348.
into one or few highly entangled pairs. However, as it [10] Kurtsiefer, C.; Zarda, P.; Halder, M.; Weinfurter, H.;
Gorman, P.M.; Tapster, P.R.; Rarity, J.G. Nature
turns out, there is no such linear optical scheme that
(London) 2002, 419, 450.
can deliver the desired results [109,110]. This rather [11] Kok, P.; Munro, W.J.; Nemoto, K.; Ralph, T.C.;
disappointing result provides yet more motivation to Dowling, J.P.; Milburn, G.J. Rev. Mod. Phys. 2007,
design even better, i.e. more efficient, single-photon 79, 135–174.
sources. In passing, we mention that a similarly [12] O’Brien, J.L. Science. 2007, 318, 1567–1570.
negative result has been found when trying to enhance [13] Knill, E.; Laflamme, R.; Milburn, G.J. Nature (London)
the detection efficiency of single-photon counters [111] 2001, 409, 46–52.
by mimicking a perfect detector by a network of poor [14] Scheel, S.; Lütkenhaus, N. New J. Phys. 2004, 6, 51.
detectors. [15] Eisert, J. Phys. Rev. Lett. 2005, 95, 040502-1–4.
[16] Scheel, S.; Audenaert, K.M.R. New J. Phys. 2005, 7,
149.
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

[17] Lloyd, S. Phys. Rev. Lett. 1995, 75, 346–349.


8. Outlook [18] Knill, E. Phys. Rev. A 2002, 66, 052306-1–5.
[19] Gilchrist, A.; Hayes, A.J.F.; Ralph, T.C. Phys. Rev. A
Sources of single photons on demand are a vital tool in
2007, 75, 052328-1–5.
the rapidly expanding field of quantum technology.
[20] Yoran, N.; Reznik, B. Phys. Rev. Lett. 2003, 91, 037903-
With the advent of new methods to coherently control 1–4.
single emitters, the quality of the generated photons [21] Nielsen, M.A. Phys. Rev. Lett. 2004, 93, 040503-1–4.
will soon be good enough that attempts at designing [22] Browne, D.E.; Rudolph, T. Phys. Rev. Lett. 2005, 95,
small-scale quantum information processing circuitry 010501-1–4.
and loophole-free tests of Bell’s inequality will be [23] Raussendorf, R.; Briegel, H.J. Phys. Rev. Lett. 2001, 86,
successful. At the same time, we will see a rapid growth 5188–5191.
in the number of applications of single photons and [24] Raussendorf, R.; Browne, D.E.; Briegel, H.J. Phys. Rev.
entangled photons in imaging, tomography and A 2003, 68, 022312-1–32.
[25] Nielsen, M.A. Rep. Math. Phys. 2006, 57, 147–161.
metrology.
[26] Walther, P.; Resch, K.J.; Rudolph, T.; Schenck, E.;
Weinfurter, H.; Vedral, V.; Aspelmeyer, M.; Zeilinger,
Acknowledgements A. Nature (London) 2005, 434, 169–176.
The author thanks S.D. Barrett, S.Y. Buhmann, E.A. Hinds, [27] Lu, C.-Y.; Zhou, X.-Q.; Gühne, O.; Gao, W.-B.; Zhang,
P. Kok, A. Kuhn, and D.-G. Welsch for stimulating J.; Yuan, Z.-S.; Goebel, A.; Yang, T.; Pan, J.-W. Nature
discussions. Physics. 2007, 3, 91–95.
[28] Kieling, K.; Rudolph, T.; Eisert, J. Phys. Rev. Lett.
2007, 99, 130501-1–4.
References [29] Varnava, M.; Browne, D.E.; Rudolph, T. Phys. Rev.
Lett. 2006, 97, 120501-1–4.
[1] Bennett, C.H.; Brassard, G. In Quantum Cryptography: [30] Dawson, C.M.; Haselgrove, H.L.; Nielsen, M.A. Phys.
Public Key Distribution and Coin Tossing, Proceedings of Rev. Lett. 2006, 96, 020501-1–4.
IEEE International Conference of Computer Systems and [31] Dawson, C.M.; Haselgrove, H.L.; Nielsen, M.A. Phys.
Signal Processing, Bangalore, India, 1984; IEEE: Rev. A 2006, 73, 052306-1–26.
New York. [32] Varnava, M.; Browne, D.E.; Rudolph, T. New J. Phys.
[2] Bennett, C.H.; Bessette, F.; Brassard, G.; Salvail, L.; 2007, 9, 203.
Smolin, J. J. Cryptology. 1992, 5, 3–28. [33] Barrett, S.D.; Kok, P. Phys. Rev. A 2005, 71, 060310-
[3] Gisin, N.; Ribordy, G.; Tittel, W.; Zbinden, H. Rev. Mod. 1–4.
Phys. 2002, 74, 145–195. [34] Lim, Y.L.; Barrett, S.D.; Beige, A.; Kok, P.; Kwek, L.C.
[4] Lütkenhaus, N. Phys. Rev. A 2000, 61, 052304. Phys. Rev. A 2006, 73, 012304-1–14.
[5] Brassard, G.; Lütkenhaus, N.; Mor, T.; Sanders, B.C. [35] BIPM, The International System of Units (S.I.), 1998.
Phys. Rev. Lett. 2000, 85, 1330–1333. [36] Cheung, J.Y.; Chunnilall, C.J.; Woolliams, E.R.;
[6] Waks, E.; Inoue, K; Santori, C.; Fattal, D.; Vučković, Fox, N.P.; Mountford, J.R.; Wang, J.; Thomas, P.J.
J.; Solomon, G.S.; Yamamoto, Y. Nature. 2002, 420, J. Mod. Opt. 2007, 54, 373–396.
762. [37] Abouraddy, A.F.; Saleh, B.E.A.; Sergienko, A.V.; Teich,
[7] Beveratos, A.; Brouri, R.; Gacoin, T.; Villing, A.; Poizat, M.C. Phys. Rev. Lett. 2001, 87, 123602-1–4.
J.-P.; Grangier, P. Phys. Rev. Lett. 2002, 89, 187901-1–4. [38] Abouraddy, A.F.; Nasr, M.B.; Saleh, B.E.A.;
[8] Inoue, K.; Waks, E.; Yamamoto, Y. Phys. Rev. Lett. Sergienko, A.V.; Teich, M.C. Phys. Rev. A 2002, 65,
2002, 89, 037902-1–3. 053817-1–6.
Journal of Modern Optics 159

[39] Belinsky, A.V.; Klyshko, D.N. Sov. Phys. JETP. 1994, K.V.; Gol’tsman, G.N. Appl. Phys. Lett. 2007, 91,
78, 259–262. 031106.
[40] Strekalov, D.V.; Sergienko, A.V.; Klyshko, D.N.; [68] Lounis, B.; Orrit, M. Rep. Prog. Phys. 2005, 68,
Shih, Y.H. Phys. Rev. Lett. 1995, 74, 3600–3603. 1129–1179.
[41] Gatti, A.; Brambilla, E.; Lugiato, L.A. Phys. Rev. Lett. [69] Purcell, E.M. Phys. Rev. 1946, 69, 681.
2003, 90, 133603-1–4. [70] Dexter, D.L. Phys. Rev. 1956, 101, 48–55.
[42] Bell, J.S. Physics (Long Island City, N.Y.) 1965, 1, [71] Kleppner, D. Phys. Rev. Lett. 1981, 47, 233–236.
195–200. [72] Yablonovitch, E. Phys. Rev. Lett. 1987, 58,
[43] Franson, J.D. Phys. Rev. Lett. 1989, 62, 2205–2208. 2059–2062.
_
[44] Aerts, S.; Kwiat, P.; Larsson, J.-Å.; Zukowski, M. Phys. [73] Barnett, S.M.; Loudon, R. Phys. Rev. Lett. 1996, 77,
Rev. Lett. 1999, 83, 2872–2875. 2444–2446.
[45] Aspect, A.; Dalibard, J.; Roger, G. Phys. Rev. Lett. [74] Altarelli, M.; Dexter, D.L.; Nussenzveig, H.M.; Smith,
1982, 49, 1804–1807. D.Y. Phys. Rev. B 1972, 6, 4502–4509.
[46] Genovese, M. Phys. Rep. 2005, 413, 319–396. [75] Scheel, S.; Knöll, L.; Welsch, D.-G.; Barnett, S.M. Phys.
[47] Pearle, P. Phys. Rev. D 1970, 2, 1418–1425. Rev. A 1999, 60, 1590–1597.
[48] Eberhard, P.H. Phys. Rev. A 1993, 47, R747–R750. [76] Scheel, S.; Knöll, L.; Welsch, D.-G. Phys. Rev. A 1999,
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

[49] Cabello, A.; Larsson, J.-Å. Phys. Rev. Lett. 2007, 98, 60, 4094–4104.
220402-1–4. [77] Ho, T.D.; Knöll, L.; Welsch, D.-G. Phys. Rev. A 2000,
[50] Brunner, N.; Gisin, N.; Scarani, V.; Simon, C. Phys. 62, 053804-1–13.
Rev. Lett. 2007, 98, 220403-1–4. [78] Chew, W.C. Waves and Fields in Inhomogeneous Media;
[51] Vogel, W.; Welsch, D.-G. Quantum Optics; Wiley-VCH: IEEE Press: New York, 1995.
Weinheim, 2006. [79] Ho, T.D.; Buhmann, S.Y.; Welsch, D.-G. Phys. Rev. A
[52] Hong, C.K.; Ou, Z.Y.; Mandel, L. Phys. Rev. Lett. 1987, 2006, 74, 023803-1–11.
59, 2044–2046. [80] Scheel, S. Phys. Rev. A 2008, 78, 013841.
[53] Campos, R.A.; Saleh, B.E.A.; Teich, M.C. Phys. Rev. A [81] Burnham, D.C.; Weinberg, D.L. Phys. Rev. Lett. 1970,
1989, 40, 1371–1384. 25, 84–87.
[54] Oxborrow, M.; Sinclair, A.G. Contemp. Phys. 2005, 46, [82] Joobeur, A.; Saleh, B.E.A.; Teich, M.C. Phys. Rev. A
173–206. 1994, 50, 3349–3361.
[55] Hanbury Brown, R.; Twiss, R.Q. Nature (London) [83] Grice, W.P.; U’Ren, A.B.; Walmsley, I.A. Phys. Rev. A
1956, 178, 1046–1048. 2001, 64, 063815-1–7.
[56] Hijlkema, M.; Weber, B.; Specht, H.P.; Webster, S.C.; [84] U’Ren, A.B.; Silberhorn, C.; Erdmann, R.; Banaszek,
Kuhn, A.; Rempe, G. Nature Physics. 2007, 3, 253–255. K.; Grice, W.P.; Walmsley, I.A.; Raymer, M.G. Laser
[57] McKeever, J.; Boca, A.; Boozer, A.D.; Miller, R.; Buck, Phys. 2005, 15, 146–161.
J.R.; Kuzmich, A.; Kimble, H.J. Science. 2004, 303, [85] Hong, C.K.; Mandel, L. Phys. Rev. Lett. 1986, 56,
1992–1994. 58–60.
[58] Keller, M.; Lange, B.; Hayasaka, K.; Lange, W.; [86] Migdall, A.L.; Branning, D.; Castelletto, S. Phys. Rev. A
Walther, H. Nature (London) 2004, 431, 1075–1078. 2002, 66, 053805-1–4.
[59] Lee, T.H.; Kumar, P.; Mehta, A.; Xu, K.; Dickson, [87] Bendickson, J.M.; Dowling, J.P.; Scalora, M. Phys. Rev.
R.M.; Barnes, M.D. Appl. Phys. Lett. 2004, 85, 100–102. A 1996, 53, 4107–4121.
[60] Lukishova, S.G.; Schmid, A.W.; McNamara, A.J.; [88] Moreau, E.; Robert, I.; Gérard, J.M.; Abram, I.; Manin,
Boyd, R.W.; Stroud Jr, C.R. IEEE J. Sel. Top. L.; Thierry-Mieg, V. Appl. Phys. Lett. 2001, 79,
Quantum Electron. 2003, 9, 1512–1518. 2865–2867.
[61] Ward, M.B.; Farrow, T.; See, P.; Yuan, Z.L.; Karimov, [89] Santori, C.; Fattal, D.; Vučković, J.; Solomon, G.S.;
O.Z.; Bennett, A.J.; Shields, A.J.; Atkinson, P.; Cooper, Yamamoto, Y. Nature (London) 2002, 419, 594–597.
K.; Ritchie, D.A. Appl. Phys. Lett. 2007, 90, 063512. [90] Ho, Y.-L.D.; Cao, T.; Ivanov, P.S.; Cryan, M.J.;
[62] Patel, R.B.; Bennett, A.J.; Cooper, K.; Atkinson, P.; Craddock, I.J.; Railton, C.R.; Rarity, J.G. IEEE J.
Nicoll, C.A.; Ritchie, D.A.; Shields, A.J. Phys. Rev. Quantum Electron. 2007, 43, 462–472.
Lett. 2008, 100, 207405-1–4. [91] Florescu, M.; Scheel, S.; Häffner, H.; Lee, H.; Strekalov,
[63] Kako, S.; Santori, C.; Hoshino, K.; Götzinger, S.; D.V.; Knight, P.L.; Dowling, J.P. Europhys. Lett. 2005,
Yamamoto, Y.; Arakawa, Y. Nature Materials. 2006, 5, 69, 945–951.
887–892. [92] Florescu, M.; Scheel, S.; Lee, H.; Knight, P.L.; Dowling,
[64] Vučković, J.; Fattal, D.; Santori, C.; Solomon, G.S.; J.P. Physica E 2006, 32, 484–487.
Yamamoto, Y. Appl. Phys. Lett. 2003, 82, 3596–3598. [93] Vahala, K.J. Nature (London) 2003, 424, 839–846.
[65] Bennett, A.J.; Unitt, D.C.; Atkinson, P.; Ritchie, [94] Ho, T.D.; Knöll, L.; Welsch, D.-G. Phys. Rev. A 2001,
D.A.; Shields, A.J. Opt. Express. 2005, 13, 50–55. 64, 013804-1–15.
[66] Becher, C.; Kiraz, A.; Michler, P.; Imamoğlu, A.; [95] Kuhr, S.; Gleyzes, S.; Guerlin, C.; Bernu, J.; Hoff, U.B.;
Schoenfeld, W.V.; Petroff, P.M.; Zhang, L.; Hu, E. Deléglise, S.; Osnaghi, S.; Brune, M.; Raimond, J.-M.
Phys. Rev. B 2001, 63, 121312-1–4. Appl. Phys. Lett. 2007, 90, 164101.
[67] Zinoni, C.; Alloing, B.; Li, L.H.; Marsili, F.; Fiore, A.; [96] Kuhn, A.; Hennrich, M.; Rempe, G. Phys. Rev. Lett.
Lunghi, L.; Gerardino, A.; Vakhtomin, Yu.B.; Smirnov, 2002, 89, 067901-1–4.
160 S. Scheel

[97] Trupke, M.; Hinds, E.A.; Eriksson, S.; Curtis, E.A.; [104] Khanbekyan, M.; Knöll, L.; Semenov, A.A.; Vogel,
Moktadir, Z.; Kukharenka, E.; Kraft, M. Appl. Phys. W.; Welsch, D.-G. Phys. Rev. A 2004, 69, 043807-1–9.
Lett. 2005, 87, 211106. [105] Khanbekyan, M.; Welsch, D.-G.; di Fidio, C.; Vogel,
[98] Jaynes, E.T.; Cummings, F.W. Proc. IEEE. 1963, 51, W. 2007, quant-ph/0709.2998.
89–109. [106] Kiss, T.; Leonhardt, U. Phys. Rev. A 1995, 52,
[99] Shore, B.W.; Knight, P.L. J. Mod. Opt. 1993, 40, 2433–2435.
1195–1238. [107] D’Ariano, G.M.; Leonhardt, U.; Paul, H. Phys. Rev. A
[100] Bergmann, K.; Theuer, H.; Shore, B.W. Rev. Mod. 1995, 52, R1801–R1804.
Phys. 1998, 70, 1003–1025. [108] Zambra, G.; Paris, M.G.A. Phys. Rev. A 2006, 74,
[101] Vitanov, N.V.; Fleischhauer, M.; Shore, B.W.; 063830-1–7.
Bergmann, K. Adv. At. Mol. Opt. Phys. 2001, 46, [109] Berry, D.W.; Scheel, S.; Sanders, B.C.; Knight, P.L.
55–190. Phys. Rev. A 2004, 69, 031806-1–4.
[102] Wilk, T.; Webster, S.C.; Specht, H.P.; Rempe, G.; [110] Berry, D.W.; Scheel, S.; Myers, C.R.; Sanders, B.C.;
Kuhn, A. Phys. Rev. Lett. 2007, 98, 063601-1–4. Knight, P.L.; Laflamme, R. New J. Phys. 2004, 6, 93.
[103] Cui, G.; Raymer, M.G. Opt. Express. 2005, 13, [111] Kok, P. IEEE Sel. Top. Quantum Electron. 2003, 9,
9660–9665. 1498–1501.
Downloaded by [University of Saskatchewan Library] at 12:43 31 August 2012

You might also like