Prog Opt Vol 60

VOLUME SIXTY
PROGRESS IN OPTICS
EDITORIAL ADVISORY BOARD
G.S. Agarwal Stillwater, USA

T. Asakura Sapporo, Japan
M.V. Berry Bristol, England
C. Brosseau Brest, France
A.T. Friberg Joensuu, Finland
F. Gori Rome, Italy
D.F.V. James Toronto, Canada
P. Knight London, England
G. Leuchs Erlangen, Germany
J.B. Pendry London, England
J. Perina Olomouc, Czech Republic
J. Pu Quanzhou, PR China
W. Schleich Ulm, Germany
T.D. Visser Amsterdam, The Netherlands
VOLUME SIXTY
PROGRESS IN OPTICS
Edited by
E. WOLF
University of Rochester, NY, USA
Contributors
Ayman Alfalou, Mario Bertolotti, Fabio Bovino,
Christian Brosseau, Natalie A. Cartwright,
Rafal Demkowicz-Dobrzanski, Marcin Jarzyna,
Brandon A. Kemp, Jan Koodynski, Yuri A. Obod,
Kurt E. Oughstun, Alexander B. Shvartsburg,
Concita Sibilia, Oleg D. Volpian
Amsterdam Boston Heidelberg London New York Oxford

Paris San Diego San Francisco Singapore Sydney Tokyo
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, UK
225 Wyman Street, Waltham, MA 02451, USA
First edition 2015
Copyright 2015, Elsevier B.V. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publishers permissions policies and our
arrangements with organizations such as the Copyright Clearance Center and the Copyright
Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by
the Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and
experience broaden our understanding, changes in research methods, professional practices,
or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described
herein. In using such information or methods they should be mindful of their own safety and
the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of
products liability, negligence or otherwise, or from any use or operation of any methods,
products, instructions, or ideas contained in the material herein.
ISBN: 978-0-12-802284-9
ISSN: 0079-6638
For information on all Elsevier publications

visit our website at store.elsevier.com
CONTENTS
Contributors ix
Preface xi
Acknowledgement xiii
1. Quantum State Engineering: Generation of Single

and Pairs of Photons 1
Mario Bertolotti, Fabio Bovino, and Concita Sibilia
1. Introduction 1
2. Fock States 3
3. The Problem of Localizing Photons 6
4. Antibunching of Single-Photon States 7
5. Photon Statistics and Spectral Purity 9
6. The Purcell Effect and the Control of Emission of Electromagnetic Radiation 12
7. Preparation of Single-Photon States: Quantum Engineering 13
8. Realization of Single-Photon Sources 20
9. Entangled States 70
10. Plasmonic Sources 82
11. Application to Quantum Information Processing 86
12. Conclusions 95
Addendum 95
References 96
2. Recent Advances in Optical Image Processing 119

Ayman Alfalou and Christian Brosseau
1. General Introduction 119
2. Part A: Digital Versus Optical Processing 121
3. Part B: OIP 169
Acknowledgments 246
References 247
3. Precursors and Dispersive Pulse Dynamics: A Century

After the SommerfeldBrillouin Theory: Part II.
The Modern Asymptotic Theory 263
Natalie A. Cartwright and Kurt E. Oughstun
1. Loudon's Energy Velocity 264
2. Analysis of the Phase Function and Its Saddle Points 267
v
vi Contents
3. Uniform Asymptotic Expansions 288

4. Extensions of the SommerfeldBrillouin Theory to the Singular
and Weak Dispersive Limits 319
5. Pulsed Plane-Wave Transmission Through an Interface 327
6. Concluding Remarks 341
Acknowledgments 343
References 343
4. Quantum Limits in Optical Interferometry 345

Rafal Demkowicz-Dobrzanski, Marcin Jarzyna, and Jan Koodynski
1. Introduction 346
2. Quantum States of Light 351
3. MZ Interferometry 362
4. Estimation Theory 373
5. Quantum Limits in Decoherence-Free Interferometry 387
6. Quantum Limits in Realistic Interferometry 401
7. Conclusions 426
Acknowledgments 427
References 427
5. Macroscopic Theory of Optical Momentum 437

Brandon A. Kemp
1. Introduction 437
2. Classical Electrodynamics 440
3. Optical Force, Momentum, and Stress 452
4. Momentum and Stress in Dispersive Media 467
5. Discussion 472
Appendix A. Scattering by a Cylinder 478
Appendix B. Scattering by a Sphere 481
Acknowledgments 485
References 485
6. Tunneling of Electromagnetic Waves in All-Dielectric

Gradient Metamaterials 489
Alexander B. Shvartsburg, Yuri A. Obod, and Oleg D. Volpian
1. Introduction: Nonlocal Dispersion of Waves in Transparent
Gradient Dielectrics 490
2. Frustrated TIR from Gradient Dielectric Layers 495
3. Interplay of Artificial and Natural Dispersion in Tunneling Phenomena 511
Contents vii
4. Tunneling of Microwaves Through Gradient TLs 529

5. Conclusions: The Mystery of Photon TunnelingParadoxes
and Perspectives 551
Acknowledgments 558
References 558
Author Index for Volume 60 565

Subject Index for Volume 60 583
Contents of Previous Volumes 60 601
Cumulative Index-Volumes 160 615
This page intentionally left blank
CONTRIBUTORS
Ayman Alfalou
Vision ISEN-BREST Laboratory L@BISEN, Institut Superieur de lElectronique et du
Numerique, Brest, France
Mario Bertolotti
Dipartimento SBAI, Universita` di Roma La Sapienza, Rome, Italy
Fabio Bovino
Quantum Technologies Laboratory, SELEX-ES, Genoa, Italy
Christian Brosseau
Laboratoire des Sciences et Techniques de lInformation, de la Communication et de la
Connaissance, Universite de Brest, Brest, France
Natalie A. Cartwright
Department of Mathematics, State University of New York, New Paltz, New York, USA
Rafal Demkowicz-Dobrza nski
Faculty of Physics, University of Warsaw, ul. Pasteura 5, Warszawa, Poland
Marcin Jarzyna
Brandon A. Kemp
College of Engineering, Arkansas State University, Jonesboro, Arkansas, USA
Jan Koody nski
Yuri A. Obod
R&D Company Fotron Auto, Moscow, Russia
Kurt E. Oughstun
College of Engineering and Mathematics, University of Vermont, Burlington, Vermont,
USA
Alexander B. Shvartsburg
Joint Institute for High Temperatures, Russian Academy of Sciences; Institute for Space
Researches Russian Academy of Sciences, Moscow, and Far Eastern Federal University,
Vladivostok, Russia
Concita Sibilia
Dipartimento SBAI, Universita` di Roma La Sapienza, Rome, Italy
Oleg D. Volpian
ix
PREFACE
In this 60th volume of Progress in Optics, six review articles are presented.
Chapter 1, contributed by Bertolotti, Bovino, and Sibilia, takes a histor-
ical approach to single-photon sources. They discuss photon statistics and
entangled states. Recently developed plasmonic sources and possible appli-
cations in quantum information processing are also described.
In Chapter 2, Alfalou and Brosseau discuss a variety of image processing
techniques. They compare the more traditional digital approach with newly
emerging, all optical setups. Because of their parallel nature, these can pro-
duce real-time results.
Chapter 3, by Cartwright and Oughstun, is a continuation of their article
in Volume 59. In this second part, they discuss recent developments in the
theory of pulse propagation through dispersive media. One of these is the
extension of the classic SommerfeldBrillouin theory to both the weak
and singular dispersion limits.
In Chapter 4, written by Demkowicz-Dobrza nski, Jarzyna, and
Koody nski, the authors describe how nonclassical light states can be used
to enhance the performance of optical interferometers. This is especially
important, for example, for the ongoing search for gravitational waves.
Chapter 5 is a contribution by Kemp. Starting with the Minkowski
Abraham controversy, different expressions for the electromagnetic force
density are analyzed. This ongoing discussion has direct consequences for
the growing number of optical trapping applications.
The final chapter, Chapter 6, is written by Shvartsburg, Obod, and
Volpian. They discuss how the classical effect of tunneling through optical
barriers takes on new and unexpected forms when the barrier consists
of a dielectric gradient metamaterial. Because of the scalability of these
effects, these also apply to microwave tunneling phenomena in trans-
mission lines.
It has been my honor to serve as editor of Progress in Optics for over
50 years. But, as I used to tell my children when they were young, All good
things must come to an end. I would like to thank the members of the
editorial advisory board for their help, advice, and friendship over these
many years. Beginning with Volume 61, Dr. Taco Visser will take over
xi
xii Preface
as editor of this series. I have complete confidence that he will keep readers
of Progress in Optics well informed of the most important advances being
made in the field. I wish him well.
EMIL WOLF
Rochester, NY
April 2015
ACKNOWLEDGEMENT
The editorial staff would like to extend a special thanks for the support and
expert assistance provided by Dr. Taco Visser during the compilation of
this work.
xiii
CHAPTER ONE
Quantum State Engineering:

Generation of Single and Pairs
of Photons
Mario Bertolotti*, Fabio Bovino, Concita Sibilia*
*Dipartimento SBAI, Universita` di Roma La Sapienza, Rome, Italy
Quantum Technologies Laboratory, SELEX-ES, Genoa, Italy
Contents
1. Introduction 1
2. Fock States 3
3. The Problem of Localizing Photons 6
4. Antibunching of Single-Photon States 7
5. Photon Statistics and Spectral Purity 9
6. The Purcell Effect and the Control of Emission of Electromagnetic Radiation 12
7. Preparation of Single-Photon States: Quantum Engineering 13
8. Realization of Single-Photon Sources 20
8.1 Strongly Attenuated Sources 21
8.2 Single-Photon Sources Emitting one Photon in a Random Way 23
8.3 Single-Photon Sources on Demand 69
9. Entangled States 70
10. Plasmonic Sources 82
11. Application to Quantum Information Processing 86
12. Conclusions 95
Addendum 95
References 96
1. INTRODUCTION
The introduction of quantization of energy to discuss the interaction
of electromagnetic radiation with matter was done in 1900 by Planck (1900a,
1900b). Einstein (1905) surmised that also free radiation had a granular struc-
ture. The name photon was later proposed by Lewis (1926). The quantization
of the electromagnetic field was made by Dirac (1926, 1927). A review of
some historical papers on the subject has been made by Keller (2007).
Progress in Optics, Volume 60 # 2015 Elsevier B.V. 1

ISSN 0079-6638 All rights reserved.
http://dx.doi.org/10.1016/bs.po.2015.02.001
2 Mario Bertolotti et al.
What a photon is, exactly nobody can say. It is the quantum represen-
tation of a mode of the electromagnetic field and is an exclusively quantum
concept. With this definition, photons have associated plane waves of def-
inite wave vector k and definite polarization s. A monochromatic wave
implies delocalization in time and space; in practice, a single photon local-
ized to some degree in time and space can be described as superposition of
monochromatic photon modes.
When Glauber (1963a, 1963b) completed the model of radiation detec-
tion, discussing from a quantum point of view the interaction of radiation
and matter, and constructed a quantum theory of coherence, a number of
interesting states of radiation received a full reconnaissance as useful and pos-
sible states: among them coherent and single-photon states are perhaps the most
interesting, together with squeezed and entangled ones. Full description of
these states may be found in many excellent textbooks, like, e.g., Mandel
and Wolf (1995).
The generation of quantum states of the radiation field started to
receive great attention from the 1980s. Single-photon states, in particular,
are studied because of possible applications in quantum communication,
quantum lithography, quantum metrology, information processing, and
quantum computing, such as quantum random-number generation,
quantum networks, secure quantum communications, and quantum cryp-
tography (see, for example, Beveratos, Brouri, et al., 2002; Cerf &
Flurasek, 2006; Dusek, Lutkenhas, & Hendrych, 2006; Gisin, Ribordy,
Tittel, & Zbinden, 2002; Gisin & Thew, 2007; Grangier & Abram, 2003;
Kilin, 2001).
For example, the security in some schemes of quantum cryptography
is based on the fact that each bit of information is coded on a single photon.
The fundamental impossibility of duplicating the complete quantum state
of a single particle (no cloning theorem; Cerf & Flurasek, 2006; Diecks,
1982; Ghirardi & Weber, 1983; Wootters & Zurek, 1982) prevents any
potential eavesdropper from intercepting the message without the receivers
noticing.
An ideal single-photon source would produce exactly one photon in a
definite quantum state, in contrast with a classical source, such as attenu-
ated laser pulses, for which the photon number follows a Poisson distribu-
tion. A more stringent request would be to have the single-photon
generation on demand, that is, at a determined time. Additional requests
could be room temperature operation, high repetition frequency, high
efficient extraction into free space or fiber, good coherence, and
Quantum State Engineering 3
Fourier-transformed linewidth. Much progress has been made in the years

toward such devices, especially in suppressing the probability of two photons
in the same pulse.
Historically, the first experiment with single photons was made using an
atomic cascading process in which an excited atomic level decayed with the
emission of two photons of different frequencies (Clauser, 1974). The detec-
tion of one of them established the presence of the other; we will describe
this experiment later.
In the following, we will give a brief introduction to Fock states, remem-
ber the problem of localization of single photons, focus on their anti-
bunching property and photon statistics, and remember the Purcell effect,
which allows a control of emission probability. We then discuss the prepa-
ration of single-photon states, the so-called quantum engineering, the different
kinds of single-photon sources, entangled states, plasmonic sources, and
applications to quantum information processing. The problem of detection
is deliberately not treated.
A number of review papers already exist on the subject such as Moerner
(2004), Lounis and Orrit (2005), Oxborrow and Sinclair (2005), Scheel
(2009), and Eisaman, Fan, Mugdall, & Polyakov (2011). Quite recently, a
Single-Photon Workshop has been held at Oak Ridge National Laboratory,
October 1518, 2013. The presented papers are available to attendees only.
In the following, we will follow an approximate historical presentation
enlightening the single contributions and the evolution of the methods to
obtain single-photon sources. The survey may not be complete; we apolo-
gize for any omission.
2. FOCK STATES
States with a prescribed number of photons are called number states or
Fock states.
They were first introduced and discussed by Fock (1932) (see also
Faddeev, Khalfin, & Komarov, 2004).
A Fock state is strictly quantum mechanical and contains a precisely def-
inite number of quanta of field excitation; hence, its phase is completely
undefined.
As well known, the Hamiltonian for the free electromagnetic field can be
written as

H ~a{ , ~a ~a{ ~a + 1=2 (1)
where is h/2 with h Plancks constant, is the frequency (pulsation) of

the field, and a and a are the annihilation and creation operators, respectively.
They obey the commutation relations
{
~a, ~a 1; ~a, ~a ~a{ , ~a{ 0 (2)
The product aa is called the photon number operator n which accounts for
the photons in the chosen spatialtemporal mode
n~ ~a{ ~a (3)
The eigenstate of the photon number operator n
n~jni njni (4)
has a perfectly fixed photon number n. Since n is a Hermitian operator, the
number n is real.
If jni is an eigenstate of n, then ajni must be an eigenstate as well, with the
eigenvalue n 1. In fact,

n~~ajni ~a{ ~a2 jni ~a~a{ ~a ~a jni n 1~ajni (5)
In a similar way, ajni is an eigenstate of n with eigenvalue n + 1. So there
are the fundamental relations
~ajni njn 1i (6)

~a{ jni n + 1jn + 1i (7)
The prefactors have been obtained using the fact that
{
nj~a ~ajn
must be equal to the eigenvalue n.
The state ~aj0i 0 exists and it is the vacuum state.
States with a prescribed number of photons can be created by applying
the creation operator to the vacuum state
jni ~a{n =n!j0i (8)

The Fock states must be complete
X
1
jnihnj 1 (9)
n0
and orthonormal
hnjn0 i nn (10)
The wave function (q) of the state j0i is

o q 1=4 exp q2 =2 (11)
In momentum representation, it is

o p 1=4 exp p2 =2 (12)
The expectation value of the electric field and its square are
hnjEjni 0 (13)
and
2
njE jn =o V n + 1=2 (14)
where o is the vacuum dielectric constant and V is the volume.

The electromagnetic wave amplitude can be represented by the quantity
Eo 2=o V 1=2 n + 1=21=2 (15)
The phase of the wave is of course completely uncertain.

The use of number states could lead to great signal to noise improvement
(Yuen, 1986).
Coherent states are the eigenstates of the annihilation operator a
~aji ji (16)
ji is an eigenstate of the annihilation operator a with eigenvalue , which

is in general a complex number since a is not a Hermitian operator.
A coherent state is the closest analogue to a classical light field and exhibits
a Poisson photon number distribution with an average photon number
j2j. Coherent states have relatively well-defined amplitude and phase,
with minimal fluctuations permitted by the Heisenberg uncertainty
principle.
Coherent states are an overcomplete set.
Because the number states are a complete set, it is possible to express
coherent states as a superposition of n states
X
ji exp1=2jj2 n =n!1=2 jni (17)
Vice versa n states may be expressed as a superposition of coherent states

Z

jni 1= exp jj2 =2 *n =n!1=2 j > d2 (18)
where d2 dRedIm:
The other states of interest here are squeezed and entangled states. The real
and imaginary part of the complex amplitude of the electromagnetic field
fluctuate with equal dispersion in a coherent state. The phenomenon of vac-
uum fluctuations is a manifestation of this effect because the vacuum state is
an example of a particular coherent state. In a squeezed state, one part of the
field fluctuates less and the other part fluctuates more than in the vacuum
state. A number state can be considered a special case of a squeezed state.
Entangled states will be discussed in Section 9.
3. THE PROBLEM OF LOCALIZING PHOTONS

The number operator n refers to the total photon number in all space.
It is therefore not expected to be accessible to direct measurement. From a
practical point of view, one could interpret the electronic signal registered
by a photodetector as due to a photon that has been localized in the detector
volume. More precisely, the counts registered by a detector whose surface is
normal to the incident field and exposed for some finite time t could be
interpreted as a measurement of the number of photons in a cylindrical vol-
ume whose base is the sensitive surface of the detector and whose height is
ct. The integral of the intensity I(r,t) over such volume can therefore be
interpreted as a configuration space number operator, at least in an approx-
imate sense, provided the linear dimensions of the volume are all large com-
pared with the wavelengths of contributing modes (see, for example,
Mandel & Wolf, 1995 for a detailed discussion).
This is of course a very simplified point of view. A photon wave function
concept in the coordinate representation was introduced originally by
Landau and Peierls (1930), connected with the Maxwell electromagnetic
field solutions. This wave function is a highly nonlocal object. Its modulus
squared has the right dimensionality to be interpreted as a probability density
to find a photon. However, as noted already by Pauli (1933), it has serious
drawbacks. A more refined theory was presented by Oppenheimer (1931),
but this theory too is not completely satisfying. A fundamental point is that a
position operator cannot exist in a relativistic quantum theory. Even for
massive particles, the localization is not perfect because it is not
relativistically invariant. Two observers in relative motion would not quite

agree as to the localization region of a relativistic particle.
We need not to enter in this matter and refer, among others, to review
papers by Acharya and Sudarshan (1960), Bialynicki-Birula (1996), Scully
and Zubairy (1997), and Keller (2007). We will assume the pragmatic posi-
tion that a single-photon state is there if a detector, that can determine the
number of incident photons with 100% accuracy, clicks.
4. ANTIBUNCHING OF SINGLE-PHOTON STATES

One-photon state presents a peculiar anticorrelation effect, which
does not exist for a classical wave. If we send a one-photon state on a beam
splitter and place photon-counting detectors on the reflected and transmit-
ted beamsa disposition first used by Hanbury Brown and Twiss (1956a,
1956b, 1958) to study intensity correlations in astronomywe never
observe any coincidence between counts measured by the two detectors,
as this would violate energy conservation. The photon cannot split to be
present at the two output ports contemporarily. As a consequence of the
principles of quantum mechanics, the wave function of the photon has to
collapse onto either one or the other of the two detectors. This absence
of coincidences of detection events on the two detectors has been dubbed
photon antibunching and is utterly irreconcilable with a classical description
of light. This effect was predicted by Carmichael and Walls (1976) and
Kimble and Mandel (1976). Antibunching can be assessed measuring the
joint probability p(tjt + ) of detecting two photoelectric pulses at time t
within t and at time t + within . The normalized quantity
ptjt + =pt hI tI t + i=hI ti2 g2 (19)

is called the second-order correlation function g2(). Antibunching means
g(2)(0) 0.
Naively the antibunching can be understood very simply. In the emission
process from a single atom or molecule, the correlation function measures
the joint probability for the arrival of a photon at time t 0 and the arrival
of a photon at t > 0. After the emission of a photon at t 0, the quantum
system is projected in its ground state since it just emitted a photon. Hence,
the simultaneous emission of a second photon is impossible because the mol-
ecule cannot emit from the ground state. It needs some time to have a finite
probability to be again in the excited state and emit a second photon. On the
average a time a half a Rabi period1 has to elapse to have a finite probability
for the molecule to be in the excited state and emit a second photon.
Antibunching was first observed in pioneering experiments (Kimble,
Dagenais, & Mandel, 1977; Walls, 1979).
In the nanosecond time regime within a single bunch of photons, the
emitted photons from a single-quantum system are expected to show anti-
bunching; that is, the probability for two photons to arrive at the same time
is zero.
To observe antibunching correlations, the second-order correlation
function g(2)() is generally measured by determining the distribution of time
delays N() between the arrival of successive photons in a dual beam detector
g2 hntnt + i=hni2 pt=t + =pt (20)
In Equation (20), n(t) is the number of photons counted at time t.

For a number state with n photons, it is
g2 0 1 1=n (21)
The Hanbury Brown and Twiss experimental arrangement is shown in

Figure 1.
Half-silvered mirror
Liquid filter Interference filter
Photomultiplier tube
Lens
Mercury arc Rectangular aperture Slide
Photomultiplier tube
Correlator
Amplifier Amplifier
Integrating motor
Figure 1 The Hanbury Brown and Twiss experimental arrangement.
1
The Rabi frequency j12 Ej/, where 12 is the transition dipole moment and E is the laser field
with which the atom is excited.
120
2.0
1.5
80
g(2)(t)
c(t)
1.0
40
0.5
0.0 0
20 0 20 40
t (ns)
Figure 2 An example of antibunching. From Beveratos, Brouri, Gacoin, Poizat, and
Grangier (2001).
An example of the result obtained when antibunching is present is shown

in Figure 2.
The correlation function never reaches exactly zero because of a number
of experimental drawbacks which are characteristics of the particular
experiment.
In any case, the time response of the detectors will determine a minimum
value of the measured g(2)(0) even with an ideal g(2)(0) 0 source.
5. PHOTON STATISTICS AND SPECTRAL PURITY

The probability distribution pm(T) of m photons in the quantum field
during an observation time T (photon number distribution) can be con-
nected to the probability of counting n photons at a detector (Mandel &
Wolf, 1995).
The photon-count distribution p(n, T, t) of the detection of n counts in
the time interval (t, t + T) by a photodetector may be written as
Z
pn, T, t W n =n! expW PN W dW (22)
where PN(W) is the probability distribution of the integrated intensity W
Z
W I t0 dt0 (23)
is the photodetection efficiency and I(t) is the intensity of light incident

on a photocathode. If 1, Equation (22) is also the probability of detecting
n photons in a time interval t, t + T. Quantum mechanics arrives at an

equation formally similar to Equation (22) (Perina, 1984; Perina,
Hradil, & Jur, 1994)
Trf : W n =n! expW : PN W dW g

pn, T , t Z
(24)
W n =n! expW PN W dW
where : : represent the normal order of the field operator, PN(W) is the
quasi-probability distribution of the normal integral intensity W, that may
have also negative values, and is the density matrix.
We briefly recall the possibility of describing the most general statistical
quantum system by means of the so-called density matrix . This is a
Hermitian operator, time independent in the Heisenberg picture, such that
the quantum and statistical expectation value of any operator A is expressed as
fhAigav TrA (25)
where the trace Tr of an operator O is defined as

X
TrO h jOj i (26)
the j is forming a set of states verifying the completeness relation

j ih j 1 (27)
It is immediately apparent that for a system in a pure state j i
corresponding to no statistical indetermination, the density matrix is given
by the operator j ih j.
The photon-counting distribution relative to chaotic fields if the cou-
nting time T is much less than the inverse frequency bandwidth of the radi-
ation is (BoseEinstein distribution)
pm T hmim =1 + hmim + 1 (28)

For light from a laser, the distribution is a Poisson distribution
pm T hmim exphmi=m! (29)
The moments of the distribution are
hni hW iN
(30)
hn2 i hW iN + 2 hW 2 iN
Quantum mechanically
hn~i ~a{ ~a
2 2 (31)
n~ < ~a{ ~a > ~a{ ~a + ~a{2 ~a2
The variance of the number of absorbed photons ( 1) is expressed by

~
n2 hn~i + W 2 N (32)
For a coherent field h(W)2iN 0 and h(n)2i n2, i.e., such a field is
Poissonian. Classical fields have h(W)2iN > 0 and they are super-Poissonian,
for example, for the chaotic field of a natural source h(W)2i hIi2T2 and

n2 hni1 + hni (33)
corresponding to the BoseEinstein distribution (28).
For quantum fields having no classical analogues, it may be h(W)2iN < 0
and h(n)2i < hni, and such light is sub-Poissonian.
The photon number distribution can be calculated using the density
matrix
pm hmjjmi (34)
From an experimental point of view p(m) is related to p(n,T,t) by the
Bernoulli transformation
X
pn, T, t m!=n!m n!n 1 mn pm (35)

Introducing the so-called Fano factor Fn n2 =hni, the fluctuation
formula (32) can be written in terms of the photocount number n and the
photon number m in the form of the quantum Burgess variance theorem
(Perina, 1984)
Fn 1 Fm 1 (36)
For the Fock state Fm 0.
Although both effects are nonclassical, are exhibited only by a quantum
field, and are often associated, sub-Poissonian statistics and antibunching are
distinct effects that need not necessarily occur together; one may occur
without the other (Singh, 1983).
However for a stationary single-photon source, the nonclassical nature of
the emitted radiation would lead to sub-Poissonian photon statistics with
g(2)(0) < 1.
Observation of sub-Poissonian photon statistics was first made by Short

and Mandel (1983) using fluorescence from a single two-level atom which
emitted single photons.
For some application, it is necessary that the single photons be indistin-
guishable. The produced packets should be fully coherent, characterized by
a Fourier-transform relationship between their temporal and spectral pro-
files, that is
1=2 (37)
where 1/(2T2) and 2T1 designate, respectively, the Lorentzian
half-width and the spontaneous decay time of the field amplitude. This
property is required for the implementation of any photon-based quantum
information processing system (Knill, Laflamme, & Milburn, 2001).
6. THE PURCELL EFFECT AND THE CONTROL OF

EMISSION OF ELECTROMAGNETIC RADIATION
Spontaneous emission of an atom is a result of the interaction between
the atom dipole and the vacuum electromagnetic fields. Therefore, it is not an
intrinsic property of an isolated emitter but rather a property of the coupled
system of the emitter and the electromagnetic modes in its environment.
Purcell (1946) first predicted that nontrivial boundary conditions of an
electromagnetic field in the vicinity of an excited atom could alter its
decay rate.
The rate for spontaneous transitions from an initial state jii with no
photons to a final state jf i with one photon is given by the well-known Fermi
golden rule (Loudon, 2000)
h2 vc jh f jHjiij2 (38)
where H is the interaction Hamiltonian and (c) is the density of states at the
transition frequency c, that for radiation in free space is
2
4vc
vc 2 (39)
c3
The spontaneous emission rate can therefore be changed if an atomic sys-
tem finds itself in a space region where (c) is modified; for example, it is
placed close to a metal surface or a dielectric interface. First experimental
demonstrations were carried out by Drexhage, Kuhn, and Schafer (1968),
Kuhn (1970), and Chance, Prock, and Silbey (1974). Approximate solutions
of the electromagnetic boundary value problem were reviewed, for exam-
ple, by Drexhage (1974). More detailed calculations are in Lukosz and Kunz
(1977a, 1977b) and references therein. In a more effective way, the radiation
emitted by a source can be altered by suitably modifying the surrounding
vacuum fields in a cavity.
The application to a small cavity for which the density of modes may be
modified was considered by Kleppner (1981). In particular, when the tran-
sition frequency c is near resonance with a mode eigenfrequency, the spon-
taneous emission rate can be considerably increased.
The effect was experimentally observed (Goy, Raimond, Gross, &
Haroche, 1983) with a sodium Rydberg atom set through a resonant sup-
erconducting cavity. Also inhibited spontaneous emission was observed by
studying the cyclotron motion of a single electron (Gabrielse & Dehmelt,
1985). Therefore, a reduction of the two-photon probability relative to a
Poisson distribution may be achieved acting on the density of modes.
Periodic dielectric structures in the form of photonic crystals can alter
the emission properties ( John, 1987; Yablonovitch, 1987). The existence
of forbidden electromagnetic frequencies inside the structure may inhibit
the emission of radiation which is, on the contrary, enhanced at the frequen-
cies where the density of modes is maximum.
Enhanced spontaneous emission by quantum semiconductor boxes in a
monolithic optical microcavity, that can be considered the ancestor of quan-
tum dots experiments, was studied, for example, by Gerard et al. (1996,
1998 and; Gerard & Gayral, 1999) (see also Bulu, Caglayan, & Ozbay,
2003). A general review of 1D photonic structures is, for example,
Bertolotti (2006). In semiconductor systems, enhanced and inhibited spon-
taneous emission from GaAs quantum well excitons was demonstrated using
a planar microcavity (Yamamoto, Machida, Horikoshi, & Igeta, 1991;
Yokoyama et al., 1990). The properties of band gap structures have been
largely studied (see, for example, Yang, Fleischhauer, & Zhu, 2003;
Yang & Zhu, 2000). The effect of nanotubes on the spontaneous emission
of a single InAs quantum dot was considered by Bleuse et al. (2011).
7. PREPARATION OF SINGLE-PHOTON STATES:

QUANTUM ENGINEERING
How to obtain an arbitrary quantum state is a task that may be accom-
plished by so-called quantum-engineering methods that allow one to prepare a
previously specified quantum state by a series of elementary operations on

the system that is to be prepared. Different approaches have been considered
to create nonclassical states. The general approach is to find an appropriate
Hamiltonian which transforms via unitary time evolution a given initial state
to the desired state. Another way is to operate on the system projecting it on
another state. In general, the creation of Fock states was considered via the
interaction of a suitably prepared two-level atom with the electromagnetic
field. A way to single out atoms or molecules and have efficient interaction
with an electromagnetic field is to use microcavities.
In the years 1980s and 1990s, a great number of proposals were done on
possible production of states with defined number of photons by using the
interaction of atoms with quantum microcavities (see, for example, Meystre,
1992 for a description of the interaction in microcavities).
One fundamental aspect in the strong coupling regime of the Jaynes
Cummings ( Jaynes & Cummings, 1963) approximation is that, at resonance,
an atom initially in an excited state jei and an empty cavity, periodically
exchange a quantum. The atomfield state wave function j (t)j oscillates
between an excited atom and no photon state je,0i and a state in which the
atom is on the ground level and a photon is in the cavity jg,1i, according to
t cos o t=2je, 0i + sin o t=2jg, 1i (40)
where o is the Rabi frequency and t is the interaction time. So, instead of
simply emitting a photon and going on its way, an excited atom in a resonant
cavity oscillates back and forth between its excited and unexcited states. The
emitted photon remains in the cavity in the vicinity of the atom and is
promptly reabsorbed. The atomcavity system oscillates between two states,
one consisting of an excited atom and no photon and the other a deexcited
atom and a photon trapped in the cavity.
The basic idea of many schemes is that measurement of the atom after its
interaction with the field in a cavity provides information about the field.
The collapse of the entangled state of the atomfield system makes the field
part of the wave function to jump into a different state from the one before
the measurement. Based on the interaction between the radiation field and
atoms, many theoretical schemes were therefore proposed for the generation
of Fock states. We consider here, without pretending to be exhaustive, a few
of them. Some of them, with suitable changes, were later experimentally
realized (see Section 8.2). Eventually, as will be shown later, the real sources
of single photons have been implemented on a much simpler and direct way
choosing with skill the emitter in solid state.
The group at Max-Planck-Institute, Garching, had great experience

with quantum cavities. Filipowicz, Javanainen, and Meystre (1986a,
1986b) were probably the first ones to make the proposal to use the inter-
action of atoms with the electromagnetic field in a microwave microcavity.
They considered a lossless, single-mode cavity in which two-level atoms
were injected at such a low rate that at most one atom at a time is present
inside the cavity. Assuming that only one atom at the time was coupled to
the field, the authors investigated the interaction classically and quantum
mechanically using the JaynesCummings model ( Jaynes & Cummings,
1963). Controlling the initial conditions, if atoms in their excited state
are considered, a highly excited Fock state of the cavity mode was predicted.
The situation in which atoms in their excited state are injected in a high-
Q micromaser cavity was considered also by Krause, Scully, and Walther
(1987), but from a completely different point of view. In their proposed
experiment, Rydberg two-level atoms in their upper state were injected into
the micromaser cavity. After they leave the cavity, they are probed with a
static electric field which ionizes all atoms which are in their upper level.
All the atoms that are not ionized have emitted a photon in the cavity. When
these atoms are counted, the total number of photons in the maser can be
inferred. Therefore by the determination of the state of the outgoing atoms,
the photon number in the field is exactly known; i.e., the state of the field is
reduced to a pure number state. Since at start no radiation is in the cavity, the
field is always in a number state when an atom enters the cavity. By the inter-
action of the atom with the field, which is in a state ni, the field state will be
changed to a superposition of states ni and n + 1i. Due to the measure-
ment of the atomic state, afterward this superposition is reduced to one of the
states ni and n 1i, depending on the result of the measurement. In an
experiment, thermal photons in the cavity have to be suppressed. It means
that the cavity should be cooled at temperatures close to absolute zero. It also
should be emphasized that the number of atoms leaving the cavity in the
lower state is equal to the number of photons in the cavity only for a lossless
cavity. Moreover, a feature central to the proposed scheme is the long life-
time of the photons in the cavity, which is demanding for very high
Q-factors. The decay time for photons of frequency f is linked to the quality
factor Q of the cavity by
Q=f (41)
Quality factors Q of the order of 1010 at frequencies in the microwave region

f 1010 Hz give long lifetimes of the order of a second.
In a second paper, the control, via an electric accelerating field, of the

interaction time of ionic Rydberg atoms with the maser field was proposed
(Krause, Scully, Walther, & Walther, 1989). If the velocity of the ions is
adjusted in such a way that every ion emits a photon, then the total number
of photons is exactly known via the total number of passing ions.
The important ingredients in this type of proposals are superconducting
cavities together with the laser preparation of highly excited atoms. These
atoms are Rydberg atoms that have quite remarkable properties: the prob-
ability of induced transitions between neighboring states of a Rydberg atom
scales as n4, where n denotes the principal quantum number. Consequently,
a single photon is enough to saturate the transition between adjacent levels.
Moreover, the spontaneous lifetime of a highly excited state is very large. In
the experiments, a transition between two neighboring Rydberg levels is
resonantly coupled to a single mode of a cavity field.
The limitation of the Krause method is that it does not predict the even-
tual Fock state, but only the probability of its occurrence. Moreover, this
technique seems to be useful to generate single-photon states, which how-
ever cannot be made free, that is to say, are not available for further
experiments.
Also at Ecole Normale Superieure in Paris, the production of Fock states
was considered. A method for their production was proposed by Raimond,
Brune, Lepape, and Haroche (1989). The scheme utilized the passage of a
two-level atom through a microwave cavity, the resonant frequency of
which is continuously tuned (during the atomic transition time) so as to pro-
duce an adiabatic transformation of the initial eigenstate describing the
atomcavity system into a final eigenstate corresponding to the cavity mode
being in a Fock state. In this application, the adiabatic theorem (see Kramers,
1957) is used which applies to a system described by a time-varying
Hamiltonian H(t). Applied to this system for the interval of time from to
to t1, provided the evolution in this time interval is sufficiently slow, the
theorem says that if the system is initially in an eigenstate of H(to), it will pass
into the eigenstate of H(t1) that derives from it by continuity.
Brune, Haroche, Lefevre, Raimond, and Zagury (1990) at Ecole
Normale Superieure, Paris, suggested a method for the preparation of a Fock
state based on the quantum nondemolition measurement. The photon num-
ber is measured in a cavity by coupling the field to a beam of Rydberg atoms
and measuring the atomic phase shift. After a sequence of atomic measure-
ments, the cavity field collapses into a Fock state with an unpredictable num-
ber of photons.
Holland, Walls, and Zoller (1991) proposed a quantum nondemolition

measurement of the photon number of a cavity field mode by measuring the
deflection of atoms passing through the field.
A variation of the scheme proposed by Brune et al. (1990), involving an
initial coherent state and single-atom interaction and detection, can generate
a Schr odinger cat state of the field (Brune, Haroche, Lefevre, Raimond, &
Zagury, 1992; Savage, Braunstein, & Walls, 1990). The approach, capable of
yielding both Schr odinger cats and Fock states, was based on the detection of
the dispersive field-induced phase shift acquired by the initially polarized
state of nonresonant atoms in the cavity (Brune et al., 1992). The random
variation of these phase shifts with the atomic transit time (velocity) in a
sequence of such measurements results in repeated splitting of the phase-
space field distribution and interference between the split parts. The first step
in the sequence yields a phase difference Schr odinger cat, while subsequent
steps gradually lead to phase diffusion and decimation (through interference)
of the photon number distribution, converging to a Fock state. Also this
method cannot determine in advance the eventual Fock state.
A method for reconstructing the average evolution of the photon num-
ber distribution of a field decaying in a high-Q cavity was later presented
(Sayrin et al., 2012).
At low temperature, steady-state macroscopic superpositions can be gen-
erated in the field of a micromaser pumped by a stream of two-level atoms
injected in a coherent superposition of their upper and lower states (Slosser,
Meystre, & Wright, 1990). Under appropriate conditions, the field in a loss-
less micromaser can evolve to pure number states.
Generation of Schr odinger cats by many atoms initially prepared (by a
/2 pulse) in a polarized state whose sequential transit through the cavity
gradually builds up an approximate trapping state of the field, described
by a superposition of the vacuum and a number state, was proposed by
Meystre, Slosser, and Wilkens (1990).
Vogel, Akulin, and Schleich (1993) discussed in general the way to con-
struct any desired quantum state of the radiation field by using a simple
Hamiltonian in a single-mode resonator. A single-mode cavity field state
could be prepared by injection of a prepared and later probed two-level
atom (see also Vogel & Wellentowitz, 2001). They considered constructing
any superposition of the first N + 1 number states from the vacuum state by
injecting N appropriately prepared atoms into a cavity and detecting all of
them in the ground state. Since the successful preparation of the desired cav-
ity field state depends on the outcomes of the electronic state measurements
on the atoms after having passed through the cavity, the technique is prob-
abilistic and attains a successful engineered quantum state only for a specific
series of measurement outcomes.
Injection of atoms with their Zeeman electronic levels initially prepared
in specific superpositions interacting adiabatically with a cavity was proposed
(Parkins, Marte, Zoller, Carnal, & Kimble, 1995; Parkins, Marte, Zoller, &
Kimble, 1993).
A scheme based upon the same mechanism described by Vogel et al.
(1993) but employing two-photon interactions in a micromaser has been
described by Garraway, Sherman, Moya-Cessa, Knight, and Kurizki
(1994) who proposed a scheme that allows the generation of a variety of
nonclassical field states with controllable (predetermined) photon number
and phase distribution. It was based on the two-photon-resonant interaction
of a single electromagnetic field mode in a high-Q cavity with initially
excited atoms crossing the cavity sequentially (one at the time). The
sequence duration should be much shorter than the cavity lifetime. The
states are generated by selecting only those sequences wherein each atom
is measured to be in the excited state after the interaction. The field distri-
bution resulting from a sequence of N such measurements is peaked about
2N positions in the phase plane. When these peaks are chosen to overlap,
part of the field can be made to interfere giving rise to decimation of the
photon number distribution and Fock states may be generated.
At variance with the practice of measuring atoms once they came out of
the cavity, Law and Eberly (1996) discussed a method in which quantum
states of the atom (source) are manipulated during the atomfield interaction
process, i.e., inside the cavity. They presented a cavity model which allows
the vacuum state to evolve to an arbitrarily prescribed superposition of Fock
states.
Later, Law and Kimble (1997) discussed the possibility to produce a pre-
scribed sequence of single-photon pulses controlling the emission of an atom
with a -type three-level structure in an optical cavity. In 2003, the Caltech
group (Duan, Kuzmich, & Kimble, 2003) extended the study of the inter-
action of an atom trapped in a cavity with hot atoms with an inhomogeneous
distribution in position and/or a time-varying location, using an adiabatic
passage technique.
The previously considered schemes find at least two types of difficulties.
First, the quantum states of the sources of radiation (the atomic states) have
to be manipulated in an arbitrarily controllable manner. Second, the source
must teach the field to evolve toward the desirable quantum state.
A new method was proposed by Kuhn, Hennrich, Bondo, and Rempe

(1999) at Garching, for the generation of single photons into a single mode
of the radiation field. The idea was based on proposals by Law and Eberly
(1996) and Law and Kimble (1997). A single atom, strongly coupled to an
optical cavity, which defines the active mode and ensure photon emission
into a well-defined direction, was used as the active medium for generating
the photon. The excitation scheme is shown in Figure 3. A -type three-
level atom with two long-lived states jui and jgi, typically two Zeeman or
hyperfine states of the atomic ground state, and an electronically excited state
jei is in a cavity with a resonant frequency close to the atomic transition jei to
jgi but far off resonance from jei to jui. Hence, only the states je,0i with zero
photon and jg,1i with one photon are coupled to the cavity mode. Initially,
the system is prepared in state ju,0i. To trigger a photon emission, the atom
is exposed to a light pulse at Rabi frequency coupling the states ju,0i and
je,0i. Provided the trigger pulse rises sufficiently slowly, an adiabatic evolu-
tion is assured. If the two-photon resonance condition is fulfilled, a STIRAP
(STImulated Raman scattering involving Adiabatic Passage)-type adiabatic
passage takes place and ju,0i goes to jg,1i generating the emission of one
photon. Recycling can then be provided.
The use of nonlinear interactions was also discussed.
Leonski and Tanas (1994) proposed a scheme in which a one-photon state
can be obtained in a cavity that is periodically kicked by a sequence of classical
light pulses and is filled with nonlinear Kerr medium. If the amplitude
between the kicks and the time between the kicks are appropriately chosen,
Ie,0>
Energy of the atomic bare states
Atomcavity
coupling g
)
AP
TIR
Trigger
pulse Emission of a
r (S
WT single photon
sfe
tran
tic
Ig,1> Ig,0>
aba
Adi
Repumping pulse
Iu,0>
Figure 3 Scheme of the atomic levels coupled by the trigger pulse, the cavity, and a
possible repumping pulse. From Kuhn et al. (1999).
the one-photon quantum state can be reached. Later, Leonski (1996)

extended the model considering the interaction of a single-mode field with
both a Kerr medium and a parametric nonlinearity of the kth order. The sys-
tem was externally driven by the parametric pump field. By applying an
appropriately chosen field frequency, and for weak nonlinear process, a
k-photon Fock state of the electromagnetic field could be achieved.
Imamoglu, Schmidt, Woods, and Deutschet (1997) considered reso-
nantly enhanced absorption-free Kerr nonlinearity in a low-density atomic
medium whose energy-level structure may be represented by a four-state
diagram contained in a cavity. They showed that it is possible to achieve
coherent control of the cavity-mode wave function using pulse for single
photons that switch the state of the cavity. The system should realize strong
antibunching and deterministic injection into a cavity mode.
Kerr effect in solids was discussed more recently by Ferretti and Gerace
(2012). They show that nanostructuring of an ordinary nonlinear medium is
able to produce very large effective nonlinear susceptibilities, ultimately sen-
sitive at the single-photon level.
Different methods were also considered. Domokos, Janszky, and Adam
(1994) proposed a single-atom interference method for generating Fock
states. Pegg, Phillips, and Barnett (1998) showed that the number-state
expansion of an optical state can be truncated so as to leave only its vacuum
and one-photon component without using a cavity. In a sense they used
entangled states. A method to have a single-photon source using laser pulse
and two-photon absorption was proposed by Jacobs, Pittman, and Franson
(2006), and a number of other proposals could be considered (for example,
Waks, Diamanti, & Yamamoto, 2006).
However, how to generate an arbitrary photon state by virtue of the
interaction between the radiation field and quantum devices in the form
of micromasers is still very challenging.
8. REALIZATION OF SINGLE-PHOTON SOURCES

The realization of a single-photon source requires a single quantum
emitter with a high quantum efficiency. The philosophy for these sources
is to excite the dipole (atom, molecule, quantum dot, etc.) and collect
the subsequently emitted photon within the lifetime. More sophisticated
systems may use a train of single photons emitted in this way.
The single-photon emission is usually claimed when photon anti-
bunching is observed. One should note, however, that photon antibunching
is a necessary but not sufficient condition for a single-photon turnstile

device, as photons are emitted at random times and not deterministically.
Therefore, a control of the emission process is required, consisting of an
excitation of the single quantum emitter with probability one and subse-
quent (in most of the cases) spontaneous emission of a photon. The trigger
event for emission could be, e.g., a pulsed excitation of the emitter. How-
ever, spontaneous emission of the quantum emitter in general is directed
into the full solid angle and thus hard to capture efficiently. When the source
is in solid state, the refractive index difference with air restricts the output as
most of the light is back-reflected by total internal reflection. The problem
therefore exists to extract the radiation efficiently and in the wanted direc-
tion. Coupling the emitter to a cavity mode, for example, allows for a direc-
tional emission and thus for using the generated photons in external
applications due to the enhanced detection efficiency. The light collection
strategies have been discussed by Barnes et al. (2002).
We may divide the single-photon sources into three main categories:
1. Thermal or laser sources are so much attenuated that one may assume
photons arrive one at a time;
2. Single-photon sources that however emit at random;
3. Single-photon sources on demand.
In the following, we will discuss all these three possibilities.
8.1 Strongly Attenuated Sources

The first experiment with probably nearly a photon at the time was per-
formed by Taylor (1909) at Trinity College, Cambridge, UK. Taylor per-
formed an experiment in which single photons are supposed to play. From
an historical point of view, it is interesting to note that although in 1909 the
Einstein paper on the quantization of light (Einstein, 1905) was already pub-
lished, Taylor did not mention it but referred to a suggestion of Thomson
(1907) writing: The phenomena of ionization by light and by Roentgen
rays have led to a theory according to which energy is distributed unevenly
over the wavefront. There are regions of maximum energy widely separated
by large undisturbed areas. When the intensity of light is reduced these
regions become more widely separated, but the amount of energy in any
one of them does not change, that is, they are indivisible units. Sir J.J. Thom-
son suggested that if the intensity of light in a diffraction pattern were so
greatly reduced that only a few of these indivisible units of energy should
occur on a Huygens zone at once the ordinary phenomena of diffraction
would be modified. Therefore, Taylor performed an experiment taking

photographs of the shadow of a needle, illuminated by light coming from
a narrow slit placed in front of a gas flame, and attenuating the intensity
by means of smoked glass screens. Five diffraction photographs were taken,
each one with a different intensity of incoming light. The last photographs
were taken with an intensity so low that exposure was 2000 h or about
3 months. In no case, there was any diminution in the sharpness of the pat-
tern. The amount of energy falling on the plate during the longest exposure
was derived to be the same as that due to a standard candle burning at a dis-
tance slightly exceeding a mile or about 5 106 erg/s on 1 cm2. If we take
this figure as for good, it corresponds to a few photons per microsecond per
cm2 so that we may assume they arrived mostly one at the time.
The result is then taken as a demonstration that each photon can produce
its diffraction pattern by itself.
However, very attenuated thermal sources suffer of the circumstance that
chaotic photons tend to bunch.
Faint laser pulses and parametric sources of entangled pairs deliver
Poisson distributions of photons (or of photon pairs), from which
multiphoton events can never be entirely suppressed. Nevertheless, such
sources are much easier to built and operate than single-photon sources.
Single-photon number states may be approximated by coherent states with
very low average photon number (n). They can easily be developed using
only standard pulsed lasers and calibrated attenuators. However, when n
is small, most pulses are empty.
With an attenuated pulsed laser source, the probability of having m 0,
1, 2, 3 or more photons present at a time is controlled by the Poisson statistics
pm hnim ehni =m! (42)

Single-photon number states may be approximated by coherent states
with a very low average photon number. One may introduce a probability
pm that a nonempty weak coherent pulse contains more than one photon
pm pn>2 =pn>1 1 p0 p1=1 p0 hni=2 (43)
where pn>1 and pn>2 are the probabilities that a pulse contains at least one and
at least two photons, respectively. The value of pm can therefore be made
arbitrary small by decreasing hni. However, when hni is small, from Equa-
tion (42), most pulses are empty:
pm 0 1 hni (44)
Increasing the pulsed laser rate could help, but in this way also dark
counts will increase and the ratio of detected photons to dark counts
decreases with hni (see Gisin et al., 2002 for more details).
Weak chaotic attenuated sources are unsatisfactory also because if the sin-
gle photon is radiated into a large solid angle, it is quite difficult to collect it
into the optical system with good efficiency. The loss of photons increases
also if filtering in polarization is needed. However, this kind of source is
often used. As an example see the experiment for a daylight quantum key
distribution over a 1.6-km path in atmosphere, performed by Buttler
et al. (2000). At least one system exists based on this type of source
(Bovino, 2013a).
8.2 Single-Photon Sources Emitting one Photon in a

Random Way
These sources are based on the property of a single emitting dipole to emit
only one photon at a time. When excited by a short and intense pulse, such
an emitter delivers one and only one photon. Unfortunately, the time of
emission cannot be predicted exactly because it stays in the lifetime of the
transition and also the emission direction is at random.
These sources are built around a single emitting nanometric object, pro-
ducing photon distributions which are very far from Poissonian. In most
cases, for example, the probability density of emitting two photons at the
same time can be completely neglected, whereas it is still high for an atten-
uated Poisson source with the same brightness. In most cases, the emission
process is spontaneous and takes place after a rapid excitation of the emitter.
The first generation of single photons was probably done by Clauser
(1974), based on a cascade transition of mercury atoms. Each excited atom
delivers a couple of two photons with different colors. A photon at one of
the wavelengths is detected after spectral filtering and is used for the condi-
tional detection of its companion at the other wavelength. Each single
photon is thus herald for the presence of its companion. We will discuss this
experiment later.
The jitter in the emission can be partially neglected in the case in which a
series of single photons are obtained by pumping with a mode-locked laser
or another source of regular pulses. Such device is sometimes called
photon gun.
Continuous streams of antibunched photons were first observed from
single atoms and ions in traps (Diedrich & Walther, 1987; Kimble
et al., 1977).
Single photons have been obtained at least in the following systems.
8.2.1 Fluorescence from Single Atoms or Molecules

Single molecules have demonstrated to be useful single-photon emitters,
using optical pumping. If one highly fluorescent molecule is optically
selected and controllably forced to emit, a single-photon source will result.
The emitted light is antibunched and consists of single photons separated by
random time intervals which depend on the excitation state lifetime and on
the pumping rate. These sources were first used to show the property of
antibunching of the radiation emitted by a single emitter, and then as
single-photon emitters. A regular stream of photons can be obtained if single
photons are emitted periodically in time.
The first theory of atomic resonance fluorescence for weak excitation of
a two-level atom was formulated by Weisskopf (1931). Later, Mollow
(1969) extended the theory to arbitrary strong monochromatic excitation
of two-level atoms. The resonance spectrum consists of a single spectral line
that develops into a triplet (Mollow triplet) state above saturation with side-
bands shifted by the Rabi frequency of the system (Flagg et al., 2009;
Mollow, 1969; Vamivakas, Zhao, Lu, & Atat ure, 2009).
Photon antibunching was obtained in resonant fluorescence on single
atoms of sodium by Kimble et al. (1977). The atoms in an atomic beam were
excited by the light beam of a tunable dye laser. The emitted fluorescence
light was collected at right angles to both the laser and the atomic beams by a
microscope objective, divided by a beam splitter and sent to two photo-
multipliers. After amplification and pulse shaping, the pulses from the
two detectors were fed to start and stop inputs of a time-to-digital converter,
where the time intervals between start and stop pulses were digitized. The
number of events n() stored at address is therefore a measure of the joint
photoelectric detection probability density p(t,t + ). The value of n()
increased with from its smallest value at 0 (Figure 4).
Subsequently, storing single atomic Mg+ ions in a radio frequency trap
by photon recoil cooling, antibunching in the resonant fluorescence was
obtained by Diedrich and Walther (1987). Later, Basche, Moerner, Orrit,
and Talon (1992) studied the correlation between fluorescence photons
emitted by an optically pumped single dye molecule (pentacene in a
p-terphenyl host) obtaining antibunching. They pumped the lowest zero-
phonon purely electronic transition at 1.5 K and detected the (long wave-
length) vibronically shifted fluorescence from the excited singlet S1 rather
than the resonant fluorescence as in single-ion experiments.
n(t)
500
400
300
200
25 50 75 100
Delay t in nanoseconds
Figure 4 The number of recorded pulse pairs n() as a function of the time delay in
nanoseconds. The growth of n() from 0 shows antibunching. The bars on one point
indicate statistical uncertainties corresponding to one standard deviation. The broken
line just outlines the trend. From Kimble et al. (1977).
Photon emission was triggered with a short excitation pulse (De Martini,
Di Giuseppe, & Marrocco, 1996) in a molecular solution of oxazine in eth-
ylene glycol.
Kitson, Jonsson, Rarity, and Tapster (1998) studied the spontaneous
emission from a thin layer of a dilute solution of fluorescent dye molecules
within an optical microcavity. They identified photon antibunching. They
however reported that true single-photon emission is limited by photo-
bleaching and Brownian motion of the molecules in and out of the excita-
tion volume. The paper discussed the disexcitation of the dye which was
modeled as a singlet and a triplet state.
Hoffges, Baldauf, Lange, and Walther (1997) measured the correlation
function g(2)(t) of fluorescence light from a single 24Mg+ ion confined in a radio
frequency trap, as a function of detuning and intensities of the incident field.
The results clearly show the antibunching effect (Figure 5). For increas-
ing intensity of the exciting laser, the function displays, as should be
expected, oscillations at the Rabi frequency (the Mollow triplet).
Figure 5 Intensity correlation for three different detunings and intensities of the inci-
dent field (a, higher intensity; c, lower intensity). From Hoffges et al. (1997).
Solid-state schemes were considered, such as single organic molecules in

solid matrices. A comprehensive review of the experimental methods of
single-molecule detection and spectroscopy in condensed phase has been
made by Moerner and Fromm (2003) (see also Moerner, 2004).
A rapid adiabatic passage was used to control the emission of a few pho-
tons from a single molecule. The method was based on the manipulation of
the absorption frequency of the molecule using the Stark effect. If the
absorption frequency of a molecule can be changed in some way, for a
period of time T, the absorption frequency of the molecule can be set in
resonance with a driving continuous wave (cw) laser field. After a time inter-
val T, the absorption frequency of the molecule is changed, in such a way
that the emitter is far from resonance, namely the molecule effectively does
not interact with the laser field. If T is not too long and not too short com-
pared with the lifetime of the excited state and other parameters of the laser
field, a set of control parameters may be designed to generate a few photons.
Such a device, at cryogenic temperatures, was obtained by Brunel, Tamarat,
Lounis, Plantard, and Orrit (1998) and Brunel, Lounis, Tamarat, and Orrit
(1999), using the Stark effect to modulate the absorption frequency of an
organic molecule of dibenzanthanthrene in a n-hexadecane matrix inter-

acting with a cw laser which was kept at constant frequency. Every time
the molecules transition crosses the laser frequency, the molecule gets
excited and subsequently emits fluorescence photons with an average delay
of 8 ns which is the lifetime of the S1 state. A measure of antibunching
showed that up to 74% of the sweeps lead to the emission of a single photon.
A small though fluctuating number of photons, i.e., typically 0, 1, or 2, was
obtained. Very narrow optical absorption lines are needed for this method,
which requires liquid helium temperature (1.8 K).
The first good attempt to have operation at room temperature was per-
formed by Fleury, Segura, Zumofen, Hecht, and Wild (2000) who consid-
ered the fluorescence of single terrylene molecules in a crystalline host
(p-terphenyl) at room temperature by scanning confocal optical microscopy.
In this case, the problem of rapid photobleaching at room temperature was
avoided because terrylene was efficiently protected by the host crystal to dif-
fusion quenchers (such as oxygen) and profited from its ability to emit host
phonons to prevent thermally induced damage. The second-order correla-
tion function was measured with the startstop method and showed anti-
bunching. They observed also sub-Poissonian statistics.
Terrylene embedded in a thin platelet of crystalline p-terphenyl was used
also by Lounis and Moerner (2000) who realized a controllable source of
single photons using optical pumping of a singe molecule in a solid, at room
temperature (Figure 6). Single photons were generated at predetermined
times within the accuracy of the emission lifetime of a few nanoseconds.
Figure 6 (a) The pertinent molecule levels, (b) chemical structures of terrylene (upper)
and p-terphenyl (lower), (c) confocal fluorescence image (10 10 m) of single terrylene
molecules excited at 1.5 W, and (d) schematic of the confocal system. From Lounis and
Moerner (2000).
They used a periodic pulsed laser source (pulse width 35 ps, repetition rate
6.25 MHz). The system worked for hours without photobleaching.
The exciting source was a frequency-doubled (532 nm) mode-locked
picosecond Nd:YAG laser with a maximum average power of 0.2 mW
which pumps the four-level scheme of the molecule from the ground singlet
state to a vibriationaly excited level of the first electronic excited singlet state.
After fast (ps) relaxation to the lowest electronic excited state, the molecule
subsequently emits a single photon (at around 579 nm). Because the mole-
cule can be pumped at high intensity, the probability of preparing the emit-
ting state can approach unity. This scheme has the further advantage of
spectrally separating the laser excitation and the fluorescence emission. Sin-
gle photons were generated at predetermined times, within the accuracy of
the emission lifetime of a few nanoseconds. The pump pulse width was very
short compared to the fluorescence lifetime.
Terrylene molecules embedded in a highly transparent polymer film of
polymethyl methacrylate (PMMA) pumped with the 514 nm line of a cw
argon-ion laser were used also by Treussart, Clouqueur, Grossman, and
Roch (2001) and studied with scanning confocal microscope at room tem-
perature. The group was among the first to publish (Figure 7) the g(2)() cor-
relation function as a function of time showing the deep at its center
( 0).
Indistinguishability through a HongOuMandel experiment was dem-
onstrated by Kiraz et al. (2005) using a single terrylene diimide molecule at
1.4 K.
Treussart et al. (2002) realized also a pulsed saturation of a new single
molecule (cyanine dye DiIC18) dispersed at a concentration of about one
Figure 7 Correlation function (left scale) and counts integrated over about 100s
(right scale). From Treussart et al. (2001).
molecule per 10 m2 into a 30-nm-thick thin polymer film (PMMA). The

fluorescence light from the single molecule, centered at 570 nm, was non-
resonantly excited at 532 nm with femtosecond pulses generated by a
frequency-doubled Ti:sapphire laser, and collected by a scanning confocal
optical microscope, at room temperature. Photon statistics was also studied
finding indication of sub-Poissonian statistics.
An extended study of terrylene molecules emission as a function of tem-
perature from 1.8 up to 300 K was carried out, immobilizing the molecules
in a quasiperiodical optical /2 microresonator and using confocal micros-
copy (Steiner, Hartschuh, Korlacki, & Meixner, 2007). At room tempera-
ture, the spectral shape of the fluorescence matched the local on-axis
transmission spectrum of the microresonator, and the spectral width of
the emission was reduced by an order of magnitude as compared to single
free molecule. The emission was still significantly narrowed at 110 K, while
at 1.8 K the fluorescence emission was dominated by a spectrally isolated
zero-phonon line. A dip in the intensity correlation showed single-photon
emission at 1.8 K, albeit with a rather high value of g(2)(0) 0.5. Also inten-
sity fluctuations and irreversible photobleaching were observed.
Photon-counting statistics of single molecule in solid matrix under a laser
and a rf field was theoretically considered, showing that the rf field could
control the single-photon emission (Han, Ji, & Zheng, 2009). However,
no experiment seems to have been done.
Dye molecules hosted in liquid crystals were studied in a series of papers
by Lukishova, Schmid, McNamara, Boyd, and Stroud (2003) and Lukishova
et al. (2004, 2009, 2007, 2006).
At first, a single-photon source made by terrylene molecules embedded
in a cholesteric liquid crystal host was considered (Lukishova et al., 2003).
Using oxygen-depleted liquid crystal, dye bleaching was avoided over the
course of more than 1 h of continuous 532 nm excitation. The liquid crystal
permits to align the molecules along a direction optimized for maximum
excitation efficiency. The technique was further discussed and extended
to the use of nematic crystal host (Lukishova et al., 2006).
Well-defined polarization fluorescence from single emitters at room
temperature was obtained (Lukishova et al., 2007) using a planar aligned,
nematic liquid crystal to provide definite alignment of single dye molecules
in a preferred direction.
The improvement in the emission obtained by the use of photonic crys-
tals was investigated by several people. Without wishing to be exhaustive,
we may mention calculations made by Busch, Vats, John, and Sanders
(2000) on radiating dipoles in photonic crystals and by Wang, Gu, Wang,

and Xu (2003) who theoretically investigated spontaneous emission from
atoms in a photonic crystal; they found that the effect of pure photonic crys-
tals may lead to the coexistence of both accelerated and inhibited decay pro-
cesses, and that decay kinetic properties depend on the atomic position in the
crystal. How to select the position of field maxima in a photonic crystal by
means of two counter-propagating pump beams was discussed by Settimi
et al. (2005).
Xu, Chen, Yamada, and Otomo (2008) studied a single emitter embed-
ded in a two-dimensional photonic crystal demonstrating that the anti-
bunching of the fluorescence from a two-level ideal system is determined
by the structure of a quasiperiodic photonic crystal and could be controlled
by changing the position or the frequency of emission.
Florescu et al. (2005) considered theoretically a photon gun based on stim-
ulated Raman adiabatic passage pumping in a photonic bandgap material.
They focussed on the possibility of modifying spontaneous emission by plac-
ing the radiation source inside a 3D dielectric photonic crystal. A Er3+ ion
could be excited using STIRAP from the ground state of the ion. With a dif-
ferent geometry, Hogele et al. reported on photon antibunching in the
photoluminescence spectra of single carbon nanotubes (Hogele, Galland,
Winger, & Imamoglu, 2008) at 4.2 and 77 K in the spectral window between
855 and 885 nm. Exciton states should be responsible for the emission.
The experimental demonstration of inhibited spontaneous emission of
quantum dots in a 3D photonic crystal was eventually given by Leistikow
et al. (2011).
Lee et al. (2011) demonstrated a dielectric planar antenna, which uses a
layered structure to tailor the angular emission of a single-oriented terrylene
molecule with a fluorescence maximum at 580 nm at room temperature.
They demonstrated a collection efficiency of 96% using a microscope objec-
tive and obtained detection rates of nearly 50 MHz.
8.2.2 Single-Photon Emitting Mesoscopic pn Junctions

A single-photon turnstile device based on a mesoscopic double-barrier pn
heterojunction was proposed in 1994 (Imamoglu & Yamamoto, 1994). If a
double-barrier mesoscopic pin heterojunction is driven by an alternating
voltage source, then Coulomb blockade and quantum confinement effects
together can allow injecting single electrons and holes into the etched dou-
ble barrier. It should therefore be possible to generate heralded single-
photon states. The idea was similar to a frequency-locked turnstile device
for single electrons proposed and demonstrated by Geerligs et al. (1990).
The Imamoglu and Yamamoto (1994) proposal was later realized (Kim,
Benson, Kan, & Yamamoto, 1999) obtaining single as well as multiple pho-
ton emission events with a repetition rate of 10 MHz at 50 mK. In the
device, single-electron and hole-charging energies must be large compared
to the thermal background energy to ensure single-photon emission. There-
fore, the device can only be operated at ultra-low temperatures (T < 1 K),
and sufficient collection efficiency to measure the photon correlation func-
tion was not obtained. Imamoglu considered also a photon blockade effect
in the context of cavity electromagnetically induced transparency
(Werner & Imamoglu, 1999).
Semiconductor quantum dots show promise as they can be embedded in
devices enabling electrical injection. Of course devices of this kind are much
more easy to handle than optically pumped systems. The realization of an
electrically driven source of single photons based on integrating InAs quan-
tum dots into a conventional semiconductor light-emitting diode structure
at 5 K was studied, obtaining emission at 1.3942 eV. Correlation measure-
ments demonstrated a large suppression of two-photon emission with the
lowest injection current (2 A) with g(2)(0) 0.34 (Yuan et al., 2002).
In GaAs-based LEDs, the high refractive index leads to a low critical
angle and only a small fraction of the emitted photons escape the front sur-
face. To enhance by a factor of 10 the efficiency with which the photons are
collected from an electrically driven single InAs/GaAs quantum dot,
Bennett et al. (2005a) used a planar semiconductor cavity. The device struc-
ture is shown in Figure 8.
Al p-ohmic contact InAs quantum dot

Emission
p+ GaAs
Insulator
3l cavity
Contact
12 period metal
n+ DBR
Substrate/buffer
n-ohmic contact n+ GaAs

Figure 8 The device was grown by molecular beam epitaxy and consisted of a doped
12-period GaAs/Al0.98Ga0.02As Bragg mirror and a three-wavelength GaAs cavity with a
low-density layer of InAs quantum dots positioned one wavelength from the Bragg mir-
ror. The upper electrical contact was made with Al with an aperture which provided to
single out a single quantum dot. from Bennett et al. (2005a).
Single-photon emission at 910 nm and 5 K was demonstrated. In a sub-

sequent paper, Bennett et al. (2005b) demonstrated that the uncertainty on
the emission time of photons emitted by the source can be reduced by elec-
trical control of the charge configuration within the quantum dot. As this
transition involves the transfer of charges into and out of the quantum
dot, it can occur at least an order of magnitude faster than radiative recom-
bination. For the device of Figure 8, the authors succeeded in reducing the
uncertainty on the time of photon emission from the exciton state of the
quantum dot to less than one-fifth of the radiative lifetime.
In the following year, (Bennett et al., 2006) discussed the state of the art
of electrically driven single-photon emitting diodes.
The use of an aperture in an aluminum oxide layer to restrict current
injection into a single self-assembled InAs quantum dot was demonstrated
by Ellis, Bennett, Shields, Atkinson, and Ritchie (2006).
Ward et al. (2007) obtained emission of single photons at 1.3 m with
InAs quantum dots in a planar cavity electrically driven. The device worked
at a temperature of 10 K.
Strauf et al. (2007) produced a vertical cavity surface emitting laser elec-
trically driven integrating high-Q GaAs/AlGaAs microcavities with embed-
ded oxide-tapered apertures. They observed single photons at record high
rates up to 31 MHz and single-photon extraction efficiencies up to 38% with
controlled polarization. Other results were obtained by Lochmann et al.
(2007).
Higher repetition rates were demonstrated with a single-photon LED
incorporating an oxide-confined microcavity with a Purcell enhancement
of the radiative decay of about 2.5 and a collection efficiency of 14%. Pulsed
electrical injection allowed single-photon emission with repetition rates up
to 0.5 GHz (Ellis et al., 2008). The outcoupling efficiency of a semiconduc-
tor source in the presence of a microcavity was an important step forward,
considering that a single quantum dot suffers from a poor outcoupling effi-
ciency when embedded in a bulk semiconductor matrix as most of the iso-
tropically emitted photons are trapped due to internal reflection.
Reischle et al. (2008) built an electrically pumped single-photon emitter
in the visible spectral range, around 1.81.9 eV, working up to 80 K, using a
self-assembled single InP quantum dot. Subsequently, they demonstrated
triggered emission in the red (1.81.9 eV) at an excitation repetition rate
up to 200 MHz at temperatures up to 80 K (Reischle et al., 2010).
Electrically driven quantum dotsmicropillar cavities at temperatures
between 23.8 and 41.8 K with g(2)(0) as low as 0.13 and an overall efficiency
around 34% were reported by Heindel et al. (2010). Pulsed electrical exci-
tation was used to excite single InP/Ga0.51In0.49P QDs.
Nowak et al. (2014) described a micropillar connected to a 25 25 m2
wide frame through one-dimensional wires. Combining with a diode struc-
ture, they demonstrated a deterministic and electrically tunable single-
photon source with an extraction efficiency around 53%. However, the
device worked at cryogenic temperatures.
The use of a quantum dot embedded in a photonic nanowire and elec-
trically pumped was discussed, claiming that output efficiencies of more than
80% could be achieved (Gregersen, Nielsen, Mrk, Claudon, & Gerard,
2010). Two-photon interference from two separate, independent quantum
dots electrically driven and tuned to the same energy applying a large bias
electric field was obtained by Patel et al. (2010).
Also a proposal for a method of producing single-photon states utilizing
the acoustoelectric effect was made (Foden, Talyanskii, Milburn,
Leadbeater, & Pepper, 2000).
8.2.3 Atoms Inside a Microcavity

With the micromaser assembled at Max-Planck-Institute in Garching,
Walthers group obtained experimentally single quantum states in the inter-
action of atoms and the field confined in the microcavity. As a result of the
many proposals made in the previous years, the Garching group studied
essentially three methods of creating number states in the one-atom micro-
maser they had constructed (Meschede, Walther, & M uller, 1985). The first
method relied on trapping states which are generated in a cw operation of the
pumping beam. A second method used entanglement between pumping
atoms and cavity field. The field is prepared by state reduction and the purity
of the generated states is investigated by a probing atom. The two methods
of preparation turned out to be in fact equivalent. The third method con-
sidered pumping the cavity with a pulsed beam and using the trapping con-
dition to stabilize the photon number, so producing Fock states on demand
(Brattke, Varcoe, & Walther, 2001a). The micromaser used for the exper-
iments is shown in Figure 9.
The concept of trapping states was central in these systems. They are a fea-
ture of low-temperature operation of the micromaser, for which the steady-
state photon distribution closely approximates a Fock state under certain
conditions, and occur in the micromaser as a direct consequence of the
quantization of the cavity field (Meystre, Rempe, & Walther, 1988;
Weidinger, Varcoe, Heerlein, & Walther, 1999; see also Brattke et al.,
Rubidium oven
Stark shifting
plates Auxilliary
detector
Reference laser
Velocity selecting
UV laser
Piezos for fine-

adjustment of the
niobium resonator
State selective
field ionization of
Rydberg atoms
Figure 9 The atoms leaving the rubidium oven are excited into the 63P3/2 Rydberg state
by means of a UV laser and enter the microcavity. The cavity is tuned to the 21.456 GHz
transition from the 63P3/2 to the 61D5/2 state, which is the lower or ground state of the
maser transition. After the cavity they are detected by a state-selective field ionization.
From Brattke et al. (2001a).
2001a). Trapped states are typical of strongly coupled systems and occur
when the atomfield coupling, , and the interaction time, tint, are chosen
such that in a cavity field with n photons each atom undergoes an integer
number k of Rabi cycles or
tint n + 1 k (45)
When Equation (45) is fulfilled, the cavity photon number is left unchanged
after the interaction of an atom and hence the photon number is trapped,
regardless of the atomic pump rate.
The first demonstration of trapping states in the maser field is described in

Weidinger et al. (1999). The relevant quantity was the atomic inversion I
defined as
I tint Pg tint Pe tint (46)
where Pe(tint) and Pg(tint) are the probability of finding an excited state and a
ground state atom, respectively, for a particular interaction time tint. The
operation of a cw micromaser leads to Fock states, with the micromaser field
evolving into a number state, provided the atomic velocity is conveniently
tuned (Weidinger et al., 1999).
A Fock state shows ideal sub-Poissonian statistics. Consequently, under
the influence of a trapping state, the atomic statistics should also be sub-
Poissonian. As an atom in the ground state must have emitted a photon into
the cavity, the lower state atom statistics is strongly related to cavity photon
statistics. When the maser is under the influence of a trapping state, an atom
cannot emit into the cavity until a photon decays from the mode. When this
occurs, the next excited state atom entering the cavity will emit a photon
with a high probability, returning the cavity to the trapping state. This results
in a regular spacing of ground state atoms and a reduction of fluctuations in
the count rate. The FanoMandel Q function

nt 2 hnti2
Q t (47)
hnti 1
where n(t) is the number of counts in an interval of length t, which may be
used to assess the kind of statistics. If Q(t) 0 the beam is Poissonian; sub-
Poissonian statistics produces Q(t) < 0, and super-Poissonian statistics has
Q(t) > 0. Indications of one-photon and two-photon states were clearly
obtained (Weidinger et al., 1999).
Using the one-atom maser, the first experimental evidence for the oper-
ation of a reliable and robust source of photon Fock states was presented by
Brattke, Varcoe, and Walther (2001b). In the experiment, the micromaser
setup of Figure 9 was used with a beam of Rubidium atoms in Rydberg
states entering the microcavity cooled at 300 mK. The group claimed a suc-
cess rate of Fock state production of 85% (see also Brattke et al., 2003).
Using the same setup of Figure 9, Varcoe, Brattke, Weidinger, and
Walther (2000) observed the buildup of number states in a high-Q cavity
by investigating the interaction dynamics of a probe atom with the field.
With a train of atoms, arbitrary multiphoton Fock states can in principle
be generated in a cavity by post selecting the events in which each atom has
emitted one photon. One- and two-photon states were generated in this
way (Varcoe et al., 2000). The states were individuated looking at the
behavior of the inversion probability (46). The number of Rabi oscillations
permitted to determine the state number. In this way, single-photon states
were obtained in a cavity in which pulsed excitation of atoms was used so
that the number of atoms passing through the cavity could be pre-
determined. The group produced number states with a purity of 99% for
the n 1 state and 95% for the n 2 state (Varcoe et al., 2000).
Also an adiabatic passage in sequence in an optical cavity relying on
STIRAP was made, using 85Rb atoms at 10 K (Hennrich, Legero,
Kuhn, & Rempe, 2000).
The work toward a single-photon source in the visible range using a sin-
gle trapped ion placed in a cavity was described by Pinkse, Fischer, Maunz,
and Rempe (2000) who reported high-finesse optical cavity experiments in
which the change in transmission induced by a single slow atom approaching
the cavity triggers an external feedback switch which traps the atom in a light
field containing about one photon on average. The source was modified
(Kuhn, Hennrich, & Rempe, 2002) still employing the emission of single
photons from a single three-level atom (Rb) strongly coupled to a high-
finesse cavity. The photons are generated by an STIRAP (see also
Hennrich et al., 2000) between two atomic ground states, with the vacuum
field of the cavity stimulating one branch of the transition, and the laser
pulses deterministically driving the other branch. Figure 10 shows the
scheme of the experiment and the pertinent 85Rb levels. A single Rb atom
is prepared in state jui and is pumped with a laser whose frequency is close to
the jui ! jei transition. Provided the pump pulse rises slowly and a large
atomcavity coupling is present, a smooth transition from state u with no
photon ju,0i to state g with one photon jg,1i is realized. This state decays
emitting a photon which leaves the cavity through the mirror. The final state
jg,0i of the coupled system is reached and a recycling laser pulse is applied
between consecutive pump pulses resonant with the jgi ! jei transition so
pumping the atom to state jei from which it decays spontaneously to the
initial state jui. The system dynamics was substantially the same as described
previously (Kuhn et al., 1999). The correlation function g(2)() was measured
finding antibunching (Figure 11). It also oscillated with the same periodicity
as the sequence of pump pulses. This indicated that photons are emitted only
during the pump pulses, and no emission occurs when recycling pulses are
Magneto-optical trap
85
Rb Correlator
e,0
2
g ()
5P3/2
85Rb atoms
Recycling laser
Pump Atom Avalanche photo diodes
laser cavity
coupling
Cavity
u,0
5S1/2
g,1 g,0 Beam splitter
Photon
emission Recycling laser
Pump laser
Figure 10 On the left the relevant energy levels of Rb. The atomic states labeled jui, jei,
and jgi are involved in the Raman process, and the states j0i and j1i denote the photon
number in the cavity. On the right side the setup. The figure is substantially the same as
from Kuhn et al. (2002) but is from Hennrich, Legero, Kuhn, and Rempe (2004).
Figure 11 The correlation function g2(t) averaged over 15,000 experimental cycles
showing antibunching. The hatched area represents correlations between photons
and detector-noise counts. From Kuhn et al. (2002).
applied. The nearly Gaussian envelope of the comblike function is a conse-

quence of the limited atomcavity interaction time.
The experiment showed that a coupled atomcavity system is able to
emit single photons on demand generating a sequence of up to seven pho-
tons from one and the same atom in a time interval of about 30 s.
These single-atom emitters are designed to work in the strong coupling
regime of cavity quantum electrodynamics, where the single photon pro-
foundly impacts the dynamics of the atomcavity system and the optical
cavity greatly enhances single-photon emission into a single spatial mode

with Gaussian transverse profile.
A summary of the first experiments with single atoms in quantum cavities
made at Garching was made by Walther (1998, 2002, 2004).
Hennrich et al. (2004) at Garching investigated the photon statistics of a
single-atom source in which the photons are emitted by shining a periodic
sequence of laser pulses on Cs and Rb single atoms falling randomly through
a high-finesse optical cavity. Photons were emitted via the same scheme used
by Kuhn (Figure 10). Strong antibunching was found but of course the sta-
tistics was not sub-Poissonian as the atoms interacting with the cavity follow
a Poissonian statistics being random. If a preselection is applied to restrict the
analysis to time intervals where the presence of an atom is assured, sub-
Poissonian statistics was obtained.
The theory of the behavior of a calcium ion in a cavity was extensively
studied by Walthers group, presenting a realistic model for the interaction
between a single 40Ca+ ion and a single mode of an optical cavity (Keller,
Lange, Hayasaka, Lange, & Walther, 2004a).
The group obtained controlled production of single photons based on
strongly localized single ion in an optical cavity. The ion was optimally
coupled to a well-defined field mode resulting in the generation of single-
photon pulses with precisely defined shape and timing (Keller, Lange,
Hayasaka, Lange, & Walther, 2004b).
The theory for a single-photon source based on the vacuum-stimulated
Raman transition of a single Ca+ ion trapped inside a high-finesse cavity was
presented by Maurer, Becher, Russo, Eschner, and Blatt (2004).
Later, an efficient single-photon source using a single Ca ion trapped
within a high-finesse cavity in a disposition somehow similar to the system
used by Keller et al. (2004a, 2004b) generated single photons with an effi-
ciency 88% (Barros et al., 2009).
Vacuum-stimulated Raman transitions driven between two magnetic
substrates of a 87Rb atom strongly coupled to an optical cavity were used
by Wilk, Webster, Specht, Rempe, and Kuhn (2007). A magnetic field lifts
the degeneracy of these states, and the atom is alternatively exposed to laser
pulses of two different frequencies. This produces a stream of single photons
with alternating circular polarization in a predetermined spatiotemporal
mode. Test on the indistinguishability of the produced photons was also
made performing a time-resolved two-photon interference experiment.
Two photons that simultaneously enter different entrance ports of a
50:50 beam splitter will always leave trough the same output port if they
are indistinguishable. The degree of indistinguishability was measured by

comparing the number of coincidences obtained when the pair of photons
have parallel polarization with the case when they are perpendicularly polar-
ized and thus completely distinguishable (Wilk et al., 2007).
Hijlkema et al. (2007) implemented a cavity-based scheme (Figure 12)
with a laser for trapping, a trigger laser for photon generation, and a recycling
laser for pumping, monitoring, and cooling Rubidium atoms. A single Rb
atom is trapped in a magneto-optical trap in a high-finesse optical
microcavity. The atomcavity system is excited by a sequence of laser pulses
and single photons emitted from the system are detected by two avalanche
photodiodes in the Hanbury Brown and Twiss configuration (Figure 12).
The system is able to hold a single atom in the cavity for up to a minute at
sub-millikelvin temperatures. The single-photon character was verified by
observing sub-Poissonian statistics and antibunching.
In France, at the Laboratoire Kastler Brossel a lEcole Normale Super-
ieure, the group lead by S. Haroche obtained a one-photon Fock state by
a quantum Rabi pulse in a microwave cavity (Matre et al., 1997). The
quantum information carried by a two-level atom was transferred to a
high-Q cavity and, after a delay, to another atom realizing a quantum mem-
ory made of a field in a superposition of 0 and 1 photon states.
Figure 12 A single 85Rb atom is trapped in a high-finesse optical microcavity by means

of a two-dimensional optical lattice. Confinement along the cavity axis and a direction
perpendicular to it is achieved with a weak cavity-stabilization laser and a strong
retroreflected dipole laser, respectively. Confinement along the third direction results
mainly from the small beam waist of the dipole laser. The atomcavity is excited by
a sequence of laser pulses incident under an angle of 45 to the dipole laser and per-
pendicular to the cavity axis. Single photons emitted from the system are sent in the
Hanbury Brown and Twiss correlator. From Hijlkema et al. (2007).
Bertet et al. (2002) at Ecole Normale Superieure reported the production

of a two-photon Fock state by a single atom in a high-Q cavity. The atom, in
a single nonlinear process, emits one photon in the initially empty target
mode, while transferring another photon from the source mode (containing
initially a small field) into the target. This third-order Raman process, res-
onant for proper atomcavity mode detuning, is highly efficient (see also
Raimond, Brune, & Haroche, 2001).
The cavity used at Ecole Normale is shown in Figure 13. A summary of
part of the work performed with it can be found in Raimond et al. (2001,
2005).
At the Norman Bridge Lab in California, a single cesium atom trapped
within the mode of an optical cavity was used to generate single photons on
demand (McKeever et al., 2004). Single cesium atoms were cooled and
loaded into an optical trap which localizes them within the mode of a
high-finesse optical cavity (Figure 14a).
The atom is then illuminated by a sequence of laser pulses made by pairs
of alternate short and long pulses. The first of the two pulses, 3, drives a
dark-state transfer between hyperfine ground states F 3 to F 4
(Figure 14b), pumping up to the level 30 . Strong coupling g with the optical
cavity induces the transition 30 to 4 and one photon is created in the cavity
mode which leaves the cavity as a freely propagating spatially Gaussian wave
packet whose temporal profile is determined by the external field 3(t). The
atom is then recycled back to the original ground state by a second laser pulse
4(t) through the steps 4 to 40 and 40 to 3.
Controlled single-photon emission from a single trapped two-level
rubidium atom in the focus of a lens was obtained by Darquie` et al.
Figure 13 Velocity-selected Rb atoms from the oven on the extreme left of the figure
are excited from lasers, interact with the superconducting cavity, and are detected in
the field ionization counter at the extreme right. From Raimond et al. (2005).
Figure 14 (a) A single Cs atom is trapped in a cavity made by two mirrors and is pumped
by external fields; (b) the relevant atomic levels; and (c) the timing sequence of pumping
pulses. From McKeever et al. (2004).
(2005). Only one atom at a time was stored in the trap; if a second atom
enters the trap, both are immediately ejected. In this regime, the atom sta-
tistics is sub-Poissonian and the trap contains either one or zero atoms. The
trapped atom was excited with a 4 ns pulse of polarized laser light at
780.2 nm at a repetition rate of 5 MHz. Fluorescence photons were pro-
duced by spontaneous emission from the upper state of the transition which
has a lifetime of 26 ns. In this way, an efficient (0.6 efficiency) triggered
source of single photons with a well-defined polarization was realized.
Photon blockade was proposed, discussed, and realized by several
authors, initially considering electrically driven devices (Imamoglu et al.,
1997; Kim et al., 1999; Smolyaninov, Zayats, Gungor, & Davis, 2002;
Werner & Imamoglu, 1999). It is possible to obtain the effect also through
photonphoton interactions in a nonlinear optical cavity using, for example,
electromagnetically induced transparency (see also Lukin & Imamoglu,
2001). Observations of photon blockade for the light transmitted by an opti-
cal cavity containing one trapped atom in the regime of strong atomcavity
coupling were reported by Birnbaum et al. (2005). Excitation of the atom
cavity system by a first photon blocks the transmission of a second photon.
Kang et al. (2011) demonstrated a high-speed controlled generation of
single photons in a coupled atomcavity system using a single 85Rb atom
pumped with a nanosecond laser. The rather complex geometry involved
dropping the Rb atoms from a magneto-optical trap into the cavity in such
a way to select a single atom and then exciting it and extract the photon from
the cavity. Rates up to 10 MHz were obtained.
Figure 15 The source utilizing Rydberg atoms. (a) Sketch of the apparatus; (b) relevant
87
Rb energy levels. From Dudin and Kuzmich (2012).
A novel kind of source has been recently implemented (Dudin &

Kuzmich, 2012) using Rb atoms excited to high number Rydberg states.
In high number Rydberg atoms, the valence electron wave function
expands over large radii and if the atoms are sufficiently near to each other,
according to Pauli principle, the levels are shifted in energy. An atom in the
Rydberg level therefore shifts the energy levels of nearby atoms, suppressing
their excitation (excitation blockade). Dudin and Kuzmich prepared a dense
ensemble of laser cooled (10 K) 87Rb atoms into a one-dimensional optical
lattice formed by a retroreflected light beam (Figure 15a). A magnetic field
was applied in the z-direction. There, two laser pulses were used for exci-
tation: one 1 to excite atoms from the ground state to an intermediate
excited state, and the second 2 to promote them from the intermediate
state to a specific Rydberg state and create an optical-frequency spin wave
between the ground and the Rydberg levels. After a delay, a readout pulse
(3) brings any Rydberg atoms back to the intermediate state, from which
they return to the ground state by emitting photons (Figure 15b).
Only one atom is excited to a Rydberg state and if the timing of the
pulses is properly chosen, the emitted photon travels in the direction oppo-
site to the 3 pulse into the measuring setup.
A very low value of the correlation function was obtained g(2)(0) 0.040.
8.2.4 Semiconductor Quantum Dots

Semiconductor quantum dots (QDs) look to be an ideal source of single
photons and have already produced promising results, albeit mostly at
low temperature (see, e.g., Shields, 2007).
The emission process is spontaneous and in most cases takes place after a
rapid optical excitation of the emitter. The main difficulty with quantum
dots is that they interact with a solid-state environment, necessitating cryo-
genic operation temperatures, and yet environment-induced decoherence is
still a problem. However, these difficulties are offset by advantages such as
being fixed in place, having large dipole moments, and allowing the possi-
bility of integration into monolithic optical microcavity structures or in pho-
tonic crystals, thereby changing the density of states of the electromagnetic
field, shortening lifetime, increasing efficiency, and allowing a better cou-
pling with the external world. In this way, quantum dots integrated in an
optical microcavity are efficient solid-state sources of single photons. At
cryogenic temperatures, they display spectrally narrow emission lines com-
patible with the control of spontaneous emission based on cavity quantum
electrodynamics and the use of the Purcell effect.
Single-photon emission from semiconductor quantum dots has been
reported for many different material systems covering different wavelengths
and maximum operation temperatures. The most studied systems have been
InAs in near infrared, InP in the red region, CdSe in the green, and GaN in
the ultraviolet.
A number of review papers have been published on the subject (see, for
example, Shields, 2007), the most recent being Buckley, Rivoire, and
Vuckovic (2012). Another excellent discussion has been done by Lodahl,
Mahmoodian, and Stobbe (2013).
First experiments were performed by Michler, Imamoglu, et al. (2000)
with CdSe/ZnS quantum dots (microcrystals) that were shown to generate
single photons in the green, at room temperature. Optical pumping was per-
formed using the circularly polarized light of the 488 nm line of a continuous
Ar+ laser. Blinking was observed and the value of the correlation function at
zero time delay was rather high: g(2)(0) 0.47.
IIIV semiconductor quantum dots looked a more attractive material to
investigate for single-photon emitters, having the advantage of being free
from photobleaching, blinking, spectral diffusion effects, or broad spectral
emission, as in the case with most other systems such as molecules or semi-
conductor microcrystals.
Therefore, Michler produced a first epitaxial self-assembled semicon-
ductor InAs QD demonstrating a single-photon turnstile device that gener-
ated a train of single-photon pulses using pulsed laser excitation of a single
quantum dot, embedded in a high-quality factor microcavity structure
(Michler, Kiraz, et al., 2000). The sample was grown by molecular beam
epitaxy on a semi-insulating GaAs substrate. Figure 16 shows the microdisk

structure, which consisted of a 5-m diameter disk and a 0.5-m
Al0.65Ga0.35As post. The disk area consisted of 100 nm of GaAs, an InAs
quantum dot layer, and 100 nm of GaAs. Optical pumping was performed
with a mode-locked femtosecond Ti:sapphire laser at 750 nm. The emitted
light was collected with a scanning optical microscope. The electronhole
pairs were mainly generated in the GaAs barriers and subsequently captured
by the QDs within a short timescale. The collected light was spectrally fil-
tered so to select the emission from the exciton ground state transition (1X).
To ensure single-photon generation at the fundamental quantum dot
exciton transition, the pump power was adjusted so that two or more elec-
tronhole pairs were captured by the quantum dot during each excitation
pulse. The energy of the photons emitted during relaxation depends signif-
icantly on the number of multiexcitons that exist in the quantum dot, due to
Coulomb interactions enhanced by strong carrier confinement. If the total
recombination time of the multiexciton quantum dot state is longer than the
recombination time of the free electronhole pairs, each excitation pulse can
lead to at most one photon emission event at the fundamental exciton tran-
sition. Therefore, regulation of the photon emission process can be
achieved. When the quantum dot fundamental transition is on resonance
with the high-Q cavity, the spontaneous emission rate is enhanced because
of the Purcell effect which reduces the time jitter in photon emission.
The emission from the fundamental exciton transition (1X) under pulsed
excitation with a pulse train from the Ti:sapphire laser showed peaks at inte-
ger multiples of Trep 12.27 ns, indicating the locking of the photon emis-
sion to the pulsed excitation (Figure 17). The peak in the correlation
function g(2)() at 0 was no longer present, showing that the probability
Figure 16 The microdisk structure. From Michler, Kiraz, et al. (2000).

Figure 17 Measured unnormalized correlation function g(2)() of (a) a mode-locked Ti:

sapphire laser that exhibits peaks at integer multiples of Trep 12.27 ns and (b) for the
1X transition of a single QD that is far detuned from all microdisk modes at T 4 K. The
measured g(2)() also shows peaks at integer multiples of Trep, indicating the locking of
the photon emission to the pulse excitation, but the peak at 0 is no longer present.
From Michler, Kiraz, et al. (2000).
of finding a second photon after the detection of the first photon at

0 vanishes. The peaks at times nTrep correspond to events in which
one photon was detected from each of two different pulses.
A single InAs quantum dot, which can be considered a semiconductor
analogue of a single-atom cavity, was positioned in a three-dimensional post
microcavity, all fabricated by molecular beam epitaxy, and coupled to a sin-
gle optical mode (Solomon, Pelton, & Yamamoto, 2001) (Figure 18).
A scanning electron microscope (SEM) image of a particular post is shown
in Figure 18a with a photoluminescence spectrum and the spectral depen-
dence of the spontaneous emission lifetime (Figure 18b).
The quantum dot (22 nm in-plane diameter, 4 nm high) was formed
from InAs in a GaAs matrix by a strain-induced islanding process.
Figure 18 (a) An SEM image of a typical device, showing the tapered etch and contra-
sting AlAs and GaAs distributed Bragg reflectors. (b) A photoluminescence spectrum
(solid line) and the spectral dependence of spontaneous emission lifetime at 4 K. From
Solomon et al. (2001).
A uniform ensemble of QDs with no spatial ordering was formed with a

density of about 75 dots/m2. The single quantum dot spontaneous emis-
sion lifetime was reduced from the noncavity value of 1.3 ns to 280 ps,
resulting in a single-mode spontaneous emission coupling efficiency of
78%. The device operated at 4 K.
Zwiller et al. (2001) studied a single InAs/GaAs self-assembled Stranski
Krastanov quantum dot. Under weak continuous excitation, where the
single-dot luminescence is due primarily to single-exciton recombinations,
antibunching was observed, although with a rather high value of g(2)(0) and
efficiency was rather poor. Under weak pulsed excitation, the number of
photons emitted by the quantum dot per pulse was close to one. The system
was operated at liquid helium temperature.
Single-mode solid-state single-photon source using micropillar
processed by electron-beam lithography and reactive ion etching from a
GaAs/AlAs planar microcavity grown by molecular epitaxy were also fab-
ricated at CNRS, Bagneux, France (Moreau, Robert, Gerard, et al.,
2001). A single InAs QD array was inserted at the center of a
wavelength-thick GaAs cavity spacer, surrounded by a 9-period (top) and
a 15-period (bottom) distributed Bragg reflectors. Emission was at
1.31 eV (about 1.26 m) cooling the sample to 8 K. Photon correlation
showed a dip in the g(2)(0) correlation function.
Pillar microcavities make an important advance in that they have a highly
directional far field and a strong Purcell enhancement of spontaneous
emission.
Single-photon emission from exciton, biexciton, and charged exciton

states was studied in InAs/GaAs QDs finding single-photon emission in
all cases at 5 K (Thompson et al., 2001).
A beautiful series of experiments were performed by Santoris group, at
Stanford, CA.
Optically active quantum dots confine electrons and holes to small
regions so that their energy levels are quantized. If several electrons or holes
are placed in the dot at the same time, they will, to a first approximation,
occupy single-particle states as allowed by the Pauli exclusion principle.
However, electrostatic interactions between the particles cause perturbation
in the eigenstates and energies. For example, if two electronhole pairs
(excitons) are created (a biexcitonic state), the first pair to recombine emits
at a slightly lower energy than the second pair, due to the net attractive inter-
action. Santori, Pelton, Solomon, Dale, and Yamamoto (2001) used this
effect to generate single photons not only through regulated absorption with
a train of pumping pulses but also through the emission property that the last
photon to be emitted after an excitation pulse has a unique wavelength and
therefore can be spectrally separated from the others. So they generated trig-
gered single photons using pulsed excitation of a single InAs quantum dot
and spectral filtering to remove all but the last emitted photon. The samples
were cooled at 5 K. Mesas containing single dots were identified by their
spectral emission spectra.
Single-photon generation using InAs quantum dots in pillar micro-
cavities was deeply studied by Santoris group, demonstrating indistinguish-
ability of the emitted photons (Santori, Fattal, Vuckovic, Solomon, &
Yamamoto, 2002) obtaining two-photon suppression factors as large as
40 (Vuckovic, Fattal, Santori, Solomon, & Yamamoto, 2003), and
improved efficiencies (Pelton et al., 2002). Also polarization properties were
studied (Santori, Fattal, Pelton, Solomon, & Yamamoto, 2002). Later,
Santori, Fattal, and Yamamoto (2010) published a book on single-photon
devices and applications.
We show an example of a single-photon source from the work of
Santori, Fattal, Vuckovic, Solomon, and Yamamoto (2004). The source
sample contained self-assembled InAs quantum dots (about 25 m2) embed-
ded in GaAs and sandwiched between distributed Bragg reflector mirrors,
grown by molecular beam epitaxy. Pillars (Figure 19a) with diameters
ranging from 0.3 to 5 m and heights of 5 m were fabricated in a random
distribution by chemically assisted ion beam etching. The QDs were
surrounded by the GaAs matrix and embedded in the micropillar optical
(a) (c)
XX*
GaAs/AlAs } x 12 XX
DBR
QDs
X*
X
GaAs/AlAs
} x 30 (Emitted
DBR
photon)
Empty
(b)
1 m
Figure 19 (a) Schematic diagram of single-photon device, (b) scanning electron micro-
scope image of actual pillar structures, and (c) optical excitation scheme. From Santori
et al. (2004).
cavity with a distributed Bragg structure. The optical microcavity was used
to modify the spontaneous emission properties of the QD through the
Purcell effects. When a radiative transition of the QD is on resonance with
a cavity mode, if the QD couples much more strongly with this mode than
with the background leaky modes, the spontaneous emission rate can
increase substantially and light is emitted mainly into the cavity mode.
The resulting microcavities, exhibiting three-dimensional photon con-
finement, had quality factors of approximately 1.000 and measured
spontaneous-emission rate enhancement factors as high as 5. Many pillars
with only one or two quantum dots on resonance with a fundamental cavity
mode were found. The sample was cooled to 37 K in a cryostat. To gen-
erate single photons, 3 ps pulses from a Ti:sapphire laser every 13 ns were
focused onto the pillars from a steep angle. The quantum dot emission
was collected and a single polarization was selected and then spectrally
filtered.
The operation scheme is shown in Figure 19c. A short (23 ps) optical
pulse generated by a tunable Ti:sapphire laser raises the quantum dot into an
excited state containing one electronhole pair. The QD then quickly
relaxes (with a timescale of the order of 10 ps) to a lowest excited state. This
state then decays through a much slower spontaneous emission process
(100300 ps) to emit a single photon.
The spontaneous emission is collected and sent through a narrowband

(0.1 nm) spectral filter. This not only removes background emission from
the sample but also protects against events in which the quantum dot
receives multiple excitations.
In these events, multiple photons are emitted, but each photon has a
unique wavelength, as a result of the electrostatic interactions between par-
ticles inside the quantum dot, leading to energy shifts of the order of meV
(Santori, Solomon, Pelton, & Yamamoto, 2002).
The efficiency was studied through photon correlation with a Hanbury
Brown and Twiss-type setup. The emission from the QD was spectrally fil-
tered and split into two paths by a beamsplitter, each path leading to a photon
counter. Coincidence-counting electronics generates a histogram
(Figure 20) of the relative delay t2 t1 between photon detection events
at counters 1 and 2. The peak at 0 corresponds to events in which two
(a)
300
Counts
200
100
0
0 50 100 150
(b)
1000
Counts
500
0
0 50 100 150
(c)
600
Counts
400
200
0
0 50 100 150
t = t1 t2 (ns)
Figure 20 The figure shows photon correlation g(2)() measured with (a) above-band
excitation at 750 and 0.4 nm collection bandwidth; (b) resonant excitation at 909
and 0.4 nm collection bandwidth; and (c) resonant excitation at 904 and <0.1 nm col-
lection bandwidth. The narrower peaks in (c) are mainly due to the improved time res-
olution of the setup used for that measurement. The g(2)(0) value in (c) was 0.05. From
Santori et al. (2004).
photons were detected in the same pulse, and thus the first goal in develop-
ing a single-photon source is to make the area of this peak as small as possible.
The peaks at times nTrep, where Trep 13 ns is the laser repetition
period, correspond to events in which one photon was detected from each
of two different pulses.
In Figure 20a, the photon correlation data obtained from a dot under
pulsed, above-band, excitation at 750 nm are shown. The light was collected
from a bright emission line at 932 nm. Figure 20b and c is obtained for res-
onant excitations at 909 and 904 nm, respectively. The performance was
considerably improved.
The two-photon suppression relative to an equivalent Poisson source
was estimated by comparing the area of the central peak with the areas of
the more distant side peaks. In this way, the probability of emitting two pho-
tons in the same pulse was approximately g(2)(0) 0.3 times that for a
Poisson-distributed source of the same intensity.
Most quantum information applications of single-photon sources require
pure photon states because they employ two-photon interference in their
schemes. The simplest example of two-photon interference is the Hong
OuMandel (Hong, Ou, & Mandel, 1987) experiment; if two independent
photons with identical wave packets collide at a 50:50 beamsplitter, they
always exit together, randomly choosing a side. They never exit in opposite
directions. However, if the photons do not have identical wave packets, they
can behave independently, and the two-photon interference effect is
reduced. The indistinguishability of the photons produced by the micro-
pillars was studied by Santori, Fattal, Vuckovic, et al. (2002). Figure 21
shows the main elements of the setup.
The single-photon source (InAs quantum dot) was excited twice every
13 ns by a pair of equally intense pulses with 2 ns separation. Two pulses,
each containing zero or one photons, emerge and are split into two arms
by a beam splitter, with one arm (2ns + t) longer than the other. The beams
then recombine at a different place on the same beam splitter and are col-
lected by two photon counters t1 and t2 as shown in Figure 21a. The
resulting histogram is shown in Figure 21b. Five peaks appear for every pair
of photons. For peaks 1 and 5 at 4 ns, the first photon follows the short
arm and the second photon the long arm and one photon goes to each coun-
ter. For peaks 2 and 4 at 2 ns, both photons follow the same arm. Peak
labeled 3 at 0 corresponds to the first photon traveling through the long
arm and the second through the short arm, so that the two photons collide
upon their second pass through the beam splitter. Only in this case can
Figure 21 (a) The main elements of the setup. (b) Histogram of the photon counts. From
Santori, Fattal, Vuckovic, et al. (2002).
two-photon interference occur, and for perfect two-photon interference,

peak 3 must vanish. The experimental data showed a satisfactory albeit
not complete indistinguishability.
These results showed photon state purities such that the mean wave
packet overlap between consecutive photons was as high as 0.8 (Santori,
Fattal, Vuckovic, et al., 2002).
Interference from dissimilar photon sources, provided the measurement
cannot distinguish between the photons, was shown by Bennett, Patel,
Nicoll, Ritchie, and Shields (2009) using a nonclassical emission from an
electrically excited InAs quantum dot and a commercially available semi-
conductor laser. Two-photon interference received further attention con-
sidering single photons from two separated semiconductor quantum dots
(Flagg et al., 2010; Patel et al., 2010).
IIIV semiconductors were employed in many different systems.
Becher et al. (2002) obtained a deterministic source for single-photon
pulses with nearly 100% efficiency based on pulsed laser excitation of a
single-self-assembled InAs quantum dot, placed in a semiconductor micro-
disk cavity, at helium temperature. The single-photon emission at the
single-exciton ground-state transition was ensured by the anharmonicity
of the multiexciton spectrum in combination with slow relaxation of a

highly excited mode of a semiconductor microcavity. For a QD far detuned
from any cavity resonance, nearly 100% of the excitation pulses lead to the
emission of a single photon with a repetition rate of 82 MHz. Other exper-
iments were done by Kiraz et al. (2002).
Hours et al. (2003) reported results on single QDs of GaAs/GaAlAs
under pulsed nonresonant excitation at 10 K. A few QDs were isolated
by etching mushroom-shaped microdisk structures in the sample. Excitation
was provided with a pulsed Ti:sapphire laser delivering 1.5 ps pulses at a rep-
etition rate of 82 MHz at 700 nm, and the emission occurred at 734.6 nm.
Compared with the coherent light pulses delivered by an attenuated laser
(having the same average photon number per pulse), the probability of emit-
ting a pair of photons was reduced by a factor of 5.
Most of the realizations we have discussed so far employed dots that form
spontaneously, e.g., via strain-driven StranskiKrastanov growth or colloi-
dal nucleation. Under such growth conditions, the QDs are positioned ran-
domly on the substrate and their size exhibits a very large dispersion. As a
consequence of the random and relatively dense positioning of the QDs
on the substrate, it is necessary to further process the sample, e.g., by depos-
iting a metallic mask with small apertures, or by etching mesas, in order to
isolate spatially a few QDs. In addition, because of the dispersion in size of
the QDs, the emission spectra are spread over the relatively wide spectral
range, and spectral filtering is also needed in order to select luminescence
from a single dot, when several dots reside within the spatial aperture.
A system to grow the QDs at prescribed positions was described by Baier
et al. (2004) who observed antibunching at 10 K, from InGaAs/AlGaAs
pyramidal QDs grown on a prepatterned substrate. The prepatterning pro-
vided seeds for the nucleation of QDs at prescribed positions, which greatly
facilitates the positioning of the single-photon emitters. Moreover, the
reproducibility of the QDs structures makes it possible to design and obtain
QDs with well-defined electronic states, which is essential for preparing
them in excitonic states suitable for single-photon emission.
Benyoucef et al. (2004) demonstrated efficient generation of triggered
photon pairs by placing a single quantum dot into a micropillar cavity.
The pillar microcavity sample was grown by molecular beam epitaxy on
a (100)-oriented undoped GaAs substrate. A GaAs cavity layer was sand-
wiched between the bottom and top distributed Bragg reflectors (23 and
20-period AlAs/GaAs layers). A 1.4-nm-thick single layer of self-assembled
InGaAs/GaAs QDs was inserted at the center of the cavity (QDs density
3 1010 cm2) (Figure 22). Figure 23 shows an SEM micrograph of two

micropillars the larger with a diameter of 6 m and the smaller with diameter
0.3 m with a height of 3.4 m. The devices operated at 4 K. The small
micropillar acted as a good single-photon source.
Antibunching from a single InGaAs/GaAs quantum dot, at temperatures
up to 120 K, was obtained in the near infrared (Mirin, 2004).
Single InAs/GaAs quantum dots in photolithographically defined pillar
microcavities showed a fourfold enhancement in the radiative decay due to
the Purcell effect and a photon collection efficiency into a lens of up to 10%
(Bennett, Unitt, Atkinson, Ritchie, & Shields, 2005) operating at 5 K.
GaAs (67 nm)
20
DBRs
AIAs (79 nm)
Self-assembled
InGaAs/ GaAs
GaAs (130 nm)
QDs (1.4 nm) 23

GaAs (130 nm)
AIAs (79 nm)
GaAs (67 nm)

DBRs
AIAs (79 nm)
GaAs-buffer (400 nm)
GaAs-substrate
Figure 22 Schematic representation of the structure of the microcavity. From Benyoucef

et al. (2004).
Figure 23 Scanning electron microscope (SEM) micrograph of two micropillars of dif-

ferent diameter (6 m in front and 0.3 m in the right back). From Benyoucef et al. (2004).
Using elliptical pillar microcavities, Unitt, Bennett, Atkinson, Ritchie,

and Shields (2005) and Unitt, Bennett, Atkinson, Cooper, et al. (2005) stud-
ied single-photon emission at 5 K, from excitons confined in single InAs
quantum dots, demonstrating a polarization dependence of the Purcell
effect. Elliptical pillars display two optical modes with orthogonal linear
polarizations aligned with the major and minor axes of the pillar, and this
asymmetry allowed to select the linear polarization.
Englund et al. (2005) observed large spontaneous emission rate modifica-
tion of individual InAs QDs in a 2D photonic crystal with a modified high-Q
single defect cavity and demonstrated photon antibunching. They showed
that by designing a suitable photonic crystal structure, the spontaneous emis-
sion rate of embedded self-assembled InAs quantum dots can be significantly
increased or decreased. They obtained eight times enhancement of sponta-
neous emission with respect to QDs in bulk GaAs. The structure was similar
to a previous one (Vuckovic et al., 2003). Temperature was taken between 6
and 40 K. The effect of the control of the local density of modes in a two-
dimensional photonic crystal on the spontaneous emission of self-assembled
quantum dots was studied by Kress et al. (2005). A similar study was made by
Laurent et al. (2005) who also demonstrated the indistinguishability of the
produced photons with a mean wave-packet overlap as high as 72%.
Chang, Chen, et al. (2006) realized an efficient single-photon source
based on low-density InGaAs quantum dots (about 3/m2) in a photonic
crystal microcavity. The nanocavity was formed by introducing three miss-
ing air holes in a line (Figure 24a) in a 2D triangular PC lattice. Operation
temperature was 58 K.
A strongly coupled single InGaAs QD with a pillar microcavity in res-
onance, at temperatures between 6 and 40 K, was studied by Press et al.
(2007) (Figure 25). At resonance g(2)(0) 0.18 was obtained. When the
quantum dot was spectrally detuned from the cavity mode, the cavity emis-
sion remained antibunched confirming that only one emitter was responsi-
ble for the emission. The device demonstrated an on-demand single-photon
source operating in the strong coupling regime, with a Purcell factor of
61 7 and quantum efficiency of 97%.
The coupling of single QDs to nanocavity modes is technologically very
challenging, requiring precise spatial positioning of the dots and spectral tun-
ing of their emission frequency to match the cavity mode. On the contrary,
the use of photonic bandgap structures greatly simplifies the problem. Even
randomly positioned QDs inside a two-dimensional PC are already very
efficient.
(a) (b)
WL
Cavity
PL intensity (a.u.)
s=0
s
r a 0.10a
y 0.12a
500 nm
0.15a
x 900 930 960 990
z
Wavelength (nm)
(c) (d)
Resonant wavelength (nm) 1000
980
960
940
920
900
0.00 0.05 0.10 0.15 0.20
1.0 0 +1.0 s (a)
Figure 24 (a) Scanning electron micrograph of the fabricated nanocavity with
a 300 nm, r 0:31a, and s 0:1a. (b) PL spectra for a series of nanocavities with
s 00:15a under Pex 5 kW/cm2. (c) The calculated electric field profile (Ey) of the shal-
low donor-type cavity mode confined in the L3-defect cavity with s 0:1a. The color
scale is normalized to maxjEyj. (d) The calculated (solid line) and the measured (solid
squares) resonant wavelengths of the cavity mode as a function of s. The inset in
(d) shows the calculated far-field radiation pattern of the cavity mode. From Chang,
Chen, et al. (2006).
Gevaux et al. (2006) reported on the control of the spontaneous emission

rates in InAs self-assembled quantum dots weakly coupled to the mode of a
defect cavity in a two-dimensional photonic crystal slab. Temperature
changes were used to spectrally tune the quantum dot exciton and the mode
of the photonic crystal defect cavity observing both an enhancement and
suppression of the spontaneous emission rate.
Zinoni et al. (2006) using an InAs/GaAs single quantum dot, embedded
in a planar microcavity, emitting in the 1300-nm telecom band, identified
the exciton and biexciton transitions. By optimizing the extraction effi-
ciency into a single-mode fiber, antibunching was obtained.
Figure 25 Scanning electron micrograph of a 1.2-m diameter pillar cavity used by

Press et al. (2007).
At the wavelength range 1.31.5 m of interest for telecommunications,

Takemoto et al. (2004) succeeded to obtain efficient generation of single-
photon pulses from an InAs/InP quantum dot. The group (Takemoto
et al., 2007) fabricated a special structure, which they defined as a single-
photon horn to achieve high photon extraction efficiency.
Kaniber et al. (2008) reported highly efficient single-photon generation
from InGaAs self-assembled QDs emitting at 15 K within a two-
dimensional photonic bandgap, consisting of a triangular lattice of air holes
(Kaniber et al., 2007). When compared to QDs in bulk GaAs, the incorpo-
ration of QDs into a PC nanostructure enhances the external quantum effi-
ciency by a factor close to 16 times and results in a directly measured absolute
extraction efficiency for QDs in the PC of QDPC 25% (in bulk GaAs it is
QDbulk 1.4%).
Rakher, Stoltz, Coldren, Petroff, and Bouwmeester (2008) demon-
strated a polarization-switchable single-photon source by embedding a self-
assembled quantum dot in a high-quality, electrically gated, oxide-apertured
micropillar cavity. Due to the noncircular aperture, the polarization degen-
eracy of the fundamental cavity mode is lifted, leaving two linearly polarized
Q 20,000 modes separated by 194 eV. An intracavity electric field,
generated by an applied bias, enables Stark shifting tuning of the quantum
dot emission over a frequency range containing both polarization modes,
switching the dominant single-photon polarization through the Purcell
effect. The group measured polarization switching up to 300 kHz,
limited by the RC time constant of the device. Emission was around

1.315 eV at 4.2 K.
Red single-photon emission from an InP/GaInP QD embedded in a pla-
nar monolithic microcavity at 5 K was obtained by Robach, Reischle,
Beirne, Jetter, and Michler (2008).
The usual approach to achieve a near-unity light extraction efficiency,
defined as the fraction of photons collected by the first lens, is to embed
the quantum dots in microcavities and utilize the Purcell effect when the
exciton emission is resonant with a cavity mode. A different approach has
been used by Claudon et al. (2010) who embedded an InAs quantum dot
in a GaAs photonic nanowire. Using optical pumping, a source efficiency
of 0.72, combined with pure single-photon emission, was obtained.
A SEM image of the device is shown in Figure 26. The structure used by
Claudon was a nanowire 2.5 m long and with a diameter of around
200 nm. The nanowire is provided with a reflecting gold surface at bottom,
and to increase the extraction efficiency, a conical tapper has been realized at
the top.
The same approach was used by Reimer et al. (2012) who, to increase
the light extraction efficiency, put the quantum dot emitter in a tailored
nanowire waveguide using bottom-up growth approach through indepen-
dent control of both the nanowire shape and quantum dot location,
obtaining perfectly positioned single quantum dots on the axis of the
Figure 26 An InAs quantum dot is embedded in the central part of a GaAs photonic
nanowire. The far-field emission collection is optimized with an integrated modal mirror
and a smooth tapering of the wire tip. From Claudon et al. (2010).
nanowire waveguide. Under the appropriate growth conditions, the authors

succeeded in positioning a single InAsP quantum dot exactly on the axis of
an InP nanowire waveguide with a very small tapering angle. In comparison
to quantum dots in nanowires without waveguide, they demonstrated a
24-fold enhancement in the single-photon flux, corresponding to a light
extraction efficiency of 42%. The g(2)(0) value (0.36) was however not so
exciting. Operation temperature was 10 K. The light excitation was at
532 nm and emission was obtained around 1.4 eV.
The coupling between a quantum dot and the cavity was studied as a
function of detuning from cavity resonance using site-selectivity epitaxy
to determinalistically control the nucleation site of a single quantum dot
complex in a photonic crystal microcavity (Dalacu et al., 2010).
Davanco, Rakher, Schuh, Badolato, and Srinivasan (2011) studied a
nanostructure composed of partially etched annular trenches in a suspended
GaAs membrane which was able to extract photons from single InAs quan-
tum dots with a measured collection efficiency of 10% at 8 K.
Systematic excitation power and temperature-dependent measurements
on the emission lines of single self-assembled InP/AlGaInP quantum dots
embedded in micropillars under pulsed excitation in a temperature range
from 4 up to 130 K showed that single-photon emission was present up
to 100 K (Bommer et al., 2011).
Ugur et al. (2012) obtained InP quantum dots within an InGaP matrix
with densities as low as 1 dot m2 using ultra-low-rate epitaxy and observed
both excitonic and biexcitonic emission at 4 K. Antibunching with g(2)(0)
0.2 was obtained.
A review paper on InAs/InGaAs appeared recently (Lodahl &
Stobbe, 2013).
Single photons were generated by He, Wei, et al. (2013) with high effi-
ciency and indistinguishability from self-assembled InGaAs quantum dots
embedded in a planar microcavity cooled at 4.2 K, pumped with 3 ps laser
pulses. The generation of highly indistinguishable and bandwidth-tunable
single photons from spin-flip Raman transitions under both cw and pulsed
excitation using the same self-assembled InGaAs quantum dots at 4.2 K
embedded in a planar microcavity, using a magnetic field to obtain a double
system, was also described (He, He, et al., 2013). The frequency,
linewidth, and lifetime of the Raman photons were tunable by varying
the driving field power and detuning.
The emission of photons with different energies has been studied (Rivas
et al., 2014) using single-photon emission from both trion and neutral exci-
ton states in single InAs quantum dots.
The green photons emitted by structures with IIVI semiconductors

have been studied in a number of papers. IIVI colloidal nanocrystals
(Brokmann, Giacobino, Dahan, & Hermier, 2004; Brokmann, Messin,
et al. 2004; Michler, Imamoglu, et al., 2000) offer room-temperature oper-
ation, but in general suffer from photobleaching and fluorescence intermit-
tency which causes serious limitation for their practical use. Brokmann had
nearly perfect antibunching both in continuous and in pulsed generation and
studied fluorescence intermittency.
It was found that embedding quantum dots in single nanowires produces
intense single-photon sources, typically an order of magnitude brighter than
self-assembled quantum dots (Borgstrom, Zwiller, Muller, & Imamoglu,
2005). Single carbon nanotube optical spectroscopy was studied using
single-molecule fluorescence spectroscopy (Hartschuh et al., 2005). Quan-
tum dots embedded in nanowires were the object of a number of studies
(Friedler et al., 2009; Heinrich et al., 2010).
In the green region, photon antibunching in single CdSe/ZnS quantum
dots was studied by Lounis, Bechtel, Gerion, Alivisatos, and Moerner (2000)
at 575 nm. Single-photon emission has been observed up to 100 K on
CdSe/Zn(S,Se) self-assembled QDs inside mesa structures (Sebald et al.,
2002), but the quantum efficiency and multiphoton reduction efficiency
at higher temperature were low, and working was possible up to 220 K
for ZnSe/CdSe QDs inside a nanowire (Tribu et al., 2008).
Single-photon generation with single CdSe/ZnSe quantum dots grown
by molecular beam epitaxy under pulsed excitation from a frequency-
doubled Ti:sapphire laser emitting 150 fs pulses with a repetition rate of
76 MHz was obtained at liquid helium temperature in the visible spectral
range 510690 nm (Aichele, Zwiller, Benson, Akimov, & Henneberger,
2003). The results were also extended to InP (Aichele, Zwiller, &
Benson, 2004).
Large ordered arrays of single-photon source based on CdSe QDs were
made by Zhang et al. (2008) obtaining single-photon emission at room tem-
perature and quite a good antibunching.
Antibunching and blinking of single CdSe/ZnS QDs were investigated
by XingSheng, Xu, Chen, and Yamada (2010), finding that the radiative rate
of photoluminescence and the antibunching effect are significantly changed
at different irradiation duration and under different excitation powers.
A completely different approach was realized using the radiative recom-
bination of single nitrogen-bound excitons in a ZnSe quantum well struc-
ture (Strauf et al., 2002). The nitrogen atoms were embedded in mesa etched
ZnSe quantum well structures and excited by nonresonant optical pumping
at 420 nm at 4 K. Electronhole pairs generated in the semiconductor were

subsequently captured by the nitrogen acceptors. The spectrum showed sev-
eral spectrally sharp lines due to recombination at individual nitrogen atoms,
and antibunching was obtained in cw operation. In addition, triggered
single-photon emission was obtained under pulsed excitation at 82 MHz.
Coupling of CdTe/ZnTe quantum dot emission to micropillar cavity
eigenmodes was demonstrated also by Jakubczyk et al. (2012) who hoped
the system could evolve in a room-temperature single-photon source.
Structures that used liquid crystals to host quantum dots of CdSe have
been studied by Lukishova et al. (2009).
Nitride-based QDs possess various unique properties which distinguish
them from other materials. Group III nitrides have the prospect that
the emission energy in the InGaN/GaN system can be engineered over
a large spectral range by changing their alloy contents. Single-photon
emission on GaN QDs operating at 200 K has been demonstrated in
GaN/AlN studied from 3.5 K to room temperature by Santoris group
in the range 260450 nm (Kako et al., 2006). The dot was excited at
266 nm in both continuous and pulsed regime, finding an autocorrelation
g(2)(0) 0.32 in cw operation. In pulsed excitation, g(2)(0) was 0.42 at
3.5 K and 0.53 at 200 K.
Jarjour et al. (2007) obtained single-photon emission for InGaN/GaN
QDs at 4 K with the QDs embedded in a distributed Bragg reflector struc-
ture grown on a GaN pseudosubstrate, and Kremling et al. (2012) obtained
single-photon emission from InGaN/GaN quantum dots up to 50 K.
Eventually, Holmes, Choi, Kako, Arita, and Arakawa (2014) succeeded in
demonstrating a source made by GaN quantum dots embedded in nanowires
operating at room temperature as single-photon sources with photon statistics
insensitive to temperature. The structure was grown by selective area metal-
organic chemical vapor deposition. The g2(0) value was 0.13 at 300 K.
Single-photon sources with single quantum dots were reviewed quite
recently (Shan, Yin, Shek & Huang, 2014).
8.2.5 Individual Nitrogen-Vacancy Color Centers in Diamond and Other

Defect Centers
This type of emitter, which shares many similarities with the emission from
molecules, has important practical advantages, since it can be operated at
room temperature and is perfectly photostable for both cw and pulsed ns
excitation.
A recent paper reviews diamond-based single-photon emitters

(Aharonovich et al., 2011).
Color centers are ubiquitous features of all crystalline materials. Diamond
hosts more than 500 luminescent centers, emitting from the deep ultraviolet
to the far infrared (Zaitsev, 2001). The luminescence from the color centers
in diamond is often associated with impurities, such as nitrogen or transition
metals. However, the chemical structure of only very few diamond color
centers is well understood. In some cases, individual color centers are spa-
tially separated in a rigid diamond structure and exhibit sufficient brightness
so that they can be observed and addressed at room temperature. These indi-
vidual color centers are studied as an independent quantum system that is
capable of generating single photons. Out of 500 about 10 so far have been
identified as bright, stable photon emitters.
Diamond color centers present the unique advantage of photostable
emission, room-temperature operation, and long coherence time up to
2 ms at room temperature.
The nitrogen-vacancy (NV) defect center is one of the most common
and studied defect centers in diamond (see, for example, Gruber et al.,
1997 for a first study of these defects). It is formed through the incorporation
of a single substitutional nitrogen atom in the diamond lattice located adja-
cent to a vacant carbon lattice site in one of the h111i crystallographic ori-
entations. It is well localized, is stable against photobleaching even at room
temperature, and can be addressed in simple experimental configurations. It
can display low inhomogeneous broadening. Its fluorescence can be excited
pumping at 532 nm, and the emission spectrum is made up of a zero-phonon
line at 637 nm and phononic sidebands extending to approximately 800 nm.
Fluorescence time is 11.6 ns. The efficiency is limited by the existence of a
shelving level. The high refractive index of bulk diamond restricts the effi-
ciency to collect the fluorescence light (typically a few 0.1%).
Usually, these centers are prepared in type Ib synthetic diamond, where
single substitutional nitrogen impurities are homogeneously dispersed. To
observe bright luminescence from the sample, additional vacancies are often
created by electron or neutron irradiation (Davies & Hamer, 1976; Hanzawa
et al., 1996), and diffusion of nitrogen atoms is obtained by annealing at
900 C. However, already untreated samples of synthetic Ib diamond pro-
vide a concentration of NV centers well suited for addressing individual cen-
ters. More recently, also femtosecond pulse laser illumination has proven
useful (Liu et al., 2013).
Other centers of interest for photon emission are nickel related, silicon,
carbon, chromium color centers, and some other unknown defects.
Brouri, Beveratos, Poizat, and Grangier (2000b) reported on the use of
single color centers in diamond as single quantum emitters. They used bulk
diamond at room temperature with nonresonant excitation from an argon-
ion laser at 514 nm with a typical power of 10 mW. The used color center
was the NV defect with a zero-phonon line at 637 nm. Impurities were
introduced irradiating with 2 MeV electrons and annealing. The irradiation
was chosen such that the density of vacancies was of the order of 1 m3.
A limitation of the system is the existence of shelving in a metastable singlet
state. Shelving leads to the observation of photon bleaching for a timescale
longer than the lifetime of the center and to a decrease of the fluorescence
rate owing to the time spent by the center in this long-lived states. This com-
plex model was discussed by Brouri, Beveratos, Poizat, and Grangier
(2000a). A simplified level structure is a three-level scheme with fast non-
radiative decay within the two upper states and within two lower states
(Figure 27). The fluorescence emission is between the 2 and 1 level. Level
3 is the shelving level. The model is an oversimplification of what is the real
physical system. Shelving in some cases may be avoided.
Figure 28a shows a confocal microscopy raster scan (5 m 5 m) of the
sample made about 10 m below the diamond surface. Figure 28b is a line
scan along the dashed line in (a). Individual bright spots that correspond to
NV centers appear clearly.
Figure 29 shows the normalized correlation function g(2)(t) corrected for
the random coincidences from the background. The data are for the NV
center circled in Figure 28a.
The total absence of photobleaching allowed Brouri to lock the laser
exciting beam onto a single center during several hours. The evaluated over-
all detection efficiency was low 0.0014.
|2 rd
bG |3
r G GM
|1
Figure 27 A three-level model for an NV center. Fluorescence occurs from j2i to j1i. The
j3i level is a metastable level. From Brouri et al. (2000a).
5
(a)
4
m
2
0
0 1 2 3 4 5
m
(b)
7000
Count rate (s1)
6000
5000
4000
0 1 2 3 4 5
Position (m)
Figure 28 (a) Confocal microscope scanning; (b) counts of the scanning along the dot-
ted line shown in (a). From Brouri et al. (2000b).
c(t)
g ( 2 )(t )
450
2
400
1
0 350
20 0 20 40
t (ns)
Figure 29 Normalized correlation function corrected for the random coincidences from
the background. The data are for the center circled in Figure 28a. From Brouri et al.
(2000b).
Single NV centers in diamond were used by Kurtsiefer, Mayer, Zarda,

and Weinfurter (2000). The diamond samples of 500 500 250 m were
illuminated by the light of a diode pumped frequency-doubled Nd:YVO4
laser at 532 nm. The light was focused through a microscope objective into
the diamond crystal and the fluorescence light was extracted with the same
microscope objective and focused into a single-mode optical fiber. The
zero-phonon line at 637 nm was clearly visible, even at room temperature.
However, additional phonon contributions were also present. The emitted
light exhibited antibunching. The correlation function g(2)(t) was measured
at different excitation powers obtaining a minimum value g(2)(0) 0.26 for a
pump power of 5 mW. Increasing the excitation power g(2)(0) increased. For
extremely high excitation power (40 mW), the correlation relaxed to a
Poisson-like statistics.
Diamond nanocrystals containing a single NV color center have been
considered by Beveratos et al. (2001) and a room-temperature stable
single-photon source was obtained (Beveratos, K uhn, et al., 2002). They
obtained a complete single-photon quantum key distribution setup by using
that source at room temperature, realizing a single-photon quantum cryp-
tography system (Beveratos, Brouri, et al., 2002).
Figure 30 shows the autocorrelation function of the emission of a single
NV center on Alices side. The exciting laser had a repetition period of
187.5 ns, a pulse width of 0.8 ns, and an average power of 0.2 mW. The
count rate was about 3.5 104 s1 on each avalanche photodiode and the
integration time was 166 s. The number above each peak in the figure rep-
resents its normalized area. The dots are experimental data.
1.27 0.07 1.27 1.16 1.12

1200
Raw coincidences
1000
800
600
400
200
0
200 0 200 400 600
t (ns)
Figure 30 The autocorrelation function of the emission from a single NV center. From
Beveratos, Brouri, et al. (2002).
Englund et al. (2010) have succeeded to couple a photonic crystal (PC)

nanocavity to a single NV center in diamond, obtaining also antibunching.
Since NV centers can form in nanocrystals as small as 10 nm, their posi-
tion can in principle be controlled with an accuracy of a few nanometers.
Van der Sar et al. (2009) demonstrated a versatile technique to position a
single NV center contained in a diamond nanocrystal to an arbitrary loca-
tion, using a scanning confocal microscope and a nanomanipulator.
The luminescent centers in bulk diamond crystals have the disadvantage
of low photon outcoupling. Babinec et al. (2010) have demonstrated a single-
photon source composed of an NV center in diamond nanowire, which pro-
duces 10 times greater flux than bulk diamond devices, while using 10 times
less power. The diamond nanowire efficiently collects the green laser source
used to optically pump the NV center and also efficiently collect the emitted
red photons and direct them vertically toward collection optics (Babinec
et al., 2011). Confocal microscopy techniques allow to image, optically
address, and characterize individual nanowires devices. In the experiment,
a piezoelectric stage scanned the nanowire devices underneath a green laser,
focused with a high numerical aperture objective. The fluorescence gener-
ated by optical pumping was collected back through the objective, collected
on a single-mode fiber, and detected with a single-photon detector. Single-
photon emission from an individual NV in a nanowire device was verified via
the intensity autocorrelation function g(2)(t). A pronounced antibunching
g(2)(0) 0.06 was measured. The system improved efficiency with respect
to simple diamond by an order of magnitude.
A study of the single-photon emission through the g(2)(t) was made at
different pump power levels (Figure 31), showing that by increasing the
pump level, the probability of multiphoton emission grows.
The measured g(2)(t) is shown in Figure 31a. Qualitatively different
dynamics are observed at different excitation powers: (a) 11 W below sat-
uration; (b) 190 W, slightly above saturation; and (c) 1.6 mW, above sat-
uration. At high pump powers, coupling to the metastable shelving state is
significant and results in bunching shoulders (g(2)(t) > 1) at intermediate
times. Red (gray in the print version) curves in Figure 31a and b are fits using
a three-level model. The decay rate of the g(2)(t) gives the fluorescence life-
time. In Figure 31d are the results. Extrapolation gives a lifetime of 14 ns for
the studied NVnanowire system.
Hausmann et al. (2011) extended the Babinecs work, using the combi-
nation of ion implantation and top-down diamond nanofabrication in dia-
mond nanopillars and diamond nanowires.
(a) 1.50 (b) 2.00
1.50
1.00
g (2)( T )
g (2)( T )
1.00
0.50
0.50
0.00 0.00
100 50 0 50 100 100 50 0 50 100
Delay (ns) Delay (ns)
(c) (d)
3.00 0.12
2.50
2.00 0.08
1/T (ns1)
g (2)( T )
1.50
1.00 0.04
0.50
0.00 0
100 50 0 50 100 0 50 100
Delay (ns) Pump power (mW)
Figure 31 Measured correlation function for different excitation powers: (a) 11 W;
(b) 190 W; (c) 1.6 mW; and (d) fluorescent time as a function of pump power. Red
curves (grey in the print version) are fits. From Babinec et al. (2010).
Schroder, Gadeke, Banholzer, and Benson (2011), to maximize the

single-photon flux, spin-coated nanodiamonds containing single NV cen-
ters on the flat surface of a ZrO2 solid immersion lens. They found stable
single-photon count rates of up to 853 kcts/s at saturation under cw exci-
tation. Figure 32 shows autocorrelation and saturation measurements of dif-
ferent NV centers.
An electrically driven single-photon source at room temperature in dia-
mond, based on single neutral NV center, was described by Mizuochi et al.
(2012). The generation of electroluminescence followed a kinetics funda-
mentally different from that of photoluminescence with intra-bandgap exci-
tation. The second-order correlation function g(2)(t) at t 0 was around 0.5.
Diamond-based pin light-emitting diodes capable of single-photon
emission in the visible spectral region at room temperature using NV defects
were described also by Lohrmann et al. (2011). Using ion implantation of
boron and phosphorus, the author fabricated a pin microdiode hosting
an individual NV (neutral) defect in the i-area, resulting in single-photon
emission in the electroluminescence regime with promising results.
Figure 32 (a) Saturation measurement of the brightest stable emitter at a saturation

intensity 88 W. The inset shows g(2)(0) < 0.3; (b) brightest blinking emitter at
2.4 Mcts/s; (c) saturation count rates of different NV centers against the saturation exci-
tation intensity. From Schroder et al. (2011).
Figure 33 Structure of the NE8 center consisting of one Ni atom surrounded by four
N atoms. From Gaebel et al. (2004).
Liebermeister et al. (2013) demonstrated the evanescent coupling of sin-

gle fluorescence photons emitted by a single NV center to the guided mode
of a tapered optical fiber. The NV center was optically selected with a con-
focal scanning microscope and positioned onto the subwavelength diameter
of the tapered optical fiber with the help of an atomic force microscope. The
authors observed an improvement of coupling efficiency.
Gaebel et al. (2004) obtained emission using an as yet unexplored new
center: NE8 (nickelnitrogen complex), whose structure is shown in
Figure 33. In the center, the nickel atom is located at equal distance of
two carbon vacancies, and it is associated with four nitrogen atoms in its first
coordination shell. The level structure is shown in Figure 34.
The emission wavelength is around 800 nm with a bandwidth of 1.2 nm
at room temperature. The measured g(2)(t) correlation function had a very
nice dip at t 0, as shown in Figure 35.
Rabeau et al. (2005) described the fabrication of single nickelnitrogen
(NE8) defect centers in diamond by chemical vapor deposition. Under cw
1
G 12
k k 1/T1 2
g 12
0
Figure 34 Energy level structure of the NE8 center. From Gaebel et al. (2004).
Figure 35 Measured correlation function with cw excitation. The red (gray in the print
version) line is the theoretical curve. From Gaebel et al. (2004).
745 nm laser excitation, single defects were induced to emit single-photon

pulses at 797 nm with a linewidth of 1.5 nm at room temperature.
A single-photon source at room temperature using a color center in the
same wavelength range as NE8 was reported by Wu et al. (2006). Then
room-temperature operation from a single NE8 color center obtained in
a chemical vapor-deposited diamond nanocrystal was obtained (Wu et al.,
2007) with emission at 793.7 nm.
Nickel centers were obtained in a controllable manner by a focused ion
beam implantation into individual diamond nanocrystals (Aharonovich
et al., 2009) obtaining a narrow emission at 768 nm with g2(0) 0.16.
A silicon vacancy in diamond produced by ion implantation emitting at
738 nm at room temperature was described by Wang et al. (2006). Si vacan-
cies were studied also by Neu et al. (2011, 2013).
A new promising center: Cr-related center (UM2) in nanodiamond has

been studied by Aharonovich et al. (2010) with emission at 749 nm, with
half-width 4 nm.
More recently, other possibilities have been investigated. Kolesov et al.
(2012) have singled out photostable Pr3+ ions in yttrium aluminum garnet
nanocrystal using patterned ion implantation and obtaining single-photon
emission in the ultraviolet at cryogenic temperatures.
More interesting looks the first observation of an isolated defect in SiC
that exhibits single-photon emission at room temperature with good per-
spectives. The center is introduced by irradiation of ultrapure SiC with
low electron fluences and subsequent annealing. The formed center arises
from the carbon antisite-vacancy pair defect CsiVc. All defects exhibiting
photoluminescence between 648 and 710 nm behave as single-photon
sources, as verified with the usual g2(0) measurement (Castelletto et al., 2014).
8.3 Single-Photon Sources on Demand

A source that has the property to emit with a high degree of certainty one
(and only one) photon at a users specific time is what is called a photon source
on demand. Such a deterministic source of single photons was also termed
turnstile device (Imamoglu & Yamamoto, 1994).
Although an excited atom can only emit a single photon, there are a
number of additional requirements. An optimal scheme should work at
room temperature and allow to transmit the emitted photon through a desir-
able channel with high probability, and the pumping mechanism should
allow that pulse shapes and time intervals for single-photon emission could
be readily controlled. Moreover, the system should have a high recycling
efficiency for generating successive photons in a bit stream. A train of clas-
sical light pulses which produce a train of single-photon pulses was named
also photon gun.
Some of the sources described in the previous paragraphs satisfy in part to
these requirements in the sense that the emission following fast excitation
occurs in a very short time. Also the emission following excitation by a train
of regular pulses could be considered a reasonable deterministic source of
single photons. The two principal problems are the suppression of multiple
photons and the uncertainty in the emission time.
Another kind of source is obtained in the case in which the emitter sends
two photons, one of which is used to herald the presence of the other. Such
sources will be discussed in the next paragraph.
While the distinction between a deterministic and a probabilistic source

is clear in the abstract, this distinction blurs in practice. An example of this is a
source classified as deterministic which has losses in the extraction of the
photon from the region where it is generated. As that emission (or extrac-
tion) losses increase, a theoretically deterministic source becomes more
probabilistic in operation.
He and Barkai (2006) considered the theory of single photon on demand
from a two-level atom or molecule source.
The necessary control of emission may be achieved, in principle, in cav-
ity quantum electrodynamics. Although this approach has been used for the
production of single photons from atoms (see Section 8.2.3), such experi-
ments are compromised by limited trapping times, fluctuating atomfield
coupling and multiatom effects.
Also some of the sources considered in Section 8.2 satisfy in part to the
previous requirements, but there is no one which could be classified as
perfect (see Scheel, 2009; Scheel et al., 2007).
The best sources at present seem to be semiconductor QDs at low tem-
perature, single NV centers in diamond, able to operate at room tempera-
ture, and twin photons sources, as described in the next paragraph.
However, none of these sources are able to fulfil all the prescriptions we
mentioned at the beginning of this section.
Other schemes have also been considered as, for example, the generation
of nonclassical photon states based on the interaction between the radiation
field and a superconductor which can act as a deterministic single-photon
source as well as generate any Fock states (Liu, Wei, & Nori, 2004).
9. ENTANGLED STATES
A beam splitter is the simplest quantum optical device in which two
incident light beams interfere to produce two output beams. Quantum
properties of the beam splitter are manifested in the ability to generate an
entangled output state from a nonclassical but unentangled input (Gerry &
Knight, 2005; Kim et al., 2002).
Entangled states may thus be introduced by considering radiation inter-
acting with a beam splitter. Classically, if a field E1 enters from the left
(Figure 36), it is alternatively reflected (E2) or transmitted (E3), with reflec-
tion r and transmission t coefficients defined through
E2 rE1 and E3 tE1 (48)
Elementi separati (modificabili)
(o)
t
t
E1 (1) r E3 (3)
r
E2 (2)
Figure 36 A beam splitter.
To describe quantum mechanically the system we may replace the com-

plex fields Ei (i 1, 2, 3) with annihilation operators ai. In doing so, we have
also to consider the vacuum field entering the system from the fourth port.
We have therefore
~a2 r~a1 + t 0 ~ao and ~a3 t~a1 + r 0 ~ao (49)
where the coefficients with primes refer to the properties of the beam splitter
for a field coming from the right.
It is easily seen that the commutation relations for a^2 and a^3 hold, if
jrj jr 0 j, jtj jt 0 j, jrj2 + jtj2 1 and r*t 0 + r 0 t* 0, and r*t + r 0 t 0 * 0.
We may now ask for a given input state to the beam splitter, what is the
output state. Remembering that all photon number states jni (and therefore
any superposition) may be constructed by the action of n powers of the
creation operator on the vacuum, we may use Equation (49) to construct
output states from the action of the transformed creation operators on the
vacuum states of the output modes.
Consider now, for example, a single-photon input state j0ioj1i1 which
we may write as a1j0ioj0i1. For a 50/50 beam splitter we have

~a{ 1 i~a{ 2 + ~a{ 3 =2 (50)
Thus,

j0io j1i1 ! 1=2 i~a{ 2 + ~a{ 3 j0i2 j0i3

1=2 ij1i2 j0i3 + j0i2 j1i3 (51)
This result says that a single-photon incident at one of the input ports of
the beam splitter, the other part containing only vacuum, will be either
transmitted or reflected with equal probability. The output state (51) is an
entangled state: it cannot be written as a simple product of states of the indi-
vidual modes 2 and 3.
A first example of entanglement was observed experimentally at a time
before the quantum theory of matter had even be formalized. In the Franck
Hertz scattering, an incoming electron scatters from an electron bound in an
atom, say hydrogen for simplicity, and an electron is observed coming out.
There is no way to tell which electron comes out, and the quantum state
describing the two is an entangled state.
Entanglement can loosely be described as a kind of correlation. But it is
quite a bit more than that and, besides, classical systems can also exhibit cor-
relations. A correlation between two systems is simply the statement that if a
measurement of one system yields the result A, then a measurement on the
second system will yield the result B, with some probability. Perfect corre-
lation occurs when the second result is certain, given the outcome of the
first. Entangled states have some correlation property more. Let us consider
the case in which two photons are entangled in polarization

1=2 j0i1 j0i2 + j1i1 j1i2 (52)
If one only looks at one of the two particles, one finds it with equal prob-
ability in state j0i or in state j1i. The state shows classical two-particle cor-
relations in that when we measure a one (0) in channel 1, this immediately
implies a one (0) in channel 2. The quantum state however shows strong
correlation for any arbitrary superposition. For instance, if we consider a
polarization-entangled state, and put j0i for H (horizontal polarization)
and j1i for V (vertical polarization), a measurement in any polarization
direction in channel 1 will be 100% correlated in channel 2. No matter
which axis is chosen for the measurement of the polarization of one of
the photons, that choice completely determines what happens when we
measure the polarization state of the other photon (Aspect, Grangier, &
Roger, 1981; see also Grangier & Abram, 2003).
Two-photon states are the most popular entangled states in fundamental
and applied quantum optics. In a two-photon system, correlated or
entangled photon states can be generated by an atomic cascade process
(Grangier, Roger, & Aspect, 1986) or by a parametric down-conversion
process (Hong & Mandel, 1986; Rarity, Tapster, & Jakeman, 1987), where
incident photons are converted into photon pairs (signal and idler). The two
photons may propagate away from each other in different directions, and the
quantum states of each single photon may differ considerably. Despite the
distance between the subsystems, the pair will keep its correlation.
The subtle difference between classical correlation and entanglement
may be appreciated looking at the radiation emitted in a parametric gener-
ation process. In such processes, the interaction of a laser beam in a nonlinear
crystal produces simultaneously two beams dubbed signal and idler
(Figure 37). In what is described as type II down-conversion,2 the signal
and idler photons have orthogonal polarizations. In the process, energy
and momentum must be conserved, and birefringence effects cause the pho-
tons to be emitted along two intersecting cones, one of the ordinary (o) ray
and the other of the extraordinary (e) ray (Figure 38).
If one photon of the pair is detected at any time, then its partner is known
to be present.
If we measure the polarization of a photon in a point of the extraordinary
cone of Figure 38, we always measure a vertical polarization and we know
that in the corresponding point for which phase matching is fulfilled, the
polarization of the incoming photon is horizontal, and if we repeat the
experiment, the two polarizations will always be vertical on the extraordi-
nary cone and horizontal on the ordinary one.
Momentum conservation
Spontaneous
parametric ks ki
downconversion
kPUMP
s (signal)
Pump
Energy conservation
ws
Nonlinear i (idler) w PUMP
c(2) crystal wi
j PUMP = j s + j i
Figure 37 Scheme of spontaneous parametric down-conversion and momentum and
energy conservation.
2
The down-conversion process is called type I if the signal and idler photons have identical polarizations.
With type II phase matching, the signal and idler photons have orthogonal polarizations. If the angle
between the optical axis of the crystal and the pump beam direction is different from zero, the two
cones tilt toward the pump and intersect along two rays.
Figure 38 (a) The spontaneous down-conversion cones present with type II phase
matching. (b) A photograph of the down-conversion photons, through an interference
filter at 702 nm. From Kwiat et al. (1995).
There are however two points clearly visible in the front photo of
Figure 38b in which the two polarizations are contemporarily present.
A measure in one of these points will yield vertical or horizontal polariza-
tion, indifferently. In these points, if we measure a vertical polarization in
one point, we know that in the other point polarization will be horizontal;
but we could also have measured a horizontal polarization instead (with a
50:50 probability) and in the other point polarization would have been ver-
tical. The wave function of the total system of the two produced photons
will collapse in one of the two situations just by virtue of the measurement,
and before it, any of the two polarizations will equally be probable.
The structure of Figure 38b repeats itself for different colors as shown in
Figure 39.
In the case of the signal-idler pair of photons emitted in spontaneous
parametric down-conversion, the two photons may be entangled in wave-
length, momentum, angular momentum, and frequency, as well as polari-
zation. Entanglement received precocious attention in a famous Einstein
paper (Einstein, Podolky, & Rosen, 1935) in which the possibility that a
measurement on a particle in some place at some time made another particle
far away at the same time to collapse is some state was taken as an indication
that quantum theory was not complete. This nonlocal property is today rec-
ognized as the most characteristics aspect of quantum mechanics.
Entangled pairs of photons may be used to have single photons. In fact,
when we detect in a given direction one photon in the pair we know, there
is a second one in the direction fulfilling the momentum and energy con-
servation rules due to the phase-matching conditions in the generating crys-
tal. Entanglement is further discussed, for example, in Strekalov, Kim, and
Shih (1999) and Bovino and Degiovanni (2008). The coupling of the down-
converted photons into an optical fiber received attention (see, for example,
Castelletto, Degiovanni, Migdall, & Ware, 2004; references therein).
Probably, Clauser (1974) was among the first to use twin pairs of photons
and study coincidences between them, although his attention was directed
to a study of the photoelectric effect and not to single-photon behavior. In
Figure 39 The intersecting beams at different colors (different gray shades in the print
version). The credit of this photo is to Paul Kwiat and Michael Reck from the University of
Innsbruck. Reprinted with permission. Copyright Faculty of Physics, University of Vienna.
his experiment, the source contained 202Hg atoms which were excited by
electron bombardment. Light was produced at 1 567.6 nm and
2 435.8 nm by the cascade 91P1 ! 73S1 ! 63P1. The light emitted on
opposite sides of an assembly of excited atoms was focused separately into
two beams by means of a beam splitter. Each resulting beam was directed
through interference filter so that the wavelengths 1 and 2 were selected,
onto an appropriate photomultiplier and coincidences were studied. The
presence of one wavelength was used to know whether the other wave-
length was present. Clauser utilized two photons of different wavelengths,
produced in a cascade process in which one photon was produced after the
other. It was not an entangled pair. However, the presence of one photon at
one wavelength assured the existence of the other at the accompanying
wavelength.
In parametric interactions each time a pump photon creates an idler pho-
ton, it must also create a signal photon. Thus, if a photon is detected at the
idler mode, one can expect the presence of a corresponding photon at the
signal mode. This was verified experimentally in parametric fluorescence
(Hong & Mandel, 1986). In this way, single photons are created and anti-
bunching should occur. Such scheme of generating antibunched light was
studied theoretical by Stoler and Yurke (1986).
For these sources, the creation of photon pairs is probabilistic, rather than
deterministic. However, because the photons are created in pairs, one pho-
ton (the heralding photon) can be used to herald the creation of the other
photon. The second detector must be activated only whenever the first one
has detected a photon and not whenever a pump pulse has been emitted,
therefore circumventing the problem of empty pulses. Phase matching
allows one to choose the wavelength and determine the bandwidth of the
down-converted photons. The latter is in general rather large and varies
from a few nanometers up to some tens of nanometers. For the non-
degenerate case, one typically gets a bandwidth of 510 nm, whereas in
the degenerate case (where the central frequency of both photons is equal)
the bandwidth can be as large as 70 nm.
This photon-pair creation is very inefficient; typically, it takes some 1010
pump photons to create one pair in a given mode. The number of photon
pairs per mode is thermally distributed within the coherence time of the
photons and follows a Poissonian distribution for larger time windows
(Walls & Milburn, 1995).
With a pump power of 1 mW about 106 pairs per second can be col-
lected in single-mode fibers. Accordingly, in a time window of roughly
1 ns, the conditional probability of finding a second pair, having already

detected one, is 106 109 0.1%. In the case of continuous pumping,
this time window is given by the detector resolution. Accepting, for exam-
ple, 1% of these multipair events, one can generate 107 pairs per second
using a realistic 10 mW pump. One must observe, however, that also in
down-conversion there is a probability that two pairs be emitted
contemporarily.
First sources of high-intensity polarization-entangled photons were pro-
duced by Kwiat et al. (1995) and Kwiat, Waks, White, Appelbaum, and
Eberhar (1999).
A number of papers discuss the down-conversion generation with
nonlinear crystals. A compact all-solid-state source of polarization-entangled
photon pairs at 856 nm was realized using a frequency-doubled laser diode as
a pump source for cavity enhanced type II spontaneous parametric down-
conversion in BBO crystal. About 10,000 coincidences per second, with
6.5 mW pump power at the second harmonic, were obtained (Volz,
Kurtsiefer, & Weinfurter, 2001). Other schemes and proposals were pres-
ented ( Jeffrey, Peters, & Kwiat, 2004).
De Dood, Irvine, and Bouwmeester (2004) showed that phase-matching
conditions used in schemes to create entangled photon pairs can be achieved
in photonic crystals. Similar simulations were made by Centini et al. (2005)
and Sciscione, Centini, Sibilia, Bertolotti, and Scalora (2006). The theory of
spontaneous parametric down-conversion from photonic crystals was done
by Vamivakas, Saleh, Sergienko, and Teich (2004).
URen, Silberhorn, Banaszek, and Walmsley (2004) designed a type II
parametric down-conversion interaction in a quasi-phase-matched KTP
waveguide leading to much improved conditional detection rates and to eas-
ily separable (by means of their polarization) photon pairs.
Most parametric down-conversion-based schemes require that the out-
put be collected into a single spatial mode defined by an optical fiber and this
point was addressed, for example, by Bovino et al. (2003).
Polarization-entangled photons through a spontaneous four-wave
mixing in a single-mode optical fiber were described by Fan, Eisaman,
and Migdall (2007).
Indistinguishable heralded photons were generated in two independent
spectrally engineered sources by Mosley et al. (2008). Goldschmidt,
Eisaman, Fan, Polyakov, and Migdall (2008) experimentally demonstrated
a wavelength-tunable heralded single-photon source based on spontaneous
four-wave mixing in a single-mode microstructure fiber.
Branczyk, Ralph, Helwig, and Silberhorn (2010) discussed theoretically

the generation of heralded pure Fock states via spontaneous parametric
down-conversion in periodically poled waveguides.
Da Cunha Pereira et al. (2013) demonstrated a symmetric, single spatial
mode single-photon heralding efficiency of 84% for a type II spontaneous
parametric down-conversion process. These are only a few examples.
Be as it is, the parametric down-conversion has two main drawbacks.
The first is, as we already said, the casual generation of the pair. The second
is that sometimes more than one pair can be generated. If a pulsed laser is
used to pump the nonlinear crystal, generating the photon pair only at cer-
tain times, the probability of generating n pairs of photons per pulse is a
Poisson distribution (Equation 42) where the mean number of pairs hni
depends on the pump power and parameters of the crystals. An ideal source
would have exactly one photon generated from each pulse, which requires
hni 1. However for hni 1, the probability to observe more than one pho-
ton is 0.26 due to the Poisson distribution. This is a strong limitation that has
been suggested to be overcome using spatial or temporal multiplexing of
several sources (Kok, 2010; Ladd et al., 2010; McCusker & Kwiat, 2009;
Migdall, Branning, & Castelletto, 2002; Pittman, Jacobs, & Franson,
2002; Shapiro & Wong, 2007). By utilizing a four-photon-pair source,
an active feed-forward technique, and an ultrafast single-photon router,
Ma, Zotter, Kofler, Jennewein, and Zeilinger (2011) obtained a fourfold
enhancement of the output photon rate, simultaneously maintaining the
quality of the output single-photon state.
An extremely low-noise heralded single-photon source was obtained by
Brida et al. (2012) characterized by a g(2)(0) 0.005(7) and a ratio of the
background photons to the total photons counted by the two detectors equal
to 0.25%.
More recently, generation of heralded single photons in a silica photonic
chip by spontaneous four-wave mixing has been demonstrated (Spring et al.,
2013) with a heralding efficiency of 40% and a single-photon purity of 0.86.
Harder et al. (2013) implemented an ultrafast pulsed type II parametric
down-conversion source in a periodically poled KTP waveguide at around
1.53 m with almost identical properties between signal and idler.
For a review on the sources of photon entanglement, see Torres,
Banasek, and Walmsley (2011).
Instead of using interaction in nonlinear crystals, one may consider a dif-
ferent approach using quantum dots.
A quantum dot biexciton (two electrons and two holes, 2X) decays
radiatively through two intermediate optically active exciton states. First,
it decays to one of two optically active single-exciton states (one electron
and one hole, 1X) by emitting one photon and then to the empty-dot state
by emitting a second photon (Moreau, Robert, Manin, et al., 2001;
Regelman et al., 2001). The proposal that the biexciton-radiative cascade
could provide a source of event-ready polarization-entangled photon pairs
was made by Benson, Santori, Pelton, and Yamamoto (2000).
A first experimental study of the polarization correlation properties of
photon pairs emitted through biexciton decay in a single InAs quantum
dot was carried out by Santori, Fattal, Pelton, et al. (2002) with no positive
result, suggesting that quantum-dot asymmetry is an obstacle to realizing an
entangled photon source. Entanglement requires two decay paths with dif-
ferent polarizations, but indistinguishable otherwise. This is the case if the
intermediate exciton states are energetically degenerate and if the final state
of the QD is independent of the decay path. We may see how to achieve this
following Akopian et al. (2006). Let us consider Figure 40. In Figure 40b,
the biexciton-radiative cascade is shown with its two collinearly polarized
photons (either H or V) and distinguishable paths. As shown in
Figure 40b, the polarizations of the two photons are determined by the
intermediate exciton states (X) which are typically split into energy due
to electronhole exchange interaction and QD confining potential anisot-
ropy. Therefore, the two decay paths are spectrally distinguishable and
therefore the polarization states cannot be entangled. Akopian et al.
(2006) showed that the which path information can be erased by spectrally
selecting only the overlap of the two polarization components, and so pho-
ton entanglement can be achieved. If the splitting is smaller than the radiative
linewidth, the two decay paths are indistinguishable, producing two verti-
cally (horizontally) polarized entangled photons. The experiment was per-
formed using InAs QDs at 20 K. In a further paper, Akopian et al. (2007)
measured the density matrix of the polarization state of the photon pair emit-
ted from a continuously pumped quantum dot and showed that by applying
a temporal window, the quantum dot becomes a source of entangled light.
Young et al. demonstrated a similar system using InAs and carefully
selecting the quantum dots below 10 K. They succeeded to obtain more
than 70% of the detected photons to be entangled at 10 K, eliminating
the polarization splitting by a suitable design of the QD structure (Young
et al., 2006; Young, Stevenson, & Shields, 2007).
(a)
XX0 20
X1
15
20 K X0 10
5
1.274 1.276 1.278 1.28

(b)
PL intensity (102 cts/s)

(c)
H 25
2G
V
Eu XX0 15
XX0
H V 5
G
1.2762 1.2764
EV (d)
D
H
EH 10
V
X0
X0 6
H V
0 2
1.2803 1.2805
PL energy (eV)
Figure 40 (a) Photoluminescence spectrum of a single QD; (b) schematic description of
the biexciton cascade in natural QDs with its two collinearly polarized photons (either
H or V) and energetically distinguishable paths. (c, d) High-resolution polarization-
sensitive photoluminescence spectra of the biexciton XX and exciton X lines, respec-
tively. From Akopian et al. (2006).
Hafenbrak et al. (2007) demonstrated reliable polarization-entanglement

with radiative biexcitonexciton decay for elevated temperatures up to
30 K.
Muller, Fang, Lawall, and Solomon (2009) obtained polarization-
entangled photon pairs from a semiconductor quantum dot, using the opti-
cal Stark effect.
Dousse obtained entangled photons realizing two cavity resonances for
both XX and X photons inserting a QD in a micropillar cavity coupled to a
second identical but empty micropillar (they dubbed it a photonic molecule)
(Figure 41). Single pairs of entangled photons with a probability near unity
were obtained at temperatures between 5 and 52 K (Dousse et al., 2010).
Another possible system was studied by Ulhaq et al. (2012) considering
cascaded single-photon emission from the Mollow triplet sidebands, by
Figure 41 Sketch of the source: two identical pillar microcavities with diameter D are
coupled. The center to center distance is labeled CC0 . A single quantum dot is inserted in
one of the pillars, k is the photon wavevector. From Dousse et al. (2010).
using a single layer of self-assembled In(Ga)As QDs, sandwiched between

GaAs/AlAs distributed Bragg reflector layers at the center of a GaAs cavity.
Figure 42 shows the Michelson interferometer setup used to separate the
Mollow sidebands from the central line. Autocorrelation on a spectrally sep-
arated Mollow sideband exhibiting clear antibunching with a normalized
value of g(2)(0) 0.18 was obtained at 5 K. It was thus demonstrated that
an individual Mollow sideband channel of the resonance fluorescence from
a single QD can act as an efficient single-photon source which, by spectrally
separating both Mollow sidebands, can act as a solid-based cascaded photon
emitter with time-ordered photon pair emission process.
Entangled photons in quantum cryptography were discussed, for exam-
ple, by Naik, Peterson, White, Berglund, and Kwiat (2000), Jennewein,
Simon, Weihs, Weinfurter, and Zeilinger (2000), Tittel, Brendel,
Zbinden, and Gisin (2000), and Bovino et al. (2005).
Generation of a pair of photons in a fiber-based parametric down-
conversion was obtained by Fan and Migdall (2007). Later, a single photon
via a four-wave mixing process in a cavity was proposed by Fan et al. (2009),
Figure 42 (a) Michelson interferometer set. (b) Mollow spectrum obtained after filtering
out the central peak. (c) High-resolution spectrum of the central Rayleigh line filtered
from the Mollow sidebands. (d) Photon autocorrelation of the central Mollow triplet
peak. Inset resonance fluorescence emission spectrum of the full Mollow triplet.
(e) Autocorrelation on the T line of the Mollow spectrum. From Ulhaq et al. (2012).
finding a probability of producing a single-photon states of nearly 50% in

simulations. A narrowband source has been implemented by Srivathsan
et al. (2013) via a four-wave mixing in a cold atomic ensemble.
10. PLASMONIC SOURCES

Many proposals have been made to use surface plasmon polaritons
(SPPs) coupled to single-photon sources (Bulu, Babinec, Hausmann,
Choy, & Loncar, 2011; Chang, Chen, et al., 2006; Gan, Hugonin, &
Lalanne, 2012; Tame et al., 2008).
Surface plasmons, or SPPs, are propagating excitations of charge-density
waves and their associated electromagnetic fields on the surface of a conduc-
tor. These collective electronic excitations can produce strong electric fields
localized to subwavelength distance scales, which make SPPs interesting for
several applications. Remarkably, these collective electron states preserve
many key quantum mechanical properties of the photons used to excite
them, including entanglement (Altewischer, Van Exter, & Woerdman,

2002; Fasel et al., 2005) and sub-Poissonian statistics (Akimov et al., 2007).
A strong theoretical activity started in the first years of 2000s promoting
the use of coupled plasmonic structures with single-photon sources. The
advantages would be an increased emission probability, due to the enhance-
ment of the electric field produced by the plasmon, and a better extraction.
Chang, Chen, et al. (2006) described a method that enables strong, coherent
coupling between individual emitters and electromagnetic excitations in
conducting nanostructures at optical frequencies, via excitation of guided
plasmons localized to nanoscale dimensions. The authors show that under
realistic conditions, optical emission can be almost entirely directed into
the plasmon modes. The method, that is discussed only theoretically, can
be used to create efficient single-photon sources.
Tame et al. (2008) discussed theoretically single-photon excitation of
SPPs, considering the quantum description and finding that remarkably high
quantum efficiencies can be reached for photon-to-surface plasmon polar-
itons transfer. The single photon to SPP transfer process may be assessed by
measuring the correlation function g(2)(0) which is shown to be unaffected
by losses.
Mallek-Zouari et al. (2010) investigated the fluorescence properties of
single CdSe/CdS nanocrystals deposited close to a semicontinuous gold
film, reporting a reduction by a factor ten of the monoexcitonic state decay
rate. A large fraction of the plasmons are converted in the far-field single
photons.
Naiki, Masuo, Machida, and Itaya (2011) obtained single-photon emis-
sion from isolated CdSe/ZnS QDs exhibiting enhanced fluorescence by
interacting with silver nanoparticles.
Bulu et al. (2011) proposed to use plasmonic resonators for enhanced
diamond NV-center single-photon sources. They showed theoretically that
in an optimized geometry, pumping can be improved by a factor of 7, the
spontaneous emission rates can be enhanced up to 50 times over NV centers
in a bulk crystal, and collection efficiencies can be as high as 40%, more than
10-fold increase over the bulk.
Gan et al. (2012) discussed a compact solid-state IIIV semiconductor
single-plasmon source. A sketch of the proposal for the semiconductor
single-plasmon source and its coupled-wave model is shown in
Figure 43. An SPP channel waveguide with the attached nanocavity is
placed near to the QD. The QD first excites the cavity mode, which in turn
decays into the SPP.
Figure 43 (a) Sketched geometry composed of a GaAs SPP channel (magenta; dark gray
in the print version) on a gold (yellow; light gray in the print version) substrate with a
dielectric nanogroove (blue; gray in the print version). The QD is assumed to be linearly
polarized along the z-direction; (b), (c1), and (c2) define physical quantities. From Gan
et al. (2012).
Akimov et al. (2007) were among the first to experimentally obtaining

generation of single optical plasmons in metallic nanowires coupled to
quantum dots.
When a single CdSe quantum dot is optically excited in close proximity
to a silver nanowire, emission from the quantum dot couples directly to
guided surface plasmons in the nanowire, causing the wires ends to light
up (Figure 44a). Photon correlation demonstrated the generation of single
quantized plasmons. Moreover, the efficient coupling was accompanied by
more than 2.5-fold enhancement of the quantum dot spontaneous emission.
The light emission at the nanowire end is a result of single, quantized surface
plasmons scattering off the ends of the nanowire. This is demonstrated by the
dip at t 0 in the photon coincidence between the free space fluorescence of
the quantum dot and emission from the wire end.
Schietinger, Barth, Aichele, and Benson (2009) presented a controlled
coupling of a single NV center to a plasmonic structure. With the help of
an atomic force microscope, a single nanodiamond containing a single
NV center and two gold nanospheres was assembled step by step. The exci-
tation rate and the radiative decay rate are enhanced by about one order of
magnitude, while the single-photon character of the emission is maintained.
Kolesov et al. (2009) showed that a single-photon source coupled to a
silver nanowire is able to excite single SPPs that exhibit wave and particle
properties, similar to those of single photons. To generate the SPP, single
color centers in diamond nanocrystals were coupled to the metallic wave-
guide at room temperature (Figure 45). Figure 45b shows the wide-field
image of a silver plasmonic waveguide when the laser beam is focused on
a single NV defect coupled to the wire. Both ends of the wire show
Figure 44 (a) A coupled quantum dot can emit either into free space or into the guided
surface plasmon of the nanowire with respective rates rad, pl. (b) Theoretical depen-
dence of the enhancement factor P (solid line) and efficiency of emission into surface
plasmons (dashed lines) on distance of the emitter from the nanowire edge. (c, d) Sim-
ulations of electric field and Poynting vector. From Akimov et al. (2007).
Figure 45 (a) Sketch of the Hanbury BrownTwiss experiment; (b) the fluorescence
image of a single quantum emitter; and (c) g(2)(t) of a single NV emitter, showing the
lowered value at t 0. From Kolesov et al. (2009).
detectable fluorescence, indicating the propagation of plasmons along the

wire. To check that single SPPs are excited by the single-photon source,
measurements of the second-order correlation function were carried out
between the two ends of the wire (Figure 45c). The lowering of the peak
at zero delay time provides experimental proof that the plasmon originates
from the fluorescence of a single quantum system. It is remarkable that even
though a surface plasmon involves a large number of electrons, it behaves
like a single quantum particle. A control was made, measuring coincidences
between the defect emission and the end of the wire which correspond to
the joint detection of a photon and a plasmon. Near perfect anticorrelation
was observed. Also a single-plasmon self-interference experiment was
successfully done.
Choy et al. (2011) presented a high-yield approach to directly embed
single NV centers into metallic nanostructures, leading to a reduction in
the spontaneous emission lifetime of the enclosed NV center. They consid-
ered plasmonic apertures consisting of cylindrical diamond nanoposts sur-
rounded by silver (Figure 46).
The device was realized and antibunching was observed.
A source of polarization-entangled photon pairs based on a cross-shaped
plasmonic nanoantenna driven by a single quantum dot has been suggested
by Maksymov, Miroshnichenko, and Kivshar (2012).
Enhancement of two-photon emission in the vicinity of metallic
nanoparticles due to surface plasmon resonances was discussed by
Poddubny, Ginzburg, Belov, Zayats, and Kivshar (2012).
11. APPLICATION TO QUANTUM INFORMATION

PROCESSING
The challenge for the development of quantum information technol-
ogies is having reliable and efficient sources to produce, distribute, and
detect entangled states (see, for example, Walmsley & Raymer, 2005).
Although sources of entanglement have been described and demonstrated
in many branches of physics, so far the most common way to distribute
entanglement is by means of pairs of photons. The most reliable source of
entanglement between photons is the spontaneous parametric down-
conversion process as discussed previously. Of great importance is to have
an accurate description of the distribution of the generated entangled states.
In realistic applications, where pure entangled states become mixed states
it is crucial to discriminate the nature of different types of noise introduced
Figure 46 (a) Three-dimensional schematic of the diamondplasmonic system. A dia-

mond nanopost is embedded in a 500-nm-thick layer of silver. NV fluorescence is
excited and collected through the bulk part of the diamond sample, (b) cross-sectional
vision, (c) simulated lateral mode profile, and (d) simulated spontaneous emission
enhancement for nanoposts with different radii. From Choy et al. (2011).
by the emission process itself, due to imperfections in the collections or in

the transmission of the entangled pairs, as discussed by Bovino (2013b),
where a method to discriminate the different noise contribution in a Quan-
tum Key distribution process has been patented.
Single-photon sources are at the basis of quantum information
processing. For this application as well as for any other photonphoton
interactions, a material system is necessary to mediate the interaction. Even
then, typical materials produce photonphoton coupling rates that are orders
of magnitude too small for single-photon interactions. Exception to this state
of affairs is cavity quantum electrodynamics, where strong interactions
between light and matter at the single-photon level were demonstrated in
a number of cases, and for this reason, many applications consider devices
and systems which use cavities. However, as shown below, many applica-
tions require simple linear optics. In the following we discuss some cases.
A number of efforts were done to obtain better sources of single photons
or to use single photons to obtain new states. Improving the quality of
single-photon sources by postprocessing with linear optics was proposed by
Berry et al. (2004). An interesting paper is the production of indistinguishable
photons from the individual Mollow triplet sidebands in Weiler, Stojanovic,
Ulrich, Jetter, and Michler (2013).
An experimental investigation of a single-photon source based on a
quantum interference by mixing a weak coherent state with a phase-locked
parametric down-conversion source on a 50/50 beam splitter was discussed
(Pittman, Franson & Jacobs, 2007). The method gives a high rate, but ran-
dom source of single photons.
The problems of interfacing single photons and single QDs with pho-
tonic nanostructures has been reviewed by Lodahl, Mahmodian, and
Stobbe (2013). Near unity coupling of a single QD and a photonic crystal
waveguide was discussed by Arcari et al. (2014). Quantum optics with sur-
face plasmons and strong coupling with single photons was the object of a
study by Chang, Srensen, Hemmer, and Lukin (2006).
A consequence of beam splitters entangling property was investigated by
Lvovsky and Mlynek (2002) who showed that a single-photon state j1i and a
coherent state ji present in two beam splitter input channels generate
entangled two-mode output. A superposition of the vacuum and one-photon
states could also be created by conditional measurements on a beam splitter.
This state is generated in one of the beam splitter output channels if a coherent
state ji and a single-photon Fock state j1i are present in the two input ports
and a single photon is registered in the other beam splitter output. The single
photon thus plays a role of a catalyst: it is explicitly present in both the input
and the output channels of the interaction and yet facilitates generation of
nonclassical states of light. By changing the amplitude of the target coherent
state, a gradual transition between a highly classical and a highly nonclassical
state of light occurs. The experiment was implemented by means of linear
optics and can be viewed as a step toward experimental realization of linear
optical quantum computation (Lvovsky & Mlynek, 2002).
Zavatta, Viciani, and Bellini (2004) reported the experimental genera-
tion of single-photon-added coherent states and their complete characteri-
zation by quantum tomography.
Cruz-Ramirez et al. (2012) obtained nondiffracting single photons, char-
acterized by a single-photon transverse intensity distribution which remains
essentially unchanged over a significant propagation distance (about 25 cm),
by properly shaping the pumping beam in a down-conversion process.
In analogy with Coulomb blockade for electrons (Fulton & Dolan, 1987;
Kastner, 1992; Likharev, 1999), photonphoton interactions in a nonlinear
optical cavity were proposed to realize photon blockade (Imamoglu et al.,
1997) for which a first photon within an optical system blocks the transmission
of a second photon, leading to an orderly output of photons one by one. The
proposal was extended by Dayan et al. (2008) using a microscopic optical res-
onator demonstrating a mechanism for the regulated transport of photons one
by one. Other proposals and experiments on single-photon blockade sources
were performed in the last 10 years (Bajcsy, Majumdar, Rundquist, &
Vuckovic, 2013; Bamba, Imamoglu, Carusotto, & Ciuti, 2011; Birnbaum
et al., 2005; Carusotto, Volz, & Imamoglu, 2010; Chang, Srensen,
Demler, & Lukin, 2007; Faraon et al., 2008; Gerace, T ureci, Imamoglu,
Giovannetti, & Fazio, 2009; He, Wei, et al., 2013; Liew & Savona, 2010;
Majumdar, Bajcsy, Rundquist, & Vuckovic, 2012; Reinhard et al., 2012;
Verger, Ciuti, & Carusotto, 2006). Eventually, Majumdar and Gerace
(2013) speculated about using a doubly resonant nanocavity filled with a
(2) material to realize a new scheme of photon blockade source.
Sotier et al. (2009) operating a single CdSe/ZnSe quantum dot on fem-
tosecond timescale, in a nonlinear regime, demonstrated the ability to
change the number of quanta in a femtosecond light pulse by exactly 1.
McCutcheon, Chang, Zhang, Lukin, and Loncar (2009) discussed an
approach to generate single photons with controllable wavelength and
bandwidth making use of an integrated nonlinear optical cavity in which
optical emission is directly frequency shifted into the desired domain using
intracavity nonlinear optical processes. Some examples were discussed with
a GaAs photonic crystal and a GaP-based design for direct coupling between
an NV center in diamond or an InAs/GaAs quantum dot.
Upconversion, in principle, can be used to transduce one photon at a
given wavelength to a photon at another wavelength. Rakher, Ma,
Slattery, Tang, and Srinivasan (2010) proposed using it to couple different
quantum systems at nominally different energies. He demonstrated efficient
upconversion of single photons generated by a quantum dot at 1.3 m to
710 nm. In particular, they measured the second-order intensity correlation
of the upconverted 710 nm signal and showed that the field was dominantly
composed of single photons. The simultaneous wavelength translation and
amplitude modulation of single photons generated by a quantum dot emit-
ting near 1300 nm was further discussed (Rakher et al., 2011).
Control of single-photon packets is an important resource for developing
hybrid quantum systems which are composed of different physical systems
interacting via photons. Rakher and Srinivasan (2011) extended this control
to triggered photons emitted by a quantum dot, temporally shaping single-
photon wave packets on timescales fast compared to their radiative decay by
electro-optic modulation. In particular, telecommunicationsband single
photons resulting from the recombination of an exciton in a quantum dot
with exponentially decaying wave packets were synchronously modulated
to create Gaussian-shaped single photon wave packets.
Shen et al. (2011) discussed several proposals to achieve, at the single-
photon level, near complete optical isolation, by coupling a quantum impu-
rity to a passive, linear waveguide which has a locally planar, circular
polarization.
Aoki et al. (2009) realized a cavity quantum electrodynamic system in the
optical domain with efficient inputoutput coupling, while still maintaining
high internal efficiency for coupling to a single atom.
In quantum information processing, a promising approach is to create an
on-chip quantum network in which, for example, NVs interact optically via
nanophotonic interconnects. Coupling NVs to optical cavities plays a crucial
role in this implementation, by enhancing the NV emission into a well-
defined optical mode, which can be efficiently coupled to waveguides
and routed on-chip. Efforts to efficiently couple NVs in nanocrystalline dia-
mond to nanophotonic structures have been made by Englund, Faraon,
Zhang, Yamamoto, and Vuckovic (2007), Faraon, Barclay, Santori, Fu,
and Beausoleil, (2011), Barclay, Fu, Santori, Faraon, and Beausoleil
(2011), Englund et al. (2010), Wolters et al. (2010), and Van der Sar
et al. (2011, 2009).
Englund et al. (2007) presented a basic building block of a quantum net-

work consisting of a quantum dot coupled to a source cavity, which in turn
was coupled to a target cavity via a waveguide. Single photons were effi-
ciently transferred into the target cavity via the waveguide. The structures
were fabricated on a 160-nm-thick GaAs membrane, grown by molecular
beam epitaxy with a central layer of self-assembled InAs quantum dots
whose photoluminescence peaked at 932 nm with an inhomogeneous
linewidth of about 60 nm and a density of about 200 quantum dots/m2.
Measurements were done at 5 K.
Faraon et al. (2011) fabricated nanophotonic devices directly from single
diamond crystals, demonstrating coupling of the zero-phonon line of indi-
vidual NV centers and the modes of microring resonators fabricated in
single-crystal diamond, below 10 K.
An alternative approach is to create photonic structures from hybrid
material systems in which a thin waveguiding layer is bonded to the surface
of a single-crystal diamond substrate. Hybrid semiconductor-diamond
devices were used by Barclay, Fu, Santori, and Beausoleil (2009) and Fu
et al. (2008). Barclay et al. (2011) obtained resonantly enhanced emission
from the zero-phonon line of a diamond NV center in single-crystal dia-
mond, using a hybrid whispering gallery mode nanocavity (Figure 47). They
demonstrated optical coupling between a nanoscale hybrid optical cavity
and a single diamond NV center.
Politi, Cryan, Rarity, Yu, and OBrien (2008) demonstrated photonic
quantum circuits using silica waveguides on a silicon chip. Optical wave-
guides on the chip provide networks of interconnected interferometers with
high visibility and with excellent control over the alignment and purity of
the optical elements. Hwang and Hinds (2011) suggested introducing a
nonlinear element into the chip. They discussed, albeit theoretically, indi-
vidual organic dye molecules, deposited close to optical waveguides on a
photonic chip as suitable single-photon source. In the wavelength range
of 780900 nm, where silicon photodiodes have high quantum efficiency
and high index waveguide materials such as SiN and GaP have good trans-
parency. a molecule of dibenzoterrylene could be used operating in the
vicinity of 785 nm at 2 K.
On-chip single-photon emission of a QD coupled to a photonic wave-
guide was demonstrated also by Schwagmann et al. (2011).
Fu et al. (2008) demonstrated the optical coupling of guided modes in a
GaP waveguide to NV centers in diamond, not considering explicitly the
single-photon case.
Figure 47 (a) SEM image of a hybrid GaP-diamond whispering gallery mode cavity,
(b) top view, and (c) cross-section of the dominant electric field component. From
Barclay et al. (2011).
Solid-state quantum optics with quantum dots in photonic nano-

structures was reviewed by Lodahl and Stobbe (2013).
Laucht et al. (2012) made a waveguide-coupled on-chip single-photon
source with individual self-assembled InGaAs QDs coupled to the guided opti-
cal mode of a GaAs photonic crystal waveguide. A well-coupled QD 7.3 m
away from the cleaved facet of the waveguide demonstrated a g(2)(0) 0.27.
An optical horn structure for single-photon source using QDs at tele-
communication wavelength was made by Takemoto et al. (2007). The basic
function of the horn was directing light toward a substrate using the reflec-
tion at the semiconductor surface.
Figure 48 shows the structure with the QD at the bottom of the horn
(Figure 48a).
An important challenge in distributed quantum information processing is
the controllable transfer of a quantum state between a flying qubit and mac-
roscopic matter. To implement quantum logic, the information should be
transferable between qubit carriers and preserved between gate operations;
this involves the existence of a quantum memory. Already in 1997, Maitre
reported the realization of a quantum memory in a cavity QED experiment
Figure 48 (a) Schematic, (b) SEM image, and (c) generated electric field intensity
(calculated). From Takemoto et al. (2007).
in which a qubit was transferred from an atomic carrier to a field one, and
then to another atom. The initial atom was either in one of its two energy
eigenstates or in the superposition of them. The mediating field was pre-
pared either in a 0 or 1 photon number state or in a superposition of the
two (Matre et al., 1997).
The principle of the quantum information transfer relied on the Rabi
precession at frequency /2 of an atom between two energy eigenstates
e and g in the cavity vacuum j0i. If the atom starts in the upper level e
and the effective resonant atomcavity interaction time t is such that t ,
the combined system evolves from the je,0i into the jg,1i state: the atomic
excitation is transferred to the field. If the atom is initially in level g, the sys-
tem starts in the jg,0i state and no evolution occurs. If the atom is initially in a
superposition jei + jgi, the linearity of quantum mechanics implies that
the combined system evolves into the state (j1i + j0i)jgi. The interaction
has transferred the quantum superposition from the atom to the field, leaving
the former in g. This information can then be transferred to a second atom
initially in g and crossing the cavity after a delay, in a process reverse to the
one experienced by the first atom.
The implementation of the memory was made using Rubidium atoms,

effusing from an oven, traversing a series of cavities.
Chen et al. (2006) presented an experimental realization of a determin-
istic and storable single-photon source employing a cold atomic ensemble of
87
Rb atoms. The writing of information is performed using Raman scatter-
ing. A read pulse retrieves the photon. The lifetime of the collective spin
excitation reached 12.5 s.
A continuous-variable quantum receiver has been realized by Julsgaard,
Sherson, Cirac, Fiurasek, and Polziket (2004). Work with quantized exci-
tations demonstrated capture and release of a single photon of fixed polar-
ization (Chaneliere et al., 2005; Eisaman et al., 2005) and coherent adiabatic
transfer of a single photon between two ensembles via an optical resonator
(Simon, Tanji, Ghosh, & Vuletic, 2007). A first system capable of function-
ing as a quantum memory was realized by Choi, Deng, Laurat, and Kimble
(2008). Tanji, Ghosh, Simon, Bloom, and Vuleti (2009) demonstrated
heralded quantum memory where a single photon announces polarization
storage in the form of a single collective spin excitation that is shared
between two spatially overlapped atomic ensembles.
In the microwave domain, coherent quantum control has been obtained
with single Rydberg atoms and single photons (Haroche, Raimond, &
Brune, 2002). A coherent quantum state transfer from a two-level atomic
system to a single photon using an optically thick cold atomic cloud is
described by Matsukevich and Kuzmich (2004).
In 1999, three groups demonstrated quantum cryptography based on
entangled photons ( Jennewein et al., 2000; Naik et al., 2000; Tittel
et al., 2000).
Beveratos, Brouri, et al. (2002) presented the first complete realization
of a quantum cryptographic key distribution using an NV center in diamond
as a single-photon source. Later, Soujaeff et al. (2007) demonstrated a
quantum key distribution at 1550 nm using a pulsed heralded single-photon
source. The source used a BBO nonlinear crystal to obtain two
photons at 521 and 1550 nm, respectively. The link was made using
40 km of fiber.
Blinov, Moehring, Duan, and Monroe (2004) showed entanglement
between a single trapped 111Cd+ ion and a single photon to be used in ideal
quantum communication channels.
Eventually, a complete system was demonstrated using a single-photon
source in NV in diamond to implement an open-air quantum key distribu-
tion (Alleaume et al., 2004).
Davanco et al. (2012) demonstrated room temperature heralded single-

photon generation in a CMOS-compatible silicon nanowire device. Photon
pairs were created by spontaneous four-wave mixing with a pump beam at
1.5496 m that created photon pairs at 1.5295 and 1.5705 m. Photon cor-
relation measurement of the signal (1.5295 m) photons heralded by the
detection of the idler (1.5705 m) showed antibunching with g(2)(0) 0.19.
Experimental demonstration of complementary with single photons has
been made by Braig, Zarda, Kurtsiefer, and Weinfurter (2003).
A logic gate operated with an ultrabright InGaAs quantum dot single-
photon source was demonstrated by Gazzano et al. (2013).
12. CONCLUSIONS
Although at present none of the sources discussed in this review could
be considered the best, a great improvement has occurred since the first
experiments. On the other hand, a perfect turnstile device can never be
made in practice. A real device suffers of unavoidable intrinsic loss and
the generation of multiple photons can never be excluded with absolute cer-
titude. At present, entangled state sources seem to be the best way to have
heralded single photons, but a real on demand source does not exist, yet.
However, for example, quantum cryptography is now matured, giving
the realization of quantum cryptographic devices, such as the
Q-KeyMaker by Bovino and Giardina (2011). Security aspects of quantum
key distribution with sub-Poisson light were examined by Waks, Santori,
and Yamamoto (2002).
The recent developments in the field of quantum technologies allow to
think to the development of quantum computers, where integrated photon-
ics sources are still one of the key elements. This review has given a pano-
rama of the different efforts to obtain single-photon sources and improve
their characteristics so as to be implemented into practical applications.
We are midway to a real satisfactory situation.
ADDENDUM
After the submission of the manuscript, a number of papers have been
published which we think deserve mention. The use of semiconductor
quantum dots has been excellently revised by Beveratos, Abram, Gerard,
and Robert-Philip (2014) discussing the tailoring of the emission from single
quantum dots to produce single photons, indistinguishable single photons,

and entangled photon pairs.
Generation of single photons by self-assembled strain-tunable quantum
dots (albeit at very low temperature 4.5 K) (Kremer et al., 2014) and on-chip
quantum dot efficiently coupled into a single-mode waveguide (Makhonin
et al., 2014) have been obtained and tunable single photons from dip-
olaritons using an unconventional photon blockade based on quantum
interference effects (Kyriienko et al., 2014) and quantum dots as sources
of polarization-entangled photon pairs resulting from the biexcitonexciton
radiative cascade have been proposed (Schmidgall et al., 2014).
The production of entangled photons, as one should expect, has received
great attention (Deshpande et al., 2014; Gogyan et al., 2014; Harris et al.,
2014; Kumar et al., 2014; Nisbet-Jones et al., 2013, 2011; Vasilev et al.,
2010; Versteegh et al., 2014).
An analysis of the imperfections of single-photon sources that can be tol-
erated in the operation of linear optics schemes for entanglement distribu-
tion which consider some possible cases has been made by Lasota and Thew
(2014).
Defects in diamond were studied by Sandstrom et al. (2014) and Rogers
et al. (2015).
Cavity-based quantum networks with single atoms and optical photons
were reviewed by Reiserer and Rempe (2014).
The advantage of using Purcell effect for plasmonic nanoantenna was
considered by Akselrod et al. (2014), and a number of applications were dis-
cussed (Filip & Zapletal, 2014; Wang, Li, et al., 2014; Wang, Santamato,
et al., 2014). Emission from single colloidal nanocrystals in a microcavity
was considered by Quartieri et al. (2009).
Finally, which probably is the most exciting paper, the nonlinear inter-
action between single photons has been achieved using a sum-frequency
generation in a periodically poled lithium niobate waveguide (Guerreiro
et al., 2014).
REFERENCES
Acharya, R., & Sudarshan, E. C. G. (1960). Front description in relativistic quantum
mechanics. Journal of Mathematical Physics, 1, 532.
Aharonovich, I., Castelletto, S., Johnson, B. C., McCallum, J. C., Simpson, D. A.,
Greentree, A. D., et al. (2010). Chromium single-photon emitters in diamond fabricated
by ion implantation. Physics Review B, 81, 121201(R).
Aharonovich, I., Castelletto, S., Simpson, D. A., Su, C. H., Greentree, A. D., & Prawer, S.
(2011). Diamond-based single-photon emitters. Reports on Progress in Physics, 74, 076501.
Aharonovich, I., Zhou, C. Y., Stacey, A., Orwa, J., Castelletto, S., Simpson, D., et al. (2009).
Enhanced single-photon emission in the near infrared from a diamond color center.
Physics Review B, 79, 235316.
Aichele, T., Zwiller, V., & Benson, O. (2004). Visible single-photon generation from semi-
conductor quantum dots. New Journal of Physics, 6, 90.
Aichele, T., Zwiller, V., Benson, O., Akimov, I., & Henneberger, F. (2003). Single CdSe
quantum dots for high-bandwidth single-photon generation. Journal of the Optical Society
of America, B20, 2189.
Akimov, A. V., Mukherjee, A. Yu., Chang, D. E., Zibrov, A. S., Hemmer, P. R., Park, H.,
et al. (2007). Generation of single optical plasmons in metallic nanowires coupled to
quantum dots. Nature, 450, 402.
Akopian, N., Lindner, N. H., Poem, E., Berlatzky, Y., Avron, J., Gershoni, D., et al. (2006).
Entangled photon pairs from semiconductor quantum dots. Physical Review Letters, 96,
130501.
Akopian, N., Lindner, N. H., Poem, E., Berlatzky, Y., Avron, J., Gershoni, D., et al. (2007).
Correlated and entangled pairs of single photons from semiconductor quantum dots.
Journal of Applied Physics, 101, 081712.
Akselrod, G. M., Argyropoulos, C., Hoang, T. B., Ciraci, C., Fang, C., Huang, J., et al.
(2014). Probing the mechanisms of large Purcell enhancement in plasmonic nano-
antennas. Nature Photonics, 8, 835.
Alleaume, R., Treussart, F., Messin, G., Dumeige, Y., Roch, J. F., Beveratos, A., et al.
(2004). Experimental open-air quantum key distribution with a single-photon source.
New Journal of Physics, 6, 92.
Altewischer, E., Van Exter, M. P., & Woerdman, J. P. (2002). Plasmon-assisted transmission
of entangled photons. Nature, 418, 304.
Aoki, T., Parkins, A. S., Alton, D. J., Regal, C. A., Dayan, B., Ostby, E., et al. (2009). Effi-
cient routing of single photons by one atom and a microtoroidal cavity. Physical Review
Letters, 102, 083601.
Arcari, M., Sollner, I., Javadi, A., Lindskov, H. S., Mahmoodian, S., & Liu, J. (2014). Near-
unity coupling efficiency of a quantum emitter to a photonic-crystal waveguide. Physical
Review Letters, 113, 093603.
Aspect, A., Grangier, P., & Roger, G. (1981). Experimental tests of realistic local theories via
Bells theorem. Physical Review Letters, 47, 460.
Babinec, T. M., Hausmann, B., Choy, J., Khan, M., Hemmer, P., & Loncar, M. (2011).
Quantum photonics with diamond. IEEE Photonics Society Newsletter, 24, 13.
Babinec, T. M., Hausmann, B. J., Khan, M., Zhang, Y., Maze, J. R., Hemmer, P. R., et al.
(2010). A diamond nanowire single-photon source. Nature Nanotechnology, 5, 195.
Baier, M. H., Pelucchi, E., Kapon, E., Varoutsis, S., Gallart, M., et al. (2004). Single photon
emission from site-controlled pyramidal quantum dots. Applied Physics Letters, 84, 648.
Bajcsy, M., Majumdar, A., Rundquist, A., & Vuckovic, J. (2013). Photon blockade with a
four-level quantum emitter coupled to a photonic-crystal nanocavity. New Journal of
Physics, 15, 025014.
Bamba, M., Imamoglu, A., Carusotto, I., & Ciuti, C. (2011). Origin of strong photon anti-
bunching in weakly nonlinear photonic molecules. Physical Review A, 83, 021802(R).
Barclay, P. E., Fu K-M, C., Santori, C., & Beausoleil, R. G. (2009). Hybrid photonic crystal
cavity and waveguide for coupling to diamond NV-centers. Optics Express, 17, 9588.
Barclay, P. E., Fu K-M, C., Santori, C., Faraon, A., & Beausoleil, R. G. (2011). Hybrid
nanocavity resonant enhancement of color center emission in diamond. Physical Review
X, 1, 011007.
Barnes, W. L., Bjork, G., Gerard, J. M., Jonsson, P., Wasey, J. A. E., Worthing, P. T., et al.
(2002). Solid-state single photon sources: Light collection strategies. European Physical
Journal D: Atomic, Molecular, Optical and Plasma Physics, 18, 197.
Barros, H. G., Stute, A., Northup, T. E., Russo, C., Schmidt, P. O., & Blatt, R. (2009).
Deterministic single-photon source from a single ion. New Journal of Physics, 11, 103004.
Basche, Th., Moerner, W. E., Orrit, M., & Talon, H. (1992). Photon antibunching in the
fluorescence of a single dye molecule trapped in a solid. Physical Review Letters, 69, 1516.
Becher, C., Kiraz, A., Michler, P., Schoenfeld, W. V., Petroff, P. M., Zhang, L., et al. (2002).
A quantum dot single-photon source. Physica E, 13, 412.
Bennett, A. J., Atkinson, P., See, P., Ward, M. B., Stevenson, R. M., Yuan, Z. L., et al.
(2006). Single-photon-emitting diodes: A review. Physica Status Solidi (b), 243, 3730.
Bennett, A. J., Patel, R. B., Nicoll, C. A., Ritchie, D. A., & Shields, A. J. (2009). Interference
of dissimilar photon sources. Nature Physics, 5, 715.
Bennett, A. J., Unitt, D. C., Atkinson, P., Ritchie, D. A., & Shields, A. J. (2005). High per-
formance single photon sources from photolithographically defined pillar microcavities.
Optics Express, 13, 50, and 7772.
Bennett, A. J., Unitt, D. C., See, P., Shields, A. J., Atkinson, P., Cooper, K., et al. (2005a).
Microcavity single-photon-emitting diode. Applied Physics Letters, 86, 181102.
Bennett, A. J., Unitt, D. C., See, P., Shields, A. J., Atkinson, P., Cooper, K., et al. (2005b).
Electrical control of the uncertainty in the time of single photon emission events. Physics
Review B, 72, 033316.
Benson, O., Santori, C., Pelton, M., & Yamamoto, Y. (2000). Regulated and entangled pho-
tons from a single quantum dot. Physical Review Letters, 84, 2513.
Benyoucef, M., Ulrich, S. M., Michler, P., Wiersig, J., Jahnke, F., & Forchel, A. (2004).
Enhanced correlated photon pair emission from a pillar microcavity. New Journal of
Physics, 6, 91.
Berry, D. W., Scheel, S., Myers, C. R., Sanders, B. C., Knight, P. L., & Laflamme, R. (2004).
Postprocessing with linear optics for improving the quality of single-photon sources. New
Journal of Physics, 6, 93.
Bertet, P., Osnaghi, S., Milman, P., Auffeves, A., Maioli, P., Brune, M., et al. (2002). Gen-
erating and probing a two-photon Fock state with a single atom in a cavity. Physical
Bertolotti, M. (2006). Wave interactions in photonic band structures: An overview. Journal of
Optics A: Pure and Applied Optics, 8, 81.
Beveratos, A., Abram, I., Gerard, J. M., & Robert-Philip, I. (2014). Quantum optics with
quantum dots. European Physical Journal D: Atomic, Molecular, Optical and Plasma Physics,
68, 377.
Beveratos, A., Brouri, R., Gacoin, T., Poizat, J.-P., & Grangier, P. (2001). Nonclassical radi-
ation from diamond nanocrystals. Physical Review A, 64, 061802(R).
Beveratos, A., Brouri, R., Gacoin, T., Villing, A., Poizat, J.-P., & Grangier, P. (2002). Single
photon quantum cryptography. Physical Review Letters, 89, 187901.
Beveratos, A., Kuhn, S., Brouri, R., Gacoin, T., Poizat, J.-P., & Grangier, P. (2002). Room
temperature stable single-photon source. The European Physical Journal D, 18, 191.
Bialynicki-Birula, I. (1996). Photon wave function. Progress in Optics, 36, 248.
Birnbaum, K. M., Boca, A., Miller, R., Boozer, A. D., Northup, T. E., & Kimble, H. J.
(2005). Photon blockade in an optical cavity with one trapped atom. Nature, 436, 87.
Bleuse, J., Claudon, J., Creasey, M., Malik, N. S., Gerard, J. M., Maksymov, I., et al. (2011).
Inhibition, enhancement, and control of spontaneous emission in photonic nanowires.
Physical Review Letters, 106, 103601.
Blinov, B. B., Moehring, D. L., Duan, L. M., & Monroe, C. (2004). Observation of entan-
glement between a single trapped atom and a single photon. Nature, 428, 153.
Bommer, M., Schulz, W. M., Rossbach, R., Jetter, M., Michler, P., Thomay, T., et al.
(2011). Triggered single-photon emission in the red spectral range from optically excited
InP/(Al, Ga)InP quantum dots embedded in micropillars up to 100 K. Journal of Applied
Physics, 110, 063108.
Borgstrom, M. T., Zwiller, V., Muller, E., & Imamoglu, A. (2005). Optically bright quan-
tum dots in single nanowires. Nano Letters, 5, 1439.
Bovino, F. A. (2013a). System for distributing cryptographic keys. Patent
EP2555466OS20130208894.
Bovino, F. A. (2013b). Interferometer and method for controlling the coalescence of a pair of
photons. Patent EP2543957USP20130176573.
Bovino, F. A., & Degiovanni, I. P. (2008). Quantum correlation bounds for quantum infor-
mation experiments optimization: The Wigner inequality case. Physical Review A, 77,
053822.
Bovino, F. A., & Giardina, M. (2011). Practical quantum cryptography on metropolitan area
network: The Q-KeyMaker. Polaris Innovation Journal, 062011.
Bovino, F. A., Varisco, P., Colla, A. M., Castagnoli, G., Di Giuseppe, G., & Sergienko, A. V.
(2003). Effective fiber-coupling of entangled photons for quantum communication.
Optics Communications, 227, 343.
Bovino, F. A., Varisco, P., Martinoli, A., De Nicolo, P., Bruzzo, S., Colla, A. M., et al.
(2005). Practical quantum key distribution using polarization entangled states. The Inter-
national Journal of Quantum Information, 3, 1.
Braig, C., Zarda, P., Kurtsiefer, C., & Weinfurter, H. (2003). Experimental demonstration of
complementarity with single photons. Applied Physics B: Lasers and Optics, 76, 113.
Branczyk, A., Ralph, T. C., Helwig, W., & Silberhorn, C. (2010). Optimized generation of
heralded Fock states using parametric down-conversion. New Journal of Physics, 12,
063001.
Brattke, S., Guthohrlein, G. R., Keller, M., Lange, W., Varcoe, B., & Walther, H. (2003).
Generation of photon number states on demand. Journal of Modern Optics, 50, 1103.
Brattke, S., Varcoe, B. T. H., & Walther, H. (2001a). Preparing Fock states in the micro-
maser. Optics Express, 8, 131.
Brattke, S., Varcoe, B. T. H., & Walther, H. (2001b). Generation of photon number states on
demand via cavity quantum electrodynamics. Physical Review Letters, 86, 3534.
Brida, G., Degiovanni, I. P., Genovese, M., Piacentini, F., Traina, P., Della Frera, A., et al.
(2012). An extremely low-noise heralded single-photon source: A breakthrough for
quantum technologies. Applied Physics Letters, 101, 221112.
Brokmann, X., Giacobino, E., Dahan, M., & Hermier, J. P. (2004). Highly efficient triggered
emission of single photons by colloidal CdSe/ZnS nanocrystals. Applied Physics Letters,
85, 712.
Brokmann, X., Messin, G., Desbiolles, P., Giacobino, E., Dahan, M., & Hermier, J. P.
(2004). Colloidal CdSe/ZnS quantum dots as single-photon sources. New Journal of
Physics, 6, 99.
Brouri, R., Beveratos, A., Poizat, J.-P., & Grangier, P. (2000a). Single-photon generation by
pulsed excitation of a single dipole. Physics Review A, 62, 063817.
Brouri, R., Beveratos, A., Poizat, J.-P., & Grangier, P. (2000b). Photon antibunching in the
fluorescence of individual color centers in diamond. Optics Letters, 25, 1294.
Brune, M., Haroche, S., Lefevre, V., Raimond, J. M., & Zagury, N. (1990). Quantum non-
demolition measurements of small photon numbers by Rydberg-atom phase-sensitive
detection. Physical Review Letters, 65, 976.
Brune, M., Haroche, S., Lefevre, V., Raimond, J. M., & Zagury, N. (1992). Manipulation of
photons in a cavity by dispersive atom-field coupling: Quantum nondemolition mea-
surements and generation of Schr odinger cat states. Physics Review A, 45, 5193.
Brunel, C., Lounis, B., Tamarat, P., & Orrit, M. (1999). Triggered source of single photons
based on controlled single molecule fluorescence. Physical Review Letters, 83, 2722.
Brunel, C., Tamarat, P., Lounis, B., Plantard, J., & Orrit, M. (1998). Driving the Bloch vector
of a single molecule: Towards a triggered single photon source. Comptes-Rendus de
lAcademie des Sciences, Paris, 326, 911.
Buckley, S., Rivoire, K., & Vuckovic, J. (2012). Engineered quantum dot single-photon
sources. Reports on Progress in Physics, 75, 126503.
Bulu, I., Babinec, T., Hausmann, B., Choy, J. T., & Loncar, M. (2011). Plasmonic resonators
for enhanced diamond NV-center single photon sources. Optics Express, 19, 5268.
Bulu, I., Caglayan, H., & Ozbay, E. (2003). Radiation properties of sources inside photonic
crystals. Physics Review B, 67, 205103.
Busch, K., Vats, N., John, S., & Sanders, B. C. (2000). Radiating dipoles in photonic crystals.
Physics Review E, 62, 4251.
Buttler, W. T., Hughes, R. J., Lamoreaux, S. K., Morgan, G. L., Nordholt, J. E., &
Peterson, C. G. (2000). Daylight quantum key distribution over 1.6 km. Physical Review
Letters, 84, 5652.
Carmichael, H. J., & Walls, D. F. (1976). Proposal for the measurement of the resonant Stark
effect by photon correlation techniques. Journal of Physics B, 9, L43.
Carusotto, I., Volz, T. B., & Imamoglu, A. (2010). Feshbach blockade: Single-photon
nonlinear optics using resonantly enhanced cavity polariton scattering from biexciton
states. Europhysics Letters, 90, 37001.
Castelletto, S., Degiovanni, I. P., Migdall, A., & Ware, M. (2004). On the measurement of
two-photon single-mode coupling efficiency in parametric downconversion photon
sources. New Journal of Physics, 6, 87.
Castelletto, S., Johnson, B. C., Ivady, V., Stavrias, N., Umeda, T., Gali, A., et al. (2014).
A silicon carbide room-temperature single-photon source. Nature Materials, 13, 151.
Centini, M., Perina, J., Jr., Sciscione, L., Sibilia, C., Scalora, M., Bloemer, M. J., et al.
(2005). Entangled photon pair generation by spontaneous parametric down
conversion in finite-length one-dimensional photonic crystals. Physics Review A, 72,
033806.
Cerf, N. J., & Flurasek, J. (2006). Optical quantum cloning. Progress in Optics, 49, 455.
Chance, R. R., Prock, A., & Silbey, R. (1974). Lifetime of an emitting molecule near a par-
tially reflecting surface. The Journal of Chemical Physics, 60, 2749.
Chaneliere, T., Matsukevich, D. N., Jenkins, S. D., Lan, S.-Y., Kennedy, T. A. B., &
Kuzmich, A. (2005). Storage and retrieval of single photons transmitted between remote
quantum memories. Nature, 438, 833.
Chang, W. H., Chen, W. Y., Chang, H.-S., Hsieh, T. P., Chyi, J.-I., & Hsu, T.-M. (2006).
Efficient single-photon sources based on low-density quantum dots in photonic-crystal
nanocavities. Physical Review Letters, 96, 117401.
Chang, D. E., Srensen, A. S., Demler, E. A., & Lukin, M. D. (2007). A single-photon tran-
sistor using nanoscale surface plasmons. Nature Physics, 3, 807.
Chang, D. E., Srensen, A. S., Hemmer, P. R., & Lukin, M. D. (2006). Quantum optics with
surface plasmons. Physical Review Letters, 97, 053002.
Chen, S., Chen, Yu.-A., Strassel, T., Yuan, Zhen-Sheng, Zhao, Bo, Schmiedmayer, J., et al.
(2006). Deterministic and storable single-photon source based on a quantum memory.
Choi, K. S., Deng, H., Laurat, J., & Kimble, H. J. (2008). Mapping photonic entanglement
into and out of a quantum memory. Nature, 452, 67.
Choy, J. T., Hausmann, B. J. M., Babinec, T. M., Bulu, I., Khan, M., Maletinsky, P., et al.
(2011). Enhanced single-photon emission from a diamond-silver aperture. Nature
Photonics, 5, 738.
Claudon, J., Bleuse, J., Malik, N. S., Bazin, M., Jaffrennou, P., Gregersen, N., et al. (2010).
A highly efficient single-photon source based on a quantum dot in a photonic nanowire.
Nature Photonics, 4, 174.
Clauser, J. F. (1974). Experimental distinction between the quantum and classical field-
theoretic predictions for the photoelectric effect. Physics Review D, 9, 853.
Cruz-Ramirez, H., Ramrez-Alarcon, R., Morelos, F. J., Quinto-Su, P. A., Gutierrez-

Vega, J. C., & URen, A. (2012). Observation of non-diffracting behavior at the
single-photon level. Optics Express, 20, 29761.
Da Cunha Pereira, M., Becerra, F. E., Glebov, B. L., Fan, J., Nam, S. W., & Migdall, A.
(2013). Demonstrating highly symmetric single-mode, single-photon heralding
efficiency in spontaneous parametric downconversion. Optics Letters, 38, 1609.
Dalacu, D., Mnaymneh, K., Sazonova, V., Poole, P. J., Aers, G. C., Lapointe, J., et al. (2010).
Deterministic emitter-cavity coupling using a single-site controlled quantum dot. Physics
Review B, 82, 033301.
Darquie`, B., Jones, M. P. A., Dingjan, J., Beugnon, J., Bergamini, S., Sortais, Y., et al.
(2005). Controlled single-photon emission from a single trapped two-level atom.
Science, 309, 454.
Davanco, M., Ong, J. R., Shehata, A. B., Tosi, A., Agha, I., Assefa, S., et al. (2012).
Telecommunications-band heralded single photons from a silicon nanophotonic chip.
Applied Physics Letters, 100, 261104.
Davanco, M., Rakher, M. T., Schuh, D., Badolato, A., & Srinivasan, K. (2011). A circular
dielectric grating for vertical extraction of single quantum dot emission. Applied Physics
Letters, 99, 041102.
Davies, G., & Hamer, M. F. (1976). Optical studies of the 1.945 eV vibronic band in
diamond. Proceedings of the Royal Society of London A, 348, 285.
Dayan, B., Parkins, A. S., Aoki, T., Ostby, E. P., Vahala, K. J., & Kimble, H. J. (2008).
A photon turnstile dynamically regulated by one atom. Science, 319, 1062.
De Dood, M. J. A., Irvine, W. T. M., & Bouwmeester, D. (2004). Nonlinear photonic crys-
tals as a source of entangled photons. Physical Review Letters, 93, 040504.
De Martini, F., Di Giuseppe, G., & Marrocco, M. (1996). Single-mode generation of
quantum photon states by excited single molecules in a microcavity trap. Physical Review
Letters, 76, 900.
Deshpande, S., Frost, T., Hazari, A., & Bhattacharya, P. (2014). Electrically pumped single-
photon emission at room temperature from a single InGaN/GaN quantum dot. Applied
Physics Letters, 105, 141109.
Diecks, D. (1982). Communication by EPR devices. Physics Letters, A92, 271.
Diedrich, F., & Walther, H. (1987). Nonclassical radiation of a single stored ion. Physical
Dirac, P. A. M. (1926). On the theory of quantum mechanics. Proceedings of the Royal Society
A, 112, 661.
Dirac, P. A. M. (1927). Quantum theory of emission and absorption of radiation. Proceedings
of the Royal Society A, 114, 243.
Domokos, P., Janszky, P., & Adam, P. (1994). Single atom interference method for gener-
ating Fock states. Physics Review A, 50, 3340.
Dousse, A., Suffczynski, J., Beveratos, A., Krebs, O., Lemaitre, A., Sagnes, I., et al. (2010).
Ultrabright source of entangled photon pairs. Nature, 466, 217.
Drexhage, K. H. (1974). Interaction of light with monomolecular dye layer. Progress in Optics,
12, 163.
Drexhage, K. H., Kuhn, H., & Schafer, F. P. (1968). Berichte der Bunsen-Gesellschaft f ur
Physikalische Chemie, 72, 329.
Duan, L. M., Kuzmich, A., & Kimble, H. J. (2003). Cavity QED and quantum-information
processing with hot trapped atoms. Physics Review A, 67, 032305.
Dudin, Y. O., & Kuzmich, A. (2012). Strongly interacting Rydberg excitations of a cold
atomic Gas. Science, 336, 887.
Dusek, M., Lutkenhas, N., & Hendrych, M. (2006). Quantum cryptography. Progress in
Optics, 49, 381.
Einstein, A. (1905). Concerning an heuristic point of view toward the emission and trans-
formation of light. Annalen der Physik, 17, 132.
Einstein, A., Podolky, B., & Rosen, N. (1935). Can quantum-mechanical description of
physical reality be considered complete? Physics Review, 47, 777.
Eisaman, M. D., Andre, A., Massou, F., Fleischhauer, M., Zibrov, A. S., & Lukin, M. D.
(2005). Electromagnetically induced transparency with tunable single-photon pulses.
Nature, 438, 837.
Eisaman, M. D., Fan, J., Mugdall, A., & Polyakov, S. V. (2011). Single photon sources and
detectors. Review of Scientific Instruments, 82, 071101.
Ellis, D. J. P., Bennett, A., Dewhurst, S., Nicoll, C. A., Ritchie, D. A., & Shields, A. J.
(2008). Cavity enhanced radiative emission rate in a single-photon emitting diode
operating at 0.5GHz. New Journal of Physics, 10, 043035.
Ellis, D. J. P., Bennett, A. J., Shields, A. J., Atkinson, P., & Ritchie, D. A. (2006). Electrically
addressing a single self-assembled quantum dot. Applied Physics Letters, 88, 133509.
Englund, D., Faraon, A., Zhang, B., Yamamoto, Y., & Vuckovic, J. (2007). Generation and
transfer of single photons on a photonic crystal chip. Optics Express, 15, 5550.
Englund, D., Fattal, D., Waks, E., Solomon, G., Zhang, B., Nakaoka, T., et al. (2005). Con-
trolling the spontaneous emission rate of single quantum dots in a two-dimensional pho-
tonic crystal. Physical Review Letters, 95, 013904.
Englund, D., Shields, B., Rivoire, K., Hatami, F., Vuckovic, J., Park, H., et al. (2010). Deter-
ministic coupling of a single nitrogen vacancy center to a photonic crystal cavity. Nano
Letters, 10, 3922.
Faddeev, L. D., Khalfin, L. A., & Komarov, I. V. (2004). Selected works V.A. Fock. London:
Chapman & Hall/CRC.
Fan, B., Duan, Z., Zhou, L., Yuan, C., Ou, Z. Y., & Zhanget, W. (2009). Generation of a single-
photon source via a four-wave mixing process in a cavity. Physics Review A, 80, 063809.
Fan, J., Eisaman, M. D., & Migdall, A. (2007). Bright phase-stable broadband fiber-based
source of polarization-entangled photon pair. Physics Review A, 76, 043836.
Fan, J., & Migdall, A. (2007). A broadband high spectral brightness fiber-based two-photon
source. Optics Express, 15, 2915.
Faraon, A., Barclay, P. E., Santori, C., Fu K-M, C., & Beausoleil, R. G. (2011). Resonant
enhancement of the zero-phonon emission from a color center in a diamond cavity.
Faraon, A., Fushman, I., Englund, D., Stoltz, N., Petroff, P., & Vucokovic, J. (2008). Coher-
ent generation of non-classical light on a chip via photon-induced tunnelling and block-
ade. Nature Physics, 4, 859.
Fasel, S., Robin, F., Moreno, E., Erni, D., Gisin, N., & Zbinden, H. (2005). Energy-time
entanglement preservation in plasmon-assisted light transmission. Physical Review Letters,
94, 110501.
Ferretti, S., & Gerace, D. (2012). Single-photon nonlinear optics with Kerr-type nanostruc-
tured materials. Physics Review B, 85, 033303.
Filip, R., & Zapletal, P. (2014). Squeezed-state generation from single-photon sources. Phys-
ics Review A, 90, 043854.
Filipowicz, P., Javanainen, J., & Meystre, P. (1986a). Quantum and semiclassical steady states
of a kicked cavity mode. Journal of the Optical Society of America, B3, 906.
Filipowicz, P., Javanainen, J., & Meystre, P. (1986b). Theory of a microscopic maser. Physics
Review A, 34, 3077.
Flagg, E. B., Muller, A., Polyakov, S. V., Ling, A., Migdall, A., et al. (2010). Interference of
single photons from two separate semiconductor quantum dots. Physical Review Letters,
104, 137401.
Flagg, E. B., Muller, A., Robertson, J. W., Founta, S., Deppe, G. D., Xiao, M., et al. (2009).
Resonantly driven coherent oscillations in a solid-state quantum emitter. Nature Physics,
5, 203.
Fleury, L., Segura, J.-M., Zumofen, G., Hecht, B., & Wild, U. P. (2000). Nonclassical pho-
ton statistics in single-molecule fluorescence at room temperature. Physical Review Letters,
84, 1148.
Florescu, M., Scheel, S., Haffner, H., Lee, H., Strekalov, D., Knight, P. L., et al. (2005).
Single photons on demand from 3D photonic band-gap structures. Europhysics Letters,
69, 945.
Fock, V. (1932). Konfigurazionsraum und zweite quantellung. Zeitschrift f ur Physik, 75, 622.
Foden, C. L., Talyanskii, V. I., Milburn, G. J., Leadbeater, M. L., & Pepper, M. (2000).
High-frequency acousto-electric single-photon source. Physics Review A, 62,
011803R.
Friedler, I., Sauvan, C., Hugonin, J. P., Lalanne, P., Claudon, J., & Gerard, J. M. (2009).
Solid-state single photon sources: The nanowire antenna. Optics Express, 17, 2095.
Fu, K. M. C., Santori, C., Barclay, P. E., Aharonovich, I., Prawer, S., Meyer, N., et al.
(2008). Coupling of nitrogen-vacancy centers in diamond to a GaP waveguide. Applied
Fulton, T. A., & Dolan, G. J. (1987). Observation of single-electron charging effects in small
tunnel junctions. Physical Review Letters, 59, 109.
Gabrielse, G., & Dehmelt, H. (1985). Observation of inhibited spontaneous emission. Phys-
ical Review Letters, 55, 67.
Gaebel, T., Popa, I., Gruber, A., Domhan, M., Jelezko, F., & Wrachtrup, J. (2004). Stable
single-photon source in the near infrared. New Journal of Physics, 6, 98.
Gan, C. H., Hugonin, J. P., & Lalanne, P. (2012). Proposal for compact solid-state III-V
single-plasmon sources. Physics Review X, 2, 021008.
Garraway, B. M., Sherman, B., Moya-Cessa, H., Knight, P. L., & Kurizki, G. (1994). Gen-
eration and detection of nonclassical field states by conditional measurements following
two-photon resonant interactions. Physics Review A, 49, 535.
Gazzano, O., Almeida, M. P., Nowak, A. K., Portalupi, S. L., Lematre, A., Sagnes, I., et al.
(2013). Entangling quantum-logic gate operated with an ultrabright semiconductor
single-photon source. Physical Review Letters, 110, 250501.
Geerligs, L. S., Anderegg, V. F., Holweg, P. A. M., Mooij, J. E., Pothier, H., Esteve, D., et al.
(1990). Frequency-locked turnstile device for single electrons. Physical Review Letters, 64,
2691.
Gerace, D., T ureci, H. E., Imamoglu, A., Giovannetti, V., & Fazio, R. (2009). The
quantum-optical Josephson interferometer. Nature Physics, 5, 281.
Gerard, J. M., Barrier, D., Marzin, J. Y., Kuszelewicz, R., Manin, L., & Costard, E. (1996).
Quantum boxes as active probes for photonic microstructures: The pillar microcavity
case. Applied Physics Letters, 69, 449.
Gerard, J. M., & Gayral, B. (1999). Strong Purcell effect for InAs quantum boxes in three-
dimensional solid-state microcavities. Journal of Lightwave Technology, 17, 2089.
Gerard, J. M., Sermage, B., Gayral, B., Legrand, B., Costard, E., & Thierry-Mieg, V. (1998).
Enhanced spontaneous emission by quantum boxes in a monolithic optical microcavity.
Gerry, C., & Knight, P. (2005). Introductory quantum optics. Cambridge: Cambridge
University Press.
Gevaux, D. G., Bennett, A. J., Stevenson, R. M., Shields, A. J., Atkinson, P., & Griffiths, J.
(2006). Enhancement and suppression of spontaneous emission by temperature tuning
InAs quantum dots to photonic crystal cavities. Applied Physics Letters, 88, 131101.
Ghirardi, G. C., & Weber, T. (1983). Quantum mechanics and faster-than-light communi-
cation: Methodological considerations. Nuovo Cimento, B78, 9.
Gisin, N., Ribordy, G., Tittel, W., & Zbinden, H. (2002). Quantum cryptography. Reviews
of Modern Physics, 74, 145.
Gisin, N., & Thew, R. (2007). Quantum communication. Nature Photonics, 1, 165.
Glauber, R. J. (1963a). The quantum theory of optical coherence. Physics Review, 130, 2529.
Glauber, R. J. (1963b). Coherent and incoherent states of the radiation field. Physics Review,
131, 2766.
Gogyan, A., Sisakyan, N., Akhmedzhanov, R., & Malakyan, Y. (2014). Heralded generation
of single photons entangled in multiple temporal modes with controllable waveforms.
Laser Physics, 24, 115204.
Goldschmidt, E. A., Eisaman, M. D., Fan, J., Polyakov, S. V., & Migdall, A. (2008).
Spectrally bright and broad fiber-based heralded single-photon source. Physics Review
A, 78, 013844.
Goy, P., Raimond, J. M., Gross, M., & Haroche, S. (1983). Observation of cavity-enhanced
single-atom spontaneous emission. Physical Review Letters, 50, 1903.
Grangier, P., & Abram, I. (2003). Single photons on demand. Physics World, 27, 31.
Grangier, P., Roger, G., & Aspect, A. (1986). Experimental evidence for a photon antic-
orrelation effect on a beam splitter: A new light on single-photon interferences.
Europhysics Letters, 1, 173.
Gregersen, N., Nielsen, T. R., Mrk, J., Claudon, J., & Gerard, J.-M. (2010). Designs for
high-efficiency electrically pumped photonic nanowire single-photon sources. Optics
Express, 18, 21204.
Gruber, A., Drabenstedt, A., Tietz, C., Fleury, L., Wrachtrup, J., & von Borczyskowski, C.
(1997). Scanning confocal optical microscopy and magnetic resonance on single defect
centers. Science, 276, 2012.
Guerreiro, T., Martin, A., Sanguinetti, B., Pelc, J. S., Langrock, C., Fejer, M. M., et al.
(2014). Nonlinear interaction between single photons. Physical Review Letters, 113,
173601.
Hafenbrak, R., Ulrich, S. M., Michler, P., Wang, L., Rastelli, A., & Schmidt, O. G. (2007).
Triggered polarization-entangled photon pairs from a single quantum dot up to 30 K.
New Journal of Physics, 9, 315.
Han, B., Ji, Z., & Zheng, Y. (2009). Photon counting statistics of single molecule in solid
matrix. The Journal of Chemical Physics, 130, 244502.
Hanbury Brown, R., & Twiss, R. Q. (1956a). Correlation between photons in two coherent
beams of light. Nature, 177, 27.
Hanbury Brown, R., & Twiss, R. Q. (1956b). Interferometry of the intensity fluctuations in
light. I. Basic theory: The correlation between photons in coherent beams of radiation.
Proceedings of the Royal Society A, 242, 300.
Hanbury Brown, R., & Twiss, R. Q. (1958). Interferometry of the intensity fluctuations in
light. II. An experimental test of the theory for partially coherent light. Proceedings of the
Royal Society A, 243, 291.
Hanzawa, H., Umemura, N., Nisida, Y., Kanda, H., Okada, M., & Kobayashi, M.
(1996). Disorder effects of nitrogen impurities, irradiation-induced defects, and
C13 isotope composition on the Raman spectrum in synthetic Ib diamond. Physics
Review B, 54, 3793.
Harder, G., Ansari, V., Brecht, B., Dirmeier, T., Marquardt, C., & Silberhorn, C. (2013). An
optimized photon pair source for quantum circuits. Optics Express, 21, 13975.
Haroche, S., Raimond, J. M., & Brune, M. (2002). Experimental quantum computation and
information. In F. de Martini & C. Monroe (Eds.), Proceedings of the international school of
physics Enrico Fermi, course CXLVIII, Amsterdam 2002 (pp. 3766).
Harris, N. C., Grassani, D., Simbula, A., Pant, M., Galli, M., Baehr-Jones, T., et al. (2014).
Integrated source of spectrally filtered correlated photons for large-scale quantum
photonic systems. Physics Review X, 4, 041047.
Hartschuh, A., Pedrosa, H. N., Peterson, J., Huang, L., Anger, P., Qian, H., et al. (2005).
Single carbon nanotube optical spectroscopy. Chemphyschem: A European Journal of
Chemical Physics and Physical Chemistry, 6, 1.
Hausmann, B. J. M., Babinec, T. M., Choy, J. T., Hodges, J. S., Hong, S., Bulu, I., et al.
(2011). Single-color centers implanted in diamond nanostructures. New Journal of Physics,
13, 045004.
He, Y., & Barkai, E. (2006). Theory of single photon on demand from a single molecule
source. Physical Chemistry Chemical Physics: PCCP, 8, 5056.
He, Y., He, Y.-M., Wei, Y.-J., Jiang, X., Chen, M.-C., Xiong, F.-L., et al. (2013). Indis-
tinguishable tunable single photons emitted by spin-flip Raman transitions in InGaAs
quantum dots. Physical Review Letters, 111, 237403.
He, Y., Wei, Y.-J., Wu, D., Atat ure, M., Schneider, C., H ofling, S., et al. (2013). On-
demand semiconductor single-photon source with near-unity indistinguishability.
Nature Nanotechnology, 8, 213.
Heindel, T., Schneider, C., Lermer, M., Kwon, S. H., Braun, T., & Reitzenstein, S. (2010).
Electrically driven quantum dot-micropillar single photon source with 34% overall effi-
ciency. Applied Physics Letters, 96, 011107.
Heinrich, J., Huggenberger, A., Heindel, T., Reitzenstein, S., H ofling, S., Worschech, L.,
et al. (2010). Single photon emission from positioned GaAs/AlGaAs photonic nano-
wires. Applied Physics Letters, 96, 211117.
Hennrich, M., Legero, T., Kuhn, A., & Rempe, G. (2000). Vacuum-stimulated Raman scat-
tering based on adiabatic passage in a high-finesse optical cavity. Physical Review Letters,
85, 4872.
Hennrich, M., Legero, T., Kuhn, A., & Rempe, G. (2004). Photon statistics of a non-
stationary periodically driven single-photon source. New Journal of Physics, 6, 86.
Hijlkema, M., Weber, B., Specht, H. P., Webster, S. C., Kuhn, A., & Rempe, G. (2007).
A single-photon server with just one atom. Nature Physics, 3, 253.
Hoffges, J. T., Baldauf, H. W., Lange, W., & Walther, H. (1997). Heterodyne measurement
of the resonance fluorescence of a single ion. Journal of Modern Optics, 44, 1999.
Hogele, A., Galland, C., Winger, M., & Imamoglu, A. (2008). Photon antibunching in the
photoluminescence spectra of a single carbon nanotube. Physical Review Letters, 100, 217401.
Holland, M. J., Walls, D. F., & Zoller, P. (1991). Quantum nondemolition measurements of
photon number by atomic beam deflection. Physical Review Letters, 67, 1716.
Holmes, M. J., Choi, K., Kako, S., Arita, M., & Arakawa, Y. (2014). Room-temperature
triggered single photon emission from a III-Nitrite site-controlled nanowire quantum
dot. Nano Letters, 14, 982.
Hong, C. K., & Mandel, L. (1986). Experimental realization of a localized one-photon state.
Hong, C. K., Ou, Z. Y., & Mandel, L. (1987). Measurement of subpicosecond time intervals
between two photons by interference. Physical Review Letters, 59, 2044.
Hours, J., Varoutsis, S., Gallart, M., Bloch, J., Robert-Philip, I., Cavanna, A., et al. (2003). Single
photon emission from individual GaAs quantum dots. Applied Physics Letters, 82, 2206.
Hwang, J., & Hinds, E. A. (2011). Dye molecules as single-photon sources and large optical
nonlinearities on a chip. New Journal of Physics, 13, 085009.
Imamoglu, A., Schmidt, H., Woods, G., & Deutschet, M. (1997). Strongly interacting pho-
tons in a nonlinear cavity. Physical Review Letters, 79, 1467.
Imamoglu, A., & Yamamoto, Y. (1994). Turnstile device for heralded single photons: Cou-
lomb blockade of electron and hole tunneling in quantum confined p-i-n hetero-
junctions. Physical Review Letters, 72, 210.
Jacobs, B. C., Pittman, B., & Franson, J. D. (2006). Single photon source using laser pulses
and two-photon absorption. Physics Review A, 74, 010303.
Jakubczyk, T., Pacuski, W., Smole nski, T., Golnik, A., Florian, M., & Jahnke, F. (2012).
Pronounced Purcell enhancement of spontaneous emission in CdTe/ZnTe quantum
dots embedded in micropillar cavities. Applied Physics Letters, 101, 132105.
Jarjour, A. F., Taylor, R. A., Oliver, R. A., Kappers, M. J., Humphreys, C. J., & Tahraoui, A.
(2007). Cavity-enhanced blue single-photon emission from a single InGaN/GaN quan-
tum dot. Applied Physics Letters, 91, 052101.
Jaynes, E. T., & Cummings, F. W. (1963). Comparison of quantum and semiclassical radi-
ation theories with application to the beam maser. Proceedings of the IEEE, 51, 89.
Jeffrey, E., Peters, N. A., & Kwiat, P. G. (2004). Towards a periodic deterministic source of
arbitrary single-photon states. New Journal of Physics, 6, 100.
Jennewein, T., Simon, C., Weihs, G., Weinfurter, H., & Zeilinger, A. (2000). Quantum
cryptography with entangled photons. Physical Review Letters, 84, 4729.
John, S. (1987). Strong localization of photons in certain disordered dielectric superlattices.
Julsgaard, B., Sherson, J., Cirac, J. I., Fiurasek, J., & Polziket, E. S. (2004). Experimental
demonstration of quantum memory for light. Nature, 432, 482.
Kako, S., Santori, C., Hoshino, K., G otzinger, S., Yamamoto, Y., & Arakawa, Y. (2006).
A gallium nitride single-photon source operating at 200 K. Nature Materials, 5, 887.
Kang, S., Lim, S., Hwang, M., Kim, W., Kim, J.-R., & An, K. (2011). Controlled generation
of single photons in a coupled atom-cavity system at a fast repetition-rate. Optics Express,
19, 2440.
Kaniber, M., Kress, A., Laucht, A., Bichler, M., Meyer, R., Amann, M. C., et al. (2007).
Efficient spatial redistribution of quantum dot spontaneous emission from two-
dimensional photonic crystals. Applied Physics Letters, 91, 061106.
Kaniber, M., Laucht, A., H urlimann, T., Bichler, M., Meyer, R., Amann, M. C., et al.
(2008). Highly efficient single-photon emission from single quantum dots within a
two-dimensional photonic band-gap. Physics Review B, 77, 073312.
Kastner, M. A. (1992). The single-electron transistor. Reviews of Modern Physics, 64, 849.
Keller, O. (2007). Historical papers on the particle concept of light. Progress in Optics, 50, 51.
Keller, M., Lange, B., Hayasaka, K., Lange, W., & Walther, H. (2004a). Continuous gen-
eration of single photons with controlled waveform in an ion-trap cavity system. Nature,
431, 1075.
Keller, M., Lange, B., Hayasaka, K., Lange, W., & Walther, H. (2004b). A calcium ion in a
cavity as a controlled single-photon source. New Journal of Physics, 6, 95.
Kilin, S. Ya. (2001). Quanta and information. Progress in Optics, 42, 1.
Kim, J., Benson, O., Kan, H., & Yamamoto, Y. (1999). A single-photon turnstile device.
Nature, 397, 500.
Kim, M. S., Son, W., Buzek, V., & Knight, P. L. (2002). Entanglement by a beam splitter:
Nonclassicality as a prerequisite for entanglement. Physics Review A, 65, 034343.
Kimble, H. J., Dagenais, M., & Mandel, L. (1977). Photon antibunching in resonance fluo-
rescence. Physical Review Letters, 39, 691.
Kimble, H. J., & Mandel, L. (1976). Theory of resonance fluorescence. Physical Review A, 13,
2123.
Kiraz, A., Ehrl, M., Hellerer, Th., M E., Brauchle, C., & Zumbusch, A. (2005).
ustecaploglu, O.
Indistinguishable photons from a single molecule. Physical Review Letters, 94, 223602.
Kiraz, A., Falth, S., Becher, C., Gayral, B., Schoenfeld, W. V., Petroff, P. M., et al. (2002).
Photon correlation spectroscopy of a single quantum dot. Physics Review B, 65, 161303R.
Kitson, S. C., Jonsson, P., Rarity, J. G., & Tapster, P. R. (1998). Intensity fluctuation spec-
troscopy of small numbers of dye molecules in a microcavity. Physics Review A, 58, 620.
Kleppner, D. (1981). Inhibited spontaneous emission. Physical Review Letters, 47, 233.
Knill, E., Laflamme, R., & Milburn, G. J. (2001). A scheme for efficient quantum compu-
tation with linear optics. Nature, 409, 46.
Kok, P. (2010). Quantum optics: Entangled photons report for duty. Nature Photonics, 4, 504.
Kolesov, R., Grotz, B., Balasubramanian, G., St ohr, R. J., Nicolet, A. A. L., Hemmer, P. R., et al.
(2009). Waveparticle duality of single surface plasmon polaritons. Nature Physics, 5, 470.
Kolesov, R., Xia, K., Reuter, R., St ohr, R., Zappe, A., Meijer, J., et al. (2012). Optical
detection of a single rare-earth ion in a crystal. Nature Communications, 3, 1029.
http://dx.doi.org/10.1038/ncomms2034.
Kramers, H. A. (1957). Quantum mechanics. Amsterdam: North-Holland.
Krause, J., Scully, M. O., & Walther, H. (1987). State reduction and jni-state preparation in a
high-Q micromaser. Physical Review A, 36, 4547.
Krause, J., Scully, M. O., Walther, T., & Walther, H. (1989). Preparation of a pure number
state and measurement of the photon statistics in a high-Q micromaser. Physics Review A,
39, 1915.
Kremer, P. E., Dada, A. C., Kumar, P., Ma, Y., Kumar, S., Clarke, E., et al. (2014).
Strain-tunable quantum dot embedded in a nanowire antenna. Physical Review B, 90,
201408(R).
Kremling, S., Tessarek, C., Dartsch, H., Figge, S., H ofling, S., Worschech, L., et al. (2012).
Single photon emission from InGaN/GaN quantum dots up to 50 K. Applied Physics Let-
ters, 100, 061115.
Kress, A., Hofbauer, F., Reinelt, N., Kaniber, M., Krenner, H. J., Meyer, R., et al. (2005).
Manipulation of the spontaneous emission dynamics of quantum dots in two-
dimensional photonic crystals. Physical Review B, 71, 241304R.
Kuhn, H. (1970). Classical aspects of energy transfer in molecular systems. The Journal of
Chemical Physics, 53, 101.
Kuhn, A., Hennrich, M., Bondo, T., & Rempe, G. (1999). Controlled generation of single
photons from a strongly coupled atom-cavity system. Applied Physics B: Lasers and Optics,
69, 373.
Kuhn, A., Hennrich, M., & Rempe, G. (2002). Deterministic single-photon source for dis-
tributed quantum networking. Physical Review Letters, 89, 067901.
Kumar, R., Ong, J. R., Savanier, M., & Mookherjea, S. (2014). Controlling the spectrum of
photons generated on a silicon nanophotonic chip. Nature Communications. http://dx.doi.
org/10.1038/ncomms6489.
Kurtsiefer, C., Mayer, S., Zarda, P., & Weinfurter, H. (2000). Stable solid-state source of
single photons. Physical Review Letters, 85, 290.
Kwiat, P. G., Mattle, K., Weinfurter, H., Zeilinger, A., Sergienko, A. V., & Shih, Y. (1995).
New high-intensity source of polarization-entangled photon pairs. Physical Review
Letters, 75, 4337.
Kwiat, P. G., Waks, E., White, A. G., Appelbaum, I., & Eberhar, P. H. (1999). Ultrabright
source of polarization-entangled photons. Physical Review A, 60, 773(R).
Kyriienko, O., Shelykh, I. A., & Liew, T. C. H. (2014). Tunable single-photon emission
from dipolaritons. Physics Review A, 90, 033807.
Ladd, T. D., Jelezko, F., Laflamme, R., Nakamura, Y., Monroe, C., & OBrien, J. L. (2010).
Quantum computers. Nature, 464, 45.
Landau, L. D., & Peierls, R. (1930). Quantenelektrodynamik im konfigurationsraum.
Zeitschrift f
ur Physik, 62, 188.
Lasota, M., & Thew, R. (2014). Linear optics schemes for entanglement distribution with
realistic single-photon sources. Physics Review A, 90, 033836.
Laucht, A., P utz, S., Gunthner, T., Hauke, N., Saive, R., Frederick, S., et al. (2012).
A waveguide-coupled on-chip single-photon source. Physics Review X, 2, 011014.
Laurent, S., Varoutsis, S., Le Gratiet, L., Lematre, A., Sagnes, I., Raineri, F., et al. (2005).
Indistinguishable single photons from a single-quantum dot in a two-dimensional pho-
tonic crystal cavity. Applied Physics Letters, 87, 163107.
Law, C. K., & Eberly, J. H. (1996). Arbitrary control of a quantum electromagnetic field.
Law, C. K., & Kimble, H. J. (1997). Deterministic generation of a bit-stream of single-
photon pulses. Journal of Modern Optics, 44, 2067.
Lee, K. G., Chen, X. W., Eghlidi, H., Kukura, P., Lettow, R., Renn, A., et al. (2011).
A planar dielectric antenna for directional single-photon emission and near-unity collec-
tion efficiency. Nature Photonics, 5, 166.
Leistikow, M. D., Mosk, A. P., Yeganegi, E., Huisman, S. R., Lagendijk, A., & Vos, W. L.
(2011). Inhibited spontaneous emission of quantum dots observed in a 3D photonic band
gap. Physical Review Letters, 107, 193903.
Leonski, W. (1996). Fock states in a Kerr medium with parametric pumping. Physics Review
A, 54, 3369.
Leonski, W., & Tanas, R. (1994). Possibility of producing the one-photon state in a kicked
cavity with a nonlinear Kerr medium. Physical Review A, 49, R20.
Lewis, G. N. (1926). The conservation of photons. Nature, 118, 874.
Liebermeister, L., Petersen, F., Munchow, A., Burchardt, D., Hermelbracht, J., Tashima, T.
(2013). Tapered fiber coupling of single photons emitted by a deterministically posi-
tioned single nitrogen vacancy center. ArXiv, 1309.0421v2.
Liew, T. C. H., & Savona, V. (2010). Single photons from coupled quantum modes. Physical
Likharev, K. K. (1999). Single-electron devices and their applications. Proceedings of the IEEE,
87, 4.
Liu, Y-X, Wei, L. F., & Nori, F. (2004). Generation of nonclassical photon states using a
superconducting qubit in a microcavity. Europhysics Letters, 67, 941.
Liu, Y., Chen, G., Song, M., Ci, X., Wu, B., Wu, E., et al. (2013). Fabrication of nitrogen
vacancy color centers by femtosecond pulse laser illumination. Optics Express, 21, 12843.
Lochmann, A., Stock, E., Schulz, O., Hopfer, F., Bimberg, D., Haisler, V. A., et al. (2007).
Electrically driven quantum dot single photon source. Physica Status Solidi (c), 4, 547.
Lodahl, P., Mahmoodian, S., & Stobbe, S. (2013). Interfacing single photons and single quan-
tum dots with photonic nanostructures. arXiv:13121079v1.
Lodahl, P., & Stobbe, S. (2013). Solid-state quantum optics with quantum dots in photonic
nanostructures. Nanophotonics, 2, 39.
Lohrmann, A., Pezzagna, S., Dobrinets, I., Spinicelli, P., Jacques, V., Roch, J.-F., et al.
(2011). Diamond based light-emitting diode for visible single-photon emission at room
temperature. Applied Physics Letters, 99, 251106.
Loudon, R. (2000). The quantum theory of light. Oxford: Oxford University Press.
Lounis, B., Bechtel, H. A., Gerion, D., Alivisatos, P., & Moerner, W. E. (2000). Photon
antibunching in single CdSe/ZnS quantum dot fluorescence. Chemical Physics Letters,
329, 399.
Lounis, B., & Moerner, W. E. (2000). Single photons on demand from a single molecule at
room temperature. Nature, 407, 491.
Lounis, B., & Orrit, M. (2005). Single-photon sources. Reports on Progress in Physics, 68, 1129.
Lukin, M. D., & Imamoglu, A. (2001). Controlling photons using electromagnetically
induced transparency. Nature, 413, 273.
Lukishova, S. G., Bissell, L. J., Menon, V. M., Valappil, N., Hahn, M. A., Evans, C. M., et al.
(2009). Organic photonic bandgap microcavities doped with semiconductor nano-
crystals for room-temperature on-demand single-photon sources. Journal of Modern
Optics, 56, 167.
Lukishova, S. G., Schmid, A. W., Knox, R., Freivald, P., Bissell, L. J., Boyd, R. W., et al.
(2007). Room temperature source of single photons of definite polarization. Journal of
Modern Optics, 54, 417.
Lukishova, S. G., Schmid, A. W., Knox, R. P., Freivald, P., McNamara, A., Boyd, R. W.,
et al. (2006). Single-photon source for quantum information based on single dye mol-
ecule fluorescence in liquid crystal host. Molecular Crystals and Liquid Crystals, 454, 403.
Lukishova, S. G., Schmid, A. W., McNamara, A. J., Boyd, R. W., & Stroud, C. R., Jr.
(2003). Room temperature single-photon source: Single-dye molecule fluorescence
in liquid crystal host. IEEE Journal of Selected Topics in Quantum Electronics, 9, 1512.
Lukishova, S. G., Schmid, A. W., Supranowitz, C. M., Lippa, N., Mcnamara, A. J.,
Boyd, R. W., et al. (2004). Dye-doped cholesteric-liquid-crystal room-temperature
single-photon source. Journal of Modern Optics, 51, 1535.
Lukosz, W., & Kunz, R. E. (1977a). Light emission by magnetic and electric dipoles close
to a plane interface. I. Total radiated power. Journal of the Optical Society of America, 67,
1607.
Lukosz, W., & Kunz, R. E. (1977b). Light emission by magnetic and electric dipoles close
to a plane dielectric interface. II. Radiation patterns of perpendicular oriented dipoles.
Journal of the Optical Society of America, 67, 1615.
Lvovsky, A. I., & Mlynek, J. (2002). Quantum-optical catalysis: Generating nonclassical
states of light by means of linear optics. Physical Review Letters, 88, 250401.
Ma, X.-S., Zotter, S., Kofler, J., Jennewein, T., & Zeilinger, A. (2011). Experimental gen-
eration of single photons via active multiplexing. Physical Review A, 83, 043814.
Matre, X., Hagley, E., Nogues, G., Wunderlich, C., Goy, P., Brune, M., et al. (1997).
Quantum memory with a single photon in a cavity. Physical Review Letters, 79, 769.
Majumdar, A., Bajcsy, M., Rundquist, A., & Vuckovic, J. (2012). Loss-enabled sub-
Poissonian light generation in a bimodal nanocavity. Physical Review Letters, 108, 183601.
Majumdar, A., & Gerace, D. (2013). Single-photon blockade in doubly resonant
nanocavities with second-order nonlinearity. Physical Review B, 87, 235319.
Makhonin, M. N., Dixon, J. E., Coles, R. J., Royall, B., Luxmoore, I. J., Clarke, E., et al.
(2014). Waveguide coupled resonance fluorescence from on-chip quantum emitter.
Nano Letters, 14, 6997.
Maksymov, I. S., Miroshnichenko, A. E., & Kivshar, Y. S. (2012). Plasmonic nanoantennas
for efficient control of polarization-entangled photon pairs. Physical Review A, 86,
011801(R).
Mallek-Zouari, I., Buil, S., Quelin, X., Mahler, B., Dubertret, B., & Hermier, J.-P. (2010).
Plasmon assisted single photon emission of CdSe/CdS nanocrystals deposited on random
gold film. Applied Physics Letters, 97, 053109.
Mandel, L., & Wolf, E. (1995). Optical coherence and quantum optics. Cambridge: Cambridge
University Press.
Matsukevich, D., & Kuzmich, A. (2004). Quantum state transfer between matter and light.
Science, 306, 663.
Maurer, C., Becher, C., Russo, C., Eschner, J., & Blatt, R. (2004). A single-photon source
based on a single Ca + ion. New Journal of Physics, 6, 94.
McCusker, K. T., & Kwiat, P. G. (2009). Efficient optical quantum state engineering. Phys-
McCutcheon, M. W., Chang, D. E., Zhang, Y., Lukin, M. D., & Loncar, M. (2009). Broad-
band frequency conversion and shaping of single photons emitted from a nonlinear cav-
ity. Optics Express, 17, 22689.
McKeever, J., Boca, A., Boozer, A. D., Miller, R., Buck, J. R., Kuzmich, A., et al. (2004).
Deterministic generation of single photons from one atom trapped in a cavity. Science,
303, 1992.
Meschede, D., Walther, H., & M uller, G. (1985). One-atom maser. Physical Review Letters,
54, 551.
Meystre, P. (1992). Cavity quantum optics and the quantum measurement process. Progress in
Optics, 30, 261.
Meystre, P., Rempe, G., & Walther, H. (1988). Very-low-temperature behavior of a mic-
romaser. Optics Letters, 13, 1078.
Meystre, P., Slosser, J. J., & Wilkens, M. (1990). Cat in a cold niobium box. Optics Commu-
nications, 79, 300.
Michler, P., Imamoglu, A., Mason, M. D., Carson, P. J., Strouse, G. F., & Buratto, S. K.
(2000). Quantum correlation among photons from a single quantum dot at room tem-
perature. Nature, 406, 968.
Michler, P., Kiraz, A., Becher, C., Schoenfeld, W. V., Petroff, P. M., Zhang, L., et al. (2000).
A quantum dot single-photon turnstile device. Science, 290, 2282.
Migdall, A. L., Branning, D., & Castelletto, S. (2002). Tailoring single-photon and
multiphoton probabilities of a single-photon on-demand source. Physics Review A, 66,
053805.
Mirin, R. P. (2004). Photon antibunching at high temperature from a single InGaAs/GaAs
quantum dot. Applied Physics Letters, 84, 1260.
Mizuochi, N., Makino, T., Kato, H., Takeuchi, D., Ogura, M., Okushi, H., et al. (2012).
Electrically driven single-photon source at room temperature in diamond. Nature Pho-
tonics, 6, 299.
Moerner, W. E. (2004). Single-photon sources based on single molecules in solids. New Jour-
nal of Physics, 6, 88.
Moerner, W. E., & Fromm, D. P. (2003). Methods of single-molecule fluorescence spectros-
copy and microscopy. The Review of Scientific Instruments, 74, 3597.
Mollow, B. R. (1969). Power spectrum of light scattered by two-level systems. Physics
Review, 188, 1969.
Moreau, E., Robert, J. M., Gerard, J. M., Abram, L., Manin, L., & Thierry-Mieg, V. (2001).
Single-mode solid-state single photon source based on isolated quantum dots in pillar
microcavities. Applied Physics Letters, 79, 2865.
Moreau, E., Robert, J. M., Manin, L., Thierry-Mieg, V., Gerard, J. M., & Abram, I. (2001).
Quantum cascade of photons in semiconductor quantum dots. Physical Review Letters, 87,
183601.
Mosley, P. J., Lundeen, J. S., Smith, B. J., Wasylezyk, P., URen, A. B., Silberhorn, C., et al.
(2008). Heralded generation of ultrafast single photons in pure quantum states. Physical
Muller, A., Fang, W., Lawall, J., & Solomon, G. S. (2009). Creating polarization-entangled
photon pairs from a semiconductor quantum dot using the optical Stark effect. Physical
Naik, D. S., Peterson, C. G., White, A. G., Berglund, A. J., & Kwiat, P. G. (2000). Entangled
state quantum cryptography: Eavesdropping on the Ekert protocol. Physical Review Let-
ters, 84, 4733.
Naiki, H., Masuo, S., Machida, S., & Itaya, A. (2011). Single-photon emission behavior of
isolated CdSe/ZnS quantum dots interacting with the localized surface plasmon reso-
nance of silver nanoparticles. Journal of Physical Chemistry C, 115, 23299.
Neu, E., Arend, C., Gross, E., Guldner, F., Hepp, C., Steinmetz, D., et al. (2011). Narrow-
band fluorescent nanodiamonds produced from chemical vapor deposition films. Applied
Neu, E., Guldner, F., Arend, C., Liang, Y., Ghodbane, S., Sternschulte, H., et al. (2013).
Low temperature investigations and surface treatments of colloidal narrowband fluores-
cent nanodiamonds. Journal of Applied Physics, 113, 203507.
Nisbet-Jones, P. B. R., Dilley, J., Ljunggren, D., & Kuhn, A. (2011). Highly efficient source
for indistinguishable single photons of controlled shape. New Journal of Physics, 13,
103036.
Nisbet-Jones, P. B. R., Dillrey, J., Holleczek, A., Barter, O., & Kuhn, A. (2013). Photonic
qubits, qutrits and ququads accurately prepared and delivered on demand. New Journal of
Physics, 15, 053007.
Nowak, A. K., Portalupi, S. L., Giesz, V., Gazzano, O., Dal Savio, C., Braun, P.-F., et al.
(2014). Deterministic and electrically tunable bright single-photon source. Nature Com-
munications, 5, 3240. http://dx.doi.org/10.1038/ncomms4240.
Oppenheimer, J. R. (1931). Note on light quanta and the electromagnetic field. Physics
Review, 38, 725.
Oxborrow, M., & Sinclair, A. G. (2005). Single-photon sources. Contemporary Physics,
46, 173.
Parkins, A. S., Marte, P., Zoller, P., Carnal, O., & Kimble, H. J. (1995). Quantum-
state mapping between multilevel atoms and cavity light fields. Physics Review A,
51, 1578.
Parkins, A. S., Marte, P., Zoller, P., & Kimble, H. J. (1993). Synthesis of arbitrary
quantum states via adiabatic transfer of Zeeman coherence. Physical Review Letters, 71,
3095.
Patel, R. B., Bennett, A. J., Farrer, I., Nicoll, C. A., Ritchie, D. A., & Shields, A. J. (2010).
Two-photon interference of the emission from electrically tunable remote quantum
dots. Nature Photonics, 4, 632.
Pauli, W. (1933). Principles of Wave Mechanics. Handbuch der Physik (Vol. 24).
Pegg, D. T., Phillips, L. S., & Barnett, S. M. (1998). Optical state truncation by projection
synthesis. Physical Review Letters, 81, 1604.
Pelton, M., Santori, C., Vuckovic, J., Zhang, B., Solomon, G. S., Plant, J., et al. (2002).
Efficient source of single photons: A single quantum dot in a micropost microcavity.
Perina, J. (1984). Quantum statistics of linear and nonlinear optical phenomena. Dordrecht:
D. Reidel Publishing Company.
Perina, J., Hradil, Z., & Jur, B. (1994). Quantum optics and fundamentals of physics. London:
Kluwer Academic Publishers.
Pinkse, P. W. H., Fischer, T., Maunz, P., & Rempe, G. (2000). Trapping an atom with single
photons. Nature, 404, 365.
Pittman, T. B., Franson, J. D., & Jacobs, B. C. (2007). Investigation of a single-photon source
based on quantum interference. New Journal of Physics, 9, 195.
Pittman, T. B., Jacobs, B. C., & Franson, J. D. (2002). Single photons on pseudodemand
from stored parametric down-conversion. Physics Review A, 66, 042303.
Planck, M. (1900a). On an improvement of Wiens equation for the spectrum. Verhandlungen
der Deutsche Physikalische Gesellschaft, 2, 202.
Planck, M. (1900b). On the theory of the energy distribution law of the normal spectrum.
Verhandlungen der Deutsche Physikalische Gesellschaft, 2, 237.
Poddubny, A. N., Ginzburg, P., Belov, P. A., Zayats, A. V., & Kivshar, Y. S. (2012). Tai-
loring and enhancing spontaneous two-photon emission using resonant plasmonic
nanostructures. Physics Review A, 86, 033826.
Politi, A., Cryan, M. J., Rarity, J. G., Yu, S., & OBrien, J. L. (2008). Silica-on-silicon wave-
guide quantum circuits. Science, 320, 646.
Press, D., Gotzinger, S., Reitzenstein, S., Hofmann, C., L offler, A., Kamp, M., et al. (2007).
Photon antibunching from a single quantum-dot-microcavity system in the strong cou-
pling regime. Physical Review Letters, 98, 117402.
Purcell, E. M. (1946). Spontaneous emission probabilities at radio frequencies. Physics Review,
69, 681.
Quartieri, A., Morello, G., Spinicelli, P., Todaro, M. T., Stomeo, T., Martiradonna, L., et al.
(2009). Nonclassical emission from single colloidal nanocrystals in a microcavity:
A route towards room temperature single photon sources. New Journal of Physics, 11,
033025.
Rabeau, J. R., Chin, Y. L., Prawer, S., Jelezko, F., Gaebel, T., & Wrachtrup, J. (2005). Fab-
rication of single nickel-nitrogen defects in diamond by chemical vapor deposition.
Raimond, J. M., Brune, M., & Haroche, S. (2001). Manipulating quantum entanglement
with atoms and photons in a cavity. Reviews of Modern Physics, 73, 565.
Raimond, J. M., Brune, M., Lepape, J., & Haroche, S. (1989). Rydberg atoms in a cavity:
A new method to generate photon number states. In M. S. Feld, J. E. Thomas, &
A. Mooradian (Eds.), Laser spectroscopy IX (p. 140). New York: Academic Press.
Raimond, J. M., Meunier, T., Bertet, P., Gleyzes, S., Maioli, P., Auffeves, A., et al. (2005).
Probing a quantum field in a photon box. Journal de Physique, B38, S535.
Rakher, M. T., Ma, L., Davanco, M., Slattery, O., Tang, X., & Srinivasan, K. (2011).
Simultaneous wavelength translation and amplitude modulation of single photons from
a quantum dot. Physical Review Letters, 107, 083602.
Rakher, M. T., Ma, L., Slattery, O., Tang, X., & Srinivasan, K. (2010). Quantum transduc-
tion of telecommunications-band single photons from a quantum dot by frequency
upconversion. Nature Photonics, 4, 786.
Rakher, M. T., & Srinivasan, K. (2011). Subnanosecond electro-optic modulation of
triggered single photons from a quantum dot. Applied Physics Letters, 98, 211103.
Rakher, M. T., Stoltz, N. G., Coldren, L. A., Petroff, P. M., & Bouwmeester, D. (2008).
Polarization-switchable single photon source using the Stark effect. Applied Physics
Letters, 93, 091118.
Rarity, J. G., Tapster, P. R., & Jakeman, E. (1987). Observation of sub-Poissonian light in
parametric downconversion. Optics Communications, 62, 201.
Regelman, D. V., Mizrahi, U., Gershoni, D., Ehrenfreund, E., Schoenfeld, W. V., &
Petroff, P. M. (2001). Semiconductor quantum dot: A quantum light source of multi-
color photons with tunable statistics. Physical Review Letters, 87, 257401.
Reimer, M. E., Bulgarini, G., Akopian, N., Hocevar, M., Bavinck, M. B., Verheijen, M. A.,
et al. (2012). Bright single-photon sources in bottom-up tailored nanowires. Nature Com-
munications, 3, 737. http://dx.doi.org/10.1038/1746(2012).
Reinhard, A., Volz, T., Winger, M., Badolato, A., Hennessy, K. J., Hu, E. L., et al. (2012).
Strongly correlated photons on a chip. Nature Photonics, 6, 93.
Reischle, M., Beirne, G. J., Schulz, W.-M., Eichfelder, M., Robach, R., Jetter, M., et al.
(2008). Electrically pumped single-photon emission in the visible spectral range up to
80 K. Optics Express, 16, 12771.
Reischle, M., Kessler, C., Schulz, W.-M., Eichfelder, M., Robach, R., Jetter, M., et al.
(2010). Triggered single-photon emission from electrically excited quantum dots in
the red spectral range. Applied Physics Letters, 97, 143513.
Reiserer, A., & Rempe, G. (2014). Cavity-based quantum networks with single atoms and
optical photons. ArXiv, 1412.2889v1.
Rivas, D., Munoz-Matutano, G., Canet-Ferrer, J., Garca-Calzada, R., Trevisi, G.,
Seravalli, L., et al. (2014). Two-color single-photon emission from InAs quantum dots:
Toward logic information management using quantum light. Nano Letters, 14, 456.
Rogers, L. J., Doherty, M. W., Barson, M. S. J., Onoda, S., Ohshima, T., & Manson, N. B.
(2015). Singlet levels of the NV-centre in diamond. New Journal of Physics, 17, 013048.
Robach, R., Reischle, M., Beirne, G. J., Jetter, M., & Michler, P. (2008). Red single-
photon emission from an InP/GaInP quantum dot embedded in a planar monolithic
microcavity. Applied Physics Letters, 92, 071105.
Sandstrom, R. G., Shimoni, O., Martin, A. A., & Aharonovich, I. (2014). Study of narrowband
single photon emitters in polycrystalline diamond films. Applied Physics Letters, 105, 181104.
Santori, C., Fattal, D., Pelton, M., Solomon, G. S., & Yamamoto, Y. (2002). Polarization-
correlated photon pairs from a single quantum dot. Physics Review B, 66, 045308.
Santori, C., Fattal, D., Vuckovic, J., Solomon, G. S., & Yamamoto, Y. (2002). Indistinguish-
able photons from a single-photon device. Nature, 419, 594.
Santori, C., Fattal, D., Vuckovic, J., Solomon, G. S., & Yamamoto, Y. (2004). Single-
photon generation with InAs quantum dots. New Journal of Physics, 6, 89.
Santori, C., Fattal, D., & Yamamoto, Y. (2010). Single-photon devices and applications. Leipzig:
Wiley.
Santori, C., Pelton, M., Solomon, G. S., Dale, Y., & Yamamoto, Y. (2001). Triggered single
photons from a quantum dot. Physical Review Letters, 86, 1502.
Santori, C., Solomon, G. S., Pelton, M., & Yamamoto, Y. (2002). Time-resolved spectroscopy
of multiexcitonic decay in an InAs quantum Dot. Physics Review B, 65, 073310.
Savage, C. M., Braunstein, S. L., & Walls, D. F. (1990). Macroscopic quantum superpositions
by means of single-atom dispersion. Optics Letters, 15, 628.
Sayrin, C., Dotsenko, I., Gleyzes, S., Brune, M., Raimond, J. M., & Haroche, S. (2012).
Optimal time-resolved photon number distribution reconstruction of a cavity field by
maximum likelihood. New Journal of Physics, 14, 115007.
Scheel, S. (2009). Single-photon sourcesAn introduction. Journal of Modern Optics,
56, 141.
Scheel, S., Florescu, M., Haffner, H., Lee, H., Strekalov, D. V., Knight, P. L., et al. (2007).
Single photons on demand from tunable 3D photonic band-gap structures. Journal of
Modern Optics, 54, 409.
Schietinger, S., Barth, M., Aichele, T., & Benson, O. (2009). Plasmon-enhanced single pho-
ton emission from a nanoassembled metal-diamond hybrid structure at room tempera-
ture. Nano Letters, 9, 1694.
Schmidgall, E. R., Schwartz, I., Gantz, L., Cogan, D., Raindel, S., & Gershoni, D. (2014).
Deterministic generation of a quantum-dot-confined triexciton and its radiative decay
via three-photon cascade. Physical Review B, 90, 241411(R).
Schroder, T., Gadeke, F., Banholzer, M. J., & Benson, O. (2011). Ultrabright and efficient
single-photon generation based on nitrogen-vacancy centres in nanodiamonds on a solid
immersion lens. New Journal of Physics, 13, 055017.
Schwagmann, A., Kalliakos, S., Farrer, I., Griffiths, J. P., Jones, G. A. C., Ritchie, D. A., et al.
(2011). On-chip single photon emission from an integrated semiconductor quantum dot
into a photonic crystal waveguide. Applied Physics Letters, 99, 261108.
Sciscione, L., Centini, M., Sibilia, C., Bertolotti, M., & Scalora, M. (2006). Entangled,
guided photon generation in (1+1)-dimensional photonic crystals. Physics Review A,
74, 013815.
Scully, M. O., & Zubairy, M. S. (1997). Quantum optics. Cambridge: Cambridge University
Press.
Sebald, K., Michler, P., Passow, T., Hommel, D., Bacher, G., & Forchel, A. (2002).
Single-photon emission of CdSe quantum dots at temperatures up to 200K. Applied
Settimi, A., Severini, S., Centini, M., Sibilia, C., Bertolotti, M., Napoli, A., et al. (2005).
Coherent control of stimulated emission inside one-dimensional photonic crystals.
Physics Review E, 71, 066606.
Shan, G.-C., Yin, Z.-Q., Shek, C. H., & Huang, W. (2014). Single photon sources with
single semiconductor quantum dots. Frontiers of Physics, 9, 170.
Shapiro, J. H., & Wong, F. N. (2007). On-demand single-photon generation using a
modular array of parametric downconverters with electro-optic polarization controls.
Optics Letters, 32, 2698.
Shen, Y., Bradford, M., & Shen, J.-T. (2011). Single-photon diode by exploiting the photon
polarization in a waveguide. Physical Review Letters, 107, 173902.
Shields, A. J. (2007). Semiconductor quantum light sources. Nature Photonics, 1, 215.
Short, R., & Mandel, L. (1983). Observation of sub-Poissonian photon statistics. Physical
Simon, J., Tanji, H., Ghosh, S., & Vuletic, V. (2007). Single-photon bus connecting spin-
wave quantum memories. Nature Physics, 3, 765.
Singh, S. (1983). Antibunching, sub-Poissonian photon statistics and finite bandwidth effects
in resonance fluorescence. Optics Communications, 44, 254.
Slosser, J. J., Meystre, P., & Wright, E. M. (1990). Generation of macroscopic superpositions
in a micromaser. Optics Letters, 15, 233.
Smolyaninov, I. I., Zayats, A. V., Gungor, A., & Davis, C. C. (2002). Single-photon tunnel-
ing via localized surface plasmons. Physical Review Letters, 88, 187402.
Solomon, G. S., Pelton, M., & Yamamoto, Y. (2001). Single-mode spontaneous emission
from a single quantum dot in a three-dimensional microcavity. Physical Review Letters,
86, 3903.
Sotier, F., Thomay, T., Hanke, T., Korger, J., Mahapatra, S., Frey, A., et al. (2009). Fem-
tosecond few-fermion dynamics and deterministic single-photon gain in a quantum dot.
Nature Physics, 5, 352.
Soujaeff, A., Nishioka, T., Hasegawa, T., Takeuchi, S., Tsurumaru, T., Sasaki, K., et al.
(2007). Quantum key distribution at 1550 nm using a pulse heralded single photon
source. Optics Express, 15, 726.
Spring, J. B., Salter, P. S., Metcalf, B. J., Humphreys, P. C., Moore, M., Thomas-Peter, N.,
et al. (2013). On-chip low loss heralded source of pure single photons. Optics Express, 21,
13522.
Srivathsan, B., Gulati, G. K., Chng, B., Maslennikov, G., Matsukevich, D., & Kurtsiefer, C.
(2013). Narrow band source of transform-limited photon pairs via four-wave mixing in a
cold atomic ensemble. Physical Review Letters, 111, 123602.
Steiner, M., Hartschuh, A., Korlacki, R., & Meixner, A. J. (2007). Highly efficient,
tunable single photon source based on single molecules. Applied Physics Letters, 90,
183122.
Stoler, D., & Yurke, B. (1986). Generating antibunched light from the output of a non-
degenerate frequency converter. Physics Review A, 34, 3143.
Strauf, S., Michler, P., Klude, M., Hommel, D., Bacher, G., & Forchel, A. (2002). Quantum
optical studies on individual acceptor bound excitons in a semiconductor. Physical Review
Letters, 89, 177403.
Strauf, S., Stoltz, N. G., Rakher, M. T., Coldren, L. A., Petroff, P. M., & Bouwmeester, D.
(2007). High frequency single photon source with polarization control. Nature Photonics,
1, 704.
Strekalov, D. V., Y-Ho, Kim, & Shih, Y. (1999). Experimental study of a subsystem in an
entangled two-photon state. Physical Review A, 60, 2685.
Takemoto, K., Sakuma, Y., Hirose, S., Usuki, T., Yokoyama, N., Miyazawa, T., et al.
(2004). Non-classical photon emission from a single InAs/InP quantum dot in the
1.3-m optical-fiber band. Japanese Journal of Applied Physics, 43, L993.
Takemoto, K., Takatsu, M., Hirose, S., Yokoyama, N., Sakuma, Y., Usuki, T., et al. (2007).
An optical horn structure for single-photon source using quantum dots at telecommu-
nication wavelengths. Journal of Applied Physics, 101, 081720.
Tame, M. S., Lee, C., Lee, J., Ballester, D., Paternostro, M., Zayats, A. V., et al. (2008). Single-
photon excitation of surface plasmon polaritons. Physical Review Letters, 101, 190504.
Tanji, H., Ghosh, S., Simon, J., Bloom, B., & Vuleti, V. (2009). Heralded single-magnon
quantum memory for photon polarization states. Physical Review Letters, 103, 043601.
Taylor, G. I. (1909). Interference fringes with feeble light. Proceedings of the Cambridge Phil-
osophical Society, 15, 114.
Thompson, R. M., Stevenson, R. M., Shields, A. J., Farrer, I., Lobo, C. J., Ritchie, D. A.,
et al. (2001). Single-photon emission from exciton complexes in individual quantum
dots. Physical Review B, 64, 201302(R).
Thomson, J. J. (1907). On the ionization of gases by ultra-violet light and on the evidence as
to the structure of light afforded by its electrical effects. Proceedings of the Cambridge Phil-
osophical Society, 14, 417.
Tittel, W., Brendel, J., Zbinden, H., & Gisin, N. (2000). Quantum cryptography using
entangled photons in energy-time bell states. Physical Review Letters, 84, 4737.
Torres, J. P., Banaszek, K., & Walmsley, I. A. (2011). Engineering nonlinear optic sources of
photonic entanglement. Progress in Optics, 56, 227.
Treussart, F., Alleaume, R., Le Floch, V., Xiao, L. T., Courty, J.-M., & Roch, J.-F. (2002).
Direct measurement of the photon statistics of a triggered single photon source. Physical
Treussart, F., Clouqueur, A., Grossman, C., & Roch, J.-F. (2001). Photon antibunching in
the fluorescence of a single dye molecule embedded in a thin polymer film. Optics Letters,
26, 1504.
Tribu, A., Sallen, G., Aichele, T., Andre, R., Poizat, J.-P., Bougerol, C., et al. (2008).
A high-temperature single-photon source from nanowire quantum dots. Nano Letters,
8, 4326.
URen, A. B., Silberhorn, C., Banaszek, K., & Walmsley, I. A. (2004). Efficient conditional
preparation of high-fidelity single photon states for fiber-optic quantum networks. Phys-
Ugur, A., Kremling, S., Hatami, F., H ofling, S., Worschech, L., Forchel, A., et al. (2012).
Single-photon emitters based on epitaxial isolated InP/InGaP quantum dots. Applied
Ulhaq, A., Weiler, S., Ulrich, S. M., Robach, R., Jetter, M., & Michler, P. (2012). Cas-
caded single-photon emission from the Mollow triplet sidebands of a quantum dot.
Unitt, D. C., Bennett, A. J., Atkinson, P., Cooper, K., See, P., Gevaux, D., et al. (2005).
Quantum dots as single-photon sources for quantum information processing. Journal
of Optics B: Quantum and Semiclassical Optics, 7, S129.
Unitt, D. C., Bennett, A. J., Atkinson, P., Ritchie, D. A., & Shields, A. J. (2005). Polarization
control of quantum dot single-photon sources via a dipole-dependent Purcell effect.
Vamivakas, A. N., Saleh, B. E. A., Sergienko, A. V., & Teich, M. C. (2004). Theory of spon-
taneous parametric down-conversion from photonic crystals. Physics Review A, 70,
043810.
Vamivakas, A. N., Zhao, Y., Lu, C.-Y., & Atat ure, M. (2009). Spin-resolved quantum-dot
resonance fluorescence. Nature Physics, 5, 198.
Van der Sar, T., Hagemeier, J., Pfaff, W., Heeres, E. C., Thon, S. M., Kim, H., et al. (2011).
Deterministic nanoassembly of a coupled quantum emitter-photonic crystal cavity sys-
tem. Applied Physics Letters, 98, 193103.
Van der Sar, T., Heeres, E. C., Dmochowski, G. M., de Lange, G., Robledo, L.,
Oosterkamp, T. H., et al. (2009). Nanopositioning of a diamond nanocrystal containing
a single nitrogen-vacancy defect center. Applied Physics Letters, 94, 173104.
Varcoe, B. T. H., Brattke, S., Weidinger, M., & Walther, H. (2000). Preparing pure photon
number states of the radiation field. Nature, 403, 743.
Vasilev, G. S., Ljunggren, D., & Kuhn, A. (2010). Single photons made-to-measure. New
Verger, A., Ciuti, C., & Carusotto, I. (2006). Polariton quantum blockade in a photonic dot.
Versteegh, M. A. M., Reimer, M. E., Joens, K. D., Dalacu, D., Poole, P. J., Gulinatti, A.,
et al. (2014). Observation of strongly entangled photon pairs from a nanowire quantum
dot. Nature Communications. http://dx.doi.org/10.1038/ncomms6298.
Vogel, K., Akulin, V. M., & Schleich, W. P. (1993). Quantum state engineering of the radi-
ation field. Physical Review Letters, 71, 1816.
Vogel, W., & Wellentowitz, S. (2001). In J. Perina (Ed.), Coherence and statistics of photons and
atoms (p. 333). New York: Wiley.
Volz, J., Kurtsiefer, C., & Weinfurter, H. (2001). Compact all-solid-state source of
polarization-entangled photon pairs. Applied Physics Letters, 79, 869.
Vuckovic, J., Fattal, D., Santori, C., Solomon, G. S., & Yamamoto, Y. (2003). Enhanced
single-photon emission from a quantum dot in a micropost microcavity. Applied Physics
Letters, 82, 3596.
Waks, E., Diamanti, E., & Yamamoto, Y. (2006). Generation of photon number states. New
Waks, E., Santori, C., & Yamamoto, Y. (2002). Security aspects of quantum key distribution
with sub-Poisson light. Physics Review A, 66, 042315.
Walls, D. F. (1979). Evidence for the quantum nature of light. Nature, 280, 451.
Walls, D. F., & Milburn, G. J. (1995). Quantum optics. Berlin: Springer.
Walmsley, I. A., & Raymer, M. G. (2005). Toward quantum-information processing with
photons. Science, 307, 1733.
Walther, H. (1998). Single atom experiments in cavities and traps. Proceedings of the Royal Soci-
ety of London, Series A, 454, 431.
Walther, H. (2002). Generation and detection of Fock states of the radiation field. Journal of
Optics B: Quantum and Semiclassical Optics, 4, S418.
Walther, H. (2004). Generation of photon number states on demand. Journal of Modern Optics,
51, 1859.
Wang, X.-H., Gu, B.-Y., Wang, R., & Xu, H.-Q. (2003). Decay kinetic properties of atoms
in photonic crystals with absolute gaps. Physical Review Letters, 91, 113904.
Wang, C., Kurtsiefer, C., Weinfurter, H., & Burchard, B. (2006). Single photon emission
from SiV centres in diamond produced by ion implantation. Journal of Physics B: Atomic,
Molecular and Optical Physics, 39, 37.
Wang, D., Li, M., Zhu, F., Yin, Z.-Q., Chen, W., Han, Z.-F., et al. (2014). Quantum key
distribution with the single-photon-added coherent source. Physics Review A, 90,
062315.
Wang, J., Santamato, A., Jiang, P., Bonneau, D., Engin, E., Silverstone, J. W., et al. (2014).
Gallium arsenide (GaAs) quantum photonic waveguide circuits. Optics Communications,
327, 49.
Ward, M. B., Farrow, T., See, P., Yuan, Z. L., Karimov, O. Z., Bennett, A. J., et al. (2007).
Electrically driven telecommunication wavelength single-photon source. Applied Physics
Letters, 90, 063512.
Weidinger, M., Varcoe, B. T. H., Heerlein, R., & Walther, H. (1999). Trapping states in the
micromaser. Physical Review Letters, 82, 3795.
Weiler, S., Stojanovic, D., Ulrich, S. M., Jetter, M., & Michler, P. (2013). Postselected indis-
tinguishable single-photon emission from the Mollow triplet sidebands of a resonantly
excited quantum dot. Physical Review B, 87, 241302(R).
Weisskopf, V. (1931). Zur theorie der resonanzfluoreszenz. Annalen der Physik (Leipzig), 9, 23.
Werner, M. J., & Imamoglu, A. (1999). Photon-photon interactions in cavity electromag-
netically induced transparency. Physical Review A, 61, 011801(R).
Wilk, T., Webster, S. C., Specht, H. P., Rempe, G., & Kuhn, A. (2007). Polarization-
controlled single photons. Physical Review Letters, 98, 063601.
Wolters, J., Schell, A. W., Kewes, G., N usse, N., Schoengen, M., D oscher, H., et al. (2010).
Enhancement of the zero phonon line emission from a single nitrogen vacancy center in a
nanodiamond via coupling to a photonic crystal cavity. Applied Physics Letters, 97, 141108.
Wootters, W. K., & Zurek, W. H. (1982). A single quantum cannot be cloned. Nature, 299, 802.
Wu, E., Jacques, V., Zeng, H., Grangier, P., Treussart, F., & Roch, J.-F. (2006). Narrow-
band single-photon emission in the near infrared for quantum key distribution. Optics
Express, 14, 1296.
Wu, E., Rabeau, J. R., Roger, G., Treussart, F., Zeng, H., Grangier, P., et al. (2007). Room
temperature triggered single-photon source in the near infrared. New Journal of Physics,
9, 434.
XingSheng, X., Xu, X., Chen, S., & Yamada, T. (2010). Antibunching and blinking from
a single colloidal CdSe quantum dot. Science China Physics, Mechanics & Astronomy,
53, 1619.
Xu, X., Chen, H., Yamada, T., & Otomo, A. (2008). Single-photon source in strong
photon-atom interaction regime in quasiperiodic photonic crystals. Europhysics Letters,
84, 67003.
Yablonovitch, E. (1987). Inhibited spontaneous emission in solid-state physics and electron-
ics. Physical Review Letters, 58, 2059.
Yamamoto, Y., Machida, S., Horikoshi, Y., & Igeta, K. (1991). Enhanced and inhibited
spontaneous emission of the free excitons in GaAs quantum wells in microcavity. Optics
Communications, 80, 337.
Yang, Y., Fleischhauer, M., & Zhu, S.-Y. (2003). Spontaneous emission from a two-level
atom in two-band anisotropic photonic crystals. Physics Review A, 68, 043805.
Yang, Y., & Zhu, S. Y. (2000). Spontaneous emission from a three-level atom in two-band
photonic crystals. Journal of Modern Optics, 47, 1513.
Yokoyama, H., Nishi, K., Anan, T., Yamada, H., Boorson, S. D., & Ippen, P. (1990).
Enhanced spontaneous emission from GaAs quantum wells in monolithic microcavities.
Young, R. J., Stevenson, R. M., Atkinson, P., Cooper, K., Ritchie, D. A., & Shields, A. J.
(2006). Improved fidelity of triggered entangled photons from single quantum dots. New
Young, R. J., Stevenson, R. M., & Shields, A. J. (2007). Entangled photons from the
biexciton cascade of quantum dots. Journal of Applied Physics, 101, 081711.
Yuan, Z., Kardynal, B. E., Stevenson, R. M., Shields, A. J., Lobo, C. J., Cooper, K., et al.
(2002). Electrically driven single-photon source. Science, 295, 102.
Yuen, H. P. (1986). Generation, detection, and application of high-intensity photon
number-eigenstate fields. Physical Review Letters, 56, 2176.
Zaitsev, A. M. (2001). Optical properties of diamond: A data handbook. Berlin, New York:
Springer Science & Business Media.
Zavatta, A., Viciani, S., & Bellini, M. (2004). Quantum-to-classical transition with single
photon-added coherent states of light. Science, 306, 660.
Zhang, Q., Dang, C., Urabe, H., Wang, J., Sun, S., & Nurmikko, A. (2008). Large ordered
arrays of single photon sources based on IIVI semiconductor colloidal quantum dot.
Optics Express, 16, 19592.
Zinoni, C., Alloing, B., Monat, C., Zwiller, V., Li, L. H., Fiore, A., et al. (2006). Time-
resolved and antibunching experiments on single quantum dots at 1300 nm. Applied
Zwiller, V., Blom, H., Jonsson, P., Panev, N., Jeppesen, S., Tsegaye, T., et al. (2001). Single
quantum dots emit single photons at a time: Antibunching experiments. Applied Physics
Letters, 78, 2476.
CHAPTER TWO
Recent Advances in Optical Image

Processing
Ayman Alfalou*, Christian Brosseau
*Vision ISEN-BREST Laboratory L@BISEN, Institut Superieur de lElectronique et du Numerique,
Brest, France
Laboratoire des Sciences et Techniques de lInformation, de la Communication et de la Connaissance,

Universite de Brest, Brest, France
Contents
1. General Introduction 119
2. Part A: Digital Versus Optical Processing 121
2.1 Introduction 121
2.2 From Light to Image Formation 122
2.3 Digital Image Processing 124
2.4 Optical Image Processing 141
3. Part B: OIP 169
3.1 Optical Fourier Transform for Medical Image Processing and Phase Contrast
Imaging 169
3.2 Optical Implementation of Mathematical Operations 180
3.3 Techniques for Increasing Image Quality 187
3.4 Using Polarization for Increasing Image Quality and Edge Detection 208
3.5 Techniques for Image Denoising, Fusion, and Reconstruction 231
3.6 Techniques for Image Transmission: Optical Compression and/or Optical
Encryption Methods 234
3.7 Summary and Future Prospects 245
Acknowledgments 246
References 247
1. GENERAL INTRODUCTION
Historically, research on image processing (IP) techniques brings
together many facets of optics and information theory and presents scientists
and engineers with an extremely important question: How can informa-
tion be encoded in an optical image? In recent years, techniques of optical

120 Ayman Alfalou and Christian Brosseau
image processing (OIP) have demonstrated the utility of their unique prop-
erties in diverse applications such as compression and encryption (Alfalou &
Brosseau, 2009), holography (Nehmetallah & Banerjee, 2012; Testorf &
Lohmann, 2008), and identification and tracking (Alfalou & Brosseau,
2010a, 2010b; Alfalou, Brosseau, & Alam, 2013; Rosen, Katz, &
Brooker, 2009). As a result, researchers and engineers in the United States
and worldwide have invested significant effort in improving optical
processing (OP) of information technology. Breakthroughs have been
achieved by the well-documented correlation technique (Alfalou &
Brosseau, 2010a, 2010b; Alfalou, Brosseau, & Alam, 2013). In spite of the
aforementioned achievements, optical processing techniques continue to
suffer from the point of view of optical implementation.
While images are originally optical, digital processing (DP) is often real-
ized to fully exploit their information content. As the resources required for
all-optical processing come within experimental reach, it is desirable to
develop a toolbox sufficiently versatile to allow the implementation of a
wide class of optical schemes. Within this context, our aim is to review
the recent progress made in the field of OP of information. More specifi-
cally, we will consider techniques allowing us to increase image quality to
render them more useful for correlation and reconstruction applications.
The overall goal of this chapter is to review, from both historical and tech-
nical perspectives, various important facts that may have escaped the attention
of those interested in this area. Another aim of the review is to highlight
aspects in understanding the developments in optical setup that are particularly
relevant to OIP. In view of the magnitude of the area, which spans nearly half
a century of technical work in the field, it was not possible to cover every con-
ceivable subject. There have in the past been attempts to review topics rele-
vant to OIP and digital image processing (DIP), and some of these contain
useful information and perspectives. All however, due to the inexorable evo-
lution of the subject as researchers have struggled toward the ultimate goal of
understanding, and due to the vastness of the task, are of limited scope, and
clearly localized in time. As far as we are aware, there has not been a substantial
review in which correlation applications of OIP are the central topic. Taking
this into account, we apologize beforehand for the omission of important and
interesting developments that are beyond the scope of this chapter. Although
it is not possible to cover every topic in complete detail, sufficient details are
provided to make the review as self-contained and pedagogical as possible.
The structure of the chapter is as follows. In the first part, techniques for
increasing image quality are described in detail. As a point of interest, those
Recent Advances in Optical Image Processing 121
dealing with color encoding are addressed. In addition, methods for den-
oising images are dealt with. The remaining material fleshes out the advan-
tages and limitations of the optical methods by providing the requisite
details. The second part deals with a discussion of some current develop-
ments in OIP and their practical implementation. We then discuss central
theoretical and experimental aspects of polarization encoding methods.
Throughout this review, many examples illustrating the performances of
these techniques are given. These results illustrate the lesson that polarization
optics coupled to OIP can make important contributions to objects detec-
tion and recognition. Finally, we hope to communicate our sense of future
vistas for progress in this rapidly expanded field.
2. PART A: DIGITAL VERSUS OPTICAL PROCESSING

2.1 Introduction
The need for reliable technologies of object recognition and for extraction
of landmarks and features is of paramount importance for numerous govern-
mental and commercial applications such as classification, access control,
biometrics, and security systems. As state-of-the art face recognition tech-
nology becomes more complex with, for instance, security and privacy
requirements, there is a need to improve on existing techniques in order
to fully satisfy the latter. The current status of theoretical and experimental
research is that many optical processing techniques for detecting and recog-
nizing objects are available, ranging from the mundane to the very sophis-
ticated. In spite of this proliferation of techniques, their all-optical
implementations are rather complex to develop. Thus, alternate hybrid
methods (numerical optical using an optoelectronic interface) have been
proposed as substitutes for all-optical techniques.
These approaches, however, are not the universal panacea and have their
drawbacks and stringent requirements, i.e., aberration effects, alignment of
components, limitation of the overall speed by how fast the information can
be updated on the input and output devices, and need of a costly optoelec-
tronic interface. Furthermore, use of all-optics cannot be justified for many
applications, especially when the target image size is small. We note also that
while an all-optical approach can be well controlled, it has also severe
limitations as was shown in a recent paper dealing with a surveillance video
system which detects various posture-based events; i.e., the target image size
should be small and the video rate should not be excessive.
Interest in the field of correlation techniques has been recently rekindled

due to their high discriminating power, due to their high robustness against
various types of noise, and because they allow us to simultaneously identify
and determine the spatial position of specific images in a scene. Among the
different kinds of correlation optimization techniques suggested in the
literature, the design of correlation filters adapted to their corresponding
application is of utmost importance. In recent years, hybrid techniques,
i.e., numerical implementation of correlation, have been considered an
alternative to all-optical methods because they show a good compromise
between performance and simplicity.
The subsequent sections in this part are arranged as follows. We give a
brief review of image formation in Section 2.2. Section 2.3 contains a dis-
cussion of some of the potential advantages versus drawbacks of the digital
method of IP. Section 2.4 is devoted to a brief presentation of the advantages
and disadvantages of OIP approaches.
2.2 From Light to Image Formation

It may be worthwhile to recall several important facts concerning image
formation and processing. These include, but are not limited to the follow-
ing: (1) light propagates in space and reflects from objects; (2) colors are a
consequence of absorption; (3) the human eye is one of the most sophis-
ticated optical detectors since it can detect around 5001000 shades of gray;
(4) objects can be seen by detecting the light deflected from them; (5)
images are two-dimensional planar projections of light intensity distribu-
tions; and (6) IP deals with images which have been recorded in the
presence of one or more degradation (noise) mechanisms such that
the recorded image no longer has a one-to-one correspondence with
the object space.
In a seminal paper, Harris (1966) showed that the evaluation of an image
depends upon the purpose for which the image was obtained and the man-
ner in which the image is to be examined. Image degradation always reduces
the information content of an image. Harris (1966) suggested a way to bypass
the issue of degradation by employing preprocessing treatments1 that are
specific to the image and the envisioned application. The pioneering
approach developed by Harris (1966) was limited to the case of image
1
Preprocessing images commonly involves removing low-frequency background noise, normalizing
the intensity of the individual particles images, removing reflections, and masking portions of images.
Image preprocessing is the technique of enhancing data images prior to computational processing.
degradations of the linear homogeneous type, i.e., those in which the point
spread function characterizing the degradation is invariant over the image
plane to be examined, and to the case in which the image spectrum is
not band-limited. Within these limitations, the degraded image Id can be
described by the convolution integral

Id x, y I x, ysI x, y I x x0 ,y y0 sI x0 , y0 dx0 dy0 , (1)
1
where is the symbol of convolution, I denotes the ideal image which

would be obtained in the absence of any degradation, and sI is the input
point spread function. In the most general sense, IP consists in solving
Equation (1) for I, where Id and sI are known. Fourier transforming
Equation (1) and taking the inverse Fourier transform (FT1) for I leads to

FTId x, y
1
I x, y FT : (2)
FTsI x, y
As emphasized by Harris (1966), a number of problems arise in attempts

to make practical use of Equation (2). One of these is that FT(sI(x, y)) may
have points or regions of zero value. Since in the presence of noise the spec-
trum of the input image does not in general have zero value at these points or
regions, Equation (2) would give an infinite ratio. Another limitation lies in
the fact that even for the case of a well-behaved input transfer function, the
noise level associated with the degraded image may be so high as to make the
image meaningless. This led Harris to conclude that it is generally not
possible to obtain a perfect restoration. This general statement remains true
notwithstanding the huge progress made in image recording and storage and
preprocessing treatments since Harris paper was published.
Another method of restoring noisy degraded images which has been
developed by Richardson (1972) treats images point spread functions, and
degraded images as probability-frequency functions and applies Bayes the-
orem. Without going into details of this iterative procedure, an explicit der-
ivation of I was provided by Richardson and reads
x +X NNy 1
NNx 1 y +X
Id i, jsi x + 1, j y + 1
I x, yr + 1 I x, yr XX
n nn (3)
ix jy
I p, qr si p + 1, j q + 1
pm qmm
and
x +X
NNx 1 y + X
NNy1
Id i, jsI i x + 1, j y + 1
I x, y1 X X
n nn , (4)
ix jy
sI i p + 1, j q + 1
pm qmm
where (Nx, Ny) is the dimension of the input image, (NNx, NNy) is the
dimension of the input point spread function, m Max f1; i k + 1g,

mm Max f1; j l + 1g, n Minfi; Nx g, nn Minf j; Ny , and r 1
denotes the iteration number. Nevertheless, the convergence issue remains
untreated in the original paper of Richardson (1972). It is noteworthy that
both Harris (1966) and Richardson (1972) pointed out the positive effect
that preprocessing treatments can have on image restoration.
2.3 Digital Image Processing

2.3.1 Merits and Drawbacks of DIP
Since the technique developed by Richardson (1972) is intrinsically iter-
ative, it is necessarily difficult to implement optically. It requires easily
reconfigurable means such as computers. In addition, digital conversion
may affect significantly the information content of the image which can-
not be entirely corrected by preprocessing techniques. DIP can be also
computationally extensive, especially given the high degree of precision
required to render meaningful the information contained in badly
degraded images.
The extreme flexibility of the DIP techniques makes a wide variety of
liner and nonlinear processes possible. For example, these techniques have
been developed for removing noise with the intent of correcting geomet-
ric distortions and enhancing certain aspect of the pictures (Billingsley,
1970). DIP is advantageous compared to OIP because it requires only
low-cost systems. A single DP system can be used for many purposes,
including parallel computing. Choosing a DIP technique is eventually
related to the noise level contained in the image. It is generally accepted
that the performances of the DIP depend strongly on the image sensor
that converts an optical image into an electronic signal, i.e., dynamic
range, signal-to-noise ratio (SNR), and low light sensitivity. For sensors
of comparable types, the SNR and dynamic range improve as the size
increases.
(a) (b)
Before After
Uncorrected intensity Corrected intensity
Figure 1 Correction of nonuniform camera-response surveyor. Billingsley (1970).
Figure 3a shows a Surveyor2 picture obtained from a camera with a non-

uniform response versus light level (Billingsley, 1970). Accurate photometry
requires that this response calibration curve be known for each pixel.
As explained by Billingsley, such curves can be obtained from corresponding
pixel points in a sequence of flat field pictures taken with varying illumina-
tion, from which complete transfer curves are generated for each pixel.
When this technique is applied to Figure 1a, the shading of the ground
brightness from left to right at the top of Figure 1a has been removed in
Figure 1b to produce a uniform ground light level.
Figure 1 exemplifies the benefits of the preprocessing techniques for cor-
recting the degradation of the target image. But DIP suffers from the fact that
it requires an a priori knowledge of the different forms of noise, which sets a fun-
damental limit to the information contained in the image. The extent to which
DIP techniques approach the optimum is unclear and is difficult to evaluate
(Yaroslavsky, 2012). Explicit strategies, e.g., correlation (Figure 2), to over-
come the issues of image rotation and enhancement for the purpose of loca-
lizing a target in a scene have been described by Yaroslavsky (2012). The
preprocessing and postprocessing steps in this scheme have limitations imposed
by the form of noise, the correlator type, and the decision-making method.
For completeness, it is worth mentioning that Yaroslavsky (2012)
considers only digital correlators, while many references deal with optical
correlators (Alfalou, Brosseau, & Alam, 2013; Alfalou, Brosseau, 2010b;
2
NASAs Surveyor program was designed to find a way to safely land on the Moon. http://science1.nasa.
gov/missions/surveyor-1-7/
Figure 2 Schematic diagram of the scheme for localizing an object in a scene.

Yaroslavsky (2012).
Figure 3 Schematic of the filtering principle. Yaroslavsky (2012).
Alfalou, Brosseau, Benkelfat, Qasmi, & Leonard, 2012; Alfalou, Brosseau,

Katz, & Alam, 2012; Alfalou, Elbouz, Mansour, & Keryer, 2010 Katz,
Alfalou, Brosseau, & Alam, 2012; Leonard, Alfalou, Alam, & Arnold-Bos,
2012; Leonard, Alfalou, & Brosseau, 2012). We will highlight recent aspects
of optical correlation (OC) which have not been emphasized in other reviews
(Section 2.4.3.1). At least two potential advantages justify the interest in OC.
On the one hand, it permits detecting, localizing, and discriminating between
classes of objects in a target image on the basis of their correlation intensities to
be done simultaneously. On the other hand, it relies on a simple comparison
between the target image and a reference one, this comparison being done in
the spectral domain.
Based on Figure 3 and the filtering principle, the spectrum of the output
image (corrected image) Iout can be written as a product of the spectrum of
the input image Iin and the transfer function of the linear filter (H FTh),
h denoting its impulse function (Scharf, 1991)
FTIout HFTIin : (5)

2.3.2 Moir Correction

To show the beneficial effect of spectral filtering in IP, let us consider the
moire perturbation of an image. A moire pattern is a secondary and visually
evident superimposed pattern created, for example, when two identical
(usually transparent) patterns on a flat or curved surface (such as closely spa-
ced straight lines drawn radiating from a point or taking the form of a grid)
are overlaid while displaced or rotated a small amount from one another
(Aggoun, 2006; Andonian, 1982; Buytaert & Dirckx, 2007, 2008; Goole,
2013; Kato, Yamaguchi, Nakamura, & Kuwashima, 1997; Khalil &
Fazio, 1973; Yokozeki & Ohnishi, 1975). Moire patterns are often an unde-
sired artifact of images produced by various digital imaging and computer
graphics techniques, for example, when scanning a halftone picture or ray
tracing a checkered plane. This can be overcome in texture mapping
through the use of mipmapping and anisotropic filtering.
Moire can be viewed as an interference problem between two similar
patterns (Figure 4). The drawing on the upper right (Figure 4a) shows a
moire pattern. More complex line moire patterns are created if the lines
are curved or not exactly parallel (Figure 4b). Let us consider two patterns
made of parallel and equidistant lines, e.g., vertical lines. The step of the first
pattern is t, and the step of the second is t + t 0 .
(a) Dark area (b) Dark area
(p1 = t) d
(P2 = t+t)
Bright area Bright area
Figure 4 Illustration of the Moir effect by superimposing (a) two patterns made of par-
allel and equidistant lines and (b) example with an image. Goole (2013).
If the lines of the patterns are superimposed at the left of Figure 4, the
shift between the lines increases when going to the right. After a given num-
ber of lines, the patterns are opposed: the lines of the second pattern are
between the lines of the first pattern. If we look from a far distance, we have
the impression of pale zones when the lines are superimposed (blue (dark
gray in the print version) zone in Figure 4) (there is white between the lines),
and of dark zones when the lines are opposed (red zone (gray in the print
version) in Figure 4). The middle of the first dark zone is when the shift
is equal to t/2. The nth line of the second pattern is shifted by nt0 compared
to the nth line of the first network. The middle of the first dark zone thus
corresponds to nt 0 t=2, that is,
t
n : (6)
2t 0
The distance d between the middle of a pale zone and a dark zone is
d nt t2 =2t0 : (7)
Thus, the bigger the step, the bigger the distance between the pale and
dark zones; the bigger the discrepancy t0 , the closer the dark and pale zones;
and a large spacing between dark and pale zones means that the patterns have
very close steps. The suppression of the undesired effects (dark and pale
zones) of this phenomenon can be done as follows (with reference to
Figure 5).
Yaroslavsky (2012) suggests applying an FT of the target image showing
the moire effect (Figure 5a). It is well known that the FT of diffraction grat-
ing results is a series of spikes (Goodman, 1966). These spikes can be filtered
out by an appropriate filter (Figure 5b). Applying the FT1 leads to
Figure 5c. Comparing Figure 5a and c indicates that the moire effect has
been attenuated by the filtering. Figure 5d shows the moire noise.
A final note is in order. The moire effect can also be beneficial; e.g., it can
be useful to estimate the depth of a 3D object. Several optical analyses and
applications of the moire effect have been considered in the literature
(Aggoun, 2006; Andonian, 1982; Bara, Jaroszewicz, Kolodziejczyk, &
Moreno, 1991; Buytaert & Dirckx, 2007; Kato et al., 1997; Khalil &
Fazio, 1973; ;Kim, Park, Jung, Kim, & Lee, 2009; Lebanon & Bruckstein,
2001; Lim, Kim, & Chung, 1989; Rasouli, Ghasemi, Tavassoly, &
Khalesifard, 2011; Simova & Stoev, 1992; Yokozeki & Ohnishi, 1975).
Figure 5 Illustration of the moir effect on a specific image. (a) Original image clearly
displaying the moir perturbation (framed in red (gray in the print version)), (b) the FT
transform showing the spikes due to the moir effect, (c) processed image for which the
characteristic moir frequencies are filtered out (framed in red (gray in the print
version)), and (d) Moir noise. Yaroslavsky (2012).
2.3.3 Wiener Filtering

Several kinds of filtering techniques have been worked out in the literature,
e.g., Kalman and Wiener filterings (Duvaut, 1994; Goodman, 1966;
Yaroslavsky, 2012). One strategy for reducing noise in a noisy image Iin
is to use the Wiener deconvolution which is recognized as being a general
way of obtaining the best reconstruction of a noisy signal. This method is
used to produce an estimate of a desired or target random process by linear
time-invariant filtering of an observed noisy process, assuming known
stationary signal and noise spectra, and additive noise.
The Wiener filter minimizes the mean square error (MSE) between the
estimated random process and the desired process. We now turn to discuss in
more detail how it works. Consider that Iin is of the form3
Iin x, y hinvar Iunknown x, y + Badd x, y, (8)
where Iunknown is the unknown noiseless image, hinvar is the impulse function
of a known invariant in system time, and Badd is an additive noise which is
independent of Iin. An estimate of Iunknown can be obtained by looking for
hWiener such that
I^x, y hWiener Iin x, y, (9)

where hWiener is the impulse function of the Wiener filter, and I denotes
XN X M
1
the estimate of Iunknown minimizing MN I^ Iunknown 2 . The FT of
Equation (9) reads
Sû, v HWiener Sin u, v, (10)

where HWiener FThWiener and Sin FTIin , S^ denotes the estimate of
the spectrum of the image Iunknown, i.e., assumed to be denoised or with
reduced noise, and (u, v) are spectral coordinates. Putting all this together
finally leads to

Hinvar PSD Iunknown
HWiener 2 , (11)
jHinvar j PSD Iunknown + PSD Badd
where HWiener is the filter transfer function, PSD Iunknown is the mean power
of spectral density of Iunknown, PSD Badd is the mean power of spectral density
3
In optics, certain traditional notations exist and will be used here. Spectral domain variables (u,v), also
termed spatial frequencies, which have units of inverse length, correspond to the space domain
variables (x, y).
of Badd, H*
invar is the conjugate FT of hinvar, and jj is the square of the mod-
2
ulus. Finally, Equation (11) can be written in terms of the SNR
PSD Iunknown
SNR (12a)
PSD Badd
!
1 jHinvar j2
HWiener : (12b)
Hinvar jHinvar j2 1=SNR
Hence, Wiener filtering can be considered as an estimation problem

which is easy to implement and very efficient when a priori knowledge about
the image to be estimated is given, i.e., PSD Iunknown , PSD Badd .
We note several interesting features from Wiener filtering. First of all, let
us consider the special case of negligible noise, i.e., SNR ! 1. Thus,
Wiener filtering can be seen as the inverse of the initial system. When noise

j2
is not negligible, the SNR is reduced, and consequently jH jHj2invar + 1=SNR
invar
decreases.
Two specific cases are of interest. Figure 6 qualitatively illustrates the
effects of denoising of the dogs image (Figure 6a). Figure 6c shows the
Figure 6 Illustration of the Wiener filtering for noise suppression: (a) original noiseless
image. (b) Noisy image. (c) Filtered image. http://upload.wikimedia.org/wikipedia/com
mons/9/9d/Wiener_filter_-_my_dog.JPG.
denoised image. As outlined in the output image shown in Figure 6c, the
different points (black and white pixels) perturbating the visibility of
the image shown in Figure 6b vanish. Noise suppression performances
depend on the type of noise, its density (SNR), and used architecture
(Baker, 1981; Berriel, Bescos, & Santisteban, 1983; Birch et al., 2001;
Hardie & Barnard, 2012; Ikeda, Sato, & Kojima, 1986; Inbar & Marom,
1995; Jiang & Sawchuk, 1986; Khoury, Gianino, & Woods, 2000;
Kondo, Ichioka, & Suzuki, 1977; Kutay & Ozaktas, 1998; Riris, Carlisle,
Warren, & Cooper, 1994; Zalevsky & Mendlovic, 1996).
A second example deals with the deblurring of image. We show in
Figure 7a an example of blurred image. The deblurred image obtained with
Wiener deconvolution is shown in Figure 7b. The comparison of Figure 7b
and c conclusively demonstrates the improvement provided by this method.
To extend our considerations, we briefly present another related archi-
tecture, i.e., the empirical Wiener filter, which can be used for image
deblurring (Yaroslavsky, 2012). Ghael, Sayeed, and Baraniuk (1997)
reported on an image denoising method using wavelets and empirical
Wiener techniques. The main phenomenon is depicted using an example
in Figure 8. Figure 8a depicts an image with Gaussian white noise of variance
set to 10. As Figure 8b illustrates this denoising technique is efficient. To
further validate this as an approach accurately denoising images, we compare
this with the noiseless original image (Figure 8c).
Figure 7 Illustration of the deblurring method based on Wiener deconvolution:

(a) blurred image, (b) deblurred image, and (c) original image. http://upload.
wikimedia.org/wikipedia/commons/e/e3/Image_restoration_%28motion_blur%2C_Wiener_
filtering%29.png.
Figure 8 Illustration of denoising via empirical Wiener filtering. (a) Image with a white
Gaussian noise with variance set to 10, (b) denoised image using a wavelet shrinkage
estimate as a means to design a wavelet-domain Wiener filter, and (c) original image
(Ghael et al., 1997).
A final note is in order. The above-mentioned techniques were designed

for numerical processing. Consequently, they require rapid algorithms and
should be adapted to field programmable gate array (FPGA)-based
processing.
2.3.4 Compression and Encryption

Two main themes in the field of IP are compression and encryption.
A common goal in this area is to decrease the quantity of data to be trans-
mitted (compression) and to protect the use of data against unauthorized
access (encryption). Figure 9 depicts the elements of a communication sys-
tem (Barlaud & Labit, 2002; Richardson, 2003; Salomon, 2004; Sayood,
2012). There are three essential parts of any communication system, the
transmitter, transmission channel, and receiver. Each part plays a particular
role in signal transmission, as follows. The transmitter processes the input
signal to produce a suitable transmitted signal suited to the characteristics
of the transmission channel. The picture to have in mind is sketched in
Figure 9a.
(a)
Iin
Input image And/or Modulation
Compression Encryption
(or input video sequence) (adaptation)
Transmission channel
(b)
Iout
Output image Decompression Decryption Demodulation
(or output video sequence)
Figure 9 Elements of a communication system including compression and/or encryp-

tion: (a) emitter and (b) receiver.
IP for transmission almost always involves compression and may also

include encryption. The transmission channel is the medium that bridges
the distance from source to destination. Every channel introduces some
amount of transmission loss or attenuation. So, the signal power progres-
sively decreases with increasing distance. The receiver operates on the out-
put signal from the channel in preparation for delivery to the transducer at
the destination (Figure 9b). Receiver operations include amplification to
compensate for transmission loss. These also include decompression and
decryption to reverse the signal procession performed at the transmitter. Fil-
tering is another important function at the receiver. Note that a modulation/
demodulation block is often required to adapt the signal to be transmitted to
the channel.
2.3.4.1 Compression
Data compression refers to the process of reducing the amount of data
required to represent, store, and transmit given quantity of information
(image). There exists a large literature concerning compression techniques
(Barlaud & Labit, 2002; Pennebaker & Mitchell, 1993; Richardson, 2003;
Salomon, 2004; Sayood, 2012). Generally speaking, these techniques aim
at reducing the redundancy of information. Three kinds of redundance
can be distinguished: spectral, spatial, and temporal (Barlaud & Labit,
2002; Richardson, 2003; Salomon, 2004; Sayood, 2012). The work in
Pennebaker and Mitchell (1993), Barlaud and Labit (2002), Richardson
(2003), Salomon (2004), and Sayood (2012) and subsequent efforts have
already paved the way for a development of many compression protocols
such as Bitmap, Portable Network Graphics, Graphics Interchange Format,
Tagged Image File Format, Joint Photographic Experts Group, JPEG 2000,
Moving Picture Experts Group, and MPEG-4 AVC (Advanced Video Cod-
ing). As the amount of information to be transmitted becomes larger and
transmission speed becomes faster, data compression is becoming a critical
challenge in video imaging. The objective of image compression is to reduce
irrelevance and redundancy of the image data in order to be able to store or
transmit data in an efficient form. Certain information has relatively less
importance for the quality of image perception. This information is said
to be psychovisually redundant.
Image compression can be classified as lossy or lossless. Lossless compres-
sion, e.g., the Lempel-Ziv-Welch technique, is preferred for archival pur-
poses and is often used for medical imaging. Lossy compression methods,
e.g., JPEG, especially when used at low bit rates, introduce compression
artifacts. Lossy methods are especially suitable for natural images such as
photographs in applications where minor (sometimes imperceptible) loss
of fidelity is acceptable to achieve a substantial reduction in bit rate. In this
work, we focus on lossy data compression. In an attempt to cope with the
problem, several studies have been conducted in the past in academic
institutions and industry, especially for telecommunications, access control,
biometrics, and security systems.
Here, we are interested in the (lossy) two-step compression technique
depicted in Figure 10. The first step consists in applying a well-defined trans-
form to the image, i.e., mapping. This is a reversible process which is not
lossy. This transform is required to prepare and package the information
(redundant or pertinent). The choice of the specific transform is linked to
the application and the compression technique (Salomon, 2004). The sec-
ond step is the quantification during which redundant information is
suppressed. In order to be effective, quantification results from a compro-
mise between compression rate and quality of the reconstructed images. This
process reduces the accuracy and hence psychovisual redundancies of a given
image. This process is irreversible and therefore lossy. This quantification
step is followed by entropic coding when a fixed or variable-length code
is used to represent mapped and quantized data sets (see, e.g., Neji, Jridi,
Alfalou, & Masmoudi, 2013). This is a reversible process. It removes coding
redundancy by assigning the shortest codes to the most frequently occurring
output values.
Constraints :
application, compression method, ...
Iin
Input image Mapping
(or input video sequence)
Arithmetic, Golomb, Huffman, ...
Entropic
Quantification Bitstream
coding
Constraints :
application, flow rate, compression rate
quality of reconstructed images, ...
Figure 10 Illustration of the compression block.
Our aim is to describe in some details the mapping and quantification

steps. As mentioned above, lossy encoding is based on the concept of
compromising the accuracy of the reconstructed image in exchange for
increased compression. The lossy compression method produces distortion
which is irreversible. The most popular transform coding methods are based
on the discrete FT (DTF) and discrete cosine transform (DCT) and map the
image into a set of transform coefficients which are then quantized and
coded. The goal of the transformation is to decorrelate the pixels of a given
image block such that most of the information is packed into the smallest
number of transform coefficients. The choice of a transform in a given appli-
cation depends on the amount of the reconstruction error that can be
tolerated and computational resources available.
DFT is defined as (Duvaut, 1994)
X
N X
1 M1 h xu yvi
SIin u, v A Iin x, yexp 2j + , (13)
y0 x0
M N
where SIin is the spectrum of the discrete image Iin of size M N pixels, and
A MN 1=2 is a normalization constant. For clarity, we will frame our
discussion around the specific one-dimensional version of Equation (13)
(Chang & Nguyen, 2008; Cooley & Tukey, 1965; Glenn, 1968; Kolba &
Parks, 1977; Oppenheim, Schafer, & Buck, 1999; Ouerhani, 2012;
Ouerhani, Jridi, & Alfalou, 2011, 2012; Sorensen, Heideman, & Burrus,
1986):
X
M 1 h xui
SIin u A Iin xexp 2j : (14)
x0
M
In matrix form, Equation (14) reads

0 1 0 10 1
SIin 0 1 1 1 Iin 0
B SIin 1 C B 1 WM1 WMM1 C B C
B C B CB Iin 1 C
B .. C AB .. .. .. .. CB .. C,
@ . A @. . . . A@ . A
M1 M1
SIin M 1 1 WMM WM Iin M 1
(15)

m
where WMm exp 2j M . The main disadvantage of DFT is that it requires
large computational resources to realize M M 1 sums and M M products.
The CooleyTukey algorithm (also called Radix-24), named after Cooley and
Tukey (1965), is the most common fast FT algorithm. It reexpresses the DFT of
an arbitrary composite size N N1N2 in terms of smaller DFTs of sizes N1 and
N2, recursively, in order to reduce the computation time to O(N log N) for
highly composite N (smooth numbers). Other algorithms permitting a decrease
in computational time and/or the complexity related to numerical implemen-
tation have been proposed, i.e., Radix-4, Radix-8, Split Radix, Prime factor,
and high-order fast Fourier transform (FFT) (Chang & Nguyen, 2008;
Cooley & Tukey, 1965; Glenn, 1968; Kolba & Parks, 1977; Oppenheim
et al., 1999; Ouerhani, 2012; Ouerhani et al., 2011, 2012; Sorensen
et al., 1986).
DCT is defined as follows (for a square size image, i.e., M N ) (Ahmed,
Natarajan, & Rio, 1974; Blinn, 1993; Chen, 2005; Chen, Smith, & Fralick,
1977; Hou, 1987; Meher, 2006; Pennebaker & Mitchell, 1993; Slawecki &
Li, 1992; Strang, 1999; Sungwook & Swartziander, 2001; White, 1989):
4
In mathematical numeral systems, the radix or base is the number of unique digits, including zero that a
positional numeral system uses to represent numbers. For example, for the decimal system (the most
common system in use today), the radix is 10 because it uses 10 digits from 0 through 9.
X X
1 N 1
1 N
2x + 1u 2y + 1v
SDCTIin u, v c uc v Iin x, ycos cos ,
4 0 0
2N 2N
(16)
p
1= 2 if i 0
where c i , SDCTIin is the DCT spectrum of Iin. For most
1 if i 0
images, much of the signal energy lies at low frequencies; these appear in the
upper left corner of the DCT. The lower right values represent higher fre-
quencies and are often small (usually small enough to be neglected with little
visible distortion). Figure 11a illustrates this point schematically.
As for FT, DCT requires a large number of arithmetic operations, i.e.,
typically N 2 N sums and N2 products. To minimize complexity,
number of arithmetic operations, computation time, and energy con-
sumption, a number of groups have proposed a variety of architectures
(Ahmed et al., 1974; Blinn, 1993; Chen, 2005; Chen et al., 1977; Hou,
1987; Meher, 2006; Pennebaker & Mitchell, 1993; Marom & Inbar,
1996; Slawecki & Li, 1992; Strang, 1999; Sungwook & Swartziander,
2001; White, 1989).
In prior work (Ouerhani, 2012), we validated FFT and DCT algorithms
for implementation on FPGA. The FFT architecture can be optimized using
the Radix-4 algorithm and presents the advantages of a higher throughput,
low area-delay product, and lower computational time. The novelty consists
on using a memory sharing and dividing technique along with parallel-in
parallel-out processing elements. The proposed architecture can perform
N-point FFT using only 4N/3 delay elements and involves a latency of
N/4 cycles. Comparison in terms of hardware complexity and area-delay
product with recent work presented in the literature and commercial IPs
was made to show the efficiency of the proposed design.
(a) (b)
Low frequency
Low frequency
Medium
Medium
High frequency
High frequency
Figure 11 Spectral representation of an image using: (a) DCT and (b) FT.
2.3.4.2 Digital Encryption

Encryption is an important topic in establishing the secure transmission of
data (Figure 12). Encryption is the process of encoding messages (or trans-
mitted information) in such a way that eavesdroppers or hackers cannot read
it, but that authorized parties can. The principal tools of encryption are
encoding and decoding keys. This is usually done with the use of an encryp-
tion key, which specifies how the message is to be encoded. Any adversary
that can see the ciphertext should not be able to determine anything about
the original message. An authorized party, however, is able to decode the
ciphertext using a decryption algorithm, which usually requires a secret
decryption key that adversaries do not have access to. For technical reasons,
an encryption scheme usually needs a key-generation algorithm to randomly
produce keys. Other techniques designed to hide information have been
proposed, e.g., steganography (Cox, Miller, Bloom, Fridrich, & Kalker,
2007; Westfeld & Pfitzmann, 2000; Westfeld & Wolf, 1998).
Steganography is the art and science of writing hidden messages in such a
way that no one, apart from the sender and intended recipient, suspects the
existence of the message, a form of security through obscurity. The advan-
tage of steganography over cryptography alone is that messages do not attract
attention to themselves. Plainly visible encrypted messages, no matter how
unbreakable, will arouse suspicion and may themselves be incriminating in
countries where encryption is illegal. Therefore, whereas cryptography pro-
tects the contents of a message, steganography can be said to protect both
messages and communicating parties.
Much work has been devoted to encryption schemes and protocols that
can be divided into the ciphering and content encryption families. Before
dealing with content encryption schemes which can be optically (at least par-
tially) implementable, we spend some time on briefly presenting ciphering
Fourier plane :
Spectral plane
Input plane :
Image plane
Light f
Figure 12 Experimental setup allowing us to realize an FT optically thanks to a

converging lens.
systems. Codes generally substitute different length strings of characters in the

output, while ciphers generally substitute the same number of characters as are
input. There are exceptions and some cipher systems may use slightly more,
or fewer, characters when output versus the number that were input. The
operation of a cipher usually depends on a piece of auxiliary information,
called a key, i.e., a cryptovariable. The encrypting procedure is varied
depending on the key, which changes the detailed operation of the algorithm.
A key must be selected before using a cipher to encrypt a message. With-
out knowledge of the key, it should be extremely difficult, if not impossible,
to decrypt the resulting ciphertext into readable plaintext. Codes operate by
substituting according to a large codebook which linked a random string of
characters or numbers to a word or a phrase. When using a cipher the orig-
inal information is known as plaintext, and the encrypted form as ciphertext.
The ciphertext message contains all the information of the plaintext message,
but is not in a format readable by a human or a computer without the proper
mechanism to decrypt it. Most modern ciphers can be categorized in several
ways: by whether they work on blocks of symbols usually of a fixed size
(block ciphers), or on a continuous stream of symbols (stream ciphers), or
by whether the same key is used for both encryption and decryption (sym-
metric key algorithms), or if a different key is used for each (asymmetric key
algorithms). If the algorithm is symmetric, the key must be known to the
recipient and sender and to no one else. If the algorithm is an asymmetric
one, the enciphering key is different from, but closely related to, the deci-
phering key. If one key cannot be deduced from the other, the asymmetric
key algorithm has the public/private key property and one of the keys may
be made public without loss of confidentiality.
One of the most used asymmetric key algorithms is the so-called RSA.
RSA stands for Rivest, Shamir, and Adleman, who first publicly described
the algorithm in 1977 (Barrett, 1987). RSA is an algorithm for public-key
cryptography that is based on the presumed difficulty of factoring large inte-
gers, the factoring problem. A user of RSA creates and then publishes the
product of two large prime numbers, along with an auxiliary value, as their
public key. The prime factors must be kept secret. Anyone can use the public
key to encrypt a message, but with currently published methods, if the public
key is large enough, only someone with knowledge of the prime factors can
feasibly decode the message. Whether breaking RSA encryption is as hard as
factoring is an open question known as the RSA problem.
A closing note is in order. Part of the problem in these numerical
methods is related to the computation time, especially for real-time
applications. Here, we will explore OIP techniques of images and show that
novel optical methods can be considered as substitutes, at least partially, to
many time-consuming numerical techniques.
2.4 Optical Image Processing

The merits and drawbacks of the existing approaches of DIP motivated new
real-time IP approaches, notably those based on coherent optics that we pre-
sent here. Fourier methods have been used to describe optics for over
100 years, but only around the 1960s were optics used to describe or perform
Fourier analysis (Goodman, 1966; Wolf, 2007; Hecht, 2001). The need for
such methods relies on (1) the technological needs for diverse applications
ranging from vision to transmission to detection; (2) the progress in applied
mathematics and computer science which opened new opportunities for
OIP; and (3) the progress in the understanding of complex phenomena such
as polarized light in underwater imaging (just to mention one example). Over
the past decade, software and hardware architectures have received increasing
attention. During that time, there has been considerable progress in develop-
ing the technological and methodological base for treating architectural
design as an engineering discipline for military and civilian applications.
Based on the ease of optically realizing an FT by means of a single convergent
lens, a number of OIP techniques are presently being pursued (Goodman,
1966), optical implementation of wavelets (DeCusatis & Das, 1995;
Mendlovic & Konforti, 1993; Shabtay, Mendlovic, & Zalevsky, 1998),
implementation of DCT (Aggoun, 2011; Alfalou & Brosseau, 2009;
Alfalou, Brosseau, Abdallah, & Jridi, 2011; Alfalou, Brosseau, Abdallah,
et al., 2013; Alfalou, Mansour, Elbouz, & Brosseau, 2011), implementation
of optical filtering (Goodman, 1966), and implementation of OC (Alfalou,
1999; Alfalou & Brosseau, 2010a, 2010b; Alfalou, Brosseau, & Alam, 2013).
Two main advantages of optics are as follows. On the one hand, because
of the optical nature of the image, its conversion into a digital form is nec-
essary in order to transmit, store, compress, and/or encrypt it. It is fair to say
that this requires an important computing time, or eventually an image qual-
ity reduction. Thus, to compress and encrypt optically an image can be a
good solution. One key advantage of optics compared to digital methods
lies in its capability to provide massively parallel operations in a 2D space.
As we will explain in some detail, these methods become more and more
complex to implement optically, especially when we are interested in color
images or video sequences; in fact every time, the rate of image encryption
increases. Carrying out a part of the processing digitally is a relevant solution

to solve this complexity while maintaining a reasonable speed. On the other
hand, the inherent parallelism of optics provides an appropriate framework
for IP techniques, e.g., by using the FFT algorithm. However, from a
practical point of view, it is worth emphasizing that the latter is largely con-
ditioned by the existence of powerful electro-optical interfaces.
Fourier optics methods consider the Fourier spectrum (space of spatial
frequencies) of the object and the transmission of the spectral components
through the optical system (Goodman, 1966).5 Optical spatial filtering is
based on the FT property of a lens (with reference to the experimental setup
shown in Figure 12). It is possible to display the 2D spatial frequency6 spec-
trum of an object in such a way that individual spatial frequencies can be
filtered. Consider the configuration with an illuminated picture in the input
plane (using a spatial light modulator (SLM)), (Casasent, 1979; Casasent &
Xia, 1986), a lens of focal length f, and an output plane (where the complex
spectrum of the target image is commonly recorded by a CCD camera) as
shown in Figure 12.
Simple mathematics allows one to prove that the f to f spacing provides an
ideal FT in the amplitude of the scalar electric field (Goodman, 1966). This
configuration has a singular position in Fourier optics, since it enables the
study and visualization of the optical FT of various complicated objects.7
Limitations due to the use of optical components are largely described in
the literature. This puts more pressure on OIP to do as good a job as it can
take full account of these limitations. For concreteness, we show an example
of SLM in Figure 13a. Using a nematic modulator (CRL Opto/VGA3)
imposes several constraints on the image resolution and size; i.e., this mod-
ulator is composed of 640 480 pixels with the size 42 42m2 .
Figure 13b shows an enlarged view of this modulator, where optically active
regions are clearly visible.
5
The optical generation of FTs requires coherent light and it was only when lasers developed in the early
1960s that we had sources that were simultaneously bright and highly coherent. It was lasers that stim-
ulated Fourier optics research.
6
Spatial spectrum brings information about the spatial distribution of intensity in the image and the dis-
tinctive directions. High spatial frequencies are responsible for fast and strong changes in image inten-
sity (details, noise, patterns, edges). Low spatial frequencies are responsible for slow and smooth changes
in image intensity (global intensity distribution, large smooth areas).
7
Spatial filtering consists of altering the image content by placing masks in the Fourier plane. These
masks are designed to affect the amplitude and/or the phase of spectral components. Low pass filters
allow low spatial frequencies to pass through the system while blocking components associated with
higher spatial frequencies; high pass filters (edge enhancement) consist of a dark spot in the center of the
Fourier plane.
(a)
(b)
Pixel
Figure 13 Illustration of a nematic modulator: (a) Picture of the SLM Seiko produced by
(a) CRL Opto/VGA3 and (b) a microscope view of the SLM Seiko.
In spite of this proliferation of techniques, their all-optical implemen-

tations are rather complex to develop. Thus, alternate hybrid methods
(numerical optical using optoelectronic interface) have been proposed as
substitutes for all-optical techniques. These approaches, however, are not
the universal panacea and have their drawbacks and stringent requirements,
i.e., aberration effects, alignment of components, limitation of the overall
speed by how fast the information can be updated on the input and output
devices, and need of a costly optoelectronic interface. Furthermore, the use
of optics cannot be justified for many applications, especially when the target
image size is small. In recent years, hybrid techniques, e.g., numerical
implementation of correlation, have been considered an alternative to all-

optical methods because they show a good compromise between perfor-
mance and simplicity.
Although hybrid techniques have generated many impressive results for
the problem of correlation, a numerical approach frequently requires some
adaptation of the correlation method in order to optimize the resources of
the numerical target (central processing unit (CPU), graphics processor unit
(GPU)). On the other hand, an all-numerical implementation leads to a
larger applicability, reconfigurability, and adaptation of the correlation
preprocessing to the application considered.
2.4.1 Imaging Systems and Applications

Imaging systems have been a recent focus of attention (see, e.g., Catrysse, Imai,
Linne von Berg, & Sheridan, 2013; Nakajima, 2013). Nakajima (2013)
showed the effectiveness of lensless coherent imaging using a nonholographic
and noniterative phase-retrieval method that allows the reconstruction of a
complex-valued object from a single diffraction intensity measured with an
aperture array. Zalevsky et al. (2013) considered also the detection of axially
moving targets by a time-multiplexing approach. Shilemay et al. (2013)
reported on the performance analysis of millimeter-wave imaging systems.
These authors showed that the image quality is strongly dependent on the
type of noise, and that the image blurring is directly related to the pixel size of
the sensor. Catrysse et al. (2013) also reviewed a number of imaging systems
for military as well as civilian applications, e.g., sensors for lightning detection
and imaging (Rolando et al., 2013), open-access algorithms for commercial
videokeratographers (Espinosa, Mas, Perez, & Roig, 2013), analysis of unitary
discrete linear canonical transforms (LCTs) (Zhao, Healy, & Sheridan, 2013),
high precision calibration for fish-eye cameras (Tu, Liu, Liu, Jin, & Tang,
2013), extending depth of focus in tomography systems (Paz, Zlotnik, &
Zalevsky, 2013), information capacity for spectral imagers (Skauli, 2013),
in vivo brain imaging systems (Atchia, Levy, Dufour, & Levi, 2013), pixel scal-
ing for infrared focal plane arrays (Catrysse & Skauli, 2013), a simple digital
in-line holographic microscope (Ryle, McDonnell, Glennon, & Sheridan,
2013), and an optical see-through head-mounted display (Wang, Cheng,
Wang, Hua, & Jin, 2013; Wang, Guo, & Lei, 2013).
2.4.1.1 Image Sensor

Rolando et al. (2013) developed a lightning detector that will be
implemented in the Meteosat Third-generation Imager satellite for the
Figure 14 Synoptic diagram of the detector developed by Rolando et al. (2013). TE

means triggered event and RAD is a radiometric detector.
European Space Agency. These authors proposed and validated a smart

CMOS8 image sensor which is appropriate for applications in lightning
detection and imaging. Figure 14 shows a schematic diagram of this detector
which is composed of two triggered event (TE) detectors, one is a circuit
specialized for processing and encoding, while the other is a row decoder
TE and row initialization sequencer. Other components are the radiometric
detectors (RADs) to characterize optical pulses, and a 256 256
pixel array.
This detector was implemented experimentally using standard 0.35 m
CMOS technology (nonoptimized image sensor technology) leading to a
front-side-illuminated CMOS image sensor, with no microlens and no coat-
ing or doping profile optimized to improve sensitivity (Rolando et al., 2013).
Most importantly, Rolando et al. (2013) showed that the system dis-
played in Figure 14 can be useful to film a given scene at video frequency
of 1 kHz. For this purpose, Rolando et al. (2013) use a smart detector
8
CMOS, complementary metal oxide semiconductor.
700
50
600
500
100
Signal (mV)
Rows
400
150
300
200
200
100
250 0
50 100 150 200 250
Columns
Figure 15 Example of an image obtained by the detector suggested by Rolando et al.
(2013).
combining an in-pixel frame-to-frame difference comparison with an

adjustable threshold and on-chip DP allowing an efficient localization of
a faint lightning pulse on the entire large format array. The good quality
of the image displayed in Figure 15 obtained with this detector demonstrates
the good performances of this smart CMOS image sensor.
2.4.1.2 Imaging Systems: In Vivo Brain Optical Imaging System Using Vertical
Cavity Surface-Emitting Lasers as a Light Source
Another application that has been a recent focus of attention concerns imag-
ing systems. As pointed out by Atchia et al. (2013), optical methods have at
least five potential advantages compared to other methods such as functional
magnetic resonance imaging, positron emission tomography, and computed
tomography: (1) functional simplicity, (2) cost effectiveness, (3) noninvasive
nature, (4) good performances with spatiotemporal resolution, and (5) min-
iaturization for portable application. Two kinds of architecture exist (Atchia
et al., 2013). The first one is termed intrinsic optical signal imaging, while
the second one is based on laser speckle contrast imaging (LSCI). Based on
the latter, Atchia et al. (2013) demonstrated an imaging technique
implementing vertical cavity lasers with extremely low transient times
for a greatly simplified realization of a multiexposure laser speckle contrast
imaging system.
The LSCI technique allows one to realize a map of relative blood veloc-
ity in vivo through the interference of coherent light (Goodman, 1976). The
underlying principle of the technique consists in calculating a speckle pattern
from scattering events with coherent light excitation. The spatial speckle
contrast ratio K is expressed as

K , (17)
hI i
where in a selected window around the considered pixel, is defined as the

spatial standard deviation and hIi is the mean intensity. It can be shown that
K depends on the velocity of the dynamic scatterers. The apparatus is
depicted in Figure 16 (Atchia et al., 2013). The light source is a (680 nm,
optical power of 1 mW) vertical-cavity surface-emitting laser (VCSEL)
(Vixar Inc.). The 4f optical setup is used and the EMC2 CCD camera con-
sists of 500 500 pixels of size 8 m/pixel. As long exposure times are
required for accurate parameter fitting in multiexposure imaging, the frame
rate is limited by exposure time.
Overall, the system proposed by Atchia et al. (2013) has good perfor-
mances. An illustrative example is shown in Figure 17. Figure 17a shows
a map obtained from VCSELs, while Figure 17b displays the relative velocity
map obtained using VCSELs.
Figure 16 Experimental apparatus proposed by Atchia et al. (2013) for in vivo brain opti-
cal imaging of small animals.
Figure 17 Illustration of experimental results obtained by Atchia et al. (2013).
We note in passing that Atchia et al. (2013) also proposed different opti-
mization schemes to overcome the errors due to defocusing, and to deal with
the coherence length which is altered with changes in the VCSEL current.
2.4.1.3 Imaging Systems: Calibration of a Digital In-line Holographic

Microscopy System: Depth of Focus and Bioprocess Analysis
A second illustrative example of imaging system has been reported by Ryle
et al. (2013) based on the digital holography (DH) method documented in
Yamaguchi and Zhang (1997) and Zhang, Pedrini, Osten, and Tiziani
(2004). Following on from this, Ryle et al. (2013) proposed the digital
in-line holographic microscopy (DIHM) system for biopharmaceutical pro-
cess measurement. Standard microscopy imaging systems have strong limi-
tations dues to their depth-of-field (DOF), i.e., refer to Figure 21. Two
analytical expressions characterize the microscope objective (MO) plane,
namely, the DOF to obtain in-focus images of two axially displaced points
through

DOF (18a)
nsin2
and the lateral resolution or resolvable resolution between two points in a
diffraction-limited incoherent imaging system which is commonly evalu-
ated from Rayleighs criterion

R 0:61 , (18b)
nsin
where is the wavelength of the light, denotes the half-angle of illumination
in the imaging system, and n denotes the refractive index of the material in
which the object/MO is immersed (Goodman, 1966). However, these sys-
tems are not robust (Ryle et al., 2013). Figure 18 shows that the DH method is
inappropriate to obtain high-resolution images; i.e., any object positioned at a
Lens
q
R
DOF
f
Figure 18 Imaging geometry of a thin lens including the imaging spot. Ryle et al. (2013).
MO2 focal plane

SAMPLE MO2
NDF MO1 SF CL
HeNe
10 CCD
20
laser
z2
z1
z
Figure 19 Schematic of the experiment reported by Ryle et al. (2013).
distance larger than the DOF cannot be resolved. To remedy this, Ryle et al.
(2013) suggested using the DIHM architecture. One basic advantage of such
system lies in its simplicity. Only standard optical components such as colli-
mated spatially coherent light source, MO, and digital camera are needed
along with a computer to process the resulting holograms.
Figure 19 illustrates schematically the optical setup composed of a HeNe
laser ( 632:8nm; 20 mW), a neutral density filter (NDF), a microscope
objective (20; MO1), a spatial filter (SF with a 15-m diameter pinhole),
a collimating lens (CL; f 20cm), the sample (located away from the focal
plane of the imaging 10(n sin )), another microscope objective
(MO2), and a CCD camera.
DIHM-based systems have been shown to overcome the limitations of
DOFs, i.e., by reconstruction of a number of images over many depths (Ryle
et al., 2013). Each image, reconstructed from a single hologram, has a limited
number of DOFs (Figure 22). Combining or stacking a series of such DOF
limited numerical reconstructions results in an image with an effectively
greater DOF which is not limited longitudinally to imaging objects within

the focal plane of the MO (Ryle et al., 2013). The analysis of Ryle et al.
(2013) provided a simple DIHM capable of imaging weakly scattering
10 m diameter microspheres as well as Hs578T cells over a depth of
1 mm; i.e., these authors demonstrated an increase by a factor of 100 over
the depth of focus of a conventional microscope.
2.4.2 Toward a Super Imaging System

Over the past two decades, an interest has burgeoned in hybrid techniques at
the confluence of numerical and optical techniques. Such techniques are
efficient for military as well as civilian applications. Reflecting its multi-
disciplinary nature, it has attracted optical physicists and engineers from a
broad range of backgrounds, and work in the field has evolved with close
connections between experiments and theory.
Any efficient OIP technique should use a simple setup, i.e. with a limited
number of components, be cost effective and compact. In some situations, a
hybrid numerical optical method is preferable. Within this context, it is
important to note that Mendlovic (2013) merged ideas from OP and IP,
proposing the super imaging scheme shown in Figure 20. Mendlovic
(2013) proposed a new architecture adapted to computational photography
for designing a super imaging system.
Figure 21 provides a schematic of the proposed method combining mul-
tiple aperture, light field photography, and multiexposure.
In closing this section, we mention three applications of imaging system.
The first deals with multiple aperture cameras. Figure 22 shows an array of
M M small rectangular apertures (B) mounted on a standard digital imager
Figure 20 Illustration of the optical setup as proposed in Mendlovic (2013). ISP means
image signal processing.
Figure 21 Schematic proposed par Mendlovic (2013) for designing a super imaging
camera.
Figure 22 Multiple aperture cameras: (A) is a microlens array, (B) is the rectangular aper-
ture channel array, (C) is a specific subchannel, and (D) is the detection imager.
Mendlovic (2013).
of P P pixels (D). The aperture layer is then covered by a microlens array

(A) so that we effectively get an array of M M low-resolution cameras each
with (N/M) (N/M) pixels (N is the total pixel count in each direction).
Each lower solution imaging system is referred to as a subchannel and
noted as (C) (Mendlovic, 2013).
The second application concerns light field photography. Figure 23
shows the creation of the 4D plenoptic function. It requires the detection
Conventional
Subject
Main lens
Photosensor
Light field
Microlens
array
Subject
Main lens
Photosensor
Figure 23 Illustration of the 4D plenoptic function. Mendlovic (2013).
Figure 24 Illustration of the multiexposure concept of fusing few exposures to create

better quality image. Mendlovic (2013).
of the information distribution among the different field rays per each image
pixel. This is usually done by inserting a lens array in front of the sensing
array and, based on the ray model optics, obtaining the various fields in dif-
ferent pixels. Figure 4 shows a conventional imaging system and a light-field
recording system. In this setup, one can increase the pixel count by a factor
of M along each direction (Mendlovic, 2013).
The third application is the multiexposure camera. The basic idea is to
collect data from a few exposures and fuse it to create a better quality image.
The concept we have in mind is illustrated in Figure 24. A necessary step in
all multiexposure/multiaperture computational photography algorithms is
the frame alignment and fusion.
2.4.3 Image Detection, Identification, and Localization

Alfalou (1999) and Alfalou, Brosseau, and Alam (2013) advocated taking
OC techniques to detect, identify, and localize a target object in a scene.
The purpose of this brief section is to consider in some detail the basic
principle of OC.
2.4.3.1 Optical Correlation

The idea of correlating the target image (image to be recognized) to a ref-
erence image (from a database) has a long history (Alam, 2013; Alfalou et al.,
1999; Alfalou & Brosseau, 2010a, 2010b; Banerjee & Datta, 2012; Yu &
Jutamulia, 1998). Two important architectures implementing correlation
are the joint transform correlator ( JTC) (Weaver & Goodman, 1966) and
Vander Lugt correlator (VLC) (Vander Lugt, 1964). They are generally per-
formed using an optical setup called 4f. The 4f setup is an optical system
composed of two convergent lenses. The 2D object is illuminated by a
monochromatic wave. A first lens performs the FT of the input object in
its image focal plane. In this focal plane, a specific filter is positioned. Next,
a second convergent lens performs the FT1 in the output plane of the sys-
tem to get the filtered image. The main difference between JTC and VLC
lies in the treatment of the input and output planes. The principle of VLC is
illustrated in Figure 25.
The protocol underlying the VLC is as follows: the optical setup of the
VLC has three planes, the input plane (target image), the Fourier plane, and
the correlation plane. Passing from one plane to the other is done through
two convergent lenses which realize the FT of the input and Fourier planes
after the multiplication of the target image spectrum with a correlation filter
(fabricated from a database of reference images). This leads to a correlation
plane by inverting the FT of this product. The fabrication methods of
correlation filters will be detailed later. The main advantages of VLC are
its simplicity compared to other algorithms; its global treatment of the target
image; its good decisional performances; and its ability to detect, localize,
and identify simultaneously a target image within a scene in contrast to
numerical methods described (see Alfalou, Brosseau, & Alam, 2013).
FT Fourier FT1 Output

Input plane plane
plane
X
FT: Fourier transform FT1: Inverse Fourier

transform
Correlation
filter
Figure 25 Synoptic diagram of the VLC. Alfalou, Brosseau, and Alam (2013).
2.4.3.2 Why Correlation Is Useful for OIP?

Great progress has been made in the field of OC. Despite this significant pro-
gress, the output of papers dealing with the subject has slowed down in recent
years. The reasons are threefold. First, many efforts have concentrated on
suggesting and validating new correlation filters. These efforts concerned
essentially treatments in the Fourier plane. However, the use of specific treat-
ments of the input and correlation planes permits significant increase in the
correlators performances. Alfalou, Brosseau, Katz, et al. (2012) suggested
and validated a denoising method of the correlation plane. This method is
found to have significantly superior correlation discrimination capability. In
Alfalou et al. (2010), another type of treatment of the input plane was able
to provide better decisional performances of the correlator. Second, a large part
of the community working on correlation has focused on all-optical imple-
mentation. However, even if these all-optical implementations allow having
in principle large cadence, they are rather complex to develop (Alfalou, 1999).
In addition, all-optical approaches have their drawbacks and stringent
requirements, i.e., aberration effects, alignment of components, limitation
of the overall speed by how fast the information can be updated on the input
and output devices, and need of a costly optoelectronic interface. Further-
more, the use of all-optics cannot be justified for many applications, espe-
cially when the target image size is small. Thus, alternate hybrid
(numerical optical using optoelectronic interface) methods have been pro-
posed as substitutes for all-optical techniques. Moreover, recent advances in
reprogrammable targets, such as the GPU, or the FPGA make it possible to
manipulate computer graphics efficiently and process large blocks of data
rapidly (Ouerhani, Jridi, Alfalou, & Brosseau, 2013; Ouerhani et al.,
2012). Our recent results (Ouerhani et al., 2012, 2013) indicate that the
GPU implementation of an all-numerical correlator leads to true recogni-
tion rates larger than 85% with a run time of under 120 ms using fixed images
and true recognition rates larger than 77% using a real video sequence with a
run time smaller than 2 frames per second and a database of 100 persons.
Third, it is our opinion that the correlation operation should be conceived
as part of a global system for making a decision. Within this context, corre-
lation should be combined with a data fusion technique (Hou, Zhouand, &
Siddique, 2012). For example, when the target object is moving, it is nec-
essary to make a decision based on correlation features of many images from
a video sequence. This principle was validated in Elbouz, Alfalou, and
Brosseau (2011) for which correlation, selected as a decision-making
operation, was part of a complex scheme based on fuzzy logic.
Figure 26 Examples of target image: (a) Barbara, (b) Lena, and (c) and (d) are the ampli-
tude spectra of Lena and Barbara, respectively.
2.4.3.3 Discussion
It is quite clear that successfully using correlation methods requires a good
knowledge of the application and basic requirements such as the kind of tar-
get images, expected levels of decision robustness and discrimination, real-
time processing, lightning conditions, and noise. Our results point out that
this set of parameters will impose the correlation architecture and the most
appropriate type of correlation filter for the application considered. Con-
sider, for example, the target images Lena and Barbara shown in
Figure 26a and b, respectively. Their respective FTs lead to significantly dif-
ferent spectra. While Lenas spectrum is almost centered (Figure 26c),
Barbaras spectrum shows a larger spatial distribution in the spectral plane.
The above-mentioned issues are extremely relevant here because the choice
of the correlation filter will impact correlators performances. For a face rec-
ognition application, the correlation filter should be able to deal with face
rotation with respect to the reference image. Specially designed correlation
filters to deal with color information have already demonstrated outstanding
performances for smart pattern application (Alfalou, Brosseau, &
Alam, 2013).
2.4.4 Numerical Implementation of a VLC Based on the Fraunhofer

Diffraction and Numerical Implementation of the Optical FT
As was shown by Mendlovic (2013) and Alfalou, Brosseau, and Alam (2013),
it can useful to use hybrid technologies which partly realize OC. Addition-
ally, it may be interesting to simulate optical operations before experimen-
tally implementing them.
2.4.4.1 Adapted All-Numerical Correlator for Face Recognition Applications

Several authors (Alfalou, Brosseau, Benkelfat, et al., 2012; Elbouz et al.,
2013) suggested and validated an all-numerical version of the VLC based
on Fraunhofer diffraction for simulating OC. On the other hand, all-optics
approaches are not the universal panacea and have their drawbacks and strin-
gent requirements, i.e., aberration effects, alignment of components, limi-
tation of the overall speed by how fast the information can be updated on
the input and output devices, and need of a costly optoelectronic interface.
Furthermore, the use of all-optics cannot be justified for many applications,
especially when the target image size is small. Indeed, several studies
(Alfalou, Brosseau, Benkelfat, et al., 2012; Elbouz et al., 2013) demonstrated
that (1) numerical methods are easily reconfigurable in sharp contrast to
purely optical methods; (2) optical correlators often need preprocessing
treatments in the input plane, e.g., target image denoising, processing in
the Fourier domain, and postprocessing in the correlation plane to make
a decision. All these treatments are realized numerically; (3) OC can be a
good alternative to numerical correlation processors.
In what follows, we describe a numerical implementation of a correlator
based on a numerical version of the FT. For that purpose, we consider a VLC
and implement numerically the optical FT (NO_FT) via the Fraunhofer dif-
fraction (Goodman, 1966; Voelz, 2011). A comparison of the results
obtained with a VLC using the FFT algorithm has confirmed the good
performances of the correlator based on a numerical implementation of
NO_FT (Alfalou, Brosseau, Benkelfat, et al., 2012).
The numerical algorithm of the VLC is shown in Figure 27b. We apply
this procedure based on Fraunhofer diffraction. The two-dimensional FT of
the input plane at point (u,v) reads as

expjkz k 2 2
Ui u, v exp j u +v
jz 2z
(19)
2
Uo x, yP x, yexp j ux + vy dxdy,
z
2
where k 2= denotes the wave vector and is the wavelength of the inci-
dent wave illuminating the transparency. If we suppose that z is chosen to be
so large that the phase factor P(x,y) 1 over the entire region of the (x,y)
plane in which Uo(x,y) is nonzero, Ui(u,v) is just the two-dimensional FT
of the aperture Uo(x,y), except for a multiplicative phase factor preceding
(a) FT 1
FT
2w 2w
P1 P2 P3
L1 L2
Identification
(b)
Target
Target image Correlation
image FT FT1 plane
Yes End
spectrum
No
Correlation filter
(Fn)
Figure 27 Illustrating the principle of the optical correlator and numerical implemen-
tation of the VLC (N_VLC). Elbouz et al. (2013)
the integral and which does not affect the intensity of the diffracted light, at

frequencies zu , zv .
In this paper, we consider the special case of a circular aperture with
diameter wl given by the pupil function
p!
x2 + y2
P x, y circ , (20)
wl
where the circle function circ(x) 1 if x < 1, 1/2 if x 1, and 0 if x > 1.

A MATLAB function can compute the Fraunhofer diffraction pattern with
wl denoting the circular aperture of the lens (Figure 29a) illuminated with
light of wavelength 633 nm. In order to show the impact of these parameters
on the correlation results, we consider a face recognition application with
horizontal rotation angles ranging from 45 to 45 and vertical rotation
angles from 10 to 10, test samples (Figure 2) from the Pointing Head
Pose Image Database (PHPID; Gourier, Hall, & Crowley, 2004). We make
use of the receiver operating characteristic (ROC) (Egan, 1975) curve
obtained from the peak-to-correlation energy (PCE) calculations
(Horner, 1992). Furthermore, a phase-only filter (POF), fabricated from
the face framed in red in Figure 28, is assumed in the VLC architecture.
Figure 29 presents ROC and area under curve (AUC) curves with dif-
ferent sets of parameters. It is implicitly assumed that Elbouz et al. (2013),
Mf =2W , where M is the number of pixels in the input plane, W is
Figure 28 The face database used for tests. The image framed in red (gray in the print
version) is used as the reference image for the POF. Elbouz et al. (2013).
Figure 29 Correlation results obtained with different sets of parameters (W, L, f ). Elbouz
et al. (2013).
the lens aperture, L is the input plane size, and f denotes the focal length of
the lens. It is easily seen that the correlators performance is significantly
affected by the values of these parameters.
As a first case, we choose W 6.25 mm, f 0.125 m, M 250, and
L 25 mm. The 4f setup (Figure 30a) is used to reconstruct the input plane
with this set of parameters. Figure 30b shows the reconstructed image at the
output using NO_FT (Equation 19). The face is limited by the aperture 2W
which takes a smaller value than the size of the input plane. Hence, the
reconstructed image do contains only partly the input image.
Application of this scheme leads to erroneous correlation results. This
can be seen in Figure 31 for several target and reference images. The auto-
correlation plane is presented in Figure 31a, while the correlation results
corresponding to a target image and its variant rotated of 45 contain a cen-
tral peak which looks like a correlation peak. However, the peak height
shown in Figure 31b is still large. This is counterintuitive since the target
and reference images are significantly different. In like fashion, this peak
f f f f
FT 1
FT
2w 2w
P1 P2 P3
L1 L2
(a) (b)
2w
L L
Figure 30 Reconstructed image using a 4f setup with W 6.25 mm, f 0.125 m,

M 250, and L 25 mm. Elbouz et al. (2013).
Figure 31 Correlation results between target and reference images when L 2W.
Elbouz et al. (2013).
is present in Figure 31c and d. It should be noted that this peak is related to
the circular aperture (Figure 30b).
As a second case, we set W 17.975 mm, f 0.125 m, M 250, and
L 25 mm. To overcome the aperture issue, we assume now that
L 2W as shown in Figure 32; a same requirement is made for the other
lens. It should be noted that the length of the diagonal of the input plane
is equal to the pupils aperture. Application of the proposed reconstruction
scheme leads to disappearance of the circular pattern shown in Figure 31c.
The corresponding correlation results are now shown in Figure 33.
Figure 33a shows the autocorrelation with a numerical implementation of
f f f f
FT 1
FT
2w 2w
P1 P2 P3
L1 L2
(a)
2W
(b) (c)
L L
Figure 32 Reconstruction of an input plane case when L 2W. Elbouz et al. (2013).
the optical FT (NO_FT). As expected from a POF, it is characterized by a

narrow and sharpened peak. Other panels in Figure 33, i.e., Figure 33bd,
do not possess any correlation peak. In the following simulations, we take
W 17.975 mm, f 0.125 m, M 250, and L 25 mm.
It is worthwhile to compare the VLC performances using the FFT algo-
rithm and the NO_FT algorithm. The tests consider the correlation of sub-
ject P2 having a close resemblance to subject P1. The results, obtained with a
specific threshold of the PCEs, are shown in Figure 34. The NO_FT algo-
rithm has the effect of decreasing the false alarm rate and increasing the good
Figure 33 Correlation results for two subjects of PHPID. W 17.975 mm, L 25 mm,
f 0.125 m, and M 250. Elbouz et al. (2013).
decisions since in that case we obtain an AUC 0.88 which is significantly

larger than that obtained with the FFT algorithm (AUC 075).
2.4.4.2 Unitary Discrete LCT: Analysis and Application

In their pioneering work, Zhao et al. (2013) suggested a numerical imple-
mentation of the LCT. This choice to implement the LCT described in
Correlation obtained with FFT Correlation obtained with the NO_FT

algorithm
(a) (b) W = 17.975 mm, L = 25 mm, and f = 0.125 m
1
Personne1/Personne 3
1
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
TPR
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
FPR FPR
AUC = 0.7558 AUC = 0.8827
Figure 34 Comparison between correlation results obtained with FFT and the
algorithm (NO_FT). Elbouz et al. (2013).
Zhao et al. (2013) is motivated by the fact that knowledge of the LCT allows
one to know important transforms used in IP, i.e., FT, fractional FT, (Ding,
2001) and Fresnel transform (Hennelly & Sheridan, 2005; Stern, 2006). The
unitary property of the LCT is of fundamental importance. The principle of
the LCT (Collins, 1970) is as follows. Consider

1 +1 j Ax2 2xy + Dy2
F y LM f x y p f xexp dx, (21)
jB 1 B
where f(x) is the input signal in the space domain, x, and F(y) is the output
signal in the LCT domain, y. Zhao et al. (2013) restrict the parameters of

A B
the ABCD matrix, M to be real numbers with det M 1.
C D
The FT can be calculated from Equation (21) by setting fA, B, C, Dg
f0,1, 1,0g in Equation (21). Thus,
s
1 +1
F y f xexpj 2xy dx: (22)
j 1
The inverse LCT can be made using

+1
1 j Ax2 2xy + Dy2
f x LM1 F y x p F yexp dy:
jB 1 B
(23)
Then, the DLCTs can be written as

X+P
F mTy WM n, mf nTx , (24)
nP
where Tx and Ty denote the periodic sampling intervals in the x space and y
LCT domains, respectively, n,m 2 P, + P , the number of sampling
points N is defined as N 2P + 1, Tx Lx =N , Lx is the extent of the input
signal f(x), and the elements of the kernel are
s
1 j n 2 o
WM n, m exp AnTx 2nTx mTy +D mTy
2
: (25)
jB B
Zhao et al. (2013) proved that a straightforward discretization of the inte-

gral transform is unitary if they use Dings sampling rate, or infinitely many
related sub-Ding sampling rates. They discussed also how this result might
be used for super-resolution and noted that it is reminiscent of the Moire
technique. Additionally, they demonstrated that a unitary property signifi-
cantly improves the convergence of phase-retrieval algorithms.
2.4.5 Extending the Depth of Focus in Tomography Systems for Glass

Lattice Three-Dimensional Mapping
Motivated by concerns to inspect glass lattices to locate various impurities
inside their volume, Paz et al. (2013) reported on the development of a com-
puted optical tomography system. They suggested using the potential ben-
efit of extended depth of focus for that application while preserving the
DOF. The quality of their optical system was tested numerically against
experimental results. It is characterized by the introduction of a phase-
affecting mask into the imaging system aperture plane (Zalevsky, Shemer,
Zlotnik, Ben Eliezer, & Marom, 2006; Zalevsky, 2010) and the Radon
transform (RT) used to reconstruct a 3D image from its angular projections
(Vest, 1974). For their application, Paz et al. (2013) considered that the best
results are obtained when the whole bulk does not exceed the DOF of the
scanning optics. As documented in Paz et al. (2013), both resolution and
DOF are important. The resolution is
R
Res 1:22 , (26a)
D
where denotes the wavelength, R is the working distance or focal length,
and D is the pupil diameter. The DOF is inversely proportional to the
resolution. Consequently, in order to improve both the DOF and resolution

of the imaging system, a modification of the aperture should be made.
For that purpose, they suggested attaching a phase-affecting element to
the aperture of the imaging optics (Figure 35). Since all the rays reach the
focal plane region, they interfere creating an extended focal channel. This
improves the resolution of the system while preserving the DOF.
The projection data that are gathered from multiple directions are used
to reconstruct the 3D information of an object by making use of the RT
through

P r, f x, yfxcos + ysin r gdxdy, (26b)
where f(x,y) denotes the distribution function, is the Dirac delta function, x
and y are Cartesian coordinates, and r and describe a line at distance r from
the center at angle . The optimized setup is displayed in Figure 36.
As described in Paz et al. (2013), the system contains a telecentric lens, to
inspect the whole depth of the lattice simultaneously. The camera is
connected to the lens. An extended depth of field (EDOF) element can
be combined inside the lens. A qualitative comparison of Figure 37a and b
indicates the good performances of this approach. This technique allows
capturing information from different perspectives, so information can be
reconstructed using the RT. However, modulation of the high frequencies
with the extended DOF is low since with this system it is difficult to enhance
the signal with a deconvolution filter, as the intensity is close to the noise
level, so the accepted deconvolved pictures are a bit noisy.
Modified wavefront
Focus channel
Incoming
wavefront
Extension of focus range

Figure 35 Synoptic diagram of the setup proposed by Paz et al. (2013) to extend the
focus range.
Telecentric EDOF element

optic
Inspected lattice
inside immersion liquid
Light source Telecentric

optic
Camera
Figure 36 Synoptic diagram of the experimental setup used for extending the depth
of focus.
Figure 37 DOF measurements with the resolution target (a) with the EDOF element at
locations 0 mm and (b) without the EDOF element at locations 0 mm.
2.4.6 Spectral Imager: Information Capacity as a Figure of Merit

for Spectral Imagers
Skauli (2013) proposed a new procedure to evaluate the information capac-
ity of a spectral imager. The analysis presented in this study incorporates the
coregistration performance in a definition of total noise, which in turn can
be used in the definition of information capacity. Particular to spectral imag-
ing is the need for spatial coregistration (Cao, Guichard, & Hornung, 2010;
Cox & Sheppard, 1986; Fales, Huck, & Samms, 1984; Fellgett & Linfoot,
1955; Piestun & Miller, 2000; Tisse, Guichard, & Cao, 2008). Skauli
(2013) suggested that (1) an imaging system can be considered as a commu-
nication channel in the information-theoretic sense, where information
about the imaged object is transferred to the digital image data output. This
Figure 38 Illustration of the trade-off between coregistration error and pixel count
(Skauli, 2013). The figure illustrates the pixel footprint for two bands, shown as red (light
gray in the print version) and blue (dark gray in the print version) rectangles, for two
cases. The left image considers small pixels and a spatial coregistration error that is a
large fraction of the pixel size. The right image deals with larger but fewer pixels.
channel can be characterized by the information capacity, which is the upper

limit on the ability of the imager to collect information about the scene; (2)
for a given amount of coregistration error, coregistration will tend to be
improved if the detector elements that define the pixel size are made larger,
but at the expense of lower spatial resolution. This point is illustrated in
Figure 38 for a simple case with two bands, whose pixel footprints in a scene
are shown in red (light gray in the print version) and blue (dark gray in the
print version) rectangles. On the left is a case with two spatial pixels and a
spatial coregistration error that is a large fraction of the pixel size. On the
right is a case where the pixel footprint is twice as large, with the same
amount of distortion. This latter case has a smaller amount of coregistration
error in the recorded spectrum, but only half as many pixels.
Skauli (2013) proposed a coregistration metric to express the effect of
coregistration error as an added noise, which can be incorporated in a model
for the information capacity of the imager. If one considers a single pixel in a
single spectral band of a spectral imager, the output signal is the number of
excited N photoelectrons for a given spectral radiance L

N LtAw , (27)
hc
where is the quantum efficiency, t is the integration time, A is the area of

the entrance pupil of the imager, w is the solid angle subtended by the pixel
in the scene, is the spectral bandwidth, is the wavelength, and /hc is the
photon energy. From Equation (27), the mean number of photoelectrons

per pixel N can be expressed as

N LtAw : (28)
hc
Skauli (2013) demonstrated that the effect of coregistration errors on the

signal can be approximately modeled by assuming proportionality with the
amount of nearest-neighbor contrast in the image. Let the mean difference
between nearest-neighbor pixels be N with 0 < < 1. The mean ampli-
tude of signal errors due to misregistration can be estimated as
Ncoreg
s N , (29)
where s is a constant (Skauli, 2013). An estimate of the information capacity

was used by Skauli (2013) as
6000
5000
es0 = 0.07
Lower capacity bound (bits/frame)
4000
3000
es0 = 0.12
2000
1000 es0 = 0.2
0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Binning factor b
Figure 39 Information capacity for one band of a spectral imager with varying sizes of
the detector pixels: b 1, 0.15, and N 20,000. Skauli (2013).
r
1 N
C N , s logN s : (30)
2 2
Using Equation (30), we show the resulting information capacity esti-
mate for one spectral band ( 0.15) in Figure 39. This graph illustrates
how the information capacity reaches an optimum for a particular pixel
count if the distortions of the optics are held fixed. It illustrates also that
coregistration error can lead to a large loss of information.
3. PART B: OIP
There has been palpable progress in the development of optical tech-
niques and tools to aid picture interpretation, improve image quality,
enhance image contours, and correct geometric distortions. In this second
part, we highlight an eclectic mix of important topics in OIP that we feel
have not been given sufficient air time. It is hoped that this collection
of excursions, though idiosyncratic, will help sharpen collective understand-
ing of the core issues to be tackled on the way to a complete and credible
understanding of the basic principles of OIP.
3.1 Optical Fourier Transform for Medical Image Processing

and Phase Contrast Imaging
Many of the archetypal features of OIP originate from the FT properties;
e.g., the Fraunhofer diffraction pattern of an aperture function is identical
to its FT (Goodman, 1966). If a lens is placed after an object, it will shorten
the image plane distance. Consequently, the Fourier spectrum is located at
the back focal plane of the lens. This is the basis of the optical FT (OFT).
Under appropriate conditions, OFT is a powerful tool for OIP
(Yelleswarapu, Kothapalli, & Rao, 2008). For our purposes, we recall the
principles of a standard optical setup used in OIP (Figure 40) which can
Fourier plane:
Spatial or phase
Filter
Collimated Amplitude Spatial Filtered/
Processed
optical or phase frequency modified
image
source object spectrum spectrum
Optical Inverse
Fourier Fourier
transform transform
Figure 40 4f OFT processing scheme. Yelleswarapu et al. (2008).
readily be applied to realize low-pass, band-pass, high-pass filtering, and

phase shift (Yelleswarapu et al., 2008), correlation (Alfalou & Brosseau,
2010a, 2010b; Alfalou, Brosseau, & Alam, 2013), and compression and
encryption (Alfalou & Brosseau, 2009).
It is important to review a pioneering paper of Yelleswarapu et al. (2008)
dealing with medical image processing with application to mammography
and phase contrast microscopy to observe biological samples. We consider
the schematic diagram displayed in Figure 40. A well-collimated optical
source is incident on the target image (amplitude or phase object) to be
processed. A first converging lens allows Fourier transformation of the light
bearing the object information, thus getting the target image spectrum. In
this plane, various techniques can be used to process different spatial fre-
quencies, e.g., low pass, high pass, and phase shift, according the specific
application. It is worth observing that low frequencies are localized near
the center of the target image, while high frequencies characterize the
contour information. Then, an FT1 with another lens is used to image
processed spatial frequencies.
Remarkably, Yelleswarapu et al. (2008) showed that OFT processing
is adapted to enhance the spectral features of both amplitude and phase
objects.
3.1.1 Spatial Filtering for Medical Image Processing

For practical demonstration of their technique, Yelleswarapu et al. (2008)
considered the detection of microcalcifications in a mammogram for early
diagnosis of breast cancer. Various spatial filtering techniques such as con-
ventional 4f filtering using a spatial mask, photoinduced polarization
rotation in photosensitive materials, Fourier holography, and nonlinear
transmission characteristics of optical materials are employed for the pro-
cessing of mammograms. The detection of microcalcifications is known
to be a difficult problem for radiologists and researchers involved in bio-
medical imaging (Garfinkel, Boring, & Heath, 1994; Kerlikowske &
Barclay, 1997; Leitch et al., 1997; Russ, 2002; Smith, 1993). In this review,
we set side all questions of the microscopic origin of the microcalcifications,
and attempt to say something about the optical modes to detect them.
As noted by Yelleswarapu et al. (2008), features such as micro-
calcifications are often buried in the background of soft dense breast tissue
(low spatial frequencies) in a mammogram resulting in low contrast between
these essential features and the background. In the Fourier spectrum of a
given mammogram image, microcalcifications correspond to high spatial
CL1
L1 L2 NDF
532 nm CCD
laser camera
NDF1 Spatial filter

SLM or film
mammogram
Figure 41 Schematic of the standard 4f configuration used for contour enhancement
of mammograms. CL, collimating lens; L1 and L2, Fourier lenses; NDF, neutral density
filter. The spatial high-pass filter for image processing has a circular shape with diameter
R. Yelleswarapu et al. (2008).
frequencies (Figure 41). For the purpose of filtering out the undesired low
spatial frequencies, Yelleswarapu et al. (2008) suggest adapting the standard
4f setup by introducing various spatial filters (Figure 41). As usual, the two
lenses correspond, respectively, to the FT and FT1 processing.
To exemplify the benefit of using the optical setup shown in Figure 37,
the processed image captured by the CCD camera displaying micro-
calcifications corresponding to the high spatial frequencies is displayed in
Figure 42b. This picture is in sharp contrast with the one shown in
Figure 42a where the dense breast tissue background is present. However,
as recognized by Yelleswarapu et al. (2008), the main drawback with this
filtering technique is that the spatial filter has to be modified when the mam-
mogram is changed because the spatial frequency spectrum at the Fourier
plane will be different. Thus, it cannot be real-time processing as the filter
is not all-optical and continually controllable. To remedy this, nonlinear
filtering techniques for self-adaptative and real-time computing such as those
developed by several workers (Babkina & Voloshinov, 2001; Kato &
Goodman, 1975) can be employed.
3.1.2 Photoinduced Anisotropy for Medical Image Processing

Another important approach in the processing of mammography images has
been reported by Yelleswarapu et al. (2008). This research is based on
exploiting the photoinduced anisotropy properties of materials. A great deal
Figure 42 Processing of mammograms using spatial filtering (Figure 41): (a) original
mammogram. (b) With spatial filtering with a circular filter of diameter adapted to
the mammogram. Yelleswarapu et al. (2008).
Figure 43 Optical setup proposed in Yelleswarapu et al. (2008) with bacteriorhodopsin

films for mammography applications. The partial filtering is accomplished by selectively
rotating the analyzer such that undesired spatial frequencies are filtered out.
of research has focused on spatial filtering using photoinduced anisotropy in

bacteriorhodopsin (bR) films for medical image processing (Egami, Suzuki,
Uemori, Sugihara, & Okamoto, 1997; Ferrari, Garbusi, Frins, & Priz, 2005;
Joseph, Aranda, Rao, Akkara, & Nakashima, 1996; Kato, Yamaguchi, &
Tanaka, 1996; Korchemskaya & Stepanchikov, 1998; Menke et al., 2006;
Shih, Shishido, Chen, Wood, & Khoo, 2000). The optical device investi-
gated by Yelleswarapu et al. (2008) for mammography applications is shown
in Figure 43.
Based on this optical setup, Yelleswarapu et al. (2008) were able to show
the correspondence between spatial frequencyintensitypolarization and
each spatial frequency band is encoded with a unique polarization. When
the analyzer is at right angles to the input beam polarization, the low-
frequency components that experience no polarization rotation are blocked
by the analyzer. On the other hand, high spatial frequency components
corresponding to edges of the object experience polarization rotation due
to their low intensity.
Yelleswarapu et al. (2008) provided an example to illustrate the above
discussion. As an illustrative example, Figure 44a shows the original mam-
mogram where microcalcification areas are not visible. Figure 44b is the
processed image clearly displaying these microcalcification areas (inside
the white rectangle) when the low spatial frequencies are filtered out.
3.1.3 Fourier Holography

We now turn our attention to Fourier holography applied to medical imag-
ing (Darakis & Soraghan, 2006, 2007; Darakis, Naughton, & Soraghan,
2007; Naughton, Frauel, Javidi, & Tajahuerce, 2002; Paturzo et al., 2008;
Yamaguchi & Zhang, 1997; Xia et al., 2013). To illustrate this, we consider
phase-shifting interferometry digital holography (PSI-DH).
3.1.3.1 Phase-Shifting Interferometry Digital Holography

PSI-DH plays a central role in three-dimensional (3D) object reconstruction.
Consequently, its characterization and optimization are important issues,
which have been extensively studied recently. However, it necessitates
Figure 44 Results obtained with the optical setup shown in Figure 39 when the target
image is a mammography image. Yelleswarapu et al. (2008).
Figure 45 PSI hologram recording setup. Darakis et al. (2007).
storage and processing of large amounts of information (Darakis & Soraghan,

2006, 2007 Darakis et al., 2007; Naughton et al., 2002; Paturzo et al., 2008;
Yamaguchi & Zhang, 1997; Xia et al., 2013). In this 3D imaging technique,
the complex light distribution at the plane of the recording camera (Figure 45)
can be calculated from several interference patterns corresponding to different
reference light phases. Since several interferograms have to be recorded, the
use of this method is limited to static scenes. The large amount of storage it
needs is also a strong limitation of this method.
The recording setup for capturing PSI holograms, shown in Figure 45, is
due to Darakis et al. (2007). The general expressions for the wavefront due
to the object and the reference wave are, respectively,
Wo x, y, t ao x, yexpio x, y expit,
(31)
WR x, y, t aR x, yexpiR x, y expit;
where a and are the amplitude and the phase of the wave, respectively.
The intensity on the CCD sensor reads
I x, y, t Wo x, y, t + WR x, y, t 2
(32)
Ii x, y + C x, yconso x, y R x, y ;
where Ii(x, y) is the mean intensity, c(x, y) defines the contrast, and o x, y
R x, y is the net phase. Since the second part of the light passes through
a half-wave and a quarter-wave, one can adjust the fast and slow axes at
angle
I x, y, t Ii x, y + C x, yconso x, y R x, y + t : (33)
The central aspect of the PSI method is to capture multiple interfero-
grams I() that each indicate the optical phase relationships ( 0, /2, )
between light returned from all sampled points on the illuminated surface
and a controlled reference beam of light that is collinear to the object beam.
From the set of these I()s, holograms are computed that contain information
defining the shape of the object. Knowledge of these intensity patterns allows
us to determine the actual phase of the wave. The four interference patterns
corresponding to phase shifts of 0, /2, and for object reconstruction
are calculated from
I x, y; 0 Ii x, y + C x, yconso x, y R x, y + 0
h i
I x, y; Ii x, y + C x, ycons o x, y R x, y + (34)
2 2
I x, y; Ii x, y + C x, yconso x, y R x, y + :
Using the PSI algorithm (Yamaguchi & Zhang, 1997), the wavefront at
the plane of the camera can be calculated from the recorded interference
patterns as
1 i hn o n oi
Wo x, y I x, y; 0 I x, y; + i I x, y; I x, y; ,
4jWR j 2 2
(35)
where WR is the amplitude of the reference wave (for a plane normally inci-
dent reference wave WR 1). Equation (35) is called a PSI hologram.
3.1.3.2 Application for Medical Image Processing

Yelleswarapu et al. (2008) and others (Chang, Hong, & Yeh, 1990;
Kothapalli, Wu, Yelleswarapu, & Rao, 2004; Ochoa, Goodman, &
Hesselink, 1985; Okamoto, Yamaguchi, & Yamagata, 1997) considered
holography applications for medical image processing. By recording and
reconstructing the Fourier hologram, one can retrieve both the amplitude
and the phase information of the target image. This technique can be
exploited for spatial filtering and edge enhancement in real time by record-
ing and reading object information gratings. Similarly, two-beam coupling
in photorefractive materials can be employed to spatially amplify the desired
frequency band of the Fourier spectrum without discarding any of the inci-
dent image information (Chang et al., 1990).
Figure 46 shows the experimental arrangement of transient Fourier holo-
graphic gratings based on photoinduced isomerization properties of bR
films used to perform spatial filtering for the detection of microcalcifications
in real time (Kothapalli et al., 2004). The important thing to note is that the
diffraction efficiency is maximal when the object (mammography or binary
object E) beam intensity matches that of the reference beam intensity.
Figure 46 Standard 4f configuration for the study of the edge enhancement using
transient Fourier hologram. M, mirror, BS, beam splitter; L, lens. Kothapalli et al. (2004).
Figure 47 Image processing of mammogram using Fourier holograph: (a) mammogram

with region of interest. (b) Processed image where microcalcification zones are clearly
identified. Kothapalli et al. (2004).
At either low- or high-intensity region of the object beam, the diffraction

efficiency decreases. Thus, the desired spatial frequency components can
be selected by matching their intensity to that of the reference beam.
An example is illustrated in Figure 47. High spatial frequency compo-
nents of the mammogram are recorded by matching the intensity of the ref-
erence beam to the intensity of the high spatial frequency band. When the
object beam is blocked, the reference beam performs the reconstruction of
the recorded hologram displaying the spatial frequencies whose intensity in
the Fourier spectrum is matched to the reference beam intensity. Thus, the
radiologist can scan for desired microcalcifications by rotating the variable
attenuator that controls the reference beam intensity.
3.1.4 Nonlinear Transmission for Medical Image Processing

Another context in which the optical processing of medical images based on
spatial filtering is considered in this section (Yelleswarapu et al., 2008). This
method uses chromophores that exhibit nonlinear absorption (Kothapalli,

Wu, Yelleswarapu, & Rao, 2005; Thoma, Oesterhelt, Hampp, &
Brauchle, 1991; Xuan et al., 1988; Yelleswarapu et al., 2006). More specif-
ically, Yelleswarapu, Wu, et al. (2006) demonstrated that a power-limiting
mechanism can potentially be used for processing and filtering spatial fre-
quencies of any two-dimensional image by incorporating a power-limiting
material (i.e., a bR film) at the Fourier plane. In general, the Fourier spectrum
of objects possesses low spatial frequencies at high intensities and high spatial
frequencies at low intensities except for some special cases like fingerprints.
These authors show that low spatial frequencies are blocked due to their
nonlinear absorption in power-limiting materials at high intensities, while
the high spatial frequencies pass through due to linear absorption at low
intensities. In addition, they show that band-block filtering can be achieved
by controlling the incident beam intensity. As a fringe benefit, when the same
experimental setup for both power-limiting experiment and OIP is used, as
in the case of image bearing intense laser beam, the sensitive detectors are
potentially protected by blocking the intense low spatial frequencies, while
detecting the essential features of the image by detecting the weak high spatial
frequencies. The main feature of this technique is to demonstrate that var-
iable high-pass filtering can be achieved by controlling the incident intensity.
Kothapalli et al. (2005) demonstrated another protocol of optical
processing for edge enhancement of phantoms and mammograms.
A schematic of the experimental setup employed is shown in Figure 48. This
real-time nonlinear optical Fourier filtering for medical image processing is
based on the standard 4f setup and exploits the light-modulating character-
istics of thin films of the biophotonic material bR using two wavelengths,
442 and 568 nm. The mammogram (or phantom) target is illuminated by
a well-collimated and spatially filtered 442 nm beam from a He-Cd laser
source. A convergent lens L1 is used to get the Fourier spectrum of the tar-
get. The bR film is placed at the Fourier plane of the blue beam. Next, the
FT1 is obtained using the convergent lens L2 and the target is imaged onto
the CCD camera placed at the back of L2. A second collimated 568 nm con-
trol beam from an Ar-Kr laser is focused on the bR film using the convergent
lens L3. The low spatial frequency components of the target, which are spa-
tially overlapped on the film, are blocked as shown in Figure 48, by proper
selection of the focal lengths of the lenses L1 and L2. NDFs are placed in the
path of both the laser beams to control the intensity.
As noted earlier, this nonlinear optical filtering technique exploits the
photocontrolled light-modulating characteristics of bR films for the early
Figure 48 Experimental setup suggested by Kothapalli et al. (2005) for medical image
processing; L1, L2, and L3 are convergent lenses, NBF is a narrow band filter to block
568 nm at the CCD plane, CL1 and CL2 are collimation lenses, and NDF1 and NDF2
are neutral density filters to control the intensity of the beams.
Figure 49 Edge enhancement of the region of interest in a mammogram using the

optical setup shown in Figure 44: (a) without filtering and (b) with filtering.
Kothapalli et al. (2005).
detection of microcalcifications. The bR, an optically switchable bistable

material, is used to process the Fourier spectrum of the mammogram con-
taining information about the microcalcifications (high spatial frequencies)
and background of soft tissues (low spatial frequencies). The narrow band
filter (NBF) is used to block the 568 nm beam at the CCD plane.
Figure 49 presents an example of results obtained by the setup dis-
played in Figure 48. Figure 49a shows the scanned version of the region
of interest of a mammogram. This region of interest is illuminated by blue

light, as shown in Figure 48. As discussed previously, the yellow control
beam induces photoisomerization in the bR film, thus reducing the trans-
mission of the low spatial frequency components. The background of
soft dense breast tissue corresponding to the low spatial frequencies in
the Fourier spectrum is filtered out at the bR plane. In contrast, the high
spatial frequencies corresponding to microcalcifications are transmitted.
Figure 49b shows the reconstructed image displaying only micro-
calcifications not visible to the naked eye in the original clinical
mammogram.
3.1.5 Phase Contrast Imaging of Live Biological Specimens

It is worth observing that Yelleswarapu et al. (2008) considered also the
important case of phase contrast imaging of live biological specimens. Phase
objects are transparent; i.e., they provide no contrast with their environment
and alter only the phase of the wave. The optical thickness of such objects
generally varies from point to point due to changes in either the refractive
index or physical thickness or both. Since the eye cannot detect the changes
in phase variations, phase objects are invisible to the naked eye. This is the
main reason why Yelleswarapu et al. (2008) explained that the standard spa-
tial filtering techniques need to be modified and proposed to include an
additional phase difference between the undeviated (low spatial frequencies)
and deviated (high spatial frequencies) light so that they can interfere,
thereby converting the phase variation into amplitude contrast. Hence,
Yelleswarapu et al. (2008) proposed similar techniques as those developed
in other references (Castillo et al., 2001; Gl uckstad, 1995; Gl uckstad,
Lading, Toyoda, & Hara, 1997; Komorowska, Miniewicz, Parka, &
Kajzar, 2002; Rehn & Kowarschik, 1998; Sendhil, Vijayan, & Kothiyal,
2005; Yelleswarapu et al., 2006; Armitage et al., 1989), and have experimen-
tally demonstrated that the idea of phase contrast imaging can be a viable
technique. Figure 50a shows a bright-field image of a transparent object
characterized with onion cells from the skin of an onion bulb. The thin layer
of cells is so translucent that the phase contrast is needed for viewing.
Figure 50a shows a bright-field image of such a preparation of onion cells;
the walls and nuclei are visible but that is greatly enhanced with the phase
contrast image seen in Figure 50b. There is reason to believe that this tech-
nique increases significantly the contrast since the veil manifestly perturbing
the left image is almost invisible in the right one.
Figure 50 Example showing the edge enhancement of a transparent object: (a) Bright-
field image and (b) Fourier phase contrast image of onion cells. Yelleswarapu et al.
(2008).
3.2 Optical Implementation of Mathematical Operations

In the following sections, we motivate the study of optical systems allowing
us to realize standard mathematical operations, i.e., mathematical
morphology,9 involved in OIP such as dilation and edge detection.
3.2.1 Optical Incoherent Setup Implementing Novel Gray Image

Processing in a Single Operation Optical Morphological Processor
Proposals so far led to three kinds of morphological IP: (1) the binary images
processing (Huang, Jenkins, & Sawchuk, 1989; Liu, Zhang, & Zhang,
1994), (2) the gray image processing with binary structure elements (Liu,
Cui, & Zhao, 1994), and (3) the gray image processing with gray structure
elements (Huang, Wu, Jin, Yao, & Yan, 1994; Shih & Mitchell, 1989).
Huang, Jin, Wu, and Yan (1996) proposed and validated a novel gray IP
using a single operation optical morphological processor. Their pragmatic
approach can simplify and reduce the number of mathematical operations
required for the morphologic processing of gray images and gray structure
elements. The optical setup uses a simple algebraic representation and is
adapted to parallel architecture. Typically, it consists in using two convolu-
tion thresholds (one is between a gray image and a binary structure element,
and another between a binary image and a gray structure element) and one
algebra operation.
Binary IP is characteristic of neighbor operation. The main aim of binary
IP is to reduce the number of operations. It relies on the sole neighbor oper-
ation, whereas the gray IP depends on both the neighbor and the algebra
9
Mathematical morphology is a set of methods for manipulating collections. In IP, it is applicable for
processing images, e.g., dilation and erosion, which are implemented by convolving an image with a
structural element and thresholding on maximum and minimum values.
operations. The technique developed by Huang et al. (1996) is a mixture of

the algebra operation of gray level between the gray images and the gray
structure elements where an interconnection condition is met, and then
followed by a nonlinear operation (the gray maximum or minimum) in a
neighborhood region. Indeed, the gray image processing consists of a single
fundamental operation, two common algebra operations (sum +, subtrac-
tion ), a union U, five element images, and an image space W. Any gray
image processing can be described by using a part (or the whole) of these
elements. Considering two gray images Fg and Rg, these authors define
the binary image of the gray interesting region as Fb, the complement of
v
the gray image as Fg , and the reflection of gray image as Fg . They also define
the single fundamental operation O{.} as

O Fg , Rg , Fb , Rb , T, Si21, 2, 3

Si1 ! O Fg , Rg , Fb , Rb , T , S1 Fg Rb T + Fb Rg T

(36)
Si2 ! O Fg , Rg , Fb , Rb , T , S1 Fg Rb T Fb Rg T ,

S ! OF , R , F , R , T , S F *R [ F *R
i3 g g b b 1 g b T b g T
where Fb and Rb denote the binary images of the gray interesting region of Fg
and Rg, T is a threshold, Si21, 2, 3 denotes the addition operation for i 1, the
subtraction operation for i 2, and the union operation for i 3. We will
not dwell on the technical aspects of the derivation of Equation (36), but pass
on at once to its intuitive understanding. In practice, Equation (36) can real-
ize the union, the addition, the subtraction, the dilatation, the erosion, the
opening, the closing, the hit-or-miss, the edge detection, and the thinning
(Huang et al., 1996). In what follows, we consider the specific example of the
dilatation of a gray image Fg with a gray structure element Rg. The dilation of
the gray image Fg by a structure element can be written as

Fg Rg O Fg , Rg ,Fb , Rb , T 0, S1 Fg Rb 0 + Fb Rg , (37)
where denotes the dilation operation, T 0 denotes a given threshold, i.e.,
a high-pass filter, and S1 is the subtraction operation. In Figure 51, we show an
example illustrating how Huang et al. (1996) implement dilation.
Let us focus on the optical setup displayed in Figure 51 which is com-
posed of an incoherent source and a first convergent lens L1. The output
light is incident on a liquid crystal TV (LCTV1) used to monitor in real time
the binary images Rb (or Fb). Next, a second liquid crystal TV (LCTV2) is
used to monitor the gray image Fg (or Rg) obtained with the camera CCD1.
Figure 51 Optical setup proposed by Huang et al. (1996) to realize the single operation
gray image processing.
Behind the convergent lens L2, a CCD camera collects the convolution
superposition images Fg Rb and Fb Rg into a computer for thresholding.
Finally, a monitor is used to display the processing results. An illustration of
this method is shown in Figure 52 which implements dilation, i.e.,
Equation (37), thanks to the setup of Figure 51. The first step of this method
consists in encoding Fg and Rg and their binary versions Fb and Rb (line 2).

After thresholding, the results Fg , Rb T 0 and Fb , Rg T 0 are shown in
the third line of Figure 52. Adding the two images leads to the results shown
in Figure 52.
3.2.2 Neighborhood Operation Binary Image Algebra for Optical

Morphological Image Processing
Optical morphological IP is becoming increasingly important for a variety of
real-time applications (Huang et al., 1993). For this purpose, optics is well
adapted to make this processing (at least, partially) at the input. Such an idea
was proposed notably by Yuan, Jin, Wu, and Yan (1996) who validated an
experimental architecture dedicated to optical morphological IP based on
the neighborhood operation binary image algebra (NOBIA). This approach
requires three optically implementable operations: convolution, nonlinear
filtering, and intersection. Yuan et al. (1996) defined the NOBIA as
NOBIA fS, Fn g, (38)
where S denotes the space image.10 Fn is the basic operation of the NOBIA
and is defined by
10
The image space S is the power set (the set of all subsets) of the universal image. Any image is a subset
of the universal image.
Figure 52 Dilation of a gray image using the single fundamental operation. Huang
et al. (1996).
Fn fX, R, NPn , Y g NPn X, R \ Y ,

NLn U X, R \ Y , (39)
NLn XR \ Y ;
where X and Y are two binary given images, R is the binary structuring ele-
ment, denotes the convolution operation, and NP(.) is the neighborhood
processing operation introduced as
(
n1
! NP1 X, R XR
NPn X, R n1 (40)
! NP2 X, R XR;
where denotes the erosion function. Consequently, Equation (39) can be

rewritten according to
( n1
! F1 X, R, NP1 , Y XR \ Y
Fn fX, R, NP , Y g
n
n1 (41)
! F2 X, R, NP2 , Y XR \ Y :
This approach enables to realize a variety of operations such as the union

of two images, difference of two images, erosion, symmetric difference of
two images, opening, closing, hitmiss transform, edge detection, and
shifting. In the following, we specifically focus our attention on edge detec-
tion. Let us define the out-object detection of a given image X by

ED0 X F2 X,I,NP2 ,F1 X, R, NP1 , W , (42)
where W denotes the universal image. Yuan et al. (1996) performed a variety
of tests and demonstrated that the NOBIA achieves better performances for
edge detection than the binary image algebra. The architecture implementing
the optical neighborhood operation binary image processor (ONOBIP) for
executing the parallel algorithms of NOBIA is shown in Figure 53.
Figure 53 contains two parts. The block framed in black is the control
unit which includes a clock, a program counter, a test and feedback control-
ler, and an instruction memory. The block framed in red is the parallel
processing unit which includes an image memory device for storing multiple
images, three frame buffers, and a CPU for executing the basic operation
Fn{.} of the NOBIA (Figure 54). The three buffers X, R, and Y are, respec-
tively, used to store the three input images for the central processing unit.
The images stored in the image memory device can be, respectively, sent
to the three frame buffers. The CPU can be controlled by the controlling
unit to perform proper functions. The ONOBIP operates as follows:
Clock Program counter
Test and feedback

Instruction memory controller
Image Frame buffer X
Image input
Memory Frame buffer R Fn {.}
Image output
Device Frame buffer Y
Parallel processing unit
Figure 53 Synoptic diagram of the optical neighborhood operation binary image

processor (ONOBIP). From Yuan et al. (1996).
(a)
IP LCLV A M2 OP
M1 BS
S
L
G
P
CB
(b)
P1 ML P2 P
a d

b
(c)
X Y OP
Collimated
beam
f+z
D
n
Figure 54 (a) Optical setup suggested in Yuan et al. (1996) for optically implementing
the optical neighborhood operation binary image processor (ONOBIP). (b) The input
plane P is illuminated by a uniform incoherent light. A mask M is put close in front
of an imaging lens L with a focal length f. The distance between the plane P and the
mask M is D. The distance between the mask M and the lens L is very small and can
be neglected. The mask M is used to display the structuring element image R. Plane
P is the imaging plane in which a clear image of the input plane P1 can be obtained.
In the output plane P2, we can get the convolution result. (c) Optical circuit for
performing the intersection operation.
(1) the corresponding images are sent to the frame buffers X, R, and Y from
the image memory device; (2) the basic operation Fn{.} is executed by
selecting the proper formats of the nonlinear function; (3) the result is out-
put, tested for program control, or sent back to the image memory device.
Then the system can be fed back to step (1). Based on Figure 53, Yuan et al.
(1996) suggested a scheme (Figure 54a) which realizes Fn{.} optically. This
setup is composed of three blocks: an optical convoluter (Figure 54b), an
optical spatial nonlinear device with two operation modes such as a liquid
crystal light valve, and an optical implementation of parallel intersection
(Figure 54c).
The diffuse screen G is put just before the input plane IP. The incoherent
uniform light (S) is used to illuminate it. The mask M of the structuring ele-
ment R is put close in front of the lens L. Plane IP, mask Mi, lens L, and
LCLV construct an incoherent optical convoluter. When the image shown
in the input plane IP passes through mask M and lens L, the convolution
resultant image V(X,R) is formed in the writing plane of LCLV.
A collimated HeNe laser beam (CB) is used as the reading light to read
the written-in image of LCLV, where P is a polarizer and A is an analyzer.
Selecting different nonlinear characteristics of LCLV to simulate the
corresponding formats of the nonlinear function, the corresponding binary
image is read out from LCLV. The output binary image passes through the
mask M2 coded by binary image Y and executes the intersection operation
with image Y. Then in the output plane OP, the operation result of Fn(X, R,
NPn, Y) can be obtained. Using different nonlinear characteristic curves of
LCLV to perform different type nonlinear filterings, the different formats of
the basic operation are implemented experimentally. Figure 55 gives the
experimental results obtained by Yuan et al. (1996).
Figure 55 Results obtained by Yuan et al. (1996) using the optical setup displayed in
Figure 50a.
Figure 55a is an input binary image X. Figure 55b shows the structuring
element R with 3 3 image points, and Figure 55c gives the image Y.
Figures 51d and 55e give the experimental results of F1{X, R, NP1, Y}
and F2{X, R, NP2, Y}. The most remarkable implication of these results
is the validation of the ONOBIP by optical means. The suggested ONOBIP
is of a high parallelism, simple structure, and potential fast processing ability
and can be used to efficiently execute the algorithm of NOBIA. But it
remains difficult to realize in practice. We postpone a discussion of the edge
detection using this method for a few sections.
3.3 Techniques for Increasing Image Quality

The scope of this section is to review the recent progress in the techniques
which allow increasing image quality by considering enlightening correc-
tion, image rotation, image blur, and moire suppression.
3.3.1 Enlightening Correction

The detrimental effect of enlightening on the image quality is particularly
critical for underwater imaging applications. Water is responsible for the
attenuation of light due to absorption. As we go deeper on a dive, more
color is absorbed by the water. Color vision is also affected by turbidity
of the water as well as particulate matter. Water preferentially absorbs red
light, and to a lesser extent, yellow and green light, so the color that is least
absorbed by water is blue light. Three factors primarily affect underwater
visibility: light penetration, biological species, and particulate matter. Not
surprisingly, these factors are often related to one another, as well as to other
environmental factors. The amount of light that penetrates the water
depends on the overall light level, the angle at which the light rays meet
the water, and the roughness of the surface. In addition to increasing under-
water visibility, light penetration provides the light energy needed for pho-
tosynthesis by drifting algae and phytoplankton. In addition, particulate
matters, e.g., suspended sand and silt, affect visibility underwater.
According to the size of these particles, they scatter forward or backward
light, resulting in reduced visibility. Attenuation can be described using the
Beer-Lambert law, which implies that the underwater images should be
taken at short range. Some further observations that seem significant: first,
scattering by particles and by the water itself affects image quality; i.e., max-
imum imaging distances range from centimeters in dirty water to more than
100 m in very clear water (Dickey, Kattawar, & Voss, 2011). Second, non-
uniform lightning conditions result in different intensity levels between
images of the same area. Figure 1 illustrates the well-known fact that the
underwater medium can significantly degrade the quality of the image. That
is, small-angle forward scattering of light and nonuniform lightning condi-
tions can cause images to blur and result in intensity differences between the
images of the same area.
Light attenuation due to absorption and scattering phenomena has been
studied before (Duntley, 1963; Hou, Gray, Weidemann, Fournier, &
Forand, 2007; Jaffe, 1990; Leonard, 2012; Wells, 1969). Following earlier
studies (Duntley, 1963; Hou et al., 2007; Jaffe, 1990; Wells, 1969), Leonard
(2012) considers attenuation as a linear combination between absorption11
and scattering.12 The idea here is to consider that the light intensity trans-
mitted at a distance d in a sample can be written as
I d I0 expa + bd , (43)
where I0 denotes the intensity of the source, a is the absorption coefficient of
the underwater medium, and b is the total scattering coefficient. Equa-
tion (43) shows that I(d) depends on the absorption of the medium.
Figure 56 illustrates the estimation of the propagation distance of light in
clear water (Sogandares & Fry, 1997) as a function of its wavelength. In addi-
tion, I(d) depends also on scattering phenomena in the medium. The latter
depends on the particle size, light incidence, and the distance between the
particles and the camera. The perturbating noise that affects the overall
image is called sea snow. Leonard (2012) considers that this noise depends
on the forward scattering of light, i.e., blurring the image, and also on back-
scattering of light, i.e., the image is affected by a veil.
It is worth observing that this issue has been comprehensively reviewed
by Leonard (2012), Leonard, Alfalou, Alam, et al. (2012), and Dubreuil et al.
(2013).
3.3.1.1 Suppression of the Blurring Effect of an Image due to the Enlightening

of an Underwater Object
Based on an earlier study by Jaffe (1990), Leonard (2012) considers that the
image of the enlightened object I(out) which is received by the camera
11
The light absorption is related to the complex refraction index of the underwater medium and also to
the presence of pigmented particles.
12
The light scattering is due to the presence of transparent biological organisms and also due to
suspending particles.
0
17 m 12 m
49 m
100
The light propagation (m) 92 m
200
300
293 m
640 nm
400 371 m 600 nm
580 nm
530 nm
500 490 nm
470 nm
529 m 400 nm
600
Figure 56 Illustrating the propagation of light in clear water; i.e., the red (dark gray in
the print version) component of light is completely absorbed in water beyond 12 m.
Sogandares and Fry (1997).
can be expressed as I out I attenuation + I forward scattering +

I backscattering, where I(attenuation) represents the light (image),
I(forward scattering) is the quantity of impinging light (image) that is
scattered forward, and I(backscattering) is the same for backscattered light.
Leonard (2012) considers that the imaged scene is plane and parallel to the
camera with small aperture; i.e., the camera is at a constant distance from the
scene. Then, I(out) can be written as
I out I d + I0 fexpcd expa + bd gIFTexp Bdf

(44)
+ I backscatter:
In Equation (44), the first term, I d I0 expa + bd , represents

the attenuated intensity at distance d. The forward scattering term
I forward scattering fexp cd exp a + bd g I0 IFTexp Bdf
corresponds to the intensity of the source I0 multiplied by the difference
between two contributions, i.e., the attenuation due to the object (exp
[cd], where c is the attenuation coefficient of the object) and the attenuation
which depends of the volume between the object and the camera
(expa + bd ). Then, the forward scattering term is convoluted with
the transfer function of water (Hou et al., 2007) described by
IFTexp Bdf , where B is a phenomenological constant and f is a given
frequency. Finally, I(backscattering) is obtained by making the assumption

that the medium between the object and the camera is stratified. In this gen-
eral analysis of the propagation of light through a stratified medium, the
medium is assumed to be divided into plane-parallel layers which attenuate
and scatter light. The light is then convoluted with the transfer function of
water. In the method developed by Leonard (2012), the transfer function of
water between the camera and the target is estimated. This permits signif-
icant reduction of the blurring effect of underwater images. However, the
position of the object with respect to the camera needs to be known pre-
cisely, along with the attenuation and scattering coefficients of the embed-
ding medium.
3.3.1.2 Colorimetric Correction

There has been an active program of developing colorimetric correction.
Before getting into the details of these methods, it is worth recalling that
colors in an underwater medium are attenuated and perturbated. Hence,
underwater images have a greenish-blue tint13 as evidenced in many studies
(Dubreuil et al., 2013; Leonard, 2012; Leonard, Alfalou, Alam, et al., 2012).
To overcome this issue, several analyses have been reported concerning res-
toration of the colors of underwater images (Leonard, 2012). Several
methods are based on Beer-Lambert law. Other color equalization algo-
rithms, in the red, green, and blue (RGB), and the hue, saturation, and
intensity (HSI) color spaces, have been suggested. We briefly review these
methods. Helan, Burie, Bouwmans, and Bazeille (2006) suggested and val-
idated a correction method for colors in images; i.e., each color is corrected
by a factor of
I d I0 expd , (45)
where I(d) is the intensity at distance d, I0 I d 0, and is the absorption
coefficient at wavelength . This method achieves good performances in
cases for which the physical parameters such as the color and indices can
13
As you go deeper under water, first red light is filtered and the scenery around you has an orange tint.
Orange light is filtered after going deeper and the scenery around you has a yellowish tint. Yellow light
is filtered going deeper than that at this point the scenery around you looks greenish-blue. Green light
is filtered after going deeper in which the scenery around you looks a dark pure blue. And then finally
blue light is filtered along with indigo and violet which makes everything pitch black. These color tints
can only be viewed by scenery at the same depth as yourself so if you look down deeper into the water,
then objects start becoming bluer even though you yourself may be at the orange depth.
be determined precisely. However, this procedure cannot be easily

implemented optically. To overcome the constraints of knowing the phys-
ical parameters, Rizzi, Gatta, and Marini (2004) have proposed and validated
an alternate method for underwater color correction which is based on
unsupervised color equalization. These authors developed a digital image
automatic enhancement algorithm called automatic color equalization
(ACE) that is able to simulate HVS mechanisms, such as color constancy
and lightness constancy. However, the optical implementation of this
technique remains difficult too.
Another perspective can be found in Rizzi et al. (2004) and Chambah,
Semani, Renouf, Courtellemont, and Rizzi (2003), where the authors pres-
ented an ACE method that enables color correction for RGB channels.
Iqbal, Salam, Osman, and Talib (2007) proposed another method based
on the synoptic diagram shown in Figure 57, which allows correcting the
perturbations of an underwater image due to light absorption and scattering
(Figure 58).
This three-step method is as follows. First, Iqbal et al. (2007) suggest
modifying the histograms of R and G channels so that they become similar
to the histogram of the B channel. Next, they transform the RGB channels
Saturation &
Contrast Transform RGB
Input image intensity Output image
stretching RGB to HSI
stretching HSI
Figure 57 Synoptic diagram suggested in Iqbal et al. (2007) for increasing the quality of
underwater images.
Figure 58 The perturbation model used in Iqbal et al. (2007).

Figure 59 Underwater image by using the method suggested in Iqbal et al. (2007):
(a) without processing and (b) with processing.
in the HSI14 color space. Then, they apply a saturation operation and realize
an intensity stretching on this HSI color model. These authors considered an
underwater scene containing a groundfish (Figure 59a) characterized by an
image of poor quality. Applying their method, they obtained the image
shown in Figure 59b which is clearer and of better quality than the original
one. In addition, one advantage of applying two stretching models is that it
helps to equalize the color contrast in the images and it also addresses the
problem of enlightening.
Alternate protocols for correcting underwater images have been
suggested. For example, Torres-Mendez and Dudek (2005) considered
the problem of color restoration using statistical inference. This is applied
to color recovery for underwater images, using an energy minimization for-
mulation. For aquatic robot tasks, the quality of the images is crucial and
14
HSV, hue, saturation, and value color space.
needed in real time. Their method enhances the color of the images by using
a Markov random field (MRF) to represent the relationship between color-
depleted and color images. The parameters of the MRF model are learned
from the training data, and then the most probable color assignment for each
pixel in the given color-depleted image is inferred by using belief propaga-
tion. This allows the system to adapt the color restoration algorithm to the
current environmental conditions and also to the task requirements. Exper-
imental results on a variety of underwater scenes demonstrate the feasibility
of the method. But this kind of method remains quite difficult to realize in
practice (Granier, Refregier, & Figue, 1996; Gurov & Sheynihovich, 2000).
3.3.1.3 Phong Reflection Model for Sun Lighting Correction

Experimental study of underwater scenes for the purpose of detecting and
recognizing a mine in shallow water has been pursued recently (Leonard,
2012). As was discussed earlier, the suns rays do penetrate shallow water,
but each color penetrates to a different extent. This light contribution
due to the sun is detrimental to the detection and recognition algorithms
performances, notably those based on correlation for edge detection.
Figure 60 presents a schematic diagram showing the reflection of a light pen-
cil on the target object (assumed to be an infinite plane).
To overcome this issue, Leonard (Leonard, 2012; Leonard, Alfalou, &
Brosseau, 2012) proposed and validated an approach based on the Phong
Figure 60 Schematic diagram used in Leonard (2009) and Leonard, Alfalou, Alam, et al.
(2012) for taking into account the sun enlightening of the underwater scene, is the
angle between the normal N at the surface and the light source SL (SL, sun light), and
denotes the angle between the reflected beam R and the position of the sensor S (CCD
camera).
reflection model (Phong, 1975) used to correct the scene considered. This
approach is an empirical model of the local illumination of points on a sur-
face. In 3D computer graphics, it is sometimes ambiguously referred to as
Phong shading, in particular if the model is used in combination with the
interpolation method of the same name and in the context of pixel shaders
or other places where a lighting calculation can be referred to as shading.
Basically, this model considers that the light reflected by an object IT with
modification before the CCD camera, i.e., ICCD. Following Leonard (2012)
and Phong (1975), the reflected intensity by the target object can be written
as the sum of three contributions as shown in Figure 60
IT IA + ID + IS , (46)
where IA represents the light intensity at the scene. The second term, ID,
characterizes the diffuse reflection, that is, the reflection of light from a sur-
face such that an incident ray is reflected at many angles rather than at just
one angle as in the case of specular reflection. Lambertian reflection is often
used as a model for diffuse reflection. The reflection is calculated by taking
the dot product of the surface normal, N, and the light direction pointing
from the surface to the light source (Figure 60),
ID cos I0 , (47)
where is the angle between the two directions. The intensity will be the
highest if the surface is perpendicular to the direction of the light, and the
lowest if the surface runs parallel with the direction of the light. The third
term, IS, is the specular reflection. Its intensity depends on the cosine of the
angle (see Figure 60) at power which characterizes the brightness of the
target objects material,
IS fcos g I0 : (48)
Once IT is estimated using Equation (46), Leonard (2012) suggests to
make use of the Beer-Lambert law to quantify the received intensity by
the CCD camera ICCD. This intensity is written as
ICCD expczI0 , (49)

where c denotes the absorption coefficient of water, and z is the distance
between the CCD camera and the target object, e.g., an underwater metal
mine. In Leonard (2012), c 0:25m1 and z 20 m. Phong reflection
model leads to good results in terms of saturation correction due to sun light
Figure 61 Example illustrating the sun enlightening correction to an underwater image

based on the method developed in Leonard (2012) and Leonard, Alfalou, Alam, et al.
(2012): (a) raw image and (b) corrected image.
as shown in Figure 61. There is a clear saturation zone in the top of

Figure 61a which can be very well corrected by making use of Phongs
model as evidenced in Figure 61b. However, this model is effective if the
geographic localization of the object, source, and measurement apparatus
are known (Leonard, 2012; Leonard, Alfalou, Alam, et al., 2012).
3.3.2 Edge Enhancement

Within this perspective and extending an earlier study (Flores & Ferrari,
2010), Flores, Ferrari, and Garcia-Torales (2011) proposed and validated
the following strategy for enhancing the edges of a target image by polari-
zation filtering. In addition, their optical setup enables a segmentation of the
scene. They present an optical processor capable to perform spatial second
derivatives of an image working with partially coherent illumination.
Consider the discrete formulation of the second-order derivative of an
image in the x- and y-directions. Specifically
8
>
> @ 2 I I x + x, y 2I x, y + I x x,y
>
< @x2 x2
: (50)
>
> @ 2 I I x, y + y 2I x, y + I x,y y
>
: 2 ,
@y y2
where I(x,y) denotes the image intensity. The first and third terms of the
right-hand side of Equation (50) are the replicas of original image translated
at a distance x (or y) along the x- or y-direction. The second term

denotes the reverted replica of the original image I centered at the origin.
Thanks to Equation (50), it is possible to implement optically second-order
partial derivatives with the optical setup displayed in Figure 62. The light
source is an LED, with a first lens L0 used to parallelize the pencil, a polarizer
P0, and an LCD screen to visualize the input image which is placed in the
focal plane of the lens L1.
Then follows a 4f optical setup with two convergent lenses (L1 and L2).
In the Fourier plane (i.e., at distance f1 behind the lens L1), Flores et al.
(2011) introduced a polarization filter M which is composed of three ana-
lyzers A0 ,A1 , and A1 . A camera is positioned at distance 0 + z
(with 0 f2 ). Flores et al. (2011) showed that the output intensity at
distance 0 can be written as
@ 2 I1 x, y
I 2a2 + b2 2a2 I1 x, y a2 , (51)
@x2
where I1 x, y cos 2 x, y, (x, y) is a real-valued function that depend on

the voltage signal applied to the liquid-crystal cell of the display (image I),
and a and b are the amplitudes of the 1 and zeroth diffraction orders,
respectively.
Flores et al. (2011) validated experimentally their approach by consider-
ing several images and two specific applications (partial second-order deriv-
ative, and segmentation and partial second-order derivative). Figure 63
shows an example (Picasso painting). The upper right picture of
Figure 63 shows the optical processed image using the setup of Figure 62
and an offset equal to 65, the lower left picture of Figure 63 shows the output
image with an offset equal 125, and the lower right picture of Figure 63
shows the output image with an offset equal 145. These images along with
those presented in Flores et al. (2011) clearly indicate that this approach is
effective for edge enhancement of images.
In stark contrast with numerical approaches, the method proposed by
Flores et al. (2011) can be applied to large size images and in real time. In
addition, their optical setup makes use of a partially coherent enlightening,
i.e., which does not require a precise alignment, i.e., a limitation of the use of
the 4f optical setup.
In closing this section, we observe that other optical methods need to be
cited for edge enhancement and calibration of images; they include Espinosa
et al. (2013) and Tu et al. (2013).
A1
DF A0
A1
y L1 L2
L0 LCD M CCD
LED
z
P0
f0 f1 f1 f2 f2 Dz
S0 S
Figure 62 Optical setup used by Flores et al. (2011) for implementing their method of
enhancement of partial second-order derivatives of an image. The proposed system uti-
lizes a liquid crystal display, onto which images to be derived are displayed. The method
is based on the capacity of these devices to generate simultaneously a contrast-reverted
replica of the image displayed on it. A positive replica of original image is obtained
when the LCD is between crossed polarizers, while the negative one is obtained with
the LCD in between parallel polarizers. Since the LCD is a diffraction element, both polar-
izer arrangements may be simultaneously implemented, for example, in a 4f optical pro-
cessor using polarization filtering of the different diffraction orders. When three images
(two replicas of original image displaced in the opposite direction and the other one
with contrast reverted and centered at the origin) are imagined across a slightly
defocused plane, one obtains an image with enhanced second derivatives in a given
direction.
3.3.3 Image Blur Correction

A deblurring method using an edge map of the noisy image has been also
suggested to have great potential for improving image quality (Lee,
Park, & Hwang, 2012). It was argued that the novelty of this technique lies
in using an edge map of the noisy image, a good estimation of the blur kernel
based on a fast total variation regularization method, and specific treatments
which are schematically depicted in Figure 64. This permits keeping visible
the edges in an image while neglecting the noise in a specific area (called
smooth region). Lee et al. (2012) first defined the motion blur, i.e.,
Equation (52a), the sharp latent images, i.e., Equation (52c), and the estima-
tion of the blur kernel, i.e., Equation (52b) as
Figure 63 Illustration of results obtained by Flores et al. (2011) for suppressing and/or
enhancing specific zones of the target images. The upper left picture shows the original
Picasso's dove; the other pictures show optical processed images.
8
>
> B KL + a
>
>
< K 0 arg min jjKL Bjj2 + jjKjj2 b
k 2 2
2 2 (52)
>
>
>
>
: L 0 arg min L jjKL Bjj22 + log pL c
2
where B is the blurred image, K is the blur kernel, L is the latent image,
denotes the additive noise in the blurred image, and and are, respectively,
the weights of the likelihood (fidelity) term and the prior (regularization)
term related to the additive noise variance and the function log p(L) is
defined in Lee et al. (2012).
Figure 65 (Lee et al., 2012) clearly shows that one needs two versions of a
given image at the input of the system (top left) in order to restore it: (1) the
blurred image and (2) the noisy image. These two images are used to
Figure 64 Synoptic diagram describing the algorithm suggested by par Lee et al.
(2012).
Proposed
method
Input images
Output image
Figure 65 Illustrating the deblurring results using the algorithm shown in Figure 60. Lee
et al. (2012).
construct the edge map (green (light gray in the print version) blocks in
Figure 64). Next, this map is used to determine the kernel estimation (purple
(light gray in the print version) blocks in Figure 64). Finally, the two input
images, i.e., the edge map and the kernel estimation, are used to get the
restored image (blue (light gray in the print version) block). Based on
the technique described schematically in Figure 64 and the two (input)
images displayed in Figure 65, the restored (output) image is presented in
Figure 65.
These simulation results demonstrate the good performances of this
approach in terms of image quality at the output. In order to understand
quantitatively the above features, it is useful to remark that the peak SNR
of restored images has an average rise of 11%, compared to other algorithms.
Yet, no optical setup permitting an all-optical implementation has been
suggested.
3.3.4 Correction of Electronic Noise via an Optical Filtering Fourier

Method (Moir Correction)
The use of CCD camera in optical setups may lead to the appearance of a
detrimental noise for applications such as object recognition.15 Read or
readout noise is a form of electronic noise added to the final signal upon
15
A CCD is most simply described as an electronic photon detector. The typical CCD camera used for
object recognition applications consists of a 2D array of photon detectors in a layer of semiconducting
material, in this case silicon, which is placed at the focal plane of a camera in order to collect an image.
Each individual detector in the array is referred to as a pixel and can vary in size from a few to several
tens of microns. The smaller the size of the pixels, the higher the resolution of the image produced for
a given camera. Each individual pixel is capable of collecting photons and storing the produced elec-
trons, which can be read out from the CCD array to a computer to produce a digital image of the
varying intensities of light detected by the CCD. In this way, digital storage, reproduction, and
processing are possible, with the digital representation of the image field containing information that
is extremely accurate and easily reproduced. CCD imagers utilize the photoelectric effect. When pho-
tons hit the detector surface of the CCD sensor, electrons are liberated and stored in the detector ele-
ments, or pixels. Each pixel in the CCD is thus an electron well and accumulates electrons in
proportion to the number of photons received. The electrons in the atoms that make up the silicon
crystal exist in discreet energy bands. The lower energy band is referred to as the valence band, while
the higher energy band is called the conduction band. The electrons, typically occupying the valence
band, can be raised into the higher energy conduction band (excited) either by heat or by absorbing a
photon. Electrons that have been excited into the conductive band are able to move freely in the
silicon crystal lattice. When this occurs, a hole is left in the valence band, which then behaves as a
positively charged carrier. The CCD introduces an electric field in order to keep the electrons from
dropping back into the valance band and filling the hole, a process known as recombination. Recom-
bination will occur if there is no electric field present. Unfortunately, electrons can be excited into the
conduction band by the heat produced by the system (thermal noise) and these electrons cannot be
distinguished from those generated by photons (the signal).
readout of the device. This additive type of noise, described in the number of
electrons per pixel, is independent of exposure time. This noise is produced
during the conversion from an analog to a digital number. In addition, the
electronics of the system will introduce spurious electrons during the readout
process, adding to the fluctuations in output. Read noise can be isolated and
removed from a CCD image via the subtraction of a bias frame. A bias frame
is an image of zero exposure time where the CCD is read out without having
been exposed to light. In this manner, thermal noise produced by the heat
generated by the devices electronics and contributions of light to the expo-
sure are at a minimum, thereby isolating the effect of read noise. Since the
read noise varies from pixel to pixel and readout to readout, a number of such
bias images (nine or more) are recorded and then averaged together.
This is accomplished using the technique of median combining to pro-
duce a master bias frame. As described previously, in the photoelectric effect,
electrons generated by the heat produced by the system (thermal noise) can-
not be distinguished from electrons generated by photons (signal). There-
fore, there are always some number of electrons stored in the pixels that
are not the result of photons hitting the detector. The electrons generated
by thermal noise exist even when light does not hit the detector surface,
and are referred to as dark current. Dark current is a multiplicative form
of noise, the level of which is proportional to the length of the exposure.
Given a long enough exposure time, the detector could be fully saturated
with electrons due in large part to thermal noise. Overall this noise will
degrade the quality of the image. Other kind of noise is due to the moire
effect. This section is devoted to the strategy of optical correction of the
moire effect by making use of the 4f optical setup. As recalled in
Section 2.3.2, the moire effect manifests itself as an additive noise character-
ized by a periodic structure which is superimposed on the image. Conse-
quently, this noise can be written as
nmoir t cos wt : (53)
Fourier transforming Equation (43) leads to a sum of two Dirac delta
functions at frequencies w and w
1 1
Nmoir w + w : (54)
2 2
According to Equation (54), the two Dirac delta functions need to be
suppressed to deal with the moire issue. For that purpose, the 4f optical setup
composed of two convergent lenses L1 and L2 (Figure 66). The first one is
Input image
L1
Fourier plane :
Filter
L2
Output plane
Figure 66 Optical setup used to suppress the moir effect.
used to realize the FT. Next, the spectrum of the noisy image is multiplied
with the appropriate filter for suppressing the moire peaks. Then, a second
convergent lens is used to get the denoised image.
A single example will illustrate the benefit of using this optical setup.
Consider the image in Figure 67a which is pertubated by a moire effect.
Fourier transforming this image leads to the spectrum shown in
Figure 67b. This spectrum displays the characteristic moire peaks (circled
in red). Choosing an appropriate filter to eliminate these peaks and realizing
the inverse FT, one gets the denoised image shown in Figure 67c.
For a more comprehensive description of the moire effect, the methods
employed to suppress it, and the applications which exploit it we refer to
Khalil and Fazio (1973), Yokozeki and Ohnishi (1975), Andonian (1982),
Lim et al. (1989), Bara et al. (1991), Simova and Stoev (1992), Kato et al.
(1997), Aggoun (2006), Buytaert and Dirckx (2007), Kim et al. (2009),
Rasouli et al. (2011), and Leonard, Alfalou, Alam, et al. (2012).
3.3.5 Image Restoration and Edge Detection

The purpose of this section is to consider several examples of optical systems
allowing us to realize experimentally image restoration and/or edge detec-
tion of a target image.
3.3.5.1 Edge Detection in a Single Operation Optical Morphological Processor

Edge detection is becoming increasingly important in IP. Realizing optically
this operation is crucial since optics offers global interconnects and massive
parallelism providing many orders of magnitude improvement in execution
(a) (b)
(c)
Figure 67 (a) Example of a pertubated underwater image containing a moir noise

(from TOPVISION project)16; (b) the FT of the noised image containing several moir
peaks; and (c) the image denoised with the 4f optical setup.
time over conventional electronic algorithms (i.e., real time). Huang et al.
(1996) presented and validated a formal procedure for edge detection which
can be implemented optically by making use of the setup displayed in
Figure 51. This approach makes use of the single fundamental operation
O{.} defined earlier by Equation (36). The edge detection of gray image
ED(Fg) can be quantified by the following governing equation:

ED Fg O Fg , Rg1 ,Fb ,Rb1 , T 0,S1 O Fg , Rg2 , Fb ,Rb2 , T 0,S1
Fg Rg1 Fg Rg2 ;
(55)
where O{.} denotes the single fundamental operation, Fg and Fb are, respec-
tively, the target gray image and its binary version, (Rg1, Rg2) and (Rb1, Rb2)
are, respectively, the two structure elements considered and their binary
16
The raw mine images contained in this publication were provided by the GESMA (Groupe dEtudes
Sous Marines de lAtlantique) within the TOPVISION project coordinated by Thales Underwater
Systems SAS. This project is related to Techno-Vision Program launched by the French Ministry
of Defense. More details can be found on the Web site: http://topvision.gesma.fr.
versions, and T 0 denotes the threshold operation defined as a high-pass

filter, OfX, Y , XX, YY , T , S1 g XYY T + XXY T . Finally, the
edge detection can be defined as the subtraction of two dilation operations
between the target image and the two considered structure elements

Fg Rg1 and Fg Rg2 . The choice of these two structure elements
(Rg1, Rg2) is of crucial importance. In their pioneering work, Huang et al.
(1996) presented several possibilities for choosing Rg1 and Rg2, e.g.,
0 1 0 1
m m m n n n
Rg1 @ m n m A and Rg2 @ n m n A , (56)
m m m n n m
for extracting the whole profile of gray images, where m and n are positive
integers, and m 6 n. For extracting the horizontal edge of gray images, they
proposed to use
0 1 0 1
m n
Rg1 @ n A and Rg2 @ m A : (57)
m n
For the detection of the vertical edge of gray images, they suggested
considering
Rg1 m n m and Rg2 m n m : (58)
For the detection of the edge with an angle of 45, they proposed
0 1 0 1
0 0 m 0 0 n
Rg1 @ 0 n 0 A and Rg2 @ 0 m 0 A , (59)
m 0 0 n 0 0
while for the detection of the edge with an angle of 135, they rec-
ommended using
0 1 0 1
m 0 0 n 0 0
Rg1 @ 0 n 0 A and Rg2 @ 0 m 0 A : (60)
0 0 m 0 0 m
Using Equations (55) and (56) and the setup displayed in Figure 47,
Huang et al. (1996) obtained the edge detection shown in Figure 68.
As an illustration of this experimental approach, Figure 68 shows the
edge detection of a given gray image. Figure 68a gives the target gray image
of a dog, and the two Rg1 and Rg2 matrices (Equation 56) used are shown in
Figure 68 Example, provided by Huang et al. (1996), of the extraction of the whole pro-
file of a given gray image.
Figure 68b. The result shown in Figure 68c indicates that this approach
allows us to optically detect the edge of an image with a good quality.
3.3.5.2 Edge Detection and Image Reconstruction from Its Spectral Phase
Several authors (Alfalou et al., 2010; Alfalou, Mansour, et al., 2011) empha-
sized the importance of the spectral phase of an image to address the issue of
optical compression of images. Oppenheim and Lim (1981) demonstrated
that the spectral phase of a target image contains the information required
for edge detection and its reconstruction. Quatieri and Oppenheim
(1981) proposed an iterative algorithm enabling reconstruction of a target
image from the knowledge of the spectrums phase.17 In addition, according
the root-mean-square (RMS) duration criterion (Papoulis, 1962), only a
part of the image spectrum is required.
Edge detection
In this section, a simple and efficient method of edge detection is presented.
This method relies on earlier work by Oppenheim and Lim (1981) and the 4f
optical setup. The synoptic diagram is illustrated in Figure 69. This approach
is based on Fourier transforming the target image with a single convergent
lens. Solely the phase information in the Fourier plane is considered. Finally,
the FT1 allows reconstructing the phase or edge image (Figure 69).
This protocol is used to deal with pure phase filter (POF) correlation
(Horner & Gianino, 1984).
Image reconstruction
We now turn to describe the principle of iterative reconstruction using only
the real spectral phase of the image such as reported by Quatieri and
17
The same is also true from the knowledge of its amplitude.
Figure 69 Schematic diagram of a simple optical setup allowing edge detection by

making use of the sole knowledge of the spectrum phase of an image.
(a) (b)
Target image Phase only

(N,N) (f)
pixels
Reconstructed
Image spectrum (2N,2N) pixels
exp(if)
Image reconstructed (N,N)
pixels
1
FT
FT Reconstructed
image (2N,2N) pixels:
FT
A exp(if)
Image spectrum
A exp(if)
Phase only
(f) A
Figure 70 Iterative algorithm principle: (a) only phase section and (b) iterative
reconstruction image.
Oppenheim (1981). This two-step iterative scheme is illustrated in

Figure 70.
This approach consists in (1) selecting the target image contours by using
only its phase component (Figure 66a) and (2) reconstructing the input plane
with an iterative method (Figure 70b). The first step (Figure 70a) is identical
to that used for edge detection. The method consists in multiplying the
spectrum of the target image with a filter which has the effect of suppressing
its amplitude. Before doing that, the size of the target image (N N pixels)
is increased by zero-padding the FT; i.e., the size becomes 2N 2N pixels.
The second step uses an iterative algorithm as is illustrated in Figure 70b.
First, the image spectrum is reconstructed by taking its amplitude A set to 1.
Then, a set of FT and FT1 is realized until the output image converges
to the original image. To do this, while conserving the phase obtained at
the beginning, we change for each cycle the amplitude of the image spec-
trum through the use of the image reconstructed on N N pixels. To
show the good performances of this reconstruction algorithm, we calculate
the MSE defined as
1 XN X
N
MSE 2
jI0 i, j In i, jj2 (61)
N
for each iteration. This criterion evaluates the difference between the ampli-
tude of the target image I0 and the reconstructed image with the considered
algorithm In, where n denotes the number of iterations and N is the number
of pixels of the target image I0.
In Figure 71, simulations were conducted with the Lena image I0. By
comparing the phase image obtained with a single iteration (top left image
in the panel of Figure 71) with the target image I0, we find that the former
has a very low quality as evidenced by the high value of the MSE
211. If
we apply our algorithm with 10 iterations, we observe a significant increase
of the image quality. This can be seen by visual inspection and a lower value
of the MSE
75. After 500 iterations, the reconstructed image shown at the
0.25
0.2
0.15
MSE
One iteration Fifty iterations

0.1
0.05
0
0 50 100 150 200 250 300 350 400 450 500
Iteration number
Ten iterations Five hundred iterations
Figure 71 Influence of the iteration number on the quality of the reconstructed image.
right bottom of Figure 67 is characterized by an MSE 1:723. This result is

consistent with the good quality of the image reconstructed by making use of
the phase information solely. As also shown in Figure 71, we obtain a rapidly
convergent algorithm. It is also worth noting that even with 100 iterations,
the algorithm converges rapidly to a good quality of the image.
Experimental and theoretical studies of image restoration and edge
detection have been pursued widely. The interested reader may wish to con-
sult Mazzaferri and Ledesma (2007), Fan, Song, and Jutamulia (2007),
Mazzaferri, Campos, Escalera, Sheppard, and Ledesma (2010), and Li,
Xu, Feng, Tao, and Zhao (2011) for reviews and many references.
3.4 Using Polarization for Increasing Image Quality

and Edge Detection
The basic principles of polarization optics (PO) with applications to scatter-
ing and imaging are given in standard texts (Born & Wolf, 1970; Brosseau,
1998). Even so, in recent years, the use of polarization for imaging optics
developed in important new ways. In this section, we provide a powerful
yet simple PO method to increase the image quality and edge detection
of objects embedded in various media (air, seawater). A good introduction
to the difficulties encountered in the issue of detection and identification of
objects by underwater video imaging can be found in Leonard, Alfalou,
Alam, et al. (2012). In this study, the authors dealt with noised images
and tested preprocessing numerical treatments before applying correlation
methods for detecting and identifying mines. However, the identification
algorithms, based on edge detection, gave limited results, i.e., detection
rate less than 60%. The bad quality of the initial images has serious limi-
tations for detection and identification of objects in strongly scattering
media. This paucity of knowledge implies that many fundamental ques-
tions that are pertinent to OIP in different fields remain unanswered.
As we will see, this section allows us to understand how PO can resolve
several of these issues.
We begin with an overview of PO methods designed to increase under-
water imaging by eliminating backscattering. The central idea is to associate
these methods with recognition techniques based on OC. We then make
detailed comparisons of a single image or multiple images of a given scene
obtained with different states of polarization. In the following, we also
describe the method developed by Schechner and Karpel (2005) which is
adapted to deal with imaging in scattering media such as seawater.
3.4.1 A Brief State of the Art

It is useful for our purposes to first describe the various steps taken in recent
years toward understanding underwater imaging. For centuries, people have
regarded oceans as an inexhaustible supply of food, a useful transport route,
and a convenient dumping groundsimply too vast to be affected by any-
thing we do. But human activity, particularly over the last few decades, has
finally pushed oceans to their limit; e.g., the worlds oceans may be turning
acidic faster today because of human carbon emissions. Another limit that
can transform in potential danger for human activity concerns naval mines
(Barat & Phlypo, 2010). Minefields designed for psychological effect are
usually placed on trade routes and are used to stop shipping reaching an
enemy nation. They are often spread thinly, to create an impression of mine-
fields existing across large areas. A single mine inserted strategically on a ship-
ping route can stop maritime movements for days, while the entire area is
swept. Sea mines continue to be important in naval warfare because they
can be used defensively or offensively and because of mines demonstrated
success and capabilities (Duncan, 1962; Levie, 1992).
Influence mines can be fitted with magnetic, acoustic, seismic, and pres-
sure sensors that can detect a ships approach, determine what kind of vessel
it is and the most effective time to detonate. Along with the variety of mines,
there are many ways to search and neutralize them from the air, the surface of
the sea and underwater. Sonar is the primary way potential mines are iden-
tified in the water when the Navy is hunting for explosives, but explosive
ordnance disposal divers, marine mammals, video cameras on mine neutral-
ization vehicles, and laser systems can also be used (Hartmann & Truver,
1991; Levie, 1992). More than 1 billion dollars was earmarked for that pur-
pose in the 2012 U.S. Navys budget (Hartmann & Truver, 1991). In recent
years, there has been much interest in designing autonomous underwater
vehicles (AUVs) equipped with vision systems for remote control.
A typical military mission for an AUV is to map an area to determine if there
are any mines, or to monitor a protected area (such as a harbor) for new uni-
dentified objects, e.g., AN/BLQ-11 (Boeing Company). However, light
attenuation effects and visibility are two important issues for image quality
acquired by these AUVs.
Water is responsible for the attenuation of light due to absorption. Three
factors primarily affect underwater visibility: light penetration, biological
species, and particulate matter (e.g., suspended sand and silt). According
to the size of these particles, they scatter forward or backward light, resulting
in reduced visibility. Since attenuation can be described using the
Beer-Lambert law, it implies that the underwater images should be taken at

short range. Some further observations seem significant: first, scattering by
particles and by the water itself affects image quality; i.e., maximum imaging
distances range from centimeters in dirty water to more than 100 m in very
clear water. Second, nonuniform lightning conditions result in different
intensity levels between images of the same area. Figure 1 illustrates the
well-known fact that the underwater medium can significantly degrade
the quality of the image. That is, small-angle forward scattering of light
and nonuniform lightning conditions can cause images to blur and result
in intensity differences between the images of the same area. Fortunately,
we can do better and use light polarization to alleviate the aforementioned lim-
itations. It has been shown (Bartolini, De Dominicis, Ferri de Collibus, et al.,
2007) that backscattering preserves the light polarization state in a linear
polarization configuration, whatever the distance (Figure 72). In a circular
polarization configuration, backscattering polarization varies with distance.
Thus, helicity varies (Bicout & Brosseau, 1992; Bicout, Brosseau,
Martinez, & Schmitt, 1994; Brosseau & Bicout, 1994; Gilbert & Pernicka,
1967; Leonard, Alfalou, & Brosseau, 2013, Lewis, Jordan, & Roberts, 1999;
McGlamery, 1975; Mie, 1908; Sabbah, Lerner, Erlick, & Shashar, 2005).
Within this context, different underwater imaging systems based on
range gating (time discrimination), laser line scanning (spatial discrimina-
tion), structured lightning, modulation/demodulation techniques, multiple
perspective image construction, and polarization discrimination have also
attracted significant interest (see, for example, Bartolini, De Dominicis,
Partially
depolarizing
target
Backscattering
Light source
Cross-polarizer
Camera
Linear polarization
Figure 72 Illustrating the backscattering and depolarization of light by an underwater
target. Leonard, Alfalou, and Brosseau (2013).
Fornetti, et al., 2007; Boffety, Galland, & Allais, 2012; Cariou, Le Jeune,
Lotrian, & Guern, 1990; Clemenceau, Dogariu, & Stryewski, 2000;
Gorman, 2011; Jaffe, 1990; Kattawar & Rakovic, 1999; Kim & Moscoso,
2002; Kohlgraf-Owens & Dogariu, 2009; Kouzoubov, Brennan, &
Thomas, 1998; Miller & Caprari, 1999; Miller & Dereniak, 2012; Mullen,
Cochenour, Rabinovich, Mahon, & Muth, 2009; Piederrie`re et al., 2005;
Treibitz & Schechner, 2009; Tyo, Goldstein, Chenault, & Shaw, 2006;
Voss & Fry, 1984; Walker, Chang, & Hopcraft, 2000; You et al., 2011 for
reviews and many references).
Combining polarization filtering and correlation techniques is justified
because it is simple to implement, and it enables simultaneous detection
and identification of target objects. Schechner and Karpel (2005) suggested
an approach to increase the visibility in water by exploiting linearly polarized
images. However, this requires an a priori knowledge on the properties of the
embedding medium, which can be difficult to obtain, and of the nature of
the target object. Hence, the degree of depolarization of the wave by the
target does not allow us to select the optimal polarization state for detection.
In order to address these shortcomings, Treibitz and Schechner (2009)
introduced a technique which is based on using linear parallel and crossed
linearly polarized images for evaluating backscattering. In the following,
we will restrict ourselves to the method suggested in Schechner and
Karpel (2005).
3.4.2 Background and Definitions

3.4.2.1 Introduction
Light propagation, focusing, and scattering are governed by the electromag-
netic properties of the media being traveled (Born & Wolf, 1970). With
reference to Figure 72, the mechanisms of light absorption and scattering
by particles are the essential mathematical ingredients for modeling the
light propagation. Scattering and absorption reduce intensity of radiation.
Their rate depends linearly on the local intensity of radiation, the local
concentration of particles, and the scattering and absorption efficiencies of
particles. The scattering has the effect of changing the direction of the elec-
tromagnetic wave. The scattering efficiency and behavior depend on the
size of scatterers relative to the wavelength of radiation. For very small
particles (Rayleigh scattering), the scattering efficiency is relatively low
and is strongly dependent on wavelength. There is nearly equal scattering
in forward and backward directions. For particles similar in size to wave-
lengths of radiation (Mie scattering), the efficiency is relatively high
POL0 R P
Block
Laser A q
Target
L1 L2 L3
Block
B
CCD
L4
Figure 73 Optical setup for polarization measurement. Details have been previously
discussed in Leonard, Alfalou, and Brosseau (2013) and Dubreuil et al. (2013).
and independent of wavelength. The scattering in forward direction

becomes more pronounced with increasing size.18
3.4.2.2 Optical Setup for Polarization Measurement

A simple optical setup used for polarization measurement is displayed in
Figure 73. The setup is made of a laser source, e.g., HeNe 633 nm and a
polarizer POL0 (Dubreuil et al., 2013; Leonard, Alfalou, & Brosseau,
2013). This polarizer allows us to control the light intensity. In
Figure 73, L1, L2, L3, and L4 are convergent lenses used to focus the col-
limated laser beam. The scatterer + =2 attenuates the speckle phenomena
which is detrimental to our measurements (Figure 74). A diaphragm is also
used in order that the Gaussian approximation is satisfied. The two blocks
A and B contain different arrangements of polarizers and quarter-wave plates
depending on the considered situations, e.g., object in the air, in water, or
behind a biological tissue.
3.4.2.3 Several Definitions

There has been much theory and experimental work outlining the many
important concepts of PO. To frame the discussion, we recall several basic
definitions (Brosseau, 1998, 1995; Brosseau & Barakat, 1991; Brosseau,
Givens, & Kostinski, 1993). First, a linear polarizer is an optical element that
attenuates the orthogonal components of a beam unequally (anisotropically).
18
Note that this simplification is not always valid. For example, the light propagating in biological tissues
encounters structures of variable size and it is not obvious to classify scattering as either Mie or
Rayleigh.
Figure 74 Speckle image obtained after scattering of light by a flat surface: (a) image
obtained without R; (b) image with R showing that the speckle is significantly
attenuated.
An ideal linear polarizer is one in which the transmission of the polarized

beam takes place along only one axis (e.g., x-axis) and there is no transmis-
sion along the other orthogonal axis (i.e., y-axis). Second, a wave plate (also
termed retarder or compensator) is a polarizing element that introduces a
phase shift between the orthogonal components of the beam. There is
a special case that is of particular importance for us, i.e., the quarter-wave
plate which transforms linearly polarized light to right or left circularly polar-

ized light 2 . Third, the Jones (or Clark Jones) and MuellerStokes for-
malisms were developed in the 1940s to describe the polarization of the
optical field, respectively, in terms of amplitudes and intensities, and its
interaction with elements that change its state of polarization by matrix alge-
bras. Table 1 lists some pertinent matrices for the purpose of illustration.
In like fashion, the Jones matrix of a rotator with angle can be written as

cos sin
R : (62)
sin cos
The action of rotating a Mueller matrix through an angle with respect
to the x-direction yields
M RMR, (63)
where M is the Mueller matrix of the optical component and R() is the
rotation matrix.
3.4.3 Stokes Polarimeter

In what follows, we quantify the polarization features by making use of the
Stokes representation. Many decades of collecting and analyzing experimen-
tal data show that this phenomenology can successively describe a large body
Table 1 Several Polarizing Elements with Their Corresponding Jones Matrix

Polarizing Element Jones Matrix

Linear horizontal polarizer 1 0
0 0

Linear vertical polarizer 0 0
0 1
0 1
Wave plate, is the phase of the propagating beama
B exp i 0 C
B 2 C
@ A
0 exp i
2

Quarter-wave plate =2 1 0
exp i
4 0 i
a
A wave plate is characterized by two orthogonal axes called the fast axis and the slow axis, respectively.
These are usually taken to be along the x-axis and y-axis. The phase shift along the fast axis is + =2 and
along the slow axis is =2, so the total phase shift between the two axes is .
Brosseau (1998).
of experimental results (Brosseau, 1998). Stokes formalism involves measur-

ing the four Stokes parameters, i.e., the intensity of each electric field com-
ponent or specific combinations of these intensities. For our purpose, we
consider the four basic images: S0, S1, S2, and S3. Within this Stokes
parameter representation, the degree of polarization (DOP) of light is
defined through
p
IPol S12 + S22 + S32
DOP : (64)
I0 S0
Several things are worth noting about Equation (64). It is evident that the
DOP ranges between 0, i.e., unpolarized light, and 1, i.e., totally polarized
p
S2 + S2
light. The linear degree of polarization is DOPL S2 0 3 , and the circular
degree of polarization is DOPC SS30 . We wish to show several examples that
illustrate the benefit of using polarimetric techniques to improve the image
quality. Another purpose is to show that PO coupled to OC techniques
provides a powerful yet simple method to detect and recognize objects.
We begin by considering target images in air. Details have been previously
discussed in Dubreuil, Alfalou, and Brosseau (2012), Leonard, Alfalou, and
Brosseau (2013), Dubreuil et al. (2013), and Leonard, Alfalou, Zallat,
Lallement, and Brosseau (2013).
3.4.4 Influence of Polarization on the Quality of Images in Air

3.4.4.1 Optical Setup
The optical setup shown in Figure 73 has been adapted to measurement in
air and is illustrated in Figure 75. An incident linearly polarized beam by
POL1 illuminates a 2D scene. The Stokes polarimeter (Figure 75a) allows
us to characterize the state of polarization of light after reflection/scattering
by the target using two quarter-wave plates (QWP1, QWP2) and a linear
polarizer POL2.
(a) POL0 R P
Block A
L1 L2 L3
POL1
0
Laser
q
Target
90
QWP1
QWP2
POL2
Block B
CCD
L4
(b)
Laser
Rotating diffuser
Pinhole
CCD camera
Optics components
Lens (before)
Optics components
(after)
Target to be analyzed
Figure 75 (a) Polarization imaging optical setup for measurement in air; (b) a picture of
the optical setup. Leonard, Alfalou, and Brosseau (2013) and Dubreuil et al. (2013).
01
Sexp0
B Sexp1 C
If the output Stokes vector Sexp B C
@ Sexp2 A is measured, then we can
Sexp3
determine the Stokes vector after reflection/scattering by the target in the
0 1
Starg0
B Starg1 C
scene to be analyzed Starg B C
@ Starg2 A. The two vectors are related to each
Starg3
other by a matrix multiplication
Sexp MStarg
(65)
M M Pol M 1=4 M 2=4 :
The real-valued Mueller matrix M is a 4 4 matrix which contains the

information on the optical elements of the Stokes polarimeter. Equation (65)
can be written as
0 1 0 10 1
Sexp0 m00 m01 m02 m03 Starg0
B Sexp1 C B m10 m11 m12 m13 CB Starg1 C
B C B CB C
@ Sexp2 A @ m20 m21 m22 m23 A@ Starg2 A: (66)
Sexp3 m30 m31 m32 m33 Starg3
For example,
Sexp0 m00 Starg0 + m01 Starg1 + m02 Starg 2 + m03 Starg 3 (67)
can be measured by the CCD camera. Each Stokes image corresponds to a
specific orientation of the quarter-wave plates (Figure 76) according to some
specific values of 1 and 2.
QWP1 QWP2
POL1 Pol2
0 90
q2 q1
Target
Figure 76 Schematic diagram showing the different angular configurations required

for measuring the Stokes image.
Then, the DOP is given by

q
2
S1exp1 2
+ S2exp 2
+ S3exp
IPolexp
DOP , (68)
I0exp S0
which follows from Equation (64). We will need this expression below.
0 1
Sexp0
B Sexp1 C
Once the four Stokes images B C
@ Sexp2 A are measured, the output Stokes
Sexp3
images Starg can be calculated thanks to Equation (66) and inversion of M
0 1 0 1
Starg0 Sexp0
B Starg1 C B C
B C M 1 B Sexp1 C: (69)
@ Starg2 A @ Sexp2 A
Starg3 Sexp3
3.4.4.2 Calculation of the DOP for a Scene in Air

It turns out that the DOP is a natural candidate to characterize a target in a
given scene since it depends on the surface state and nature of the materials
target. The calculation of the DOP can be done by calibration with refer-
ence samples to ensure stability and repeatability of the measurements. From
Equation (69), matrix inversion allows us to get the four Stokes parameters
from four sets of angles 1 and 2. Here, we benchmark our approach
through the analysis of a scene (Figure 77a) composed of a metal coin, a
Nylon piece, and a plastic piece. As expected, the metal coin reflects light
Figure 77 (a) Intensity image; (b) DOP image.

with a high DOP in sharp contrast with the plastic piece which is charac-
terized by a small DOP. Two kinds of materials with similar reflectance
but otherwise different depolarization properties will appear identical in
standard (intensity) imaging, but PO offers a mechanism of image contrast
as illustrated in Figure 77b.
In addition, PO makes it possible to get away from the nonuniformity
of the lighting of the scene, i.e., compare Figure 78a and b. It also renders
easier the application of correlation techniques. As a quantitative check,
we show the correlation planes, respectively, with or without taking into
account polarization (respectively, Figure 78c and d) along with the values
of the PCE which is the basic descriptor typically used to characterize the
correlation.
On the experimental side, we conclude that correlation associated with
polarization imaging is well adapted when the lighting condition is weak and
nonuniform, e.g., Figure 78c. PO can be implemented in a complete form
using the Mueller matrix formalism to characterize each pixel of an image.
The correlation peak is more or less intense, depending on the degree of
(a) (b)
(c) (d)
PCE = 0.0772 PCE = 0.0039

Figure 78 Stokes imaging for nonuniform lighting. (a) Metal coin imaged using polar-
ization, (b) as in (a) for the intensity image, (c) and (d) the corresponding correlation
planes. The PCE is defined as the energy of the peak correlation normalized to the total
energy of the correlation plane. Dubreuil et al. (2012).
similarity between the target image and a reference image. If correlation is

used as detection criterion in order to recognize a pattern in a complex
scene, PO is much more efficient than standard optics based on the measure-
ment of the sole intensity.
3.4.4.3 Linear Polarimeter in Air

In Figure 75, two linear polarizers were placed, respectively, in front of and
behind the object. In the schematic of the experiment, two configurations
(crossed and parallel) are considered as shown in Figure 79.
The crossed configuration, shown in Figure 79a, leads to the measure-
ment of I? and allows us to identify the change in the polarization state or
even light depolarization, while the parallel configuration leads to the mea-
surement of Ik and characterizes light with invariant linear polarization.
Then, as done in many references (Bartolini, De Dominicis, Fornetti,
et al., 2007; Boffety et al., 2012; Born & Wolf, 1970; Brosseau, 1998;
Cariou et al., 1990; Clemenceau et al., 2000; Dubreuil et al., 2012, 2013;
Gorman, 2011; Jaffe, 1990; Kattawar & Rakovic, 1999; Kim & Moscoso,
2002; Kohlgraf-Owens & Dogariu, 2009; Kouzoubov et al., 1998;
Leonard, Alfalou, & Brosseau, 2013; Miller & Caprari, 1999; Miller &
Dereniak, 2012; Mullen et al., 2009; Piederrie`re et al., 2005; Treibitz &
(a)
POL1
POL2
Target
(b) POL2
POL1
Target
Figure 79 Linear polarimeter with either (a) crossed polarizers or (b) parallel polarizers.
Schechner, 2009; Tyo et al., 2006; Voss & Fry, 1984; Walker et al., 2000;
You et al., 2011), the DOPL can be written as
Ik I?
DOPL : (70)
Ik + I?
Figure 80a and b shows, respectively, I? and Ik for a metal coin in air.
Then, by making use of Equation (70), we get the DOPL image shown
in Figure 80c. This permits enhancement of the external edges of the coin.
These DOPL measurements are consistent with the low depolarizing feature
of metal.
3.4.4.4 Circular Polarimeter in Air

Using the optical setup shown in Figure 81, I +45 and I45 are measured. In
Figure 81a, I45 is obtained when the quarter-wave plate QWP(2) is rotated
by 90 compared to the quarter-wave plate QWP(1). The latter QWP(1) is
rotated by 45 compared to the polarizer POL(1). This measurement per-
mits to evaluate the quantity of light which is purely linearly polarized.
Figure 80 Experimental results obtained with a metal coin: (a) intensity I? (crossed
polarizers), (b) intensity Ik (parallel polarizers), and (c) corresponding DOPL.
(a) QWP1
POL1 POL2
Target QWp2
(b) QWP1
QWp2
POL2
POL1
Target
Figure 81 Circular polarimeter used for obtaining: (a) I45 and (b) I + 45
Correspondingly, the measurement described in Figure 81b allows us to

identify the quantity of light which remains circularly polarized. As shown
in many sources (Bartolini, De Dominicis, Fornetti, et al., 2007; Boffety
et al., 2012; Cariou et al., 1990; Clemenceau et al., 2000; Dubreuil et al.,
2012, 2013; Gorman, 2011; Jaffe, 1990; Kattawar & Rakovic, 1999;
Kim & Moscoso, 2002; Kohlgraf-Owens & Dogariu, 2009; Kouzoubov
et al., 1998; Leonard, Alfalou, & Brosseau, 2013; Miller & Caprari, 1999;
Miller & Dereniak, 2012; Mullen et al., 2009; Piederrie`re et al., 2005;
Treibitz & Schechner, 2009; Tyo et al., 2006; Voss & Fry, 1984; Walker
et al., 2000; You et al., 2011), the DOPc can be expressed as
I +45 I45
DOPc : (71)
I +45 + I45
The same scene as in Figure 80 is used. The measurements of I45 and
I +45 are shown, respectively, in Figure 82a and b. The corresponding DOPc
is shown in Figure 82c.
3.4.5 Detection of a Metal Object

3.4.5.1 Detection of a Metal Coin Placed Behind Diffusive Pieces of Scotch
The transparent adhesive Scotch (3M, Minneapolis, MN) tape is used as a
scattering medium, i.e., we placed the metal coin behind one or several
Scotch pieces. The DOP can be easily measured by making use of the optical
setup shown in Figure 75. The key feature is evident: using the Stokes polar-
imeter shown in Figure 83a gives better results compared to the intensity
image (adding several Scotch layers) obtained by using the Stokes polarim-
eter displayed in Figure 83b.
Here, our objective is to enhance the coins edge in order to compare it
with a circle. Figure 84 shows the PCE values as a function of the number of
Scotch pieces. Consequently, our measurements confirm the benefit of
using polarization for discriminating materials.
Figure 82 Experimental results with a metal coin: (a) intensity I45 (parallel polarizers
and quarter-wave plates placed with 90), (b) intensity I + 45 (parallel polarizers and par-
allel quarter-wave plates), and (c) corresponding DOPc.
Figure 83 Images of a metal coin placed behind a scattering medium (Scotch piece):
(a) intensity image and (b) DOP.
0.0018
DOP
0.0016
I0
0.0014
0.0012
0.001
PCE
0.0008
0.0006
0.0004
0.0002
0
0 2 4 6 8 10 12
Scotch pieces
Figure 84 A comparison of correlation tests of the DOP and intensity image as a func-
tion of the number of Scotch pieces added to the target.
3.4.5.2 Discriminating Between Different Types of Disk-Shaped Materials

Polarization imaging is able to discriminate between different types of mate-
rials with identical shape (Figure 85a). This scene contains four types of
materials: (c) metal, (d) cardboard, (e) absorbing paper, and (f ) white paper.
The corresponding DOP image (Figure 85b) shows that the metal is weakly
depolarizing in contrast to cardboard and paper. This enables exploitation of
this phenomenon through a broader set of materials.
3.4.6 Exploiting Polarization for Underwater Imaging

Here, the focus is on active polarization imaging and OC-based methods to
experimentally characterize underwater target detection. Experiments are
Figure 85 (a) Scene with different types of disk-shaped materials ((c) is a metal coin,
(d) is piece of cardboard, (e) is a piece of absorbing paper, and (f ) is a piece of white
paper) and (b) corresponding DOP images.
Figure 86 Image of an object on a paper sheet: (a) in clear water and (b) in water
with milk.
conducted in a glass tank filled either with a scattering medium (milk dis-
persed in water) or with seawater.
3.4.6.1 Experiments in a Glass Tank

We now discuss the influence of the veil, due to backscattering of light
impinging on an object in a scene, on an image as shown in Figure 86.
As mentioned earlier, the exploitation of the polarization features of
the backscattered light allows us to minimize the effect of backscattering
and hence to recover a neat image. Dubreuil et al. (2013) reported on
an active polarization imaging and OC-based method to experimentally
characterize underwater target detection. Active polarization imaging was
performed to suppress the veil in the scene, whereas correlation was used
for detection criterion and target localization. In case of a depolarizing
object, performing a single measurement using adapted polarization
states (cross-linear or co-circular) leads to better performance than intensity
imaging because it filters the major part of the backscatter and only a
small part of the target signal.
For polarization-maintaining objects, performing only one measurement
is insufficient to improve target detection because an important part of the
signal is lost. However, whatever the target is, performing two measure-
ments with two orthogonal polarization states improves recovery of the tar-
get signal by suppressing the backscatter that veils the object. The contours
of the target are thus well defined which enhances correlation performance.
Experiments performed in different scattering media and different targets
showed that there is always an improvement in target detection when target
estimation is considered. In this study, no difference in decision performance
was found between linear and circular polarized lights.
The panels of Figure 87 show images of a plastic plug for skimmed milk
dispersed in water (20 mL milk to 10 L water) giving s 0.08 cm1.
Typically, the distance that light travels before reaching the target is
15 cm, corresponding to an optical thickness of 1.2. Images taken with
I and CRL modes, linear estimation, are compared to an image taken in pure
water. Differences between these images merit discussion.
Figure 87 Images of the plastic plug. Experimentally, 20 mL of skimmed milk has been
added to 10 L of tap water. (a) Intensity image. The red (gray in the print version) square
indicates the area over which the degree of polarization is calculated, (b) cross-linear
image, (c) signal estimation image (linear), and (d) pure water image. Dubreuil et al.
(2013).
It is interesting to observe that the backscatter clearly appears in the

intensity image (Figure 87a) in its left part. Another interesting feature is that
the contours of the plastic plug are not well defined in the intensity image. In
the CRL image shown in Figure 87b, most of the backscatter is suppressed.
We observe that the backscatter has the same polarization as the incident
polarized light so imaging in the opposite polarization mode is a way to limit
the backscatter. However, the backscatter is not fully polarized and a sub-
stantial part of it is still present in the CRL image. In the estimation signal,
the backscatter is totally removed and the image is very close to that obtained
in pure water. Based on these observations, it appears that performing two
measurements with two orthogonally polarized states leads to better perfor-
mance than a single measurement.
Figure 88 shows a comparison between correlation planes and PCE for
each image of Figure 87. The reference image is taken as the image in clear
water, i.e., Figure 88d, leading to the autocorrelation plane.
We point out that Figure 88a shows a very noisy correlation plane
because the backscatter creates additional contours that correlate with the
reference image ones. Consequently, PCE is small. The correlation plane
(a) (b)
PCE = 6.4
PCE = 2.2
1
1 0.8
0.6
0.5 0.4
0.2
0 0
600 400 600

400 400
400 200
200 200 200
0 0
(c) (d)
PCE = 15.6 PCE = 30.6
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
400 600 400 600

400 400
200 200 200 200
0 0
Figure 88 Correlation planes corresponding to the images shown in Figure 83. Dubreuil
et al. (2013).
Figure 89 (a) PCE versus optical thickness 0 for a plastic plug and the different imaging
modes shown in the inset. (b) As in (a) for painted metal. (c) As in (a) for rusted metal.
The scattering medium is composed of skimmed milk diluted in tap water. Dubreuil et al.
(2013).
looks better in the CRL image (Figure 88b) and even better for the target
estimation (Figure 88c). The PCE parameter provides valuable information
about images and can be useful for automatic target recognition.
In order to validate the advantage supplied by polarization imaging, a
series of experiments were conducted with varying milk concentration
and for different targets (plastic plug, painted and rusted metal). Figure 89
illustrates that as the optical thickness increases in skimmed milk, the
PCE tends to decrease for I, COL and CRL, COC and CRC imaging
modes, and target estimations for linear (ESTL) and circular (ESTC) polar-
izations. The panels in Figure 90 show the corresponding results for exper-
iments in semi-skimmed milk. It should be noted that the PCE values are
normalized by the highest value for one set of experiment (corresponding
to a specific target). Hence, these values can safely be compared for different
polarizations states.
We find that all curves have the same behavior. For depolarizing media
(plastic plug and painted metal), the PCE values (and thus the detection
Figure 90 As in Figure 5 for semi-skimmed milk diluted in tap water. Dubreuil et al.
(2013).
efficiency) are lower for COL and CRC than for I. Imaging with such polar-
ization modes does not allow for backscatter suppression. Moreover, a part
of the target signal is filtered with these modes. The PCE increases for COC
and CRL because of backscatter suppression. Larger PCEs are obtained for
ESTL and ESTC for which full backscatter is suppressed. For a polarization-
maintaining medium (rusted metal), one should note that imaging with
COC and CRL is not beneficial because a large part of the target signal is
filtered as well as the backscatter. However, target estimation leads to better
detection efficiency than for I, CRC, and COL. We note also that no sig-
nificant differences can be seen between imaging with circularly and linearly
polarized light. We further find that higher optical thickness can be achieved
in the case of semi-skimmed milk. This agrees with the fact that in Mie scat-
tering regime, the backscatter is smaller than in Rayleigh scattering regime,
and thus it requires more milk concentration to obtain a significant veil.
To gain more insight into realistic environment of target detection, e.g.,

underwater mine, experiments using turbid seawater (from Brest Bay area)
and lake water (Penfeld, Brest) were conducted. Painted metal like that used
in experiments with milk was taken as the target. As the turbidity of the
water is fixed, the target was progressively moved to vary optical thickness.
Figure 91a and b presents PCE as a function of target depth for seawater and
Figure 91 (a) PCE as a function of target depth for painted metal in ocean water.
(b) Same as in (a) for lake water. Dubreuil et al. (2013).
lake water, respectively. The reference images are the painted metal in pure
water taken at different depths.
Likewise, our analysis suggests that adapted polarization imaging (CRL
and COC) improves target detection compared to intensity imaging. Target
estimation is the best way to achieve good detection when the turbidity of
the embedding medium is large. As the turbidity of water is fixed and given
the dimensions of our glass tank, smaller optical thicknesses can be achieved
in these experiments as compared to our previous experiments with diluted
milk. The turbidity of the lake water was found to be much smaller than that
of seawater.
3.4.6.2 Tests in a Water Tank (6 m3)

Leonard, Alfalou, and Brosseau (2013) reported an underwater polarimetric
imaging setup depicted in Figure 92.
Figure 92 consists of a commercially available digital underwater camera
(Sealife DC1400 Pro Video Set). Its main specifications are as follows: image
sensor of 14 mega pixel high-resolution CCD, 26-mm-wide angle lens,
internal memory of 20 MB, underwater field of view of 51, and depth rated
to 60 m. The scene of interest is illuminated by three white LEDs. Full con-
trol of the incident polarization state is achieved by fixing a linear polarizer in
front of the light source. A second polarizer in front of the camera allows us
to analyze the reflected light. The large volume tank (Figure 92) is filled with
spring water (6000 L) and semi-skimmed milk. The experiments are
designed to mimic in situ conditions, allowing optical measurements on
unperturbed suspended particles with a set of instruments that can be nor-
mally employed at sea. Milk is composed of spherical particles of different
sizes (casein molecules 0.040.3 m and fat globules 120 m). Semi-
skimmed milk contains more fat globules resulting in a Mie scattering
regime. The estimated scattering coefficient s for semi-skimmed milk is
Figure 92 Underwater imaging system used by Leonard, Alfalou, and Brosseau (2013).
1.40c cm1, where c is the concentration of milk in water. The attenuation

length (or optical thickness) 0 is defined as the product of the scattering
coefficient s by the physical depth d, so that 0 sd. For the wavelength
considered here, the absorption coefficient of milk can be safely ignored
with respect to the scattering coefficient. Extinction coefficient data ranging
between 0.1 and 4 m1 were reported earlier for seawater. Milk concentra-
tion was adjusted to obtain an extinction coefficient in this range. When 1 L
of skimmed milk is dispersed in 6000 L of water, the scattering coefficient is
s 0.007 cm1.
In Figure 93, we present the object visibility changing with polarization,
milk addition (1 L of skimmed milk dispersed in the tank containing 6000 L
1m80
1m60
1m40
1m20
1m00
Figure 93 From the right, the first column indicates the object-to-camera distance; the
second column is the unpolarized image of the spherical object, the third column cor-
responds to the same object imaged with LPP, and the fourth column corresponds to
LCP. 1 L of skimmed milk was dispersed in the tank containing 6000 L of water.
of water), and varying the distance (number in the left column) of the camera
to the target (painted metal sphere) (Leonard, Alfalou, & Brosseau, 2013). In
the second column of Figure 93, we present the unpolarized image, whereas
the third and fourth columns of Figure 93 show, respectively, the
corresponding contrasts for the linearly parallel polarized (LPP) and linearly
crossed polarized (LCP) images. Polarization is shown to a useful tool for sig-
nificantly increasing the visibility of the object which is practically invisible in
the unpolarized image. There are many points to consider. First of all and as
shown in Figure 93, decreasing the object-to-camera distance increases the
recognition rate of the object because the visibility is increased. Second,
Figure 93 also indicates that the use of LCP allows better identification of per-
formances than LPP. It is also worth observing that Leonard, Alfalou, and
Brosseau (2013) presented analogous results dealing with color images for
an object-to-camera distance set to 1m80 (Figure 94) and milk addition.
As a final remark, we would like to point out that these experimental
results underlined the difficulties of dealing with real-time measurement
on moving targets or when the turbidity of the water is rapidly varying. Cur-
rent developments involving optical procedures using liquid crystals permit-
ting the obtention of fast polarization measurements and microlenses
permitting imaging the scene on different areas of the sensor for different
states of polarization are underway.
3.5 Techniques for Image Denoising, Fusion, and

Reconstruction
The purpose of this section is to present an illustrative example of denoising
by using the principle of the single operation gray image processing and its
Figure 94 Color image intensity (left) and DOP (right) of the target in water with milk
addition. The object-to-camera distance is set to 1m80.
optical setup shown in Figure 51. As discussed earlier, Huang et al. (1996)
described and implemented a method for removing a hole using an optical
setup based on the use of the single fundamental operation. They first
showed that the noise in a gray image can be suppressed (or attenuated)
by applying opening and/or closing operations to the image based on the
single fundamental operation O{.} defined by Equation (36). The reader
is referred to Huang et al. (1996) for the full details of the image processing.
The following example is obtained by making use of the optical setup
displayed in Figure 51 to suppress (or at least reduce) optically noise in a gray
image. The aim of Huang et al. (1996) was to remove the hole appearing in
the gray image shown in Figure 95a. For that specific purpose, these authors
used a choice of structure element with 2 2 pixel displayed in Figure 95b.
The result of the noise removing from Figure 95a after closing is displayed in
Figure 95c. The experimental results show that the single operation mor-
phological processor has good performances. In addition, it can be optically
implemented.
3.5.1 Wiener Filtering

Other denoising techniques of images have been described in the literature,
e.g., Wiener filtering (Duvaut, 1994; Wiener, 1949). Details on the Wiener
filtering and its application in OIP can be found in Kondo et al. (1977),
Baker (1981), Berriel et al. (1983), Jiang and Sawchuk (1986), Ikeda et al.
(1986), Inbar and Marom (1995), Zalevsky and Mendlovic (1996), Riris
et al. (1994), Kutay and Ozaktas (1998), Shabtay et al. (1998), Khoury
et al. (2000), Birch et al. (2001), Qian, Hu, Cheng, and Jin (2011),
Hardie and Barnard (2012), and Sizikov and E kzemplyarov (2013).
Figure 95 Example of noise removing based on the single fundamental operation

described by Huang et al. (1996).
3.5.2 Fusion
Fu and Zhang (2012) described and implemented numerically an efficient
method of fringe pattern denoising using averaging based on nonlocal
self-similarity. Related work includes the multifocus image fusion and den-
oising scheme based on homogeneity similarity reported by Li, Chai, Yin,
and Liu (2012), and a fusion method based on the shearlet transform
described in Miao, Shi, Xu, Yang, and Shi (2011).
3.5.3 Phase-Retrieval Algorithms

A problem of broad interest in optics deals with phase-retrieval algorithms
(PRAs). PRAs adapted to OIP have been described by Fienup (1982, 2013).
An example of application of PRAs is shown in Figure 96. Nakajima (2013)
reported another application of deterministic PRAs for the analysis of
coherent diffractive imaging beyond the Fresnel approximation with an
aperture-array filter.
Figure 96 Illustration of image reconstruction using PRA. (a) Nondegraded object

(a digitized photograph of a satellite); (b) and (c) examples of degraded images simu-
lated to include the effects of atmospheric turbulence and photon noise; (d) Fourier
modulus estimate computed from the degraded images; and (e) reconstructed image
using the iterative algorithm. Fienup (1982).
3.5.4 Binary Morphological Operation IP

In Wang, Liu, Liang, and Wang (1995), authors demonstrated that many
morphological operation IP can be realized optically. The results are shown
in Figure 96.
Other groups have reported studies on mathematical operations on
images, e.g., Benkelfat, El Wardi, Zghal, and Alfalou (2012), Ebersole
(1975), Liu and Chao (1986), Ebersole and Wyant (1976), Yu and Tai
(1979), Ghosh and Murata (1988), Shoop, Pezeshki, Goodman, and
Harris (1992), Denz, Dellwig, Lembcke, and Tschudi (1996), Fleck, Jahn,
and Wenke (1998), Pati, Unnikrishnan, and Singh (1998), and Shih,
Shishido, and Khoo (2001).
3.6 Techniques for Image Transmission: Optical Compression

and/or Optical Encryption Methods
Resources for storing, packaging, and/or transmitting big data (sound,
image, and video) in a secure manner confront engineers and physicists of
optical processing with a challenge. As the information to be transmitted
becomes of a larger amount and the transmission rate becomes faster, data
compression is becoming a critical challenge in video imaging. Conceptu-
ally, one ultimate goal of image compression is to reduce irrelevance and
redundancy of the image data in order to be able to store or transmit data
in an efficient form. Image compression can be classified as lossy or lossless.
Lossless compression is preferred for archival purposes and is often used for
medical imaging. Lossy compression methods, especially when used at low
bit rates, introduce compression artifacts. Lossy methods are especially suit-
able for natural images such as photographs in applications where minor
(sometimes imperceptible) loss of fidelity is acceptable to achieve a substan-
tial reduction in bit rate. In addition, the compression method used should
be adapted to the data. Here, emphasis will put on image compression tech-
niques by optical means (Alfalou & Brosseau, 2009).
The global economic infrastructure is becoming increasingly dependent
upon information technology, with computer and communication technol-
ogy being essential and vital components of government facilities, power
plant systems, medical infrastructures, financial centers, and military instal-
lations to name but a few. Finding effective ways to protect information
systems, networks, and sensitive data within the critical information infra-
structure is challenging even with the most advanced technology and trained
professionals.
The increasing number of information security-related incidents, orga-

nized crimes, and phishing scams mean that securing information is becom-
ing a major issue in the current information-based economy. To secure
information, many research directions have been suggested in the last
decade. Some security systems rely on the protocols secret, for the algo-
rithm encoding the information or cryptography (software approaches),
and for architecture aspects (hardware approaches). In fact, electronic
devices consume power, take time to compute, and emit electromagnetic
radiations highly correlated with the decoding processing. This brings a
number of fundamental questions on encryption to the fore; these are related
to the effects of that choice and choice of the encryption key. Symmetric key
algorithms are a class of algorithms for cryptography that use the same cryp-
tographic keys for both encryption of plaintext and decryption of ciphertext.
The keys may be identical or there may be a simple transformation to go
between the two keys.
The keys, in practice, represent a shared secret between two or more
parties that can be used to maintain a private information link. This require-
ment that both parties have access to the secret key is one of the main draw-
backs of symmetric key encryption, in comparison to public-key
encryption. Asymmetric cryptography or public-key cryptography is cryp-
tography in which a pair of keys is used to encrypt and decrypt a message so
that it arrives securely. Initially, a network user receives a public and private
key pair from a certificate authority. Any other user who wants to send an
encrypted message can get the intended recipients public key from a public
directory. They use this key to encrypt the message, and they send it to the
recipient. When the recipient gets the message, they decrypt it with their
private key, which no one else should have access to.
This section is organized as follows. We begin by giving a brief introduc-
tion to image compression and encryption. Several techniques allowing us to
realize separately or simultaneously these operations are reported. The
schemes described in the first and second rows of Figure 97 present two ways
for realizing these operations in cascade. The first scheme leads to the better
quality of the reconstructed image at the output of the system. But the
requested information size is not optimized since the encryption step
increases the information amount.
In comparison, the second scheme leads to a smaller size, but the quality
of the reconstructed image obtained is lesser. A significant body of literature
has amassed since those schemes were first considered. But now, several
schemes have been suggested to realize both operations simultaneously
Image
Compression Encryption Increase size
Image
Encryption Compression Lose of information
Good compromise
Image
Compression & encryption simultaneousely Compression/encryption
Figure 97 Fast and secure image transmission realized by compression/encryption,

encryption/compression, or compression and encryption realized simultaneously.
and independently (Alfalou & Brosseau, 2009; Alfalou, Mansour, et al.,

2011). Alfalou, Mansour, et al. (2011) and Alfalou and Brosseau (2009) gave
also details on the opportunities to realize these schemes by optical means.
3.6.1 Compression and Encryption Techniques

Compression
This review is intended to describe only lossy compression techniques. It is
possible to compress many types of digital data in a way that reduces the size
of a computer file needed to store it, or the bandwidth needed to stream it,
with no loss of the full information contained in the original file. A picture,
for example, is converted to a digital file by considering it to be an array of
dots and specifying the color and brightness of each dot. The original con-
tains a certain amount of information; there is a lower limit to the size of file
that can carry all the information. In many cases, files or data streams contain
more information than is needed for a particular purpose. For example, a
picture may have more detail than the eye can distinguish when reproduced
at the largest size intended; likewise, an audio file does not need a lot of fine
detail during a very loud passage. Developing lossy compression techniques
as closely matched to human perception as possible is a complex task. Some-
times, the ideal is a file that provides exactly the same perception as the orig-
inal, with as much digital information as possible removed; other times,
perceptible loss of quality is considered a valid trade-off for the reduced data.
More generally, some forms of lossy compression can be thought of as an
application of transform codingin the case of multimedia data, perceptual
coding: it transforms the raw data to a domain that more accurately reflects
the information content. Three classes of lossy compression have been
developed: those by prediction, e.g., ADPCM,19 those by transform, e.g.,
JPEG,20 JPEG 2000,21 MPEG,22 and those based on fractal properties.
Decompression is the inverse operation of compression.
Encryption
In cryptography, encryption is the process of encoding messages (or infor-
mation) in such a way that only authorized parties can read it. Encryption
does not prevent hacking but it reduces the likelihood that the hacker will
be able to read the data that is encrypted. In an encryption scheme, the mes-
sage or information (referred to as plaintext) is encrypted using an encryp-
tion algorithm, turning it into an unreadable ciphertext. This is usually done
with the use of an encryption key, which specifies how the message is to be
encoded. Any adversary that can see the ciphertext should not be able to
determine anything about the original message. An authorized party, how-
ever, is able to decode the ciphertext using a decryption algorithm (that usu-
ally requires a secret decryption key) which adversaries do not have access to.
For technical reasons, an encryption scheme usually needs a key-generation
algorithm to randomly produce keys. As mentioned briefly earlier, in sym-
metric key schemes, the encryption and decryption keys are the same. Thus,
communicating parties must agree on a secret key before they wish to com-
municate. In public-key encryption schemes, the encryption key is publi-
shed for anyone to use and encrypt messages.
However, only the receiving party has access to the decryption key and is
capable of reading the encrypted messages. Public-key encryption is a rela-
tively recent invention: historically, all encryption schemes have been sym-
metric key (also called private key) schemes. Encryption, by itself, can
19
ADPCM (Adaptive Differential Pulse-Code Modulation) is a variant of differential pulse-code mod-
ulation that varies the size of the quantization step, to allow further reduction of the required band-
width for a given SNR.
20
JPEG (Joint Photographic Experts Group) is a group of experts that develops and maintains standards
for a suite of compression algorithms for computer image files.
21
JPEG 2000 is a standard which was meant as an update of the widespread JPEG image standard. The
lossy image compression mode of JPEG 2000 is based on the so-called discrete wavelet transformation
which is supposed to be more efficient than the JPEG algorithm based on the cosine transformation.
22
MPEG (Moving Picture Experts Group) is a working group with the mission to develop standards for
coded representation of digital audio and video and related data.
protect the confidentiality of messages, but other techniques are still needed
to protect the integrity and authenticity of a message; for example, verifica-
tion of a message authentication code or a digital signature. In cryptography,
Kerckhoffs principle was stated by Auguste Kerckhoffs in the nineteenth
century: a cryptosystem should be secure even if everything about the sys-
tem, except the key, is public knowledge. Kerckhoffs principle23 was
reformulated (or perhaps independently formulated) by Claude Shannon as
the enemy knows the system. In contrast to security through obscurity,
it is widely embraced by cryptographers. Stated simply, the security of a
cryptosystem should depend solely on the secrecy of the key and the private
randomizer. Another way of putting it is that a method of secretly coding
and transmitting information should be secure even if everyone knows
how it works. Of course, despite the attackers familiarity with the system
in question, the attacker lacks knowledge as to which of all possible instances
is being presently observed.
The DES24 (Shamir, 1985) is a previously predominant symmetric-key
algorithm for the encryption of electronic data. DES is now considered to
be insecure for many applications. This is chiefly due to the 56-bit key size
being too small. The algorithm is believed to be practically secure in the form
of triple DES, although there are theoretical attacks. In recent years, the cipher
has been superseded by the advanced encryption standard. RSA25 (Biham &
Shamir 1999; Daemen & Rijmen, 2002) is a cryptosystem, which is known
as one of the first practicable public-key cryptosystems and is widely used for
secure data transmission. In such a cryptosystem, the encryption key is public
and differs from the decryption key which is kept secret.
In RSA, this asymmetry is based on the practical difficulty of factoring
the product of two large prime numbers, the factoring problem. However,
in recent years, other weak encryption methods have been used, e.g.,
ROT13 (rotation of 13 characters, without key), Caesars code (or Caesar
shift) which is a type of substitution cipher in which each letter in the
23
The system must be practically, if not mathematically, indecipherable. It must not be required to be
secret, and it must be able to fall into the hands of the enemy without inconvenience. Its key must be
communicable and retainable without the help of written notes, and changeable or modifiable at the
will of the correspondents. It must be applicable to telegraphic correspondence. It must be portable,
and its usage and function must not require the concourse of several people. Finally, it is necessary,
given the circumstances that command its application, that the system be easy to use, requiring neither
mental strain nor the knowledge of a long series of rules to observe.
24
DES, data encryption standard.
25
RSA stands for Ron Rivest, Adi Shamir, and Leonard Adleman, who first publicly described the algo-
rithm in 1977. Clifford Cocks, an English mathematician, had developed an equivalent system in
1973, but it was not declassified until 1997.
plaintext is replaced by a letter some fixed number of positions down the

alphabet; Vigene`res cipher which is a method of encrypting alphabetic text
by using a series of different Caesar ciphers based on the letters of a keyword.
3.6.2 Optical Compression

A first method, based on them 4f optical setup, was developed in Alfalou and
Brosseau (2010a, 2010b), Katz et al. (2012), and Alfalou, Brosseau, and Alam
(2013). This technique uses a specific fusion without overlapping of the tar-
get image spectra (Alfalou & Brosseau, 2009, 2010a, 2010b). The basic prin-
ciple is based on three concepts: (1) a local choice of relevant spectral
information coming from each target image; (2) the shift of the different
spectra according to a criterion calculating the minimum size of a given
spectrum (RSM duration); and (3) a fusion of different relevant spectral
information, without overlapping, to carry out a good compression and
encryption. However, this local selection depends directly on the different
target image spectra and may cause a deterioration of quality of the
reconstructed output images, especially if the spectra are similar. This phe-
nomenon is due to the appearance of isolated pixels; i.e., an isolated pixel is a
pixel assigned to a given spectrum, but it is surrounded by pixels coming
from other images. While a lot of effort was devoted to devise encryption
algorithms (Alfalou & Brosseau, 2009; Liu et al., 2011, 2013; Rajput &
Nishchal, 2013; Refregier & Javidi, 1995; Wang, 2012; Wang, Guo, &
Lei, 2013), it seems that the compression counterparts did not receive yet
the amount of effort they deserve.
A wealth of studies has appeared to reduce the size of the useful infor-
mation required for reconstruction of the target image by holographic tech-
niques (Darakis & Soraghan, 2007; Liu et al., 2002; Naughton et al., 2002;
Paturzo et al., 2008; Tahara et al., 2010; Xia et al., 2013). However, there is a
paucity of methods dealing with simultaneous compression and encryption
of multiple images which resemble each other, e.g., images in a video
sequence. Within this context, it is interesting to refer to Paturzo et al.
(2008), which deals with a small part of a specific spectrum and can be used
in the optical encryption domain. Recently, Alfalou and Brosseau (2013)
presented a method of compression and encryption based on DCT that
makes it possible to multiplex digital holograms.
This spectral multiple-image fusion analysis is based on the DCT and a
specific spectral filtering method. In order to decrease the size of the mul-
tiplexed file, they suggested a procedure of compression which is based on an
adapted spectral quantization. Each frequency is encoded with an optimized
number of bits according to its importance and position in the DC domain.

This fusion and compression scheme constitutes a first level of encryption.
A supplementary level of encryption is realized by making use of biometric
information. Several implementations of this analysis by experimenting with
sequences of gray scale images showed that it provides a viable solution for
simultaneous compression and encryption of multiple images.
For video applications, Mosso, Barrera, Tebaldi, Bolognini, and Torroba
(2011) suggested an encryption method of video images inspired from the
DRP method (Refregier & Javidi, 1995) and a specific spectral technique
in the output plane. However, this technique is not optimized from the
compression standpoint. The method requires a large number of bits to code
different output values, which eventually leads to an increase in the amount
of information to transmit and/or store.
We would like to highlight that other compression techniques which can
be implemented by optical means or devised specifically for OIP were pro-
posed in the archival literature as reviewed in Alfalou and Brosseau (2009).
Let us begin with the bandwidth compression scheme for 2D images devel-
oped by Smith and Barrett (1983). For the purpose of reducing the size of the
information necessary to reconstruct an image, this technique uses an RT,
along with filtering, thresholding, and quantification steps. The RT lends
itself to FT compression for at least three reasons: the coding process can
be performed with state-of-the-art 1D devices, the large dynamic range typ-
ical of the components of the FT is significantly reduced by the filtering
operation, and one line through the center of the 2D FT can be examined
at a time and adaptively compressed. Darakis and Soraghan (2007) described
a method for compressing interference patterns based on off-axis (phase-
shifting) DH using the PSI-DH hologram compression scheme.
In this work, Darakis and Soraghan showed clearly that if a compression
is carried out in the reconstructed plane, one gets a better normalized root-
mean-square error than that obtained with other techniques. For an appli-
cation of 3D object recognition using the PSI-DH technique, Naughton,
Frauel, Matoba, Javidi, and Tajahuerce (2002) suggest a digital hologram
compression technique based on Fourier-domain processing. The principle
of this operation consists in recording the digital holograms with an optical
system based on a MachZehnder interferometer. Naughton et al. (2002)
showed that the compression performances (the metric is the correlation
peak height between the original and reconstructed images) are rather poor
and proposed new lossy compression schemes; i.e., resizing of a hologram by
a resampling technique is first performed, then a quantization is directly
applied to the complex-valued holographic pixels, followed by a selective

removal of DTF coefficients. In addition, they demonstrated that digital
holograms are very sensitive to resampling; this is probably due to the
speckle effect which can be reduced thanks to a high degree of median
filtering.
Ding, Yan, Xue, and Jin (2003) proposed and validated an optical image
recognition system based on the volume holography and wavelet packet
compression methods. Volume holographic associative storage in a photo-
refractive crystal has some special properties, which can provide a suitable
mechanism to develop an OC system for image recognition. The method
of wavelet packet was introduced in this optical system to reduce the number
of images stored in the crystal. Through the best wavelet packet basis selec-
tion, a set of best eigen images, which are stored in the crystal as the reference
images for recognition, are extracted from a large number of training
images. Naughton, Joh, McDonald, and Javidi (2003) proposed and vali-
dated a new scheme of multiple-client compression and transmission of
3D objects. These authors used the PSI-DH technique for recording mul-
tiple views of a 3D object (Fresnel fields).
The digital Fresnel fields have the dimensions of 2028 2044 pixels
and in their native store 8 bytes of amplitude information and 8 bytes of
phase information for each pixel. Then, they constructed an Internet-based
Fresnel field compression application to measure reliably and accurately the
interaction between compression times and transmission times. This client-
server application and associated compression algorithms were written with
Java allowing them to develop a platform-independent environment for
experimentation over the Internet.
3.6.3 Optical Encryption

During the last decade, there has been a growing level of interest for increasing
security in communication systems. Many encryption algorithms have been
devised to improve the security level. But, in general, this requires large com-
putational time and resources. Second, in terms of fundamental studies, the
digital encryption step breaks the optical chain of measurements. Several sem-
inal papers for optical encryption have been presented (Alfalou & Brosseau,
2009; Refregier & Javidi, 1995). There is also tremendous interest in quantum
cryptography (Duraffourg, Merolla, Goedgebuer, Mazurenko, & Rhodes,
2001) and chaotic cryptography (Guglielmi et al., 2002).
During the last decade, there has been a growing level of interest in pro-
posing new algorithms of image encryption with random phase (Alfalou &
Brosseau, 2009) and image coding with an optical FT (Refregier & Javidi,
1995) or a fractional FT (Sahin, Ozaktas, & Mendlovic, 1995). There is a
wealth of research on encryption methods based on FT (Alfalou &
Brosseau, 2009). However, alternative approaches using the properties of
the DCT spectral fusion method are known to be very useful for encryption.
For example, Liu et al. (2011) proposed a color image encryption algorithm
by making use of the Arnold transform and DCT. The RGB components of
the color image are scrambled by the Arnold transform at the level of pixel
sequence. The scrambled RGB components are exchanged and mixed ran-
domly under the control of a matrix defined by a random angle. DCT is
employed for changing the pixel values of color image. In this encryption
scheme, the operations mentioned above are performed twice continuously.
The parameters of the Arnold transform and the random angle serve as the
key of the color image encryption method. Some numerical simulations are
made to test the validity and capability of the color encryption algorithm.
Numerical simulations have been reported to verify the color coding ability
of this algorithm (Liu et al., 2011).
The principle of other methods is briefly described.
Color image compression algorithm based on the DCT transform com-
bined to an adaptive block scanning
Remarkable success has been achieved in the low-loss compression of color
images (Douak, Benzid, & Benoudjit, 2011). After a preprocessing step
(mean removing and transforming the RGB components in a specific decor-
relating transform), the DCT transform is applied and followed by an iter-
ative phase (using the bisection method) including the thresholding,
quantization, dequantization, inverse DCT, inverse transform to get the
RGB components, and mean recovering. The efficiency of this method
was demonstrated with a series of experiments.
Key management of the double random-phase-encoding method using
public-key encryption
Saini and Sinha (2010) suggested that public-key encryption can be been
used to encode the key of the encryption process. In their technique, an
input image is encrypted by using the DRP encoding method using
extended fractional FT. The key of the encryption process is encoded by
using the RSA public-key encryption algorithm. The encoded is then trans-
mitted to the receiver side along with the encrypted image. In the decryp-
tion process, first the encoded key is decrypted using the secret key and then
the encrypted image is decrypted by using the retrieved key parameters. The
proposed technique has an advantage over DRP encoding method because
the problem associated with the transmission of the key has been eliminated
by using public-key encryption. The robust capability of this technique has
been illustrated by many computer simulations.
Image encryption based on the multiple-parameter discrete fractional FT
and chaos function
Lang, Tao, and Wang (2010) reported on a fractional Fourier method
for image encryption based on the multiple-parameter discrete fractional
Fourier transform and chaotic logistic maps in order to meet the require-
ments of the secure image transmission. In the proposed image encryption
scheme, the image is encrypted by juxtaposition of sections of the image in
the multiple-parameter discrete fractional Fourier domains and the align-
ment of sections is determined by chaotic logistic maps. This method does
not require the use of phase keys. The new method has been compared with
several existing methods and shows comparable or superior robustness to
blind decryption.
Multiple-image parallel optical encryption
Yong-Liang, Xin, Sheng, and Yao-Yao (2010) presented a method that
enables multiple-image optical encryption, in which a set of parallel plain-
texts can be extracted from the same designed ciphertext. This approach uses
a 6f optical setup. The principle of random phase encoding is utilized, and
the phase keys corresponding to different plaintexts are achieved indepen-
dently from the same designed ciphertext by cascade phase-retrieval algo-
rithm. Based on this general idea and on fusion, this method leads to an
independent decryption. The advantages of the approach could be con-
cluded as implementing decryption without cross talk, infinite encrypted
capacity and simple architecture. The plaintexts extracted mode can be also
extended from peer-to-peer to peer-to-multipeer.
One-step multiplexing optical encryption
Barrera and Torroba (2010) presented a combination of a free propagation
scheme and a classical 4f encrypting architecture to design a technique to
simultaneously encrypt multiple input objects. Unlike existing multiplexing
methods, encryption can be performed without setup modifications. An
interesting point of the approach presented in this paper is that the fusion
and encryption operations can be achieved in a single step. Moreover, it
requires only a single light source. The first encrypting masks are not placed
at the object plane, thus serving as an extra safety parameter. Additionally, the
multiplexing encryption scheme does not require setup modifications.
Authorized users need a master mask and the individual mask for the allowed
object, as well as the involved Fresnel distances. Due to the above described
characteristics, this procedure is suited for data safety handling connecting

multiple users.
Optical multi-image encryption based on the frequency shift
We first discuss the encryption algorithm of Liu, Zhang, Zhao, Ashfaq
Ahmad, and Liu (2010). This algorithm, based on fractional FT, allows
encrypting simultaneously several images. Lower frequency parts of the
original images are selected, frequency shifted, and encrypted by using dou-
ble phase encoding in fractional Fourier domains. Multiple images are
encrypted together into a single one. A simple optical setup is given to
implement the proposed algorithm. This scheme has features of enhance-
ment in decryption accuracy and high optical efficiency. The numerical
results demonstrated that this method permits to increase significantly the
encryption rate while maintaining the image quality if the number of input
images is reasonable. It is interesting to discuss a specific example.
Let us consider a 256 256 pixel image (Figure 98a) and its
corresponding Fourier spectrum of size 256 256 pixels (Figure 98b). If
we only cropped the center part of its Fourier spectrum, the photo of a
man and a camera (cameraman) with 256 256 pixels is shown in
Figure 98a. Its Fourier spectrum, shown in Figure 98b, is obtained from
FT. If we only use the central part (128 128 pixels, indicated by the white
square lines) of the spectrum, we recover an image as shown in Figure 98c.
The visual difference between input image and recovered image is very small.
Only some higher frequency details are lost. In some applications of image
encryption, such a high-frequency component lost can be acceptable.
3.6.4 Simultaneous Compression and Encryption

Only in the last few years, quantitative techniques have been developed for
simultaneous compression and encryption of images, and we have made
contributions to this development (Alfalou & Brosseau, 2010a, 2010b;
(a) (b) (c)
Figure 98 Illustration of the Liu et al. (2010) algorithm: (a) original image; (b) its Fourier
transform; and (c) image after decryption. Liu et al. (2010).
Alfalou, Brosseau, et al., 2011). These techniques open the way to a fuller
control of the reconstruction of target images. In the following, we will
focus our OIP analysis on the 4f system. The justification for such an
approach follows from the fact that image spectra can be manipulated
between the reading and formation steps of the process. Using such an
approach, the correlation between images has most recently been studied
in numerical simulations and experimental observations for face recognition
applications (Alfalou, Brosseau, & Alam, 2013; Katz et al., 2012). Very
recently, we pointed out that the 4f system provides a consistent way to
compress and encrypt an image with a specific filter (Alfalou & Brosseau,
2009). On the one hand, redundant information can be suppressed. On
the other hand, changing the distribution of the frequencies in images
can dramatically change the representation of the data and allows us to ren-
der them useless to a hacker (Alfalou & Brosseau, 2009).
Here, we restrict to a brief presentation of the approach reported by
Aldossari, Alfalou, and Brosseau (2013) to encrypt and compress simulta-
neously multiple images from a video sequence. The idea here is to use
an adaptation of a spectral fusion technique developed in Alfalou and
Brosseau (2010a, 2010b). On the one hand, the compression is based on
the RMS duration criterion of the spectrums image. That is, we determine
the size of the useful spectra for each target image by exploiting the RMS
criterion. This parameter is used to determine the allowed area of each target
image within the compressed spectrum. Moreover, this parameter is adapted
in order to minimize overlapping between the different spectra. Then, the
spectra are merged together by making use of a segmentation criterion. The
latter compares the local energy relative to each pixel for each spectrum.
Furthermore, it optimizes assignment of the considered pixel by taking into
account the adjacent areas to the considered pixel. This permits avoiding the
presence of isolated areas and small-sized areas (less than 10 pixels). On the
other hand, the encryption part is optimized by using and merging several
encryption keys (according to the number of multiple images considered).
The validation of this approach was realized by using a VLC architecture,
phase-only correlation filter, and peak-to-correlation energy as decision cri-
terion. The reader is referred to Aldossari et al. (2013) for greater detail about
the application of this technique for analysis of video sequences.
3.7 Summary and Future Prospects

The primary goal of the review presented herein was to give a brief account
of some of the theoretical and experimental developments that have taken
place in the area of OIP. With apologies, we have omitted several aspects of
this subject, and for the topics discussed here, we primarily used examples
from our research group. We focused our attention on optically
implementable methods which can be efficiently and rapidly processed.
From the experimental side, we showed that these methods are adapted
to denoise image, enhance edges, and realize filtering. The study of spatial
spectrum brings information about spatial distribution of intensity in the
image and their distinctive directions; i.e., high spatial frequencies are
responsible for fast and strong changes in image intensity (noise, edges),
while low spatial frequencies are responsible for smooth and slow changes
in image intensity. These techniques are not only helpful in 2D but also
allow us to manipulate 3D holograms. Furthermore, we showed several
examples which demonstrate that these techniques are of major importance
to the current concerns on the breast cancer problems. Taken together, the
works described in this review have provided a pleasingly complete analysis
of underwater polarization-filtered images. A recent application is the detec-
tion and classification of microcalcifications in a mammogram for early diag-
nosis of breast cancer. The achievement of these experiments in image
compression and encryption is also outstanding.
Overall, OIP is useful in applications in which a high parallelism and real-
time processing can be effectively realized. OIP is still in its early days, and
there are a number of directions into which the field is likely to move in the
coming years. PO associated with correlation techniques and Fourier optics
has particular sensitivity to detect and recognize objects in scattering or
poor-visibility media, addressing the venerable questions that persist in
the field of OIP, i.e., can we provide optimized optical techniques for den-
oising (deblurring, etc.) images in order to extract more and more features?
while introducing new questions, e.g., how optical neural networks, e.g.,
optical implementation of Hopfield model, can be done? What can we
do with all of this high speed processing? From an experimental perspective,
it would be equally interesting to use recent progress made in snapshot
polarization imaging to obtain measurements instantaneously (Kohlgraf-
Owens & Dogariu, 2009; Tyo et al., 2006). The diversity of this review sug-
gests that a rich spectrum of additional optical techniques of OIP is awaiting
discovery.
ACKNOWLEDGMENTS
Having a group working directly together or on related problems has been a source of
strength for us and the congenital atmosphere over the years has made our efforts so
enjoyable. We have been lucky to work with N. Abdallah, A. Arnold-Bos, B.-E. Benkelfat,
F. Bouzidi, P. Delrot, M. Dubreuil, M. Elbouz, M. Jridi, P. Katz, I. Leonard, A. Mansour,
Y. Ouerhani, Neji Nihel, and S. Qasmi. It was also great pleasure to work with A. Dogariu,
M. S. Alam, and J. Zallat. We also appreciate the support of Lab-STICC (Lab-STICC is
UMR CNRS 6285).
REFERENCES
Aggoun, A. (2006). Pre-processing of integral images for 3-D displays. Journal of Display Tech-
nology, 2, 393400. http://www.opticsinfobase.org/jdt/abstract.cfm?URIjdt-2-4-393.
Aggoun, A. (2011). Compression of 3D integral images using 3D wavelet transform. Journal of
Display Technology, 7, 586. http://www.opticsinfobase.org/jdt/abstract.cfm?URIjdt-
7-11-586.
Ahmed, N., Natarajan, T., & Rio, K. R. (1974). Discrete cosine transform. IEEE Transactions
on Computers, C-23, 90.
Alam, M. S. (2013). Optical pattern recognition. In M. Dekker (Ed.), Encyclopedia of optical
engineering (pp. 17781797). New York: Marcel Dekker.
Aldossari, M., Alfalou, A., & Brosseau, C. (2013). Optimized fusion method based on
adaptation of the RMS time-frequency criterion for simultaneous compression and
encryption of multiple images. In Proceedings of SPIE 8748, optical pattern recognition
XXIV, 87480B. http://dx.doi.org/10.1117/12.2014465.
Alfalou, A. (1999). Implementation of optical multichannel correlators: Application to pattern recog-
nition. Ph.D. thesis, Rennes-France: Universite de Rennes 1ENST Bretagne.
Alfalou, A., & Brosseau, C. (2009). Optical image compression and encryption methods.
Advances in Optics and Photonics, 1, 589. http://dx.doi.org/10.1364/AOP.1.000589.
Alfalou, A., & Brosseau, C. (2010a). Exploiting root-mean-square time-frequency structure
for multiple-image optical compression and encryption. Optics Letters, 35, 19141916.
http://dx.doi.org/10.1364/OL.35.001914.
Alfalou, A., & Brosseau, C. (2010b). Understanding correlation techniques for face recog-
nition: From basics to applications. In M. Oravec (Ed.), Face recognition book (p. 354):
India: INTECH. http://sciyo.com/articles/show/title/understanding-correlation-
techniques-for-face-recognition-from-basics-to-applications.
Alfalou, A., & Brosseau, C. (2013). Implementing compression and encryption of phase-
shifting digital holograms for three-dimensional object reconstruction. Optics Communi-
cation, 307, 6772.
Alfalou, A., Brosseau, C., Abdallah, N., & Jridi, M. (2011). Simultaneous fusion, compres-
sion, and encryption of multiple images. Optics Express, 19, 24023. http://dx.doi.org/10.
1364/OE.19.024023.
Alfalou, A., Brosseau, C., Abdallah, N., & Jridi, M. (2013). Assessing the performance of a
method of simultaneous compression and encryption of multiple images and its resistance
against various attacks. Optics Express, 21, 8025. http://dx.doi.org/10.1364/OE.21.
008025.
Alfalou, A., Brosseau, C., & Alam, M. S. (2013). Smart pattern recognition. In Proceedings
of SPIE 8748, optical pattern recognition XXIV, 874809. http://dx.doi.org/
10.1117/12.2018249.
Alfalou, A., Brosseau, C., Benkelfat, B.-E., Qasmi, S., & Leonard, I. (2012). Towards all-
numerical implementation of correlation. In Proceedings of SPIE 8398, optical pattern rec-
ognition XXIII, 839809. http://dx.doi.org/10.1117/12.919378.
Alfalou, A., Brosseau, C., Katz, P., & Alam, M. S. (2012). Decision optimization for face
recognition based on an alternate correlation plane quantification metric. Optics Letters,
37, 1562. http://dx.doi.org/10.1364/OL.37.001562.
Alfalou, A., Elbouz, M., Mansour, A., & Keryer, G. (2010). New spectral image compression
method based on an optimal phase coding and the RMS duration principle. Journal of
Optics, 12, 115403. http://dx.doi.org/10.1088/2040-8978/12/11/115403.
Alfalou, A., Keryer, G., & De Bougrenet De la Tocnaye, J. L. (1999). Optical implementa-
tion of segmented composite filtering. Applied Optics, 38, 61296135. http://www.
opticsinfobase.org/ao/abstract.cfm?URIao-38-29-6129.
Alfalou, A., Mansour, A., Elbouz, M., & Brosseau, C. (2011). Optical compression scheme to
multiplex & simultaneously encode images. In G. Cristobal, P. Schelkens, &
H. Thienpont (Eds.), Optical and digital image processing fundamentals and applications
(p. 463): Hoboken, NJ (USA): Wiley, ISBN: 978-3-527-40956-3.
Andonian, A. T. (1982). Correction for rigid body motion in moire topography. Applied
Optics, 21, 12531258. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-21-
7-1253.
Armitage, D., Thackara, J. l., & Eades, W. D. (1989). Photoaddressed liquid crystal spatial
light modulators. Applied Optics, 28, 4763. http://www.opticsinfobase.org/ao/abstract.
cfm?URIao-28-22-4763.
Atchia, Y., Levy, H., Dufour, S., & Levi, O. (2013). Rapid multiexposure in vivo brain
imaging system using vertical cavity surface emitting lasers as a light source. Applied
Optics, 52, C64C71. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-
C64.
Babkina, T. M., & Voloshinov, V. B. (2001). A new method of acousto-optic image
processing and edge enhancement. Journal of Optics A: Pure and Applied Optics, 3, S54.
http://dx.doi.org/10.1088/1464-4258/3/4/359.
Baker, A. (1981). Image detection and enhancement. Applied Optics, 20, 3917. http://www.
Banerjee, P. K., & Datta, A. K. (2012). Techniques of frequency domain correlation for
face recognition and its photonic implementation. In Q. Adamo & C. M. Epifano
(Eds.), Face recognition: Methods, applications and technology (pp. 165186): New York
(USA): Nova Publisher. https://www.novapublishers.com/catalog/product_info.php?
products_id28370 (Chapter 9).
Bara, S., Jaroszewicz, Z., Kolodziejczyk, A., & Moreno, V. (1991). Determination of basic
grids for subtractive moire patterns. Applied Optics, 30, 12581262. http://www.
Barat, C., & Phlypo, R. (2010). A fully automated method to detect and segment a man-
ufactured object in an underwater color image. EURASIP Journal on Advances in Signal
Processing, 2010, 110. http://dx.doi.org/10.1155/2010/568092.
Barlaud, M., & Labit, C. (2002). Compression et codage des images et des videos. Paris, France: Ed.
Hermes-Lavoisier.
Barrera, J. F., & Torroba, R. (2010). One step multiplexing optical encryption. Optics
Communication, 283, 12681272. http://dx.doi.org/10.1016/j.optcom.2009.11.083.
Barrett, P. (1987). Implementing the Rivest Shamir and Adleman public key encryption
algorithm on a standard digital signal processor. In Lecture notes in computer science: 263.
Proceedings of advances in cryptology-Crypto85 (p. 311): London (UK): Springer-Verlag.
Bartolini, L., De Dominicis, L., Ferri de Collibus, M., Fornetti, G., Francucci, M.,
Guarneri, M., et al. (2007). Polarimetry as tool to improve phase measurement in
amplitude modulated laser for submarine archaeological sites inspection. In O3A: Optics
for arts, architecture, and archaeology. Proceedings of SPIE 6618 (p.66180I). http://dx.doi.org/
10.1117/12.725320.
Bartolini, L., De Dominicis, L., Fornetti, G., Francucci, M., Guarneri, M., Poggi, C., et al.
(2007). Improvement in underwater phase measurement of an amplitude-modulated
laser beam by polarimetric techniques. Optics Letters, 32, 14021404. http://www.
opticsinfobase.org/ol/abstract.cfm?URIol-32-11-1402.
Benkelfat, B.-E., El Wardi, S., Zghal, M., & Alfalou, A. (2012). 2D real-time arithmetic
operations using optical coherence properties: Image processing applications. Proceedings
of SPIE, 8397, 83970D.
Berriel, R., Bescos, J., & Santisteban, A. (1983). Image restoration for a defocused optical
system. Applied Optics, 22, 2772. http://www.opticsinfobase.org/ao/abstract.cfm?
URIao-22-18-2772.
Bicout, D., & Brosseau, C. (1992). Multiply scattered waves through a spatially random
medium: Entropy production and depolarization. Journal of Physics I, 2, 20472063.
http://dx.doi.org/10.1051/jp1:1992266.
Bicout, D., Brosseau, C., Martinez, A. S., & Schmitt, J. M. (1994). Depolarization of mul-
tiply scattered waves by spherical diffusers: Influence of the size parameter. Physical
Review E, 49, 17671770. http://link.aps.org/doi/10.1103/PhysRevE.49.1767.
Biham, E., & Shamir, A. (1999). Power Analysis of the Key Scheduling of the AES Candi-
dates. In Proceedings of the second AES candidate conference, printed by The National Institute of
Standards and Technology, Gaithersburg, MD (p. 115).
Billingsley, F. C. (1970). Applications of digital image processing. Applied Optics, 9, 289299.
Birch, P., Tan, S., Young, R., Koukoulas, T., Claret-Tournier, F., Budgett, D., et al. (2001).
Experimental implementation of a Wiener filter in a hybrid digitaloptical correlator. Optics
Letters, 26, 494. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-26-8-494.
Blinn, J. F. (1993). Whats the deal with the DCT? IEEE Computer Graphics and Applications,
13, 78.
Boffety, M., Galland, F., & Allais, A. G. (2012). Influence of polarization filtering on image
registration precision in underwater conditions. Optics Letters, 37, 32733275. http://
www.opticsinfobase.org/ol/abstract.cfm?URIol-37-15-3273.
Born, M., & Wolf, E. (1970). Principles of optics (4th ed.). Oxford: Pergamon Press.
Brosseau, C. (1995). Statistics of the normalized Stokes parameters for a Gaussian stochastic
plane wave field. Applied Optics, 34, 47884793. http://dx.doi.org/10.1364/AO.34.
004788.
Brosseau, C. (1998). Fundamentals of polarized light: A statistical optics approach. New York:
Wiley, ISBN: 0471143022.
Brosseau, C., & Barakat, R. (1991). Jones and Mueller polarization matrices for random media,
Optics Communication, 84, 127133. http://dx.doi.org/10.1016/0030-4018(91)90214-X.
Brosseau, C., & Bicout, D. (1994). Entropy production in multiple scattering of light by a
spatially random medium. Physical Review E, 50, 49975005. http://link.aps.org/doi/
10.1103/PhysRevE.50.4997.
Brosseau, C., Givens, C. R., & Kostinski, A. B. (1993). Generalized trace condition on the
Mueller-Jones polarization matrix. Journal of the Optical Society of America. A, 10,
22482251. http://dx.doi.org/10.1364/JOSAA.10.002248.
Buytaert, J. A. N., & Dirckx, J. J. J. (2007). Design considerations in projection phase-shift
moire topography based on theoretical analysis of fringe formation. Journal of the Optical
Society of America. A, 24, 20032013. http://www.opticsinfobase.org/josaa/abstract.cfm?
URIjosaa-24-7-2003.
Buytaert, J. A. N., & Dirckx, J. J. J. (2008). Moire profilometry using liquid crystals for pro-
jection and demodulation. Optics Express, 16, 179193. http://www.opticsinfobase.org/
oe/abstract.cfm?URIoe-16-1-179.
Cao, F., Guichard, F., & Hornung, H. (2010). Information capacity: A measure of potential
image quality of a digital camera. Proceedings of SPIE, 7537, 75370F.
Cariou, J., Le Jeune, B., Lotrian, J., & Guern, Y. (1990). Polarization effects of seawater and
underwater targets. Applied Optics, 29, 16891695. http://www.opticsinfobase.org/ao/
abstract.cfm?URIao-29-11-1689.
Casasent, D. (1979). Performance evaluation of spatial light modulators. Applied Optics, 18,
2445. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-18-14-2445.
Casasent, D., & Xia, S. F. (1986). Phase correction of light modulators. Optics Letters, 11, 398.
http://www.opticsinfobase.org/ol/abstract.cfm?URIol-11-6-398.
Castillo, I. M. D., Sanchez-de-la-Llave, D. D., Garca, R. R., Olivos-Perez, L. I.,
Gonzalez, L. A., & Rodrguez-Ortiz, M. M. (2001). Real-time self-induced nonlinear
optical Zernike-type filter in a bacteriorhodopsin film. Optical Engineering, 40,
23672368. http://dx.doi.org/10.1117/1.1412425.
Catrysse, P. B., Imai, F. H., Linne von Berg, D. C., & Sheridan, J. T. (2013). Imaging systems
and applications. Applied Optics, 52, ISA1ISA3. http://www.opticsinfobase.org/ao/
abstract.cfm?URIao-52-7-ISA1.
Catrysse, P. B., & Skauli, T. (2013). Pixel scaling in infrared focal plane arrays. Applied Optics,
52, C72C77. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C72.
Chambah, M., Semani, D., Renouf, A., Courtellemont, P., & Rizzi, A. (2003). Underwater
color constancy: Enhancement of automatic live fish recognition. In R. Eschbach, &
G. G. Marcu (Eds.), Proceedings of SPIE: Vol. 5293. Color imaging IX: Processing, hardcopy,
and applications (pp. 157168).
Chang, T. Y., Hong, J. H., & Yeh, P. (1990). Spatial amplification: An image-
processing technique using the selective amplification of spatial frequencies. Optics
Chang, W. H., & Nguyen, T. Q. (2008). On the fixed-point accuracy analysis of FFT
algorithms. IEEE Transactions on Signal Processing, 56, 4673.
Chen, H. C. (2005). A memory-efficient realization of cyclic convolution and its application
to discrete cosine transform. IEEE Transactions on Circuits and Systems for Video Technology,
15, 445.
Chen, W., Smith, C., & Fralick, S. (1977). A fast computational algorithm for the discrete
cosine transform. IEEE Transactions on Communications, 25, 1004.
Clemenceau, P., Dogariu, A., & Stryewski, J. (2000). Polarization active imaging. In Laser
radar technology and applications V: Vol. 4035. Proceedings of SPIE (pp. 401409). USA.
Collins, S. A. (1970). Lens-system diffraction integral written in terms of matrix optics. Journal
of the Optical Society of America, 60, 11681177.
Cooley, J. W., & Tukey, J. W. (1965). An algorithm for the machine calculation of complex
Fourier series. Mathematics of Computation, 19, 297.
Cox, I., Miller, M., Bloom, J., Fridrich, J., & Kalker, T. (2007). Digital watermarking and steg-
anography (2nd ed.). San Francisco, CA: Elsevier.
Cox, I. J., & Sheppard, C. J. R. (1986). Information capacity and resolution in an optical
system. Journal of the Optical Society of America. A, 3, 11521158.
Daemen, J., & Rijmen, V. (2002). The design of Rijndael: AESThe advanced encryption
standard. London (UK): Springer.
Darakis, E., Naughton, T., & Soraghan, J. J. (2007). Compression defects in different recon-
struction from phase-shifting digital holographic data. Applied Optics, 46, 45794586.
http://www.opticsinfobase.org/ao/abstract.cfm?URIao-46-21-4579.
Darakis, E., & Soraghan, J. J. (2006). Compression of interference patterns with application
to phase-shifting digital holography. Applied Optics, 45, 24372443. http://www.
Darakis, E., & Soraghan, J. J. (2007a). Reconstruction domain compression of phase-shifting
digital holograms. Applied Optics, 46, 351356. http://www.opticsinfobase.org/ao/
DeCusatis, C., & Das, P. (1995). Signal processing using wavelet transforms. Optics &
Photonics News, 6, 39. http://www.opticsinfobase.org/opn/abstract.cfm?URIopn-6-
12-39.
Denz, C., Dellwig, T., Lembcke, J., & Tschudi, T. (1996). Parallel optical image addition and
subtraction in a dynamic photorefractive memory by phase-code multiplexing. Optics
Letters, 21, 278280.
Dickey, T. D., Kattawar, G. W., & Voss, K. J. (2011). Shedding new light on light in the
ocean. Physics Today, 64, 4449.
Ding, J. J. (2001). Research of fractional Fourier transform and linear canonical transform. Taiwan:
National Taiwan University, Ph.D. dissertation.
Ding, L., Yan, Y., Xue, Q., & Jin, G. (2003). Wavelet packet compression for volume
holographic image recognition. Optics Communication, 216, 105113.
Douak, F., Benzid, R., & Benoudjit, N. (2011). Color image compression algorithm based
on the DCT transform combined to an adaptive block scanning. AEU: International
Journal of Electronics and Communications, 65, 1626. http://dx.doi.org/10.1016/j.aeue.
2010.03.003.
Dubreuil, M., Alfalou, A., & Brosseau, C. (2012). Imaging polarimetry and correlation tech-
niques for target detection. In Proceedings of Frontiers in optics 2012/Laser Science XXVIII,
OSA Technical Digest (online) (Optical Society of America, 2012), paper FW3A.30. http://
www.opticsinfobase.org/abstract.cfm?URIFiO-2012-FW3A.30.
Dubreuil, M., Delrot, P., Leonard, I., Alfalou, A., Brosseau, C., & Dogariu, A. (2013).
Exploring underwater target detection by imaging polarimetry and correlation tech-
niques. Applied Optics, 52, 9971005. http://www.opticsinfobase.org/ao/abstract.cfm?
URIao-52-5-997.
Duncan, R. C. (1962). Americas use of sea mines. White Oak: United States Naval Ordinance
Laboratory.
Duntley, S. Q. (1963). Light in the sea. Journal of the Optical Society of America, 53, 214233.
http://dx.doi.org/10.1364/JOSA.53.000214.
Duraffourg, L., Merolla, J. M., Goedgebuer, J. P., Mazurenko, Y., & Rhodes, W. T. (2001).
Compact transmission system using single-sideband modulation of light for quantum
cryptography. Optics Letters, 26, 1427.
Duvaut, P. (1994). Traitement du signal: Concepts et applications. Paris: Herme`s.
Ebersole, J. F. (1975). Optical image subtraction. Optical Engineering, 14, 436447.
Ebersole, J. F., & Wyant, J. C. (1976). Real-time optical subtraction of photographic imagery
for difference detection. Applied Optics, 15, 871876.
Egami, C., Suzuki, Y., Uemori, T., Sugihara, O., & Okamoto, N. (1997). Self-adaptive
spatial filtering by use of azo chromophores doped in low glass-transition-temperature
polymers. Optics Letters, 22, 14241426. http://www.opticsinfobase.org/ol/abstract.
cfm?URIol-22-18-1424.
Egan, J. P. (1975). Signal detection theory and ROC analysis. New York: Academic Press Series
in Cognition and Perception.
Elbouz, M., Alfalou, A., & Brosseau, C. (2011). Fuzzy logic and optical correlation-based
face recognition method for patient monitoring application in home video surveillance.
Optical Engineering, 50, 067003. http://dx.doi.org/10.1117/1.3582861.
Elbouz, M., Bouzidi, F., Alfalou, A., Brosseau, C., Leonard, I., & Benkelfat, B.-E. (2013).
Adapted all-numerical correlator for face recognition applications. In Optical pattern rec-
ognition XXIV: Vol. 8748. Proceedings of SPIE (p. 874807). USA. http://dx.doi.org/10.
1117/12.2014383.
Espinosa, J., Mas, D., Perez, J., & Roig, A. B. (2013). Open-access operating algorithms for
commercial videokeratographer and improvement of corneal sampling. Applied Optics,
52, C24C29. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C24.
Fales, C. L., Huck, F. O., & Samms, R. W. (1984). Imaging system design for improved
information capacity. Applied Optics, 23, 872888.
Fan, L., Song, F., & Jutamulia, S. (2007). Edge detection with large depth of focus using
differential HaarGaussian wavelet transform. Optics Communications, 270, 169175.
http://www.sciencedirect.com/science/article/pii/S0030401806009485.
Fellgett, P. B., & Linfoot, E. H. (1955). On the assessment of optical images. Philosophical
Transactions of the Royal Society A, 247, 369407.
Ferrari, J. A., Garbusi, E., Frins, E. M., & Priz, G. (2005). Spatial self-filtering with polarizer
sheets. Applied Optics, 44, 4146. http://www.opticsinfobase.org/ao/abstract.cfm?
URIao-44-1-41.
Fienup, J. R. (1982). Phase retrieval algorithms: A comparison. Applied Optics, 21,
Fienup, J. R. (2013). Phase retrieval algorithms: A personal tour [Invited]. Applied Optics, 52,
Fleck, B., Jahn, D., & Wenke, L. (1998). A nonlinear optical element for incoherent image
subtraction based on photoanisotropy. Optical Communications, 154, 339344.
Flores, J. L., & Ferrari, J. A. (2010). Orientation-selective edge detection and enhancement
using the irradiance transport equation. Applied Optics, 49, 619624. http://www.
Flores, J. L., Ferrari, J. A., & Garcia-Torales, G. (2011). Partially coherent optical processor
for enhancement of partial second order derivatives of an image: Analysis and implemen-
tation. Optical Communications, 284, 56305635. http://www.sciencedirect.com/sci
ence/article/pii/S0030401811009485.
Fu, S., & Zhang, C. (2012). Fringe pattern denoising using averaging based on nonlocal self-
similarity. Optics Communications, 285, 25412544. http://dx.doi.org/10.1016/j.
optcom.2012.01.059.
Garfinkel, L., Boring, C. C., & Heath, C. W., Jr. (1994). Changing trends. An overview of
breast cancer incidence and mortality. Cancer, 74, 222227.
Ghael, S. P., Sayeed, A. M., & Baraniuk, R. G. (1997). Improved wavelet denoising via
empirical Wiener filtering. In Wavelet applications in signal and image processing V:
Vol. 3169. Proceedings of SPIE (p. 389). USA. http://dx.doi.org/10.1117/12.292799.
Ghosh, A., & Murata, K. (1988). Real-time addition and subtraction of two images by vector
wave interferometry. Optics and Laser Technology, 20, 211215.
Gilbert, G. D., & Pernicka, J. C. (1967). Improvement of underwater visibility by reduction
of backscatter with a circular polarization technique. Applied Optics, 6, 741746. http://
www.opticsinfobase.org/ao/abstract.cfm?URIao-6-4-741.
Glenn, D. B. (1968). A fast Fourier transform algorithm for teal-valued series. Communications
of the ACM, 11, 703.
Gl
uckstad, J. (1995). Adaptive array illumination and structured light generated by spatial
zero-order self-phase modulation in a Kerr medium. Optics Communication, 120,
194203.
uckstad, J., Lading, L., Toyoda, H., & Hara, T. (1997). Lossless light projection. Optics Letters,
Gl
22, 13731375. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-22-18-1373.
Goodman, J. W. (1966). Introduction to Fourier optics (2nd ed.). New York: McGraw-Hill.
Goodman, J. (1976). Some fundamental properties of speckle. Journal of the Optical Society of
America, 66, 11451150.
Goole (2013). http://fr.wikipedia.org/wiki/Moir%C3%A9_%28effet_de_contraste%29.
Gorman, G. A. (2011). Field deployable dynamic lightning system for turbid water imaging. M.S.
Thesis, Cambridge, USA: Massachusetts Institute of Technology.
Gourier, N., Hall, D., & Crowley, J. L. (2004). Estimating face orientation from robust
detection of salient facial features. In ICPR, International workshop on visual observation
of deictic gestures. Proceedings of Pointing 2004.
Granier, B., Refregier, P., & Figue, J. (1996). Restoration of images degraded by atmospheric
turbulence by a least-squares method and a Markov process. Optics Letters, 21, 423425.
http://www.opticsinfobase.org/ol/abstract.cfm?URIol-21-6-423.
Guglielmi, V., Fournier-Prunaret, D., Taha, A. K., Pinel, P., Rouabhi, S., & Beneteau, L.
(2002). Encryption using chaotic dynamics of two noninvertible maps. In Proceedings of
workshop on nonlinear dynamics of electronic systems (NDES02), Izmir, Turquie.
Gurov, I. P., & Sheynihovich, D. V. (2000). Interferometric data analysis based on Markov
nonlinear filtering methodology. Journal of the Optical Society of America. A, 17, 2127.
http://www.opticsinfobase.org/josaa/abstract.cfm?URIjosaa-17-1-21.
Hardie, R. C., & Barnard, K. J. (2012). Fast super-resolution using an adaptive Wiener filter
with robustness to local motion. Optics Express, 20, 21053.
Harris, J. L., Sr. (1966). Image evaluation and restoration. Journal of the Optical Society of
America, 56, 569. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-30-10-1258.
Hartmann, G. K., & Truver, S. (1991). Weapons that wait: Mine warfare in the United States
Navy. Annapolis: Naval Institute Press.
Hecht, E. (2001). Optics (4th ed.). New York: Addison-Wesley.
Helan, S., Burie, J., Bouwmans, T., & Bazeille, S. (2006). Object detection in underwater
images. In Proceedings of CMM06, Caracterisation du milieu marin, 1619/10.
Hennelly, B. M., & Sheridan, J. T. (2005). Generalizing, optimizing, and inventing numer-
ical algorithms for the fractional Fourier, Fresnel, and linear canonical transforms. Journal
of the Optical Society of America. A, 22, 917927. http://dx.doi.org/10.1364/JOSAA.22.
000917.
Horner, J. L. (1992). Metrics for assessing pattern-recognition performance. Applied Optics,
31, 165166. http://dx.doi.org/10.1364/AO.31.000165.
Horner, J. L., & Gianino, P. D. (1984). Phase-only matched filtering. Applied Optics, 23,
812816.
Hou, H. (1987). A fast recursive algorithm for computing the discrete cosine transform. IEEE
Transactions on Acoustics, Speech, and Signal Processing, 35, 1455.
Hou, W., Gray, D., Weidemann, A., Fournier, G., & Forand, J. (2007). Automated under-
water image restoration and retrieval of related optical properties. In Geoscience and remote
sensing symposium. Proceedings of IEEE IGARSS (pp. 18891892). http://ieeexplore.ieee.
org/stamp/stamp.jsp?tp&arnumber4423193&isnumber4422708.
Hou, S., Zhouand, S., & Siddique, M. A. (2012). Query by example video based on fuzzy
c-means initialized by fixed clustering center. Optical Engineering, 51, 047405. http://dx.
doi.org/10.1117/1.OE.51.4.047405.
Huang, K. S., Jenkins, B. K., & Sawchuk, A. A. (1989). Image algebra representation of
parallel optical binary arithmetic. Applied Optics, 28, 12631278. http://www.
Huang, G., Jin, G., Wu, M., & Yan, Y. (1996). Novel gray image processing in a single oper-
ation optical morphological processor. Optics Communication, 130, 143152. http://
www.sciencedirect.com/science/article/pii/0030401896002155.
Huang, K. S., Sawchuk, A. A., Jenkins, B. K., Chavel, P., Wang, J. M., Weber, A. G., et al.
(1993). Digital optical cellular image processor (DOCIP): Experimental implementation.
Applied Optics, 32, 166173. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-
32-2-166.
Huang, G., Wu, M., Jin, G., Yao, Z., & Yan, Y. (1994). The parallel processor for the char-
acter extraction of digitizing image. Chinese Journal of Lasers, 283288. http://www.
oriprobe.com/journals/caod_483/1994_3.html.
Ikeda, O., Sato, T., & Kojima, H. (1986). Construction of a Wiener filter using a phase-
conjugate filter. Journal of the Optical Society of America. A, 3, 645. http://www.
opticsinfobase.org/josaa/abstract.cfm?URIjosaa-3-5-645.
Inbar, H., & Marom, E. (1995). A priori and adaptive Wiener filtering with joint transform
correlators. Optics Letters, 20, 1050. http://www.opticsinfobase.org/ol/abstract.cfm?
URIol-20-9-1050.
Iqbal, K., Salam, R., Osman, A., & Talib, A. (2007). Underwater image enhancement using
an integrated colour model. IAENG International Journal of Computer Science, 1819-
656X.32, 529534. http://eprints.usm.my/id/eprint/9675.
Jaffe, J. (1990). Computer modeling and the design of optimal underwater imaging systems.
IEEE Journal of Oceanic Engineering, 15, 101111. http://jaffeweb.ucsd.edu/files/
JaffeIEEEJOE.pdf.
Jiang, S. S., & Sawchuk, A. A. (1986). Noise updating repeated Wiener filter and other adap-
tive noise smoothing filters using local image statistics. Applied Optics, 25, 2326. http://
Joseph, J., Aranda, F. J., Rao, D. V. G. L. N., Akkara, J. A., & Nakashima, M. (1996). Optical
Fourier processing using photoinduced dichroism in a bacteriorhodopsin film. Optics
Letters, 21, 14991501. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-21-
18-1499.
Kato, H., & Goodman, J. W. (1975). Nonlinear filtering in coherent optical systems through
halftone screen processes. Applied Optics, 14, 18131824. http://www.opticsinfobase.
org/ao/abstract.cfm?URIao-14-8-1813.
Kato, J. I., Yamaguchi, I., Nakamura, T., & Kuwashima, S. (1997). Video-rate fringe analyzer
based on phase-shifting electronic moire patterns. Applied Optics, 36, 84038412, http://
Kato, U. I., Yamaguchi, I., & Tanaka, H. (1996). Nonlinear spatial filtering with a dye-doped
liquid-crystal cell. Optics Letters, 21, 767769. http://www.opticsinfobase.org/ol/
abstract.cfm?URIol-21-11-767.
Kattawar, G. W., & Rakovic, M. J. (1999). Virtues of Mueller matrix imaging for underwater
target detection. Applied Optics, 38, 64316438. http://www.opticsinfobase.org/ao/
Katz, P., Alfalou, A., Brosseau, C., & Alam, M. S. (2012). Correlation and independent
component analysis based approaches for biometric recognition. In A. Quaglia, &
C. M. Epifano (Eds.), Face recognition: Methods, applications and technology
(pp. 201229): New York (USA): NOVA Publishers. https://www.novapublishers.
com/catalog/product_info.php?products_id28370.
Kerlikowske, K., & Barclay, J. (1997). Outcomes of modern screening mammography.
Journal of the National Cancer Institute. Monographs, 22, 105111.
Khalil, A. M., & Fazio, P. (1973). Lens distortion equation using moire. Applied Optics, 12,
Khoury, J., Gianino, P. D., & Woods, C. L. (2000). Wiener-like correlation filters.
Applied Optics, 39, 231. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-39-
2-231.
Kim, A. D., & Moscoso, M. (2002). Backscattering of circularly polarized pulses. Optics
Letters, 27, 15891591. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-27-
18-1589.
Kim, Y., Park, G., Jung, J. H., Kim, J., & Lee, B. (2009). Color moire pattern simulation and
analysis in three-dimensional integral imaging for finding the moire-reduced tilted angle
of a lens array. Applied Optics, 48, 21782187. http://www.opticsinfobase.org/ao/
Kohlgraf-Owens, T., & Dogariu, A. (2009). Spatially resolved scattering polarimeter. Optics
Letters, 34, 13211323. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-34-9-
1321.
Kolba, D., & Parks, T. (1977). A prime factor FFT algorithm using high-speed convolution.
IEEE Transactions on Acoustics, Speech, and Signal Processing, 25, 281.
Komorowska, K., Miniewicz, A., Parka, J., & Kajzar, F. (2002). Self-induced nonlinear
Zernike filter realized with optically addressed liquid crystal spatial light modulator.
Journal of Applied Physics, 92, 56355642. http://dx.doi.org/10.1063/1.1515949.
Kondo, K., Ichioka, Y., & Suzuki, T. (1977). Image restoration by Wiener filtering in
the presence of signal-dependent noise. Applied Optics, 16, 2554. http://www.
Korchemskaya, E. Y., & Stepanchikov, D. A. (1998). Analysis of object motion by dynamic

holography recording in bacteriorhodopsin films. Proceedings of SPIE. 3486, (p. 156).
Kothapalli, S. R., Wu, P., Yelleswarapu, C., & Rao, D. V. G. L. N. (2004). Medical image
processing using transient Fourier holography in bacteriorhodopsin films. Applied Physics
Letters, 85, 58365838. http://dx.doi.org/10.1063/1.1833567.
Kothapalli, S. R., Wu, P., Yelleswarapu, C. S., & Rao, D. V. G. L. N. (2005). Nonlinear
optical Fourier filtering technique for medical image processing. Journal of Biomedical
Optics, 10, 0440281. http://dx.doi.org/10.1117/1.1953287.
Kouzoubov, A., Brennan, M. J., & Thomas, J. C. (1998). Treatment of polarization in laser
remote sensing of ocean water. Applied Optics, 37, 38733885. http://www.
Kutay, M. A., & Ozaktas, H. M. (1998). Optimal image restoration with the fractional Fou-
rier transform. Journal of the Optical Society of America. A, 15, 825. http://www.
opticsinfobase.org/josaa/abstract.cfm?URIjosaa-15-4-825.
Lang, J., Tao, R., & Wang, Y. (2010). Image encryption based on the multiple-parameter
discrete fractional Fourier transform and chaos function. Optics Communications, 283,
20922096. http://dx.doi.org/10.1016/j.optcom.2010.01.060.
Lebanon, G., & Bruckstein, A. M. (2001). Variational approach to moire pattern synthesis.
Journal of the Optical Society of America. A, 18, 1371. http://www.opticsinfobase.org/josaa/
abstract.cfm?URIjosaa-18-6-1371.
Lee, S. H., Park, H. M., & Hwang, S. Y. (2012). Motion deblurring using edge map with
blurred/noisy image pairs. Optics Communications, 285, 17771786. http://www.
sciencedirect.com/science/article/pii/S0030401811014404.
Leitch, A. M., Dodd, G. D., Constanza, M., Linver, M., Pressman, P., McGinis, L., et al.
(1997). American Cancer Society guidelines for the early detection of breast cancer:
Update 1997. CA: A Cancer Journal for Clinicians, 47, 150153.
Leonard, I. (2012). Underwater mine recognition. Ph.D. Thesis, Brest, France: UBO.
Leonard, I., Alfalou, A., Alam, M. S., & Arnold-Bos, A. (2012). Adaptive nonlinear fringe-
adjusted joint transform correlator. Optical Engineering, 51(9), 098201-1098201-14.
http://dx.doi.org/10.1117/1.OE.51.9.098201.
Leonard, I., Alfalou, A., & Brosseau, C. (2012). Face recognition based on
composite correlation filters: Analysis of their performances. In A. Quaglia, &
C. M. Epifano (Eds.), Face recognition: Methods, applications and technology (pp. 5780):
New York (USA): NOVA Publishers.
Leonard, I., Alfalou, A., & Brosseau, C. (2013). Sensitive test for sea mines identification
based on polarization-aided image processing. Optics Express, 21, 2928329297.
Leonard, I., Alfalou, A., Zallat, J., Lallement, A., & Brosseau, C. (2013). Sensitive test for
object identification based on polarization imaging and correlation. In I. Kang,
D. Reitze, N. Alic, & D. Hagan (Eds.), Proceedings of Frontiers in optics 2013, OSA Tech-
nical Digest (online), paper FTh3D.4: Optical Society of America. http://www.
opticsinfobase.org/abstract.cfm?URIFiO-2013-FTh3D.4.
Levie, H. S. (1992). Mine warfare at sea. Dordrecht: Martinus Nijhoff Publishers.
Lewis, G. D., Jordan, D. L., & Roberts, P. J. (1999). Backscattering target detection in a tur-
bid medium by polarization discrimination. Applied Optics, 38, 39373944. http://www.
Li, H., Chai, Y., Yin, H., & Liu, G. (2012). Multifocus image fusion and denoising scheme
based on homogeneity similarity. Optics Communications, 285, 91100. http://dx.doi.
org/10.1016/j.optcom.2011.08.078.
Li, Q., Xu, Z., Feng, H., Tao, X., & Zhao, J. (2011). Analysis of image restoration and eval-
uation for diffraction-degraded remote sensing image. In International symposium on pho-
toelectronic detection and imaging 2011: Space exploration technologies and applications:
Vol. 8196. Proceedings of SPIE (p. 819607). http://dx.doi.org/10.1117/12.897423.
Lim, J. S., Kim, J., & Chung, M. S. (1989). Additive type moire with computer image
processing. Applied Optics, 28, 26772680. http://www.opticsinfobase.org/ao/
Liu, Z., Centurion, M., Panotopoulos, G., Hong, J., & Psaltis, D. (2002). Holographic
recording of fast events on a CCD camera. Optics Letters, 27, 22. http://www.
Liu, H. K., & Chao, T. H. (1986). Optical image subtraction techniques, 1975-1985.
Proceedings of SPIE, 638, 5565.
Liu, L., Cui, B., & Zhao, L. (1994). Optical cellular fuzzy logic gray-scale image proces-
sor. Optics Communications, 107, 453. http://dx.doi.org/10.1016/0030-4018(94)
90363-8.
Liu, Z., Xu, L., Liu, T., Chen, H., Li, P., Lin, C., et al. (2011). Color image encryption by
using Arnold transform and color-blend operation in discrete cosine transform domains.
Optics Communications, 284, 123.
Liu, Z., Zhang, Y., Li, S., Liu, W., Liu, W., Wang, Y., et al. (2013). Double image
encryption scheme by using random phase encoding and pixel exchanging in the gyrator
transform domains. Optics & Laser Technology, 47, 152.
Liu, L., Zhang, Z., & Zhang, X. (1994). One-operation image algebra and optoelectronic
cellular two-layer logic array. Journal of the Optical Society of America. A, 11,
17891797. http://www.opticsinfobase.org/josaa/abstract.cfm?URIjosaa-11-6-1789.
Liu, Z., Zhang, Y., Zhao, H., Ashfaq Ahmad, M., & Liu, S. (2010). Optical multi-image
encryption based on frequency shift. Optik, 122, 10101013. http://dx.doi.org/
10.1016/j.ijleo.2010.06.039.
Marom, E., & Inbar, H. (1996). New interpretations of Wiener filters for image recognition.
Journal of the Optical Society of America. A, 13, 1325. http://www.opticsinfobase.org/josaa/
abstract.cfm?URIjosaa-13-7-1325.
Mazzaferri, J., Campos, J., Escalera, J. C., Sheppard, C. J. R., & Ledesma, S. (2010). Edge
detector tolerant to object defocusing. Optics Communications, 283, 36393645. http://
www.sciencedirect.com/science/article/pii/S0030401810004839.
Mazzaferri, J., & Ledesma, S. (2007). Rotation invariant real-time optical edge detector.
Optics Communications, 272, 367376. http://www.sciencedirect.com/science/article/
pii/S0030401806012971.
McGlamery, B. L. (1975). Computer analysis and simulation of underwater camera system
performance. San Diego: University of California. Scripps Institution of Oceanography,
Visibility Laboratory, San Diego, California 92152, http://misclab.umeoce.maine.
edu/education/VisibilityLab/reports/SIO_75-2.pdf.
Meher, P. K. (2006). Systolic designs for DCT using a low-complexity concurrent con-
volutional formulation. IEEE Transactions on Circuits and Systems for Video Technology,
16, 1041.
Mendlovic, D. (2013). Toward a super imaging system. Applied Optics, 52, 561566. http://
Mendlovic, D., & Konforti, N. (1993). Optical realization of the wavelet transform for two-
dimensional objects. Applied Optics, 32, 6542. http://www.opticsinfobase.org/ao/
Menke, N., Yao, B., Wang, Y., Zheng, Y., Lei, M., Ren, L., et al. (2006). Optical image
processing using the photoinduced anisotropy of pyrrylfulgide. Journal of the Optical
Society of America. A, 23, 267271. http://www.opticsinfobase.org/josaa/abstract.cfm?
URIjosaa-23-2-267.
Miao, Q. G., Shi, C., Xu, P. F., Yang, M., & Shi, Y. b. (2011). A novel algorithm of image
fusion using shearlets. Optics Communications, 284, 15401547. http://dx.doi.org/
10.1016/j.optcom.2010.11.048.
Mie, G. (1908). Beitrage zur optik tr osungen. Annalen
uber medien, speziell kolloidaler metall
der Physik, 330, 377445. http://dx.doi.org/10.1002/andp.19083300302.
Miller, P. C., & Caprari, R. S. (1999). Demonstration of improved automatic target-

recognition performance by moment analysis of correlation peaks. Applied Optics, 38,
Miller, D. A., & Dereniak, E. L. (2012). Selective polarization imager for contrast enhance-
ments in remote scattering media. Applied Optics, 51, 40924102. http://www.
Mosso, F., Barrera, J. F., Tebaldi, M., Bolognini, N., & Torroba, R. (2011). All-optical
encrypted movie. Optics Express, 19, 5706. http://dx.doi.org/10.1364/OE.19.013779.
Mullen, L., Cochenour, B., Rabinovich, W., Mahon, R., & Muth, J. (2009). Backscatter
suppression for underwater modulating retroreflector links using polarization discrimi-
nation. Applied Optics, 48, 328337. http://www.opticsinfobase.org/ao/abstract.cfm?
URIao-48-2-328.
Nakajima, N. (2013). Coherent diffractive imaging beyond the Fresnel approximation using
a deterministic phase-retrieval method with an aperture-array filter. Applied Optics, 52,
C1C10.
Naughton, T. J., Frauel, Y., Javidi, B., & Tajahuerce, E. (2002a). Compression of interference
digital holograms for three-dimensional object reconstruction and recognition. Applied
20-4124.
Naughton, T. J., Frauel, Y., Matoba, O., Javidi, B., & Tajahuerce, E. (2002). Compression of
digital holograms for three-dimensional video. In B. Javidi, & F. Okano (Eds.), Three-
dimensional television, video, and display technologies (pp. 273295): Berlin, Germany:
Springer-Verlag.
Naughton, T. J., Joh, J., McDonald, B., & Javidi, B. (2003). Efficient compression of Fresnel
fields for Internet transmission of three-dimensional images. Applied Optics, 42, 47584764.
Nehmetallah, G., & Banerjee, P. P. (2012). Applications of digital and analog holography in
three-dimensional imaging. Advances in Optics and Photonics, 4, 472553. http://www.
opticsinfobase.org/aop/abstract.cfm?URIaop-4-4-472.
Neji, N., Jridi, M., Alfalou, A., & Masmoudi, N. (2013). A CABAC codec of H.264AVC
with secure arithmetic coding. In Real-time image and video processing 2013: Vol. 8656.
Proceedings of SPIE (p. 86560G). USA. http://dx.doi.org/10.1117/12.2006351.
Ochoa, E., Goodman, J. W., & Hesselink, L. (1985). Real-time enhancement of defects in a
periodic mask using photorefractive Bi12SiO20. Optics Letters, 10, 430432. http://www.
Okamoto, T., Yamaguchi, I., & Yamagata, K. (1997). Real-time enhancement of defects in
periodic patterns by use of a bacteriorhodopsin film. Optics Letters, 22, 337339. http://
Oppenheim, A. V., & Lim, J. S. (1981). The importance of phase in signals. Proceedings of the
IEEE, 69(5), 529541.
Oppenheim, A. V., Schafer, R. W., & Buck, J. R. (1999). Discrete-time signal processing
(2nd ed.). Upper Saddle River, NJ: Prentice-Hall Signal Processing Series.
Ouerhani, Y. (2012). Contribution a` la definition, a` loptimisation et a` limplantation dIP de
traitement du signal et des donnees en temps reel sur cibles programmables. Ph.D. thesis,
France: ISEN-Brest, Univesite de Bretagen Occidentale, Brest, France.
Ouerhani, Y., Jridi, M., & Alfalou, A. (2011). Implementation techniques of high-order FFT
into low-cost FPGA. In 2011 IEEE 54th international midwest symposium on circuits and
systems (MWSCAS) (pp. 14). http://dx.doi.org/10.1109/MWSCAS.2011.6026390.
Ouerhani, Y., Jridi, M., & Alfalou, A. (2012). Area-delay efficient architecture using parallel
processing and new memory sharing technique. Journal of Circuits, Systems, and Computers,
21, 1240018. http://dx.doi.org/10.1142/S021812661240018X.
Ouerhani, Y., Jridi, M., Alfalou, A., & Brosseau, C. (2013). Graphics processor unit imple-
mentation of correlation technique using a segmented phase only composite filter. Optics
Communications, 289, 3344.
Papoulis, A. (1962a). The Fourier integral and its applications. New York: McGraw-Hill.
Pati, G. S., Unnikrishnan, G., & Singh, K. (1998). Multichannel image addition and subtrac-
tion using joint-transform correlator architecture. Optics Communications, 150, 3337.
Paturzo, M., Memmolo, P., Miccio, L., Finizio, A., Ferraro, P., Tulino, A., et al. (2008a).
Numerical multiplexing and demultiplexing of digital holographic information for
remote reconstruction in amplitude and phase. Optics Letters, 33, 26292631. http://
Paz, S., Zlotnik, A., & Zalevsky, Z. (2013). Extending the depth of focus in tomography
systems for glass lattice three dimensional mapping. Applied Optics, 52, C50C57.
http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C50.
Pennebaker, W. B., & Mitchell, J. L. (1993). JPEG: Still image data compression standard.
New York: International Thompson Publishing.
Phong, B. (1975). Illumination for computer generated pictures. Communications of the ACM,
18, 311317. http://dl.acm.org/citation.cfm?id360839.
Piederrie`re, Y., Boulvert, F., Cariou, J., Le Jeune, B., Guern, Y., & Le Brun, G. (2005).
Backscattered speckle size as a function of polarization: Influence of particle-size and
-concentration. Optics Express, 13, 50305039. http://www.opticsinfobase.org/oe/
abstract.cfm?URIoe-13-13-5030.
Piestun, R., & Miller, D. A. B. (2000). Electromagnetic degrees of freedom of an optical
system. Journal of the Optical Society of America. A, 17, 892902.
Qian, Y., Hu, F., Cheng, X., & Jin, W. (2011). Real-time image deblurring by optoelec-
tronic hybrid processing. Applied Optics, 50, 61846188.
Quatieri, T., Jr., & Oppenheim, A. V. (1981). Iterative techniques for minimum phase signal
reconstruction from phase or magnitude. IEEE Transactions on Acoustics, Speech, and Signal
Processing, 29, 11871193.
Rajput, S. K., & Nishchal, N. K. (2013). Known-plaintext attack-based optical cryptosystem
using phase-truncated Fresnel transform. Applied Optics, 52, 871.
Rasouli, S., Ghasemi, H., Tavassoly, M. T., & Khalesifard, H. R. (2011). Application of
parallel moire deflectometry and the single beam Z-scan technique in the measure-
ment of the nonlinear refractive index. Applied Optics, 50, 23562360. http://www.
Refregier, P., & Javidi, B. (1995). Optical image encryption based on input plane and Fourier
plane random encoding. Optics Letters, 20, 767.
Rehn, H., & Kowarschik, R. (1998). Real-time non-linear spatial filtering with a leaky
OASLM. Optics & Laser Technology, 30, 3947. http://www.sciencedirect.com/sci
ence/article/pii/S0030399298000115.
Richardson, W. H. (1972). Bayesian-based iterative method of image restoration. Journal of
the Optical Society of America, 62, 55.
Richardson, E. G. (2003). H.264 and MPEG-4 video compression: Video coding for next-
generation multimedia. Hoboken, NJ (USA): John Wiley & Sons Ltd.
Riris, H., Carlisle, C. B., Warren, R. E., & Cooper, D. E. (1994). Signal-to-noise ratio
enhancement in frequency-modulation spectrometers by digital signal processing. Optics
Rizzi, A., Gatta, C., & Marini, D. (2004). From retinex to automatic color equalization:
Issues in developing a new algorithm for unsupervised color equalization. Journal of Elec-
tronic Imaging, 13, 7584.
Rolando, S., Goiffon, V., Magnan, P., Corbie`re, E., Molina, R., Tulet, M., et al. (2013).
Smart CMOS image sensor for lightning detection and imaging. Applied Optics, 52,
C16C23. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C16.
Rosen, J., Katz, B., & Brooker, G. (2009). Fresnel incoherent correlation hologramA
review. Chinese Optics Letters, 7, 11341141. http://www.opticsinfobase.org/col/
abstract.cfm?URIcol-7-12-1134.
Russ, J. C. (2002). The image processing hand book (4th ed.). USA: CRC Press.
Ryle, J. P., McDonnell, S., Glennon, B., & Sheridan, J. T. (2013). Calibration of a digital
in-line holographic microscopy system: Depth of focus and bioprocess analysis. Applied
Optics, 52, C78C87.
Sabbah, S., Lerner, A., Erlick, C., & Shashar, N. (2005). Underwater polarization visionA
physical examination. In S. G. Pandalai (Ed.), Recent research and developments in experimen-
tal & theoretical biology (pp. 123176). Kerala, India: Transworld Research Network.
Sahin, A., Ozaktas, H. M., & Mendlovic, D. (1995). Optics Communications, 120, 128134.
Saini, N., & Sinha, A. (2010). Key management of the double random-phase-encoding
method using public-key encryption. Optical Laser Engineering, 48, 329334.
Salomon, D. (2004). Data compression: The complete reference (3rd ed.). New York: Springer.
Sayood, S. (2012). Introduction to data compression. Morgan Kaufmann series in multimedia
information and systems. (4th ed.): Elsevier.
Scharf, L. L. (1991). Statistical signal processing: Detection, estimation, and time series analysis.
Prentice Hall.
Schechner, Y., & Karpel, N. (2005). Recovery of underwater visibility and structure by
polarization analysis. IEEE Journal of Oceanic Engineering, 30, 570587. http://dx.doi.
org/10.1109/JOE.2005.850871.
Sendhil, K., Vijayan, C., & Kothiyal, M. P. (2005). Spatial phase filtering with a porphyrin
derivative as phase filter in an optical image processor. Optics Communications, 251,
292298. http://dx.doi.org/10.1016/j.optcom.2005.03.014.
Shabtay, G., Mendlovic, D., & Zalevsky, Z. (1998). Optical implementation of the contin-
uous wavelet transform. Applied Optics, 37, 2964. http://www.opticsinfobase.org/ao/
Shamir, A. (1985). On the security of DES. In Lecture notes in computer science: Vol. 218.
Advances in cryptologyCrypto85 (p. 280). New York: Springer-Verlag.
Shih, F., & Mitchell, O. (1989). Threshold decomposition of gray-scale morphology
into binary morphology. IEEE Transactions in Pattern Analysis Machine Intelligence, 11,
3142.
Shih, M. Y., Shishido, A., Chen, P. H., Wood, M. V., & Khoo, I. C. (2000). All-optical
image processing with a supranonlinear dye-doped liquid-crystal film. Optics Letters,
25, 978980. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-25-13-978.
Shih, M. Y., Shishido, A., & Khoo, I. C. (2001). All-optical image processing by means of a
photosensitive nonlinear liquid-crystal film: Edge enhancement and image addition-
subtraction. Optics Letters, 26, 11401142.
Shilemay, M., Rozban, D., Levanon, A., Yitzhaky, Y., Kopeika, N. S., Yadid-Pecht, O.,
et al. (2013). Performance quantification of a millimeter-wavelength imaging system
based on inexpensive glow-discharge-detector focal-plane array. Applied Optics, 52,
C43C49. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C43.
Shoop, B. L., Pezeshki, B., Goodman, J. W., & Harris, J. S., Jr. (1992). Noninterferometric
optical subtraction using reflection-electroabsorption modulators. Optics Letters, 17,
5860.
Simova, E. S., & Stoev, K. N. (1992). Phase-stepping automatic fringe analysis in holographic
moire. Applied Optics, 31, 59655974. http://www.opticsinfobase.org/ao/abstract.cfm?
URIao-31-28-5965.
Sizikov, V. S., & E kzemplyarov, R. A. (2013). Operating sequence when noise is being
filtered on distorted images. Journal of Optical Technology, 80(1), 2834.
Skauli, T. (2013). Information capacity as a figure of merit for spectral imagers: The trade-off
between resolution and coregistration. Applied Optics, 52, C58C63. http://www.
opticsinfobase.org/ao/abstract.cfm?URIao-52-7-C58.
Slawecki, D., & Li, W. (1992). DCT/IDCT processor design for high data rate image coding.
IEEE Transactions on Circuits and Systems for Video Technology, 2, 135.
Smith, R. A. (1993). In A. G. Haus, & M. J. Yaffe (Eds.), A categorical course in physics, technical
aspects of breast imaging. Oak Brook, IL: Radiological Society of North America.
Smith, W. E., & Barrett, H. H. (1983). Radon transform and bandwidth compression.
Optics Letters, 8, 395397. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-8-
7-395.
Sogandares, F. M., & Fry, E. S. (1997). Absorption spectrum (340640 nm) of pure water.
I. Photothermal measurements. Applied Optics, 36, 86998709. http://www.
Sorensen, H., Heideman, M., & Burrus, C. (1986). On computing the split-radix FFT. IEEE
Transactions on Acoustics, Speech, and Signal Processing, 34, 152.
Stern, A. (2006). Sampling of linear canonical transformed signals. Signal Processing, 86,
14211425.
Strang, G. (1999). The discrete cosine transform. SIAM Review, 41, 135.
Sungwook, Y., & Swartziander, E. (2001). DCT implementation with distributed arith-
metic. IEEE Transactions on Computers, 50, 985.
Tahara, T., Ito, K., Kakue, T., Fujii, M., Shimozato, Y., Awatsuji, Y., et al. (2010). Parallel
phase-shifting digital holographic microscopy. Biomedical Optics Express, 1, 610.
Testorf, M., & Lohmann, A. W. (2008). Holography in phase space. Applied Optics, 47,
A70A77. http://www.opticsinfobase.org/ao/abstract.cfm?URIao-47-4-A70.
Thoma, R., Oesterhelt, D., Hampp, N., & Brauchle, C. (1991). Bacteriorhodopsin films as
spatial light modulators for nonlinear-optical filtering. Optics Letters, 16, 651653. http://
Tisse, C.-L., Guichard, F., & Cao, F. (2008). Does resolution really increase image quality?
Proceedings of SPIE, 6817, 68170Q.
Torres-Mendez, L., & Dudek, G. (2005). Color correction of underwater images for aquatic
robot inspection. In A. Rangarajan, B. Vemuri, & A. L. Yuille (Eds.), Energy minimization
methods in computer vision and pattern recognition (pp. 6073): Berlin/Heidelberg: Springer,
ISBN: 978-3-540-30287-2. http://link.springer.com/chapter/10.1007%2F11585978_5.
Treibitz, T., & Schechner, Y. Y. (2009). Active polarization descattering. IEEE Transactions
on Pattern Analysis and Machine Intelligence, 31, 385399.
Tu, B., Liu, L., Liu, Y., Jin, Y., & Tang, J. (2013). High precision two-step calibration
method for the fish-eye camera. Applied Optics, 52, C37C42.
Tyo, J. S., Goldstein, D. L., Chenault, D. B., & Shaw, J. A. (2006). Review of passive
imaging polarimetry for remote sensing applications. Applied Optics, 45, 54535469.
http://www.opticsinfobase.org/ao/abstract.cfm?URIao-45-22-5453, https://www.
novapublishers.com/catalog/product_info.php?products_id28370.
Vander Lugt, A. (1964). Signal detection by complex spatial filtering. IEEE Transactions on
Information Theory, IT-10, 139.
Vest, C. M. (1974). Formation of images from projections: Radon and Abel transforms. Jour-
nal of the Optical Society of America, 64, 12151218.
Voelz, D. G. (2011). Computational Fourier optics: A MATLAB tutorial, SPIE tutorial texts
TT89. Bellingham, WA: SPIE Press.
Voss, K. J., & Fry, E. S. (1984). Measurement of the Mueller matrix for ocean water. Applied
23-4427.
Walker, J. G., Chang, P. C. Y., & Hopcraft, K. I. (2000). Visibility depth improvement in
active polarization imaging in scattering media. Applied Optics, 39, 49334941. http://
Wang, Q. (2012). Optical image encryption with silhouette removal based on interference
and phase blend processing. Optics Communications, 285, 4294.
Wang, Q., Cheng, D., Wang, Y., Hua, H., & Jin, G. (2013). Design, tolerance, and fabri-
cation of an optical see-through head-mounted display with free-form surface elements.
Applied Optics, 52, C88C99.
Wang, Q., Guo, Q., & Lei, L. (2013). Asymmetric multiple-image hiding using phase
retrieval technique based on amplitude- and phase-truncation in fractional Fourier
domain. Optik, 124, 38983902. http://dx.doi.org/10.1016/j.ijleo.2012.12.018.
Wang, Q., Liu, L., Liang, F., & Wang, Z. (1995). Optical cellular logic two-level networks
for morphological image processing. Optics Communications, 115, 251255. http://dx.
doi.org/10.1016/0030-4018(94)00700-5.
Weaver, C. S., & Goodman, J. W. (1966). A technique for optically convolving two func-
tions. Applied Optics, 5, 1248.
Wells, W. H. (1969). Loss of resolution in water as a result of multiple small-angle scattering.
Journal of the Optical Society of America, 59, 686691. http://www.opticsinfobase.org/josa/
abstract.cfm?URIjosa-59-6-686.
Westfeld, A., & Pfitzmann, A. (2000). Attacks on steganographic systems. In A. Pfitzmann
(Ed.), Information hiding Third international workshop, LNCS 1768 (p. 61). Berlin,
Heidelberg: Springer-Verlag.
Westfeld, A., & Wolf, G. (1998). Steganography in a video conferencing system.
In D. Aucsmith (Ed.), Information hiding LNCS 1525. Berlin, Heidelberg: Springer-
Verlag.
White, S. (1989). Applications of distributed arithmetic to digital signal processing: A tutorial
review. IEEE ASSP Magazine, 6, 4.
Wiener, N. (1949). Extrapolation, interpolation, and smoothing of stationary time series.
New York: Wiley, ISBN: 0-262-73005-7,
Wolf, E. (2007). Introduction to the theory of coherence and polarization of light. Cambridge:
Cambridge University Press.
Xia, P., Shimozato, Y., Tahara, T., Kakue, T., Awatsuji, Y., Nishio, K., et al. (2013a).
Image reconstruction algorithm for recovering high-frequency information in parallel
phase-shifting digital holography [Invited]. Applied Optics, 52, A210A215. http://
www.opticsinfobase.org/ao/abstract.cfm?URIao-52-1-A210.
Xuan, N. P., Ferrier, J. L., Gazengel, J., Rivoire, G., Breki-Iovskhikh, G. L.,
Kudriavtseva, A. D., et al. (1988). Changes in the space structures of light beams induced
by nonlinear optical phenomena: Application to phase contrast and image processing.
Optics Communications, 68, 244250. http://www.sciencedirect.com/science/article/
pii/0030401888903938.
Yamaguchi, I., & Zhang, T. (1997). Phase-shifting digital holography. Optics Letters, 22,
12681270. http://www.opticsinfobase.org/ol/abstract.cfm?URIol-22-16-1268.
Yaroslavsky, L. P. (2012). Theoretical foundations of digital imaging using MATLAB. Boca
Raton, FL: CRC Press Taylor & Francis Group.
Yelleswarapu, C. S., Kothapalli, S.-R., & Rao, D. V. G. L. N. (2008). Optical Fourier tech-
niques for medical image processing and phase contrast imaging. Optics Communications,
281, 18761888.
Yelleswarapu, C. S., Kothapalli, S. R., Vaillancourt, Y., Aranda, F. J., Kimball, B., &
Rao, D. V. G. L. N. (2006). Phase contrast imaging using photo-thermally induced phase
transitions in liquid crystals. Applied Physics Letters, 89, 2111161. http://dx.doi.org/
10.1063/1.4737942.
Yelleswarapu, C., Wu, P., Kothapalli, S. R., Rao, D. V. G. L. N., Kimball, B.,
Sai, S. S., et al. (2006). All-optical spatial filtering with power limiting materials.
Optics Express, 14, 14511457. http://www.opticsinfobase.org/oe/abstract.cfm?
URIoe-14-4-1451.
Yokozeki, S., & Ohnishi, K. (1975). Spherical aberration measurement with shearing inter-
ferometer using Fourier imaging and moire method. Applied Optics, 14, 623627. http://
Yong-Liang, X., Xin, Z., Sheng, Y., & Yao-Yao, C. (2010). Multiple-image parallel optical
encryption. Optics Communication, 283, 27892793. http://dx.doi.org/10.1016/
j.optcom.2010.03.003.
You, Y., Tonizzo, A., Gilerson, A. A., Cummings, M. E., Brady, P., Sullivan, J. M., et al.
(2011). Measurements and simulations of polarization states of underwater light in clear
oceanic waters. Applied Optics, 50, 48734893. http://www.opticsinfobase.org/ao/
Yu, F. T. S., & Jutamulia, S. (1998). Optical pattern recognition. Cambridge (UK): Cambridge
University Press.
Yu, F. T. S., & Tai, A. (1979). Incoherent image addition and subtraction: A technique.
Applied Optics, 18, 27052707.
Yuan, S., Jin, G., Wu, M., & Yan, Y. (1996). Neighborhood operation binary image algebra
for optical morphological image processing. Optics Communication, 123, 705715. http://
dx.doi.org/10.1016/0030-4018(95)00317-7.
Zalevsky, Z. (2010). Extended depth of focus imaging: A review. SPIE Reviews, 1, 018001.
Zalevsky, Z., Gaffling, S., Hutter, J., Chen, L., Iff, W., Tobisch, A., et al. (2013). Passive
time-multiplexing super-resolved technique for axially moving targets. Applied Optics,
52, C11C15.
Zalevsky, Z., & Mendlovic, D. (1996). Fractional Wiener filter. Applied Optics, 35, 3930.
http://www.opticsinfobase.org/ao/abstract.cfm?URIao-35-20-3930.
Zalevsky, Z., Shemer, A., Zlotnik, A., Ben Eliezer, E., & Marom, E. (2006). All-optical axial
super resolving imaging using a low-frequency binary-phase mask. Optics Express, 14,
26312643.
Zhang, Y., Pedrini, G., Osten, W., & Tiziani, H. J. (2004). Reconstruction of in-line digital
holograms from two intensity measurements. Optics Letters, 29, 17871789.
Zhao, L., Healy, J. J., & Sheridan, J. T. (2013). Unitary discrete linear canonical transform:
Analysis and application. Applied Optics, 52, C30C36. http://www.opticsinfobase.org/
ao/abstract.cfm?URIao-52-7-C30.
CHAPTER THREE
Precursors and Dispersive Pulse

Dynamics: A Century After
the SommerfeldBrillouin
Theory: Part II. The Modern
Asymptotic Theory
Natalie A. Cartwright*, Kurt E. Oughstun
*Department of Mathematics, State University of New York, New Paltz, New York, USA
College of Engineering and Mathematics, University of Vermont, Burlington, Vermont, USA
Contents
1. Loudon's Energy Velocity 264
2. Analysis of the Phase Function and Its Saddle Points 267
2.1 Numerical Investigation 268
2.2 Location of the Saddle Points and Approximation of the Phase Function 274
3. Uniform Asymptotic Expansions 288
3.1 Procedure for the Asymptotic Analysis of the Propagated Signal 289
3.2 The Sommerfeld Precursor Field 295
3.3 The Brillouin Precursor 300
3.4 The Signal Contribution 308
3.5 The Total Field 312
3.6 The Signal Arrival and the Signal Velocity 312
4. Extensions of the SommerfeldBrillouin Theory to the Singular and Weak
Dispersive Limits 319
4.1 Dispersive Signal Propagation in the Singular Dispersion Limit 320
4.2 Dispersive Signal Propagation in the Weak Dispersion Limit 323
5. Pulsed Plane-Wave Transmission Through an Interface 327
5.1 Formulation 329
5.2 Asymptotic Description 332
5.3 Transmitted Signal Evolution 338
6. Concluding Remarks 341
Acknowledgments 343
References 343

264 Natalie A. Cartwright and Kurt E. Oughstun
1. LOUDON'S ENERGY VELOCITY

The modern asymptotic theory of the SommerfeldBrillouin problem
has its origin in the classical theory presented in Part I of this two-part
review, and was initiated, in part, with the publication of the first correction
to Brillouins analysis of the energy transport velocity in a Lorentz model
dielectric (see Section 2.2.2 of Part I) by Loudon1 in 1970. As stated by
Loudon (1970), the true energy velocity vE for an electromagnetic wave
can be defined as the rate of energy flow, determined by the Poynting vec-
tor, divided by the stored energy density of the wave... The Poynting vector
is easily calculated, and the significant step in deriving vE is the calculation of
the energy density associated with the wave. A previous calculation of vE by
Brillouin (1960) gives an incorrect result for the energy density, and more
recently Neufeld (1970a, 1970b) has suggested that classical dispersion
theory is incapable of producing a satisfactory expression for the energy den-
sity... an acceptable expression for the energy density can in fact be obtained.
When the frequency of the electromagnetic wave is close to the oscillator
frequency of the dielectric, a substantial part of the energy of the wave resides
in the excited oscillators. This part of the energy must be added to the
electromagnetic field energy to obtain a correct expression for the energy
density.
The analysis begins with the differential form of Poyntings theorem for a
dispersive dielectric (Oughstun, 2006; Stratton, 1941)

1 @H @E @P
rS H +E + 4E (1)
4 @t @t @t
in cgs units, where

c
S EH (2)
4
1
Rodney Loudon received his PhD in physics from the University of Oxford in 1958, a Postdoctoral
Fellow at the University of California at Berkeley from 1959 to 1960, and finally professor at the
University of Essex. He is a Fellow of the Royal Society and was awarded the Max Born Award
in 1992.
Part II. The Modern Asymptotic Theory 265
is the Poynting vector and where P Nqer is the macroscopic polarization

density of the simple polarizable dielectric [see Equation (1.3) in Part I],
where [cf. Equation (1.20) in Part I]
qe =m
r E (3)
20 + 2i0
2
is the (phasor) solution of Equation (1.19) in Part I. The last term in

Equation (1) is then given by
@P 1 @ 2
E mN _ + 20 jrj2 + 20 mNjrj
jrj _ 2, (4)
@t 2 @t
which is the sum of a perfect time-differential and a term in 0 corresponding
to dissipation in the dielectric. Loudon (1970) then defines the electromag-
netic energy density stored in both the field and the medium as
1 2 1 2
U tot U em + U rev jEj + jHj2 + mN jrj
_ + 20 jrj2 , (5)
8 2
where
1 2
U em jEj + jHj2 (6)
8
is the energy density stored in the electromagnetic field alone, and where
1 2
_ + 20 jrj2
U rev mN jrj (7)
2
is the reversible energy density stored in the Lorentz medium. With this
identification, Poyntings theorem (1) becomes
_ 2 U_ tot :
r S + 20 mNjrj (8)
Integration of this expression over a region R of space bounded by the sur-
face S with outward normal n^ followed by application of the divergence
theorem then yields
I Z Z
2 3 d
S nd
^ r + 20 mN jrj
2
_ d r U tot d3 r, (9)
S R dt R
which expresses the conservation of energy in the dispersive Lorentz model

dielectric. The two terms on the left-hand side represent the rate of energy
loss in the region R through flow across its surface S and by dissipation in the
medium contained inside S, respectively, and the integral on the right-hand

side represents the rate of change of the total electromagnetic energy stored
within the region R.
The time-average value of the magnitude of the Poynting vector S is
given by
c
S nr jEj2 (10)
8
and the time-average value of the electromagnetic energy density U tot is
given by

1
U tot nr ni + nr jEj2 ,
2
(11)
8 0
where the expressions (Loudon, 1970)
b2 2 20
n2r n2i 1 , (12)
2 20 2 + 420 2
b2 0
nr ni (13)
2 20 2 + 420 2
relating the real and imaginary parts of the complex index of refraction
n() nr() + ini() for a Lorentz model dielectric have been used in deriv-
ing these expressions. The time-average energy velocity in a single-resonance
Lorentz model dielectric is then given by the ratio of these two time-averaged
quantities as
S c
vE ,
U tot ni (14)
nr +
0
which is Loudons classic result. Because nr + ni

0
1 for all real 0,
this energy velocity yields results that are in agreement with the special theory
of relativity. Furthermore, this expression reduces to that obtained for
the group velocity in a single-resonance Lorentz model dielectric when
the frequency is bounded away from the region of anomalous dispersion
about the resonance frequency 0, as seen in Figure 1. Notice further that
the energy velocity attains a minimum near the medium resonance frequency
0, in agreement with Baerwalds description of the signal velocity (see
Figure 8 in Part I).
2.5
Relative wave velocity, v/c
1.5
vp /c
1 vg /c
0.5
vE /c
0
0 0.2 w0 0.6 0.8 1 1.2 1.4 1.6 1.8 2
w r/s 1017
Figure 1 Comparison of the frequency dependence of the relative phase vp()/c
(dashed curve), relative group vg()/c (dotted curve), and relative energy vE()/c (solid
curve) velocities in a single-resonance Lorentz model dielectric with medium parame-
p
ters 0 4 1016 r/s, b 20 1016 r=s, and 0 0.28 1016 r/s.
Although the energy velocity has been derived only for the special case of
a time-harmonic wave field, it does find direct application to the more gen-
eral case of an electromagnetic pulse. In particular, Oughstun and Sherman
(1980) and Sherman and Oughstun (1981, 1995) have derived a physical
description of the precursor fields in a Lorentz model dielectric that is based
solely upon Loudons energy velocity and the attenuation of the dispersive
medium. This physical description then represents a generalization of the
group velocity description to dispersive attenuative media.
2. ANALYSIS OF THE PHASE FUNCTION AND ITS

SADDLE POINTS
The asymptotic expansions obtained by Sommerfeld (1914) and
Brillouin (1914), as presented in Part One, have limited regions of validity.
Although these limitations were noted by Brillouin (1960), he was unable to
provide uniform asymptotic expansions for the two forerunners and the sig-
nal that remain valid for all 1. Evaluation of the integral representation of
the propagated pulse [Eq. (1.58) in Part I] remained at this point until early
1974. At that time, a complete reinvestigation of the SommerfeldBrillouin
theory was initiated by Oughstun2 and Sherman3 . Together, they found
more accurate expressions for the saddle point locations and applied newly
developed uniform asymptotic methods that enabled them to provide uni-
form asymptotic expansions for each of the three components of the prop-
agated field. This research relied upon the then recently published theorem
due to Olver (1970) which proved that the asymptotic expansion provided
by the method of steepest descents was independent of the steepest descent
path provided (among other things) that the path through the relevant saddle
point was inside the region with greater exponential decay than that at that
saddle point. This relaxation from the earlier steepest descent requirement,
referred to here as the saddle point method, had far-reaching implications
with regard to the signal velocity. In this section, we present the analysis
of the phase function and its saddle points as determined by Oughstun
and Sherman. Application of these results to well-defined uniform asymp-
totic methods is presented in the following section.
2.1 Numerical Investigation

The analysis of the behavior of the complex index of refraction n() and the
complex phase function (, ) i(n() ) remains unchanged from that
given by Brillouin (1914, 1960), as described in Part I, Section 2 of this review.
This simple sketch was first completed by Oughstun (1978) through
computer-generated isotimic contour plots of (, ) Re{(, )} in
the right-half of the complex -plane at several fixed spacetime points
using Brillouins choice of the medium parameters, reproduced in
Figures 27 for several values of the spacetime parameter increasing away
from unity.
In each of these contour plots, those isotimic contours of (, ) that are
greater than or equal to the value at the dominant saddle point [the saddle
2
Kurt E. Oughstun received his PhD in optics in 1978 from The Institute of Optics under the super-
vision of George C. Sherman. His thesis Propagation of Optical Pulses in Dispersive Media forms the basis
for Chapters 59 of the book by Oughstun and Sherman (1994). He is a Fellow of the Optical Society
of America.
3
George C. Sherman received his PhD from UCLA in 1969 under the supervision of Zdenek Sekera
and Emil Wolf (University of Rochester). He then joined The Institute of Optics at the University of
Rochester and was a faculty member there until 1978. He is a Fellow of the Optical Society of America.
w q=1
(1016 r/s)
5
SPn+
(1.2 1016)
w0 5
0
w+ w + 10 15 w
(0) (1016 r/s)
SPn
5 (+1.2 1016)
Figure 2 Isotimic contours of the real part (, ) Re{(, )} of the complex phase
function in the right-half of the complex -plane at the fixed spacetime point 1.
Notice that the distant saddle points SP
d are symmetrically located at 1 20 i at this
luminal spacetime point. Reprinted with kind permission of Springer Science + Business
Media from Oughstun (2009), Fig. 12.4, p. 275.
q = 1.25
w
(1016 r/s)
(+1.0 1016)
5
SPn+
(1.0 1016)
w0 5 (0)
0
w+
10 15 (1016 r/s)
w+ SPd+
(+1.0 1016)
SPn
5
(1.0 1016)
function in the right-half of the complex -plane at the fixed spacetime point 1.25.
Reprinted with kind permission of Springer Science + Business Media from Oughstun (2009),
Fig. 12.5, p. 275.
q = q SB 1.33425
w
(1016 r/s)
5
(+1.0 1016)
SPn+ (1.0 1016)

w0 5
0 (0) w
w+ 10 15 (1016 r/s)
w+ SPd +

SPn (+1.0 1016)
(1.0 1016)
5
function in the right-half of the complex -plane at the fixed spacetime point
SB 1.33425 when the upper near and distant saddle points are of equal
dominance, i.e., (+SPn, SB) (
SPd, SB). Reprinted with kind permission of Springer
Science + Business Media from Oughstun (2009), Fig. 12.6, p. 276.
w q q1 1.501
(1016 r/s)
5
(+1.0 1016)
(1.0 1016)
w0 5 (0)
0 SPn+ w
SPn
w+ 10 15 (1016 r/s)
(0) w+ SPd+
(+1.0 1016)
(1.0 1016)
function in the right-half of the complex -plane at the fixed spacetime point 1.501
just prior to the coalescence of the two near saddle points SP n into a single second-
order saddle point when 1. Reprinted with kind permission of Springer Science
+ Business Media from Oughstun (2009), Fig. 12.7, p. 276.
q = 1.65
w
(1016 r/s)
5
(+1.0 1016)
(1.0 1016)
w0 5 (0)
w
0
SPn+ w+ 10 15 (1016 r/s)
w+ SPd+
(+1.0 1016)
(1.0 1016)
Reprinted with kind permission of Springer Science+Business Media from Oughstun (2009),
Fig. 12.8, p. 277.
q = 5.0
w
(1016 r/s)
(2.0 1016)
(3.4 1016) (0 )
0 w
w0 5 (1016 r/s)
w+
SPn+ w+ SPd+
(2.0 1016)
1 (3.4 1016)
Reprinted with kind permission of Springer Science+Business Media from Oughstun (2009),
Fig. 12.9, p. 278.
point with negative (, ) with the smallest absolute value] are indicated by
solid curves, whereas those contours which are less than the value at the
dominant saddle point are indicated by dashed curves. In addition, the
region of the complex -plane where (, ) is less than that at the dominant
saddle point has been shaded in light gray and the region where (, ) is less
than that at the nondominant saddle point, which is contained inside the
light gray-shaded region, is shaded in dark gray. Finally, the (, ) values
of the maximum and minimum isotimic contours, as well as the zero contour
(, ) 0, are indicated in each figure by the number contained in paren-
thesis that is either adjacent to or on top of the appropriate contour.
As predicted by Brillouin, for all values of the spacetime parameter 1
except one, there are four first-order saddle points of the complex phase
function (, ) for a single-resonance Lorentz model dielectric, symmetri-
cally located about the imaginary axis. Two of these saddle points, called near
saddle points, reside in the region jj 0 about the origin of the complex
-plane and the other two, called distant saddle points, reside in the region
jj 1 of the complex -plane that is removed from the origin.
The two near first-order saddle points SP n are seen in Figures 25 to lie
along the imaginary axis over the initial spacetime domain 2 [1, 1],
approaching each other as increases to 1, the upper near saddle point
p
SP+n crossing the origin at 0, where 0 1 + b2 =20 1:5 for
Brillouins choice of the medium parameters, and then coalescing into a sin-
gle second-order saddle point at 1, where the numerical value of this
critical spacetime point for Brillouins choice of the medium parameters is
just slightly larger than the spacetime value 1.501 illustrated in Figure 5.
As increases above 1, the two near first-order saddle points separate such
that they are symmetrically situated about the imaginary axis, approaching
the inner branch points , respectively, as ! 1.
The two distant saddle points SP d , on the other hand, are located in the
lower-half of the complex -plane for all 1 and are located at 1 i0
at the luminal spacetime point 1. As increases from unity, these two
saddle points symmetrically move in from infinity and approach the respec-
tive outer branch points 0 as increases to infinity, as evident in
Figures 27.
Initially, the distant saddle points SP d have less exponential decay asso-
ciated with them than does the upper near saddle point SP+n , as seen in
Figures 2 and 3, so that there exists some spacetime point SB such that

SPd , > SPn , when 1 < SB :
+
(15)
Because the original contour of integration C appearing in the integral rep-

resentation of the propagated plane wave signal given in Equation (1.58) in
Part I is not deformable through the lower near saddle point SP n over the
initial spacetime domain 1 < 1, that saddle point is irrelevant for the
present analysis for below and bounded away from 1. At the critical space
time point SB 1.33425, illustrated in Figure 4, the upper near saddle
point SP+n has precisely the same exponential decay associated with it as do
the two distant saddle points, so that

SPd , SB SPn , SB when SB :
+
(16)
Consequently, at SB, those three saddle points (SP+d , SP +
d , and SPn ) are
of equal importance (or dominance) in the asymptotic description of the
propagated wave field.
The remaining Figures 57 show that for values of 2 (SB, 1), the
upper near saddle point SP+n is dominant over the two distant saddle points
SPd , so that
SP
+
n
, >
SPd , when SB < < 1 : (17)
At 1, the two near first-order saddle points SP
n have coalesced into a
single second-order saddle point SPn which is dominant over the distant
saddle point pair, so that
SPn , 1 >
SPd , 1 when 1 : (18)
Finally, for all > 1, the near saddle points SP
n are dominant over the

distant saddle points SPd , so that

SPn , > SPd , when > 1 , (19)
as evident in Figures 67. Notice the change in scale of the real and imag-
inary coordinate axes in Figure 7, demonstrating how the topography of
0
(, ) becomes increasingly concentrated about the branch cuts
0
and ++ as increases above the critical spacetime value 1 with the near
saddle points SPn approaching the lower branch points , respectively, and
the distant saddle points SPd approaching the upper branch points .
0
These detailed numerical results demonstrate the necessity of obtaining

approximate analytic expressions for both the near and distant saddle point
locations that accurately describe their dynamical evolution in the complex
-plane for all 1 as well as the complex phase behavior at them. In
addition, accurate analytic expressions are needed for each of the critical
spacetime points encountered here. These include the critical value SB at

which the upper near and distant saddle points are of equal importance (see
Figure 4) and the critical value 1 at which the two near first-order saddle
points coalesce into a single second-order saddle point (see Figure 5).
2.2 Location of the Saddle Points and Approximation

of the Phase Function
In order to obtain the asymptotic expansion of the propagated signal AH(z, t)
for large values of the propagation distance z > 0, where AH(z, t) is given by
the FourierLaplace integral representation in Equation (1.58) in Part I, the
saddle points of the complex phase function (, ) must be located in the
complex -plane and the behavior of (, ) at these critical points deter-
mined as a function of 1. The condition that (, ) is stationary at a
saddle point is
0, in + in0 0, (20)
where the prime denotes differentiation with respect to . The resultant
saddle point equation is then given by
n + n0 : (21)
Approximate analytic expressions for AH(z,t) then require approximate ana-
lytic solutions of Equation (21) for the dynamical evolution of the saddle
point locations with , as well as for the complex phase behavior at each
of these saddle points.
For a single-resonance Lorentz model dielectric with complex index of
refraction
1=2
b2
n 1 2 , (22)
20 + 2i0
the saddle point Equation (21) becomes (Oughstun & Sherman, 1988)
2
b2 + i0
1 + 2i0 + 2
2 2
20 + 2i0 (23)
2
1 + 2i0 0 + 2i0 ,
2 2 2 2
where 21 20 + b2 . Notice that two exact roots of this saddle point

Equation (23) are readily obtained in the limit as ! 1. In that limit, either
2 20 + 2i0 0, yielding the roots
q
! 20 20 i0 , as ! 1, (24)
or n() 0, yielding the roots

q
0
! 21 20 i0 , as ! 1: (25)
Thus, the saddle points move into the branch points + and +0 in the right-
0
half plane and and in the left-half plane in the limit as ! 1.
An exact polynomial equation describing the location of the saddle
points can be obtained from Equation (23) as
2
1 8 + 8i0 2 1 7 4 20 + 620 2 1 + b2 2 6

2i0 1220 + 3b2 + 1620 2 1 5

+ 640 + 4820 + 2b2 20 + 12b2 20 + 1640 2 1

+ b2 20 2 1220
4

+ 4i0 640 + 4b2 20 + 820 20 + 420 b2 2 1
(26)
+ 20 + 220 220 2 b2
3

20 440 + 3b2 20 + 2420 20 + 1220 b2 2 1

b2 40 + 2020 20 + 920 b2
2

2i0 20 420 + 3b2 20 2 1 b2

+ 40 20 + b2 20 2 1 b2 0:
Because this eighth-order polynomial is extremely formidable, approximate
solutions of the saddle point Equation (23) were then developed for both the
distant and near saddle points that are a characteristic of a single-resonance
Lorentz model dielectric.
2.2.1 The Region Removed from the Origin (jj 1)

In order to obtain a more accurate description of the distant saddle point
locations than that given by Brillouin (1914, 1960), particularly for large
values of , the exact saddle point equation given in Equation (23) was first
rewritten by Oughstun and Sherman (1988) as
2
1 b2 + i0
1 + 2i0 + 2
2 2
2 2 20 + 2i0 20 + 2i0 (27)
2 21 + 2i0 :
This particular form of the saddle point equation explicitly displays the
desired limiting behavior as ! 1 that is given in Equation (25). With this
limiting behavior in mind, the rational function appearing in the squared
term of Equation (27) may be approximated as
b2 + i0 0 b2
b 2
i
2 20 + 2i0
provided that jj 0 and jj 0. With this substitution, the saddle point

equation given in Equation (27) becomes
2 2 0 b2 2 2 b4
20 + 2i0 2i 20 + 2i0 0 2

2 2
0 + 2i0 1 + 2i0 :
2 2 2
The term 20 b4 =2 may be neglected in comparison to the other two terms

on the left-hand side of this equation with the result

b2 2 0 b2
+ 2i0 0 + 2
3 2 2
+ 2i 2 0: (28)
1 1
The distant saddle point locations satisfying this cubic equation are found to
be expressed as
SPd i0 1 + (29)
with the approximate expressions (Oughstun, 2009)

r
b2 2
20 20 + 2 , (30a)
1
b2
2 : (30b)
0 20 2 1 + b2 2

Notice that b2 = 2 12 to this order of approximation.
For values of close to unity, the above expressions simplify to !
p
b and ! 1 so that this approximation reduces to Brillouins
21
approximation [cf. Equation (2.18) in Part I] in this limit. In particular, in
the limit as approaches unity from above
lim SPd 1 2i0 : (31)

!1 +
On the other hand, for sufficiently large values of , Equation (30a) and
q
(30b) have the limiting behaviors ! 21 20 and ! 0, so that
in the limit as approaches infinity
q
lim SPd 21 20 i0 0 , (32)
!1
and the distant saddle points SP d , respectively, approach the outer branch
0
points . This approximation to the distant saddle point locations (due
to Oughstun, 1978) then captures the exact limiting behavior in the two
opposite extremes at either 1 or 1. A sketch of the respective paths
followed by these two distant saddle points in the complex -plane is
presented in Figure 8.
The approximate local behavior of the phase function (, ) about
the distant saddle point SP+d in the right-half of the complex -plane, where
SP+d () + rei, is given by Oughstun and Sherman (1988) as
w branch w d w+ branch w+
cut cut
+
SPd SPd
2d
Figure 8 Behavior of the distant saddle points SP d in the complex -plane for a single-
resonance Lorentz model dielectric. The dotted curves indicate the respective directed
paths that these saddle points follow as increases to infinity. The dashed lines through
each saddle point indicate the local behavior of the isotimic contour (, ) (SPd , )
through that saddle point, the shaded region indicates the local region about each sad-
dle point where the inequality (, ) < (SPd , ) is satisfied, and the arrows indicate
local direction of ascent along the lines of steepest descent and ascent through each
saddle point. Reprinted with kind permission of Springer Science+Business Media from
Oughstun (2009), Fig. 12.39, p. 325.
r,, 1 i + 0 1 + + irei
b2 i0 1 + rei
i ,
2 2 + 20 1 + 2r cos + 20 1 r sin + r 2
(33)
with real part
r,, 1 0 1 + r sin
b2 0 1 + r sin
:
2 2 + 20 1 + 2r cos + 20 1 r sin + r 2
(34)
It is then seen that (r, , ) attains its maximum variation with r about the
distant saddle point SP+d when /4, 3/4, 5/4, 7/4. Consequently,
in the right-half plane, the lines of steepest descent through the distant sad-
dle point SP+d are at the angles 3/4 and 7/4, and the lines of
steepest ascent are at /4 and 5/4. Because of the even symmetry
of (, ) about the imaginary axis, the reverse holds true for the distant
saddle point SP d in the left-half plane. This local behavior about the distant
saddle points SP d is depicted in Figure 8, where the vectors indicate the
direction of ascent along the lines of steepest descent and ascent through
each respective distant saddle point, which are at angles of 45 to the coor-
dinate axes, and the shaded areas indicate the local regions about each saddle
point wherein the inequality (, ) < (SPd , ) is satisfied and in which
the path of steepest descent from the respective saddle point lies. These
results are in complete agreement with the numerical results presented in
Figures 27.
2.2.2 The Region About the Origin (jj 0)

In order to obtain a more accurate description of the near saddle point
locations than that given by Brillouin (1914, 1960), particularly for
> 1, the exact saddle point equation given in Equation (23) was rewritten
by Oughstun and Sherman (1988) as
+ i0
2 2 20 + 2i0 2 21 + 2i0 + 2b2
2 20 + 2i0
(35)
2 + i0 2
+ b4 2:
2 21 + 2i0 2 20 + 2i0
For jj small in comparison to 0, the two expansions

+ i0 1 20 0 20
2 i0 + 1 2 2 + i 2 3 4 2 3
2
2 20 + 2i0 0 0 0 0
and
2 + i0 2 2
2
2 21 + 2i0 2 20 + 2i0 21 40

23
20 2i0 + i 2 0 2 221 + 20
1 0
are used in Equation (35), which then assumes the approximate polynomial
form

2 0 b2 b2 20 20 b2 2 3
0 + 2i0 2i 4 3 + 2 4 2 4 2 21 + 0
2 2 2
0 1 0 1 0

b2
2
2 b2
+ 1 2 2 4 02 20 2 2
0 0 1 0

b2
+ 2i0 1 2 20 b2 :
0
Because the coefficient of the cubic term in is small in comparison to the
other terms appearing in this expression, it may be neglected. The approx-
imate saddle point equation for the near saddle point locations then becomes
(Oughstun, 1978)
b2
2 20 + 2
20 20 2 20
+ 2i0
2
0, (36)
b2 b2
0 + 3 2
2 2
0 + 3 2
2 2
0 0
where
s
b2
0 n0 1 + 2 , (37)
0
and
20 2 2
1 41 + b 1: (38)
320 21
Notice that this expression reduces to Brillouins approximation [cf.

Equation (2.21) in Part I] for values of very close to 0, showing that
Brillouins approximation of the near saddle point locations is valid only
in the immediate spacetime region about the value 0.
The near saddle point locations may then be expressed as (Oughstun,
1978; Oughstun & Sherman, 1988)
2
SPn i0 (39)
3
with
2 0 2
12 31=2

2 2
2 2 20 + 2 b 2
4 2 0 2 0 2@
2 0
0 A 5
, (40a)
0 + 3 b 2 2 20 + 3 b 2
2
0 0
2
b
2 20 + 2
3 20
: (40b)
2 b2
0 + 3 2
2 2
0
In the limit as approaches infinity
q
lim SPn 20 20 i0 (41)
!1
and the near saddle points SP n , respectively, approach the inner branch
points , in agreement with the numerical results presented in
Figures 27. The critical spacetime point 1 at which the two near
first-order saddle points coalesce into a single second-order saddle point is
found from this expression to be given by
220 b2
1 0 + , (42)
20 320 420
which reduces to the approximate expression given by Brillouins analysis
through neglect of the term 420 in comparison to 320 in the denominator.
With Brillouins choice of the material parameters, this critical value is
numerically found (Oughstun, 1978) to lie in the range 1.50275 < 1 <
1.50300 while Equation (42) gives the value 1 1.50414.
The dynamical behavior of the near saddle points and the local complex
phase behavior about them, as described by Oughstuns approximation (39)
with
b2
, i 0 + 2 i 20 (43)
20 40
is now considered for the three separate cases 1 < 1, 1, and > 1.
Case 1 (1 < 1): Over this initial spacetime domain, the near saddle
point locations are given by

2
SPn i o 0 (44)
3
with the approximate expressions
2 0 2 12 31=2
2 20 + 2 b 2

2 2
2
o 420 @ 2 0 A
20 2 0 5
, (45)
20 + 3 b 2 0 + 3 b 2
2 2
0 0
2
b
2 20 + 2
3 20
, (46)
2 2 b2
0 + 3 2
2
0
that are appropriate over this domain. As depicted in Figure 9, the near sad-
dle points SP n are located along the imaginary axis, symmetrically situated
about the point 00 23 0 , where () varies slowly over this space
time domain.
The approximate local behavior of (, ) Re{(, )} about each
near saddle point SP 00
n , where i () + re
i
with 00
23 0 , is given by (Oughstun & Sherman, 1988)
r,, 00 + r sin 0
b2
00 3 00 2 00 2 00
+ + 20 + 3 + 40 r sin (47)
20 40
300 + 20 r 2 cos 2 r 3 sin3
,
from which it is seen that (r, , ) attains its maximum variation about each
near saddle point when 0, /2, , 3/2. The lines of steepest descent and
ascent through the near saddle points are then parallel to the coordinate axes,
as depicted in Figure 9, where the vectors indicate the direction of ascent
along these lines. The dashed lines through each saddle point indicate the
local behavior of the isotimic contour (, ) (SPn , ) through that
+
SPn
w
d

SPn
Figure 9 Phase behavior about the near saddle points SP n situated along the imaginary
axis of the complex -plane for a single-resonance Lorentz model dielectric over the
initial spacetime domain 1 < 1. The dashed lines through each saddle point indi-
cate the local behavior of the isotimic contour (, ) (SPn, ) through that saddle
point, the shaded region indicates the local region about each saddle point where the
inequality (, ) < (SPn , ) is satisfied, and the arrows indicate local direction of
ascent along the lines of steepest descent and ascent through each saddle point.
Reprinted with kind permission of Springer Science+Business Media from Oughstun
(2009), Fig. 12.40, p. 332.
saddle point, and the shaded region indicates the local region about each
saddle point where the inequality (, ) < (SPn , ) is satisfied. Notice
that the paths of steepest descent through the upper near saddle point
SP+n are at the angles 0, , whereas the paths of steepest descent through
the lower near saddle point SP n are at the angles /2, 2/2, as indicated
in the figure.
Notice that at the critical spacetime point 0 (where 1 < 0 < 1),
the near saddle points are located at SP+n (0) 0 and SPn 0 i 4 3 ,
0
where the solution for the upper near saddle point at this value is exact.
Furthermore
SPn+ 0 , 0 0 SPn+ 0 ,0 0 (48)
exactly. It is this latter property that makes this saddle point so important in
the subsequent asymptotic field behavior.
Case 2 ( 1): At the critical spacetime point 1, Equations (39)

(40a) yield a single near saddle point located along the negative imaginary
axis at
2 20
SPn 1 i0 1 i, (49)
3 3
as depicted in Figure 10. Both the first and second derivatives of the complex
phase function vanish at this saddle point, viz.,
0 SPn 1 ,1 00 SPn 1 ,1 0, (50)
and the two first-order near saddle points have coalesced into a single
second-order saddle point.
From Equation (4.34) with 00 20/(3), the local behavior of
(, 1) about this second-order near saddle point is found to be given by
3
b2 40 3
r, ,1 + r sin 3 : (51)
0 40 272 2
SPn wSPn(q1) = -i 2d/(3a)
Figure 10 Phase behavior about the near saddle point SPn in the complex -plane for a
single-resonance Lorentz model dielectric at the critical spacetime point 1 when
the two first-order saddle points SP
n have coalesced into a single second-order saddle
point. The dashed lines indicate the local behavior of the isotimic contour (, )
(SPn, ) through this second-order saddle point, the shaded area indicates the local
region about the saddle point where the inequality (, ) < (SPn, ) is satisfied,
and the arrows indicate local direction of ascent along the lines of steepest descent
and ascent through the saddle point. Reprinted with kind permission of Springer
Science + Business Media from Oughstun (2009), Fig. 12.41, p. 334.
Hence, (r, , 1) attains its maximum variation about this second-order

near saddle point when /6, /2, 5/6, 7/6, 3/2, 11/6. The lines
of steepest descent from this saddle point are at /6, 5/6, 3/2, and
the lines of steepest ascent are at /2, 7/6, 11/6, as depicted in
Figure 10.
Case 3 ( > 1): Over this final spacetime domain, the near saddle point
locations are given by
2
SPn i0 , (52)
3
where () and () are both real-valued and are given by Equation (40a).
As increases away from 1, the second-order saddle point SPn separates into
two near first-order saddle points SPn that move off of the imaginary axis
into the lower-half of the complex -plane, symmetrically situated about
the imaginary axis, as depicted in Figure 11. In the limit as ! 1, these
two near saddle points SP n approach the inner branch points ,
respectively.
_
2d

3a
SPn SPn+
w
branch
cut w
d w+
branch
cut w+

Figure 11 Behavior of the near saddle points SP n in the complex -plane for a single-
resonance Lorentz model dielectric over the final spacetime domain > 1. The dotted
curves indicate the respective directed paths that these first-order saddle points follow
as increases to infinity. The dashed lines through each saddle point indicate the local
behavior of the isotimic contour (, ) (SPn , ) through that saddle point, the
shaded region indicates the local region about each saddle point where the inequality
(, ) < (SPn, ) is satisfied, and the arrows indicate local direction of ascent along
the lines of steepest descent and ascent through each saddle point. Reprinted with kind
permission of Springer Science+Business Media from Oughstun (2009), Fig. 12.42, p. 335.
The approximate local behavior of the real phase function (, ) in the

vicinity of these near saddle points, where SPn+ + rei
i 23 0 + rei for the near saddle point SP+n for > 1, is found
to be given by (Oughstun & Sherman, 1988)

2 b2 8 3 2
r, , rsin 0 0 + 0 1
3 20 0 9
4
3
+ 20 1: + 40 1rcos
2

4 2
3 + 0 2 rsin
2
3
o
+ 20 1r 2 cos 2 3r 2 sin2 r 3 sin 3 :
(53)
It is then seen that (r, , ) attains its maximum variation about the near
saddle point SP+n when /4,3 /4, 5/4, 7/4. The lines of steepest
descent through this saddle point are at /4, 7/4 and the lines of
steepest ascent are at 3/4, 7/4. Because of the even symmetry of
(,) about the imaginary axis, the lines of steepest ascent and descent
through the near saddle point SP n are reversed, as illustrated in Figure 11.
2.2.3 Determination of the Dominant Saddle Points

The asymptotic description of dispersive pulse dynamics in a given medium
relies upon the determination of the saddle point (or points) that give the
least exponential decay as the propagation distance z ! 1. Such a saddle
point SP at which (SP, ) Re{(SP, )} is least negative or zero is
called the dominant saddle point. The analytic determination of the spacetime
domains over which the distant or near saddle points are dominant for a
single-resonance Lorentz model dielectric is now considered based upon
the approximate expressions for the complex phase behavior at these saddle
points (Oughstun, 1978; Oughstun & Sherman, 1988).
The real and imaginary parts of the complex phase behavior at the distant
saddle points SPd are described by the pair of approximate expressions
b2 0 1
SPd , 0 1 + 1 , (54)
2 + 20 1 2
2
and
" #
b2 =2
SPd , 1 + 2 , (55)
+ 20 1 2
respectively, for all 1. It is then seen that at the luminal spacetime point
1, (SPd , 1) 0 (exactly), and that as increases away from unity,
(SPd , ) is negative with monotonically increasing magnitude, where
lim !1 SPd , 1 as the distant saddle points SP d approach the
0
outer branch points , respectively.
The real part of the complex phase behavior at the near saddle points is
described by the set of approximate expressions
2
2 b2 2
SPn , o 0 0 + o 0
3 20 40 3

2
o 0 + 20 ; 1 1 ,
3
(56)
3 2
40 b
SPn ,1 ; 1 , (57)
272 0 40
2 b2
SPn , 0 0 +
3 20 40
8 3 2 h i
1 + 2 1
; 1 :
2
9 0 3
(58)
The imaginary part of the complex phase behavior at the near saddle
points SPn identically vanishes over the spacetime domain 1 1,
and for 1 is given by

b2 4 2
SPn , 0 2 + :
2
20 40 3
(59)
These expressions then show that at the upper near saddle point SP+n ,
(SP+n , ) is initially negative over the spacetime domain 2 [1, 0),
increasing to zero as increases to 0, identically vanishes at 0, and then
grows negative monotonically over the short spacetime interval 2
(0, 1]. At the lower near saddle point SP n , (SPn, ) is initially positive
and monotonically decreases to the approximate value 43 b2 =272 0 40
as increases to 1. Because the original contour of integration is not

deformable into an Olver-type path through this lower near saddle point
over this spacetime domain, this saddle point may be ignored in the present
analysis. Finally, for 1, during which the two near first-order saddle
points first emerge from their coalescence as a single second-order saddle
point at 1 and then move off symmetrically into the complex -plane,
the quantity (SPn,) continues to decrease monotonically with increasing
, where lim !1 SPn , 1 as the near saddle points SP n approach
the inner branch points , respectively.
It is then seen that the distant saddle points SP
d are at first dominant over
+
the upper near saddle point SPn but that, for some critical spacetime value
SB between unity and 0, the upper near saddle point SP+n becomes
dominant and remains dominant for all later values of , as depicted in
Figure 12. This critical spacetime value is then defined by the expression
SPd , SB SPn+ ,SB , (60)
where 1 < SB < 0, with the approximate solution (Oughstun, 2009;

Oughstun & Sherman, 1988)
2 2 1=3
420 b2 0 b
SB 0 3 0 1 2
, (61)
30 40 20 40
(w SP ,q)
1 q SB q0
0 q
-
q1 (wSP+- ,q)
n
(wSP+ ,q) (wSP+- ,q)

n d
Figure 12 Sketch of the behavior of the real part (, ) of the complex phase function
(, ) at the relevant saddle points for a single-resonance Lorentz model dielectric.
Reprinted with kind permission of Springer Science+Business Media from Oughstun
(2009), Fig. 12.43, p. 339.
2
provided that 20 b2 =40 1. With Brillouins choice of the material
parameters, this critical value is numerically found (Oughstun, 1978) to
be SB 1.334 while Equation (61) estimates the value as SB 1.255.
A related quantity of interest is the value of the real angular frequency
q
SB 21 20 that is defined by the relation
SPd+ , SB SB , (62)
where (0 ) 0 ni(0 ) along the real axis. According to this definition,
the angular frequency SB is the real coordinate value at which the isotimic
contour (, ) (SP+d , ) through the distant saddle point SP+d crosses the
positive 0 -axis when SB. Unfortunately, finding the solution of
Equation (62) for SB is an extremely formidable, if not impossible, task.
However, because the isotimic contour through SP+d at the angle /2 to
the 0 -axis remains at essentially this angle when it intersects the 0 -axis,
as seen in Figure 4, an approximate expression for SB is given by the real
coordinate value of the distant saddle point SP+d at SB, so that
(Oughstun & Sherman, 1988)
s
b2 52
SB SB 0 2 + 2 + 02 : (63)
0 30
p
Notice that SB 20 + b2 =40 when 0 > b 0, and that
SB b + 20 =b when b > 0 0. With Brillouins choice of the material
parameters, this critical frequency value is numerically found (Oughstun,
1978) to be SB 8.70 1016 r/s while the final approximate expression
above gives SB 7.22 1016 r/s whereas SB (SB) 8.41 1016 r/s.
3. UNIFORM ASYMPTOTIC EXPANSIONS

The theory of uniform asymptotic expansions of integrals saw many
advances during the 1950s and 1960s. In particular, uniform expansions that
arise in the analysis of precursors (Handelsman & Bleistein, 1969), the uni-
form expansion of an integral with a stationary point near a pole (L. B.
Felsen, 1963; L. B. Felsen & Marcuvitz, 1959), the uniform expansion of
an integral with a stationary point near an algebraic singularity (Bleistein,
1966), and the uniform expansion of an integral with two nearby stationary
points (Chester, Friedman, & Ursell, 1957) were all developed during this
period. Equipped with these theories, Oughstun and Sherman (1988, 1989,
1994) were able to provide uniform asymptotic expansions of the propa-

gated field [(1.58) in Part I] that remain valid for all > 1. In this section,
we present the procedure for the asymptotic analysis of the propagated sig-
nal, followed by these uniform asymptotic theories and their application to
the step-modulated sine wave.
3.1 Procedure for the Asymptotic Analysis of the Propagated

Signal
With detailed knowledge of the topography of the complex phase function
(, ) and the spacetime dynamics of its saddle points for a single-
resonance Lorentz model dielectric, a detailed procedure to perform the
asymptotic analysis of the propagated signal AH(z,t) given by
Equation (1.58) in Part I was then developed by Oughstun and Sherman
(1988, 1994), which is now described.
The first step in the asymptotic analysis of the signal wave field AH(z, t) is
to express its FourierLaplace integral representation [(1.58) in Part I] in
terms of an integral I(z, ) with the same integrand but with a new contour
of integration P() to which the original contour of integration C may be
deformed. In the present application, the simple pole of the spectral function
u~H c that is crossed when the original contour C is deformed to P()
is encircled in the process, in the clockwise sense. Hence, according to
Cauchys residue theorem, the integral representation of AH(z,t) and the
integral I(z, ) are related by
AH z, t Iz, Ref2ig, (64)
where

i z=c, 1
Resc u~ c e ez=cc , (65)
2 2
is the residue of the pole at c along the positive real frequency axis that
is crossed, and where I(z,) is defined as
(Z )
1 1 z=c,
Iz, Re e d (66)
2 P c
for all z 0. Notice that () changes discontinuously with the spacetime

parameter as the path P() crosses over the pole. However, this disconti-
nuity is canceled by a corresponding discontinuous change in I(z, ).
Because the distant saddle points SP d are dominant for some values of
and either the upper near saddle point SP+n or both of the near saddle points
SPn are dominant for other values of , there is no single path P() that is an
Olver-type path (see Oughstun, 1978, 2009) with respect to a single saddle
point and that evolves continuously with for all 1. In order to obtain an
asymptotic representation that remains uniformly valid for all 1, the con-
tour P() must evolve continuously for all 1 and, in the vicinity of the
spacetime point SB when the saddle point dominance changes, the
path must pass through both the dominant and nondominant saddle points
involved in the dominance change. Moreover, the path P() must be divis-
ible into a sum of subpaths Pj(), each of which is an Olver-type path with
respect to one of the saddle points.
For spacetime values in the range 2 [1, 1) during which the two near
saddle points SP 00
n are situated along the imaginary -axis, the lower near

saddle point SPn is dominant over all of the other saddle points (see
Figures 24). That saddle point is not useful, however, because the
Olver-type paths with respect to it are not deformable to the original con-
tour C (and vice versa) due to the presence of the branch cuts and is hence-
forth ignored.
There are many paths having the required properties that pass through both
the upper near saddle point SP+n and the distant saddle points SP d for space
time points in the domain 2 [1,1). There are also many paths having the
required properties that pass through both of the near saddle points SP n and

the distant saddle points SPd for > 1. Finally, there are many paths having
the required properties that pass through the single second-order near saddle
point SPn and both of the distant saddle points SP d when 1. As a conse-
quence, the contour P() can always be chosen so that it passes through the
upper near saddle point SP+n and the distant saddle points SP d for 2 [1, 1]
and through all four saddle points for > 1 so that it evolves in a continuous
manner as varies over the entire subluminal spacetime domain 1 and can
be divided into the desired subpaths with respect to each relevant saddle point.
An example of such a path P() and its component subpaths Pj() is illus-
trated in Figure 13. For values of in the range 2 [1,1], the component
subpaths (from left to right in the figure) are P + +
d (), Pn (), and Pd (), and for

> 1 the component subpaths are Pd (), Pn (), Pn (), and Pd (). The sub-
+ +
paths P
d () and Pn () are Olver-type paths with respect to the saddle points

SPd and SPn , respectively.
Provided that the path P() and its component subpaths Pj() satisfy the
above constraints, it is unimportant which particular paths are used; by
w''
C
+
SPn
Pd Pn+
w
w w w+ w+
SPn
SP SPd+
d Pd+
q < qSB
1<
w
Original contour of integration C
SPn+
w
Pd Pn+
w w w+ w+
SPn
SP d
SPd+
Pd+
qSB < q < q1

w
C
w
Pd SPn SPn+ Pn+
w w w+ w+
Pn
SPd SPd+
Pd+
q1 < q
Figure 13 Olver-type paths through the relevant saddle points of a single-resonance

Lorentz model dielectric. The shaded area indicates the region of the complex -plane
wherein the inequality (, ) < (SP>, ) is satisfied, where SP> denotes the dominant
saddle point (or points) and the darker shaded area indicates the region of the complex
-plane wherein the inequality (, ) < (SP<, ) is satisfied, where SP< denotes the
nondominant saddle point (or points) over the indicated spacetime interval. Reprinted
with kind permission of Springer Science+Business Media from Oughstun (2009), Fig. 12.66,
p. 379.
Olvers theorem (Olver, 1970; Oughstun, 1978, 2009), the asymptotic

results are independent of the choice. Some particular choices may be more
convenient, however, in that they reduce the computation required. In the
analysis that follows, the paths are sometimes taken to follow the path of
steepest descent in the vicinity of each saddle point in order to simplify
the determination of the appropriate values of the multivalued functions

appearing in the asymptotic expressions.
The deformed contour of integration employed by Brillouin (1914,
1960) followed along the entire paths of steepest descent through the distant
saddle points SPd and the entire steepest descent path through the upper
near saddle point SP+n for 1 1 and through both near saddle points
SPn for > 1, the various individual paths being connected along the
branch cuts. Although this is a perfectly valid deformed contour of integra-
tion, it is unnecessarily complicated and places unnecessary importance to
the steepest descent path in the resultant asymptotic description.
The integral I(z, ) given in Equation (66) is then expressed as the sum of
integrals with the same integrand over the various subpaths, so that for a
single-resonance Lorentz model dielectric
Iz, Id z, + In+ z, + Id+ z,; for 1 1 , (67a)

Iz, Id z, + In z, + In+ z, + Id+ z, ; for > 1 , (67b)
where I
d (z, ) and In (z, ) denote the contour integrals taken over the
Olver-type paths P
d and Pn , respectively. In order to obtain an asymptotic
approximation of the propagated signal AH(z,t) in a Lorentz model dielectric,
it now only remains to obtain asymptotic approximations of the various con-
tour integrals appearing on the right-hand sides of Equation (67a) and (67b).
If the distant saddle points SP
d do not pass too near to the pole at c,
then an asymptotic approximation of the quantity I d (z, ) + Id (z, ) in the
+
form (Oughstun & Sherman, 1988)
Id z, + Id+ z, As z, t + Rd z,, (68)
is obtained, where As(z, t) is the contribution due to the two distant saddle
points alone, referred to as the first, or Sommerfeld, precursor, and Rd is the
remainder.
If the near saddle points SP+n for 1 < 1 and SP n for > 1 do not
pass too close to the pole singularity at c > 0, then asymptotic approxima-
tions of I+n (z, ) for 1 < 1 and In (z,) + In (z,) for > 1 in the form
+
In+ z, Ab z, t + Rn z,; for 1 1 , (69a)

In z, + In+ z, Ab z, t + Rn z,; for > 1 , (69b)
are obtained, where Ab(z,t) is the contribution due to the first upper near
saddle point and then both near saddle points, referred to as the second,
or Brillouin, precursor, and Rn is the remainder.
Consider now the above absorption band situation when the distant sad-
q
dle point SP+d approaches (as varies) the pole at c > 21 20 that is
bounded away from + 1. An asymptotic approximation of the quantity
Id(z,) + Id+(z,) in the form
Id z, + Id+ z, As z, t + Cd+ z,t + Rd z, (70)
is then obtained, where As(z, t) is given by the same expression as is As(z, t) in
Equation (68) for all 1. The quantity C+d (z, t) appearing in Equation (70)
is the contribution due to the proximity of the saddle point to the pole,
which is asymptotically negligible if the distant saddle points SP+d do not
approach the pole at c so that Equation (70) reduces to Equation (68).
Similarly, in the below absorption band situation when near saddle
point SP+n approaches (as varies) the simple pole at c with
q
0 < c < 20 20 , asymptotic approximations of the form I+n (z, ) for
1 < < 1 and I
n (z, ) + In (z, ) for > 1 in the form
+
In+ z, Ab z,t + Cn+ z,t + Rn z, , (71a)

for 1 < 1, and
In z, + In+ z, Ab z,t + Cn+ z,t + Rn z, , (71b)
for > 1 are obtained, where the expression for Ab(z, t) is the same as that in
Equation (69a) and (69b). The quantity C+n (z, t), which represents the con-
tribution due to the proximity of the pole to the saddle point, is asymptot-
ically negligible if the near saddle point SP+n does not approach the pole at c,
in which case the expressions in Equation (71a) and (71b) reduce to the
corresponding expressions in Equation (69a) and (69b).
q
Finally, when c is within the absorption band 20 20 c
q
21 20 , both the distant saddle point SP+d and the near saddle point
SP+n approach the pole at c for finite , the distant saddle point from
above and the near saddle point from below. In that situation, a linear com-
bination of Equations (70) and (71a) and (71b) applies.
Combination of Equations (64), (67a) and (67b), and (70)(71b) results
in the general expression
AH z,t As z,t + Ab z, t + Ac z,t + Rz, (72)
for the asymptotic approximation of the integral representation of the
propagated signal wave field AH(z,t) as z ! 1 in a single-resonance
Lorentz model dielectric. This approximation is uniformly valid for all sub-
luminal spacetime points 1, provided that the simple pole singularity
at c is bounded away from the limiting locations taken by the distant
saddle points SP d as ! 1 and by SPn as ! 1 from below. The con-
+ +
tribution Ac(z, t) appearing in Equation (72) is obtained by adding all of the

terms that involve the pole, so that
Ac z, t Ref2ig + Cd z, t + Cd+ z, t + Cn+ z, t (73)
for 1 1, and
Ac z, t Ref2ig + Cd z,t + Cd+ z,t + Cn z,t + Cn+ z,t

(74)
for > 1. An estimate of the remainder term R(z,) as z ! 1 is obtained

by taking the largest estimate of the remainder terms appearing in
Equations (70)(71b).
An important feature of the general expression given in Equation (72) is
that the asymptotic behavior of the propagated signal wave field AH(z,t) in a
Lorentz model medium is expressed as the sum of three terms which are
essentially uncoupled so that they can be treated independently of one
another. Each term is determined both by the dynamical behavior of specific
saddle points that are a characteristic of the dispersive medium as well as
by the analytic behavior of the input pulse spectrum, as described in the
paragraphs to follow.
The dynamic behavior of As(z, t) is determined by the dynamical evolu-
tion of the distant saddle points SP d and the value of the input pulse spec-
trum at these saddle points. Because the distant saddle points are dominant
over the initial spacetime domain 2 [1,SB), the propagated wave field
component As(z, t) describes the dynamical spacetime behavior of the
first or Sommerfeld precursor field. This first precursor field is asymptotically
negligible during most of the remaining field evolution.
The dynamic behavior of Ab(z, t) is determined by the dynamical evo-
lution of the near saddle points SP n and the value of the input pulse spectrum
at these saddle points. Because the near saddle points are dominant imme-
diately following the distant saddle point dominance, the propagated wave
field component Ab(z, t) describes the dynamical spacetime behavior of the
second or Brillouin precursor field. This second precursor field is asymptotically
negligible during most of the first precursor and remaining field evolution.
The dynamic behavior of Ac(z, t) is determined by the pole at c and

the dynamics of the saddle points that interact with them. The implication of
nonzero Ac(z, t) is that the wave field AH(z, t) oscillates with angular frequency
c for positive time t on the plane z 0 and will tend to do the same at larger
values of z for sufficiently large time t. As a result, the propagated wave field
component Ac(z, t) describes the dynamic behavior of the signal contribution
oscillating with angular frequency c. This contribution to the total field evo-
lution is negligible during most of the precursor field evolution.
For most values of , only one of the terms, As(z, t), Ab(z, t), and Ac(z, t)
appearing in Equation (72), is important at a time. There are short space
time intervals in , however, during which two or more of these terms
are significant for fixed values of z. These spacetime intervals mark the
transition periods when the wave field is changing its character from one
form to another and the presence of both terms in the expression leads to
a continuous transition in the spacetime behavior of the propagated wave
field. As a result, Equation (72) displays the entire evolution of the field
through its various forms in a continuous manner.
3.2 The Sommerfeld Precursor Field

The contribution to the integral [(1.58) in Part I] from the distant
saddle points D gave rise to Brillouins first forerunner. As noted by
Brillouin, his expression [(2.28) in Part I] for the first forerunner is invalid
as ! 1+ and as ! 1. This limited validity occurs for two reasons: First,
Brillouins approximate locations of the distant saddle points [(2.18) in Part I]
lose their validity as ! 1, as the distant saddle points approach the outer
0
branch cuts . Second, as ! 1+, the distant saddle points become
infinite-order saddle points. Hence, the method of steepest descent applied
to a first-order isolated saddle point [see Equation (2.25) in Part I] is not
valid. Improved approximate locations of the distant saddle points were pro-
vided by Oughstun and Sherman (1988), as shown in Section 2. Here, the
uniform asymptotic method of Handelsman and Bleistein (1969) is applied
to obtain the asymptotic expansion of the integral representation of the
propagated field [(1.58) in Part I] about the distant saddle points SPd ().
Uniform methods consist of a transformation of the phase function and
an expansion of the amplitude function into a finite number of terms plus a
remainder term, the remainder term being regular and equal to zero at the
saddle points of the transformed phase function. The transformation is such
that the saddle points of the new phase function retain the limiting values of the
original saddle points. Of significance here is the work of Handelsman and

Bleistein (1969) that provides a uniform method appropriate for a pair of saddle
points, symmetrically located about the imaginary axis, that become infinite-
order saddle points in the limit as ! 1+, as their real parts become infinite,
precisely the case for the pair of distant saddle points D() (see Figures 2
and 8). In this uniform method, the phase function is transformed as
1
, 2 s + + , (75)
4s
and the amplitude function is expanded as

1 d 1
0 + 1 s + 2 H0 s,:
2
(76)
c ds 4s
Here, , , 0, 1 are functions to be determined, and H0(s, ) is regular and
equal to zero at s (1/). Substitution of (75) and (76) into (1.58) in Part
I results in a uniform asymptotic description of the first forerunner, or
Sommerfeld precursor, As(z, t),
n h z ih z z io
As z,t Re exp i 0 J0 + 2ei=2 1 J1 ,
c c c
(77)
to leading order as z ! 1 for all 1 (Oughstun & Sherman, 1994). The
coefficients appearing in (77) are given by
ih i n o
SPD+ , SPD , Im SPD+ , , (78a)
2
ih i n o
SPD+ , + SPD , iRe SPD+ , , (78b)
22
!1=2
14 i 1 43
0
2 SPD+ c 2 i2 SPD+ ,
!1=2 3 (78c)
i 1 4
3
5,
+ 2
SPD c 2 i SPD ,
2 !1=2

1 4 i 1 43
1
4 SPD+ c 2 i2 SPD+ ,
!1=2 3 (78d)
i 1 4
3
5:
2
SPD c 2 i SPD ,
The Sommerfeld precursor field As(z, t) for the step-function-modulated

sine wave AH(z, t) with below-resonance-applied signal frequency c 0.5
0 at an observation distance of z 1.2 106 m 3.8 zd into the single-
resonance Lorentz dielectric is illustrated in Figure 14. Characteristic of the
Sommerfeld precursor is an amplitude that starts at zero when 1,
increases rapidly to a peak occurring shortly after 1, and then monoton-
ically decreases with increasing . The instantaneous angular frequency of
oscillation of the Sommerfeld precursor, defined as the time derivative of
the oscillatory phase (Brillouin, 1914, 1960; Oughstun & Sherman,
1994), may be approximated as the real part of the distant saddle point loca-
n o
tion in the right-half plane Re SPd+ . Hence, the instantaneous
angular frequency of oscillation of As(z, t) begins at infinity (when the field
amplitude is zero) and then rapidly decreases with increasing , approaching
q
the angular frequency value 20 + 2p 20 associated with the upper edge
of the material absorption band as ! 1 (Oughstun & Sherman, 1994;
Wyns, Foty, & Oughstun, 1988).
Although numerically determined distant saddle point positions (as a
function of ) may always be used in the exact expressions for the complex
phase function and its second derivative appearing in Equation (78a)(78d)
103
8
2
As (z,t)
8
1 1.05 1.1 1.15 1.2 1.25 1.3
q = ct/z
Figure 14 Dynamic behavior of the Sommerfeld precursor As(z, t) for the step- function-
modulated sine wave with below-resonance-applied signal frequency c 0.50 rad/s
at a relative observation distance z/zd 3.8 into the single-resonance Lorentz dielectric
with Brillouin's choice of material parameters as described by the uniform asymptotic
theory with numerically determined saddle point locations.
in order to obtain the precise asymptotic behavior of the Sommerfeld

precursor for a given input pulse, approximate analytic expressions of these
quantities are useful in their own right, particularly in connection with
earlier results. The approximate behavior of the complex phase function
(,) i(n() ) in the region jj > 1 of the complex -plane above
2
the absorption band is given by , i1 i 2 +b 2i0 so that at
SPd (), one obtains
" #
b2 =21
SPd , 0 1 + 1 + 2
+ 20 1 2
" # (79)
b2 =2
i 1 + 2 ,
+ 20 1 2
and
b2
00 SPd , i : (80)
+ i0 1
3
Substitution of Equations (79) and (80) into (78a)(78d) gives the approx-
imate expression for the Sommerfeld precursor
" #1=2
b2 =2
As z,t 1+ 2
2b + 20 1 2
( " #)
z b2 =21
exp 0 1 + 1 + 2
c + 20 1 2
( " #
0 5 + 3c 1 5 3c 1

2 + c 2 + 20 1 + 2 c 2 + 20 1 + 2
" !#
z b2 =2
J0 1 + 2
c + 20 1 2
2 3
3 3
c 20 1 2 + c 20 1 2
6 2 2 7
+4 5
c + 0 1 +
2 2 2
+ c + 0 1 +
2 2 2
" !#)
z b2 =2
J1 1 + 2
c + 20 1 2
(81)
as z ! 1 for all 1.
For values of > 1 bounded away from unity, the two Bessel functions
appearing in (81) may be replaced by their large argument asymptotic
approximation
r
2
J cos (82)
2 4
as ! 1 with jarg()j < . Substitution of (82) into (81) results in the non-
uniform asymptotic approximation of the Sommerfeld precursor (Oughstun &
Sherman, 1988, 1994)
r ( " #)
1 c z b2 =21
As z, t exp 0 1 + 1 + 2
b 2z c + 20 1 2
82 3
> 3 3
< c 20 1 2 + c 20 1 2
6 2 2 7
4 2 5
>
: c 2
+ 2
0 1 + 2
+ c 2
+ 2
0 1 +
" ! #
z b2 =2
cos 1 + 2 +
c + 20 1 2 4
" #
0 5 + 3c 1 5 3c 1
+
2 + c 2 + 20 1 + 2 c 2 + 20 1 + 2
" ! #)
z b2 =2
sin 1 + 2 + ,
c + 20 1 2 4
(83)
as z ! 1 with 1 + with > 0.

This nonuniform asymptotic expression is not a valid asymptotic approx-
imation of the first precursor field in the limit as ! 1+ for fixed values of
the propagation distance z > 0. In order to establish a connection with the
now classical result obtained by Brillouin (1914, 1960) for the first forerun-
ner, however, the behavior of this expression in that limit is now examined.
For spacetime values approximately equal to but greater than unity,
Equation (83) simplifies to
r
bc
2 13=4 e20 c 1
z
As z,t
(" 2z
b2 11=2 c

b2 11=2 c 2 + 420
#
b2 11=2 + c z p
cos b 2 1 + (84)
b2 11=2 + c 2 + 420 c 4
"
1
+ 20 1=2
b2 1 c 2 + 420
# )
1 z p
sin b 2 1 + ,
b2 11=2 + c 2 + 420 c 4
as z ! 1 with subluminal 1. This expression can be further simplified

by noting that for spacetime values very close to unity, any finite angular
carrier frequency c > 0 will be negligible in comparison to the quantity
p
b= 2 1. Hence, in the limit as ! 1+, the nonuniform asymptotic
approximation given in Equation (84) simplifies to
r
2bc c 2 11=4 20 z1 z p
As z,t e c cos b 2 1 + (85)
z b2 + 820 1 c 4
as z ! 1 and ! 1+. This expression is precisely Brillouins result for the
first forerunner given in Equation (2.28) in Part I. Hence, Brillouins asymp-
totic approximation of the first precursor field is an approximation, valid for
near 1, of an expression that is not valid for near 1.
3.3 The Brillouin Precursor

The contribution to the integral [Equation (1.58) in Part I] from the near sad-
dle points N gave rise to Brillouins second forerunner. As noted by
Brillouin, his expressions (2.33) and (2.36c) in Part I for the second forerunner
are invalid as the two near saddle points approach each other, coalesce, and
then separate. Again, this limited validity occurs for two reasons: First, the
approximate locations of the near saddle points [Equation (2.21) in Part I]
are accurate only in the region about the spacetime point 0 at which the
saddle point +N crosses the origin. Second, the method of steepest descent
applied to a first-order isolated saddle point [see Equation (2.30) in Part I] loses
its validity as two first-order saddle points come within close proximity and
coalesce into a second-order saddle point. Improved approximate locations of
the near saddle points were provided by Oughstun and Sherman, as shown in
Section 2. Here, the uniform asymptotic method of Chester et al. (1957) is
applied to obtain the asymptotic expansion of the integral representation of
the propagated field [(1.58) in Part I] about the near saddle points SPn ().
The asymptotic expansion of AH(z, t) as z ! 1 about the upper first-
order near saddle point SP+N() for 1 < < 1, about the second-order near
saddle point SPN for 1, and about both first-order near saddle points
SPN() for > 1 (see Figures 27) yields the dynamical evolution of the
second forerunner, or Brillouin precursor, Ab(z, t). A direct application of
the method of steepest descent does not result in a uniform approximation
to the propagated field because the order of the saddle points changes dis-
continuously at 1. A uniform expansion, valid for all 1, is obtained
through use of the method due to Chester et al. (1957). Here, the transfor-
mation that retains the behavior of the saddle points about 1 is
1 3
1 0 + , 0, (86)
3
where 0, 1 are functions to be determined, and the branch of the transfor-
mation (86) to be used is the one that defines a conformal mapping of some
disc that contains both saddle points SPn () (see Bleistein & Handelsman,
1975). The amplitude function is expanded as a finite number of terms plus a
remainder that is regular and equal to zero at the transformed saddle points
1 d
G0 , h1 + h2 + 2 1 H0 , , (87)
c d
where h1,h2, and H0 are to be determined. Substitution of (86) and (87) into
(1.58) in Part I results in the uniform asymptotic expansion of the Brillouin
precursor Ab(z,t) for the step-function-modulated sine wave (Oughstun &
Sherman, 1988, 1994)
h i c 1=3 2=3
z i2=3 i2=3 c
Ab z, t Re exp 0 e Ai 1 e
( " c z #) z
1 i i
h + + h
2 SPN+ c SPN c
c 2=3 2=3
i2=3 1 i2=3 c
+ e Ai 1 e
(z " z #)))
1 i i
h + h ,
2 SPN+ c SPN c
1=2
1
(88)
as z ! 1 for all 1, to leading order. Here, Ai() and Ai(1)() denote the
Airy function and its first derivative, respectively, and the coefficients appe-
aring in Equation (88) are given by
1h i
0 SPN+ , + SPN , , (89a)
2
h i1=3
1=2 3
1 SPN+ , SPN , , (89b)
4
!1=2
1=2
21
h 2 , (89c)
SPN+ ,
for 6 1. At the critical value 1 when the two near saddle points coalesce
into a single second-order saddle point, these coefficients take on the lim-
iting values
!1=3
2
lim h 3 hs 1 , (90a)
!1 SPN , 1
" #
1 i i i
lim h + + h hs 1 , (90b)
!1 2 SP + c SPN c SPN c
N
" #
1 i i d i
lim 1=2 h + h h2s 1 ,
!1 2
1
SPN+ c SPN c d SPN c
(90c)
and 0(1) remains given by Equation (89a).

The Brillouin precursor field Ab(z, t) for the step-function-modulated sine
wave AH(z,t) with below-resonance-applied signal frequency c 0.5 0
at an observation distance of z 1.2 106 m 3.8 zd into the
dispersive material is depicted in Figure 15. Characteristic of this Brillouin
precursor is an amplitude that peaks near the spacetime point 0
n(0) with this peak becoming more pronounced as z ! 1 due to its alge-
braic decay (as z1/2) with increasing propagation distance. The instanta-
neous angular frequency of oscillation of the Brillouin precursor, which
n o
may be approximated as Re SPn+ , starts at zero when
q
0 and monotonically increases to approach the value 20 20 with
increasing . However, notice that the instantaneous oscillation frequency
0.2
0.15
0.1
Ab(z,t)
0.05
0.05
0.1
1 1.5 2 2.5
q = ct/z
Figure 15 Dynamic behavior of the Brillouin precursor field Ab(z, t) for the step-function-
modulated sine wave with below-resonance-applied signal frequency c 0.5 0 at a
relative observation distance z 3.8 zd into the single-resonance Lorentz dielectric with
Brillouin's choice of material parameters as described by the uniform asymptotic theory
with numerically determined near saddle point locations.
is zero only at the single spacetime point ct/z 0; this does not mean
that the field is static about this point (L. Felsen & Marcuvitz, 1973).
A physically meaningful effective oscillation frequency has been defined
by Oughstun (2009) over the initial spacetime domain 2 (1, 1] by the
temporal width of the build-up to the peak amplitude point that occurs
between the spacetime points SB and 0. The difference between
these two spacetime points is given by 0 0 SB 420 b2 =30 40 .
This difference corresponds to the effective half-period Teff 2/eff of the
field over this spacetime domain through the relation 0 (c/z)(Teff/2),
so that
30 40 c
eff 0 : (91)
420 b2 z
Notice that this effective angular oscillation frequency of the Brillouin pre-
cursor at 0 asymptotically approaches zero as z ! 1, in agreement
with the limiting behavior of Re{SP+n ()} as ! 1+ .
Although numerically determined distant saddle point positions (as a
function of ) may always be used in the exact expressions for the complex
phase function and its second derivative appearing in Equation (88) in order
to obtain the precise asymptotic behavior of the Brillouin precursor for a
given input pulse, approximate analytic expressions of these quantities are

useful in their own right. Because the near saddle point behavior separates
into the three separate spacetime domains 1 < < 1, 1, and > 1,
the asymptotic behavior of the propagated signal for each case is now
separately examined.
3.3.1 Case 1: 1 < < 1

For values 1 < < 1, the near saddle points are two first-order saddle points
located on the imaginary axis. The approximate behavior of the complex
q
phase function (,) in the below-resonance region jj < 20 20 about
2
the origin is given by , i0 + 2b0 4 2 i 20 so that at
0
the position SPn+ i 0 23 0 of the upper near saddle
point,
1
SPn+ , 20 3 0 0
3
b2
+ 20 3 0 2 20 3 + 3 0
,
540 40
(92)
2
b
00 SPn+ , 20 1 + 3 0
: (93)
0 40
For 1 < 1, the spectrum of the envelope function at the near saddle
points is given by
1
i 3 0 20 ic
u~H SPn c 3
SPn c 1
2c + 3 0 20 2
9
(94)
Thus, for values of bounded away from 1 such that (z/c)2/3j1()j 1,

substitution of (92)(94), together with the large argument asymptotic
expansion of the Airy function and its derivative, into the uniform expansion
(88) reduces to the nonuniform expansion valid for spacetime values 2
(1,1) with bounded away from 1 (Oughstun, 2009; Oughstun &
Sherman, 1988, 1994)
1=2
20 c 0 c=z
Ab z,t
1 40 1 + 6 0
b 2c + 3 0 20 2
9
nz h
exp 20 3 0 0
3c
b 2 h io
+ 20 3 0 20 3 3 0 ,
180 40
(95)
as z ! 1. For 0, this nonuniform asymptotic approximation simplifies
somewhat to
1=2
20 c 0 c
Ab z, t " !2 #
20 6z
b 2c +
3

z 20 b2 20 4
exp 0 + + 0 ,
c 3 20 40 3 3
(96)
as z ! 1, where
1=2
420 20 40
+ 0 :
92 3b2
The nonuniform asymptotic approximation of the second precursor field for

the unit step-function-modulated signal given in Equation (96) is the clas-
sical result for the second forerunner obtained by Brillouin (1914, 1960) for
2 (1, 1), given in Equation (2.33) in Part I. That result is then seen to be
accurate only for values of near 0, becoming invalid as approaches 1.
At the critical spacetime point 0 ct0/z, at which 0(0) (0)
20/3, both Equations (95) and (96) simplify to

20 0 c 1=2
Ab z,t0 , (97)
bc 40 z
at which point the Brillouin precursor varies with the propagation distance
z > 0 only as z1/2 as z ! 1, making this spacetime point in the field
evolution entirely unique.
3.3.2 Case 2: 5 1
At 1, the two near saddle points have coalesced into a single
second-order saddle point SPn located at SPn 1 i 2 3 with
0
0 00
1 0 41 + b =30 1 1, where (SPn, 1) (SPn, 1)
2 2 2 2 2
0 with 000 (SPn, 1) 6 0. The approximate behavior of the complex phase

function (SPn) and its third derivative at the second-order saddle point
are found to be

20 420 b2
SPn ,1 0 1 + , (98a)
3 90 40
b2
000 SPn , 1 3i : (98b)
0 40
The spectrum of the envelope function at the near saddle point at 1 is
given by the approximate expression
20
ic
u~H SPn 1 c 3 : (99)
42
2c + 02
9
Thus, for values of bounded away from 1 such that (z/c)2/3j1()j 1,
substitution of (98a),(98b) and (99), together with the large argument
asymptotic expansion of the Airy function and its derivative, into the uni-
form expansion (88) reduces to the nonuniform expansion valid for the space
time value 1

13 0 c 20 0 c 1=3 20 z 420 b2
Ab z, t1 p exp 0 1 +
2 3 2 420 b2 z 3c 90 40
c + 2
9
(100)
as z ! 1 with 1 ct1/z. From Equation (42), it is seen that
420 b2 220 b2
0 1 + ,
90 40 90 40
and the second precursor field attenuates exponentially with increasing
propagation distance z > 0 at the fixed spacetime point 1.
In the special (limiting) case when 0 0, it follows that 1 0 exactly
and the two first-order near saddle points coalesce into a single second-order
saddle point at the origin. The peak amplitude in the Brillouin precursor
then decays with the propagation distance z > 0 only as z1/3 [see (3.26)
in Part I] instead of the usual z1/2 behavior described by Equation (97)
when 0 > 0 (Oughstun, Cartwright, Gauthier, & Jeong, 2010).
3.3.3 Case 3: > 1

For > 1, the near saddle points have split into two first-order saddle points
in the lower half of the complex -plane. The complex phase function and
its second derivative at the two first-order near saddle point locations
SPn i 23 0 are found to be given by

2 b2
SPn , 0 0 + 4 1
2
3 0
0
4 2 2 1
+ 0 1
9 3
b2 4 2
i 0 + 2 + 2
,
20 40 3 0
(101)
2
b
00 SPn , 20 1 3i
: (102)
0 40
One then has that
h i1=2 2 0
1=2
00
SPn , 0

b 20 1 3i
s (103)
2 0
0 ei=4 ,
b 3
where the final approximation is valid for all > 1 such that 3()
20(() 1). The spectrum of the envelope function at the near saddle
points is given by
2
i 0 + i c
u~H SPn c 3
SPn c 4
c 2 + 20 2
9
(104)
for 1. Thus, for values of bounded away from 1 such that
(z/c)2/3j1()j 1, substitution of Equations (101)(104), together with
the large argument asymptotic expansion of the Airy function and its deriv-
ative, into the uniform expansion (88) reduces to the nonuniform expansion
given by Oughstun and Sherman (1988, 1994) valid for spacetime values
> 1
r
20 c
0 c
2
3z
Ab z, t
4 2 2 4 2 2
b c + 0 + c + 0
2 2
9 9
z 2 b2
exp 0 0 + 4 1
2
c 3

0
4 0
4
+ 20 2 1 2c + 20 2 2
9 3 9
2 2 #
z 0 b 42 2
cos 0 + + +
c 20 04
3 02
4
2 2
4 z b 42
+ 0 sin 0 + 0 4
3 #) c 20 0 3

2

+ +
20 4
(105)
as z ! 1. This nonuniform asymptotic approximation reduces to the clas-
sical result [see Equation (2.36c) in Part I] given by Brillouin (1914, 1960) if
() and () are further approximated by expressions that are valid only for
spacetime points not too distant from 0. Hence, Brillouins classical
expression for the second precursor field over the spacetime domain
> 1 is an approximation, valid for near 1, of an expression that becomes
invalid as approaches 1 from above. As a result, Brillouins expression
for the asymptotic behavior of the second precursor field (or second forerun-
ner) for the unit step-function-modulated signal over the spacetime
domain > 1 is not applicable.
3.4 The Signal Contribution

The signal contribution to the propagated field of the unit step-function-
modulated sine wave is due to the simple pole singularity at c
appearing in the integrand of (1.58) in Part I. In Brillouins analysis, the pole
at c > 0 was assumed to lie either well above or well below the absorption
band of the material so that only the distant saddle point +D() or the near
saddle point +N(), respectively, comes within close proximity of the pole,
thus giving rise to the signal contribution. With this assumption, Brillouin
reasoned that the signal contribution describes a smooth transition from the
second forerunner to the residue contribution [(2.37) in Part I]. Brillouins
analysis of the signal contribution is limited not only by its restrictions on the
carrier frequency c, but also because a uniform asymptotic method is
required to account for a saddle point in the vicinity of a singularity of
the amplitude function. Such a uniform method provides the smooth tran-
sition of the field into the residue contribution that occurs at the spacetime
point s when the path of integration crosses the pole.
Let s > 1 denote the spacetime point at which the path of integration
crosses the simple pole located at c, determined by the equation
c , s SP , s , (106)
where Im{} and SP denotes either +D() or +N(). The uniform
expansion method developed by L. B. Felsen and Marcuvitz (1959) and
L. B. Felsen (1963) for a saddle point in the vicinity of a simple pole, and
later generalized by Bleistein (1966) for a saddle point in the vicinity of a
singularity, is obtained using the same transformation as for the method
of steepest descent
, SP , s2 , (107)
but with an expansion of the amplitude function as
1 d
+ Ts: (108)
c ds s b
Here, s is real-valued along the path of steepest descent, is the residue of the
amplitude function at c, b is the location of the transformed pole, and T(s) is
regular at both s 0 (the transformed saddle point) and at s b.
Substitution of Equations (107) and (108) into the integral representation
of the propagated field given in Equation (1.58) in Part I yields
( )
1 hz iZ 1 z=cs2
AH z,t Re exp SP , e ds
2 c P s b
( ) (109)
1 hz iZ
T sez=cs ds ,
2
Re exp SP ,
2 c P
where P() is the path of integration that has been deformed through the
valleys of the accessible saddle points. The second term in Equation (109)
is the method of steepest descent applied to a first-order saddle point located
at SP. This contribution to the integral [(1.58) in Part I] is not considered as

part of the signal contribution as it has already been accounted for with either
the Sommerfeld or Brillouin precursor for SP +D() or SP +N(),
respectively. The first term in Equation (109) is the effect of the pole located
at c on either the distant or near saddle points located in the right-half plane
and may be evaluated in terms of the complementary error function. This
term, together with the appropriate residue contribution, yields the pole
contribution to the propagated field.
A uniform expansion of the signal contribution for one first-order saddle
point in the vicinity of a pole, valid for all 1, was obtained by Oughstun
and Sherman (1989, 1994) by application of the method of L. B. Felsen and
Marcuvitz (1959) and L. B. Felsen (1963). An extension of this method to
two isolated saddle points in the vicinity of a pole was later provided by
Cartwright (2004)4 and Cartwright and Oughstun (2006). This extension
improves the accuracy of the uniform approximation when the pole lies
within the absorption band of the material where the pole may influence
the contribution of both the distant +D() and near +N() saddle points.
In this case, the contribution due to the influence of the pole on both saddle
points is given by the first term in (109) for SP +D() and SP +N(),
and s is defined by the one saddle point that satisfies Equation (106). Thus,
the asymptotic approximation of the signal contribution is given by
r h i
1 z z
Ac z, t Re i i sgnD erfc i sgnD D exp c ;
2 r c c
1 c h z i
+ exp SPD+ ,
D z c r h
z z i
+ i i sgnN erfc i sgnN N exp c ,
r c
c
1 c h z i
+ exp + ,
N z n c h SPN io
z
+ u s Re i exp c , ,
c
(110)
as z ! 1 for all 1, to leading order. Here, erfc() is the complementary

error function
4
Natalie Cartwright received her PhD in mathematics in 2004 from the University of Vermont under
the supervision of Kurt Oughstun.
Z 1
2
ex dx,
2
erfc p (111)

and the function D,N() is defined as

h i1=2
D;N D;N
+
, c , : (112)
The sign function sgnD,N 1 if Im D;N 0, respectively, and u( s)

is the unit step function that turns on at the spacetime point(s) s, defined by
either Im D 0 or Im N 0, depending upon the location of c.
The signal contribution Ac(z, t) for the step-function-modulated sine
wave AH(z, t) with below-resonance-applied signal frequency c
0.5 0 at an observation distance of z 1.2 106 m 3.8 zd into the
dispersive material with Brillouins choice of material parameters is depicted
in Figure 16. Characteristic of this signal contribution is a smooth transition
in frequency from that of the precursor field to the carrier frequency c. The
asterisk in Figure 16 denotes the spacetime value s at which the steepest
descent path P() crosses c. Notice the difference between this result, valid
for a dispersive attenuative medium, and that given in Figure 10 of Part I for
a loss-free medium.
0.03
0.02
0.01
Ac(z,t)
0.01
0.02
0.03
1 1.5 2 2.5
q = ct/z
Figure 16 The uniform asymptotic approximation of the signal contribution for a step-
function-modulated sine wave with below absorption band frequency c 0.5 0 at an
observation distance of z 1.2 106 m 3.8 zd into the single-resonance Lorentz
dielectric with Brillouin's choice of material parameters. The asterisk denotes the
spacetime value s at which the path P() crosses c.
The nonuniform asymptotic approximation for the simple pole contri-

bution at c only accounts for the residue contribution once the path
P() crosses the c. Hence, it is immediately obtained from the integral rep-
resentation of the propagated field [(1.58) in Part I] as
Ac z, t 0, < s ,
1
Ac z, t ezc sin c z c ts , s , (113)
2
Ac z, t ezc sinc z c t, > s ,
as z ! 1. It is then seen that the simple pole contribution results in a field
contribution that is oscillating with fixed angular frequency c and with an
amplitude that is attenuated with propagation distance z > 0 at a constant
time-independent rate given by the attenuation coefficient (c) at that
fixed frequency c. This is the result [(2.37) in Part I] given by Brillouin.
3.5 The Total Field

A uniform asymptotic description of the dynamical evolution of a step-
modulated sine wave of fixed carrier frequency c at a distance of z > 0 into
a Lorentz dielectric is then given by the superposition (72) of the three com-
ponents given in Equations (77), (88), and (110). As an example, the total
propagated field AH(z, t) for the step-function-modulated sine wave with
below-resonance-applied signal frequency c 0.5 0 at an observation
distance of z 1.2 106 m 3.8 zd into the dispersive material with
Brillouins choice of material parameters as a function of the spacetime var-
iable is depicted in Figure 17. The transition from the Sommerfeld pre-
cursor evolution to the Brillouin precursor evolution and then to the
signal contribution is evident.
3.6 The Signal Arrival and the Signal Velocity

The contribution of the simple pole singularity at c occurs when the
original contour of integration C, which extends along the straight line from
ia 1 to ia + 1 in the upper-half of the complex -plane, lies on the
opposite side of the pole singularity than does the Olver-type path P()
through the accessible saddle points. That is, P() and the original integra-
tion contour C lie on the same side of the pole when < s and lie on oppo-
site sides when > s. Consequently, for < s the pole is not crossed when
the original contour is deformed to P() and there is no residue contribution,
whereas for > s the pole is crossed in deforming the contour C to P() and
there is a residue contribution to the asymptotic behavior of the propagated
0.2
0.15
0.1
AH (z,t)
0.05
0.05
0.1
1 1.5 2 2.5
q = ct/z
Figure 17 The uniform asymptotic approximation of the dynamical evolution of the
propagated field contribution for a step-function-modulated sine wave with below
absorption band frequency c 0.5 0 at an observation distance of z 1.2 106
m 3.8 zd into the single-resonance Lorentz dielectric with Brillouin's choice of material
parameters.
wave field. The value of s depends upon which Olver-type path is chosen
for P(). If that path is taken to lie along the path of steepest descent through
the saddle point nearest the pole, then, because (, ) Im{(, )} is
constant along the path of steepest descent, it follows that the value of s
is defined by the expression (Brillouin, 1914, 1960)
sp ,s c , s , (114)
where sp sp() denotes the saddle point that interacts with the pole sin-
gularity. At s, however, the pole contribution is asymptotically negli-
gible in comparison to the saddle point contribution because P() is an
Olver-type path with respect to that saddle point. Consequently, the
particular value of s at which the pole contribution occurs is of little or no
significance to the asymptotic behavior of the propagated wave field
AH(z, t). An example of such an Olver-type path at a fixed spacetime point
> 1 when the two near saddle points SP n are dominant over the two distant
saddle points SPd in a single-resonance Lorentz model dielectric is depicted in
Figure 18. The path P() through the pair of near saddle points SP n and SPn
+
can lie anywhere within the shaded region of the figure. With the path P()
shown in this figure, < s if the pole p lies in the angular frequency interval
c2 < p < c4, > s if p lies within either of the angular frequency inter-
vals 0 p < c2 or p > c4, and s if either p c2 or p c4. If the
Original contour of integration
wc1 wc2 wc4 wc6

w
wc3 wc5
SPn SPn+
w w w+ w+
SPd SPd+
P(q)
Figure 18 A deformed contour of integration P() passing through both the near and
distant saddle points for a fixed spacetime value > 0. This contour is an Olver-type
path with respect to the near saddle point SP+n in the right-half of the complex -plane,
and is an Olver-type path with respect to the near saddle point SP n in the left-half of the
complex -plane. The lighter shaded area indicates the region of the complex -plane
wherein the inequality (, ) < (SPn , ) is satisfied and the darker shaded area indi-
cates the region of the complex -plane wherein the inequality (, ) < (SPd, ) is
satisfied. Reprinted with kind permission of Springer Science+Business Media from
Oughstun (2009), Fig. 15.1, p. 510.
path P() was chosen to be completely in the lower-half of the complex -

plane (such an Olver-type path is possible for the situation illustrated in
Figure 18), then > s for all p 0.
The pole contribution at c is the dominant contribution to the
asymptotic behavior of the propagated wave field when > c > s, where
c is defined as the spacetime value that satisfies the relation (Oughstun,
2009; Oughstun & Sherman, 1988, 1994)
sp , c c , (115)
where sp sp(c) denotes the dominant saddle point at c. Notice

that (c) is independent of the value of because c is real-valued. For
spacetime values < c such that the inequality (sp, ) > (c) is sat-
isfied, the saddle point is the dominant contribution to the asymptotic
behavior of the propagated wave field and the pole contribution is asymp-
totically negligible by comparison. For spacetime values > c such that
the inequality (sp, ) < (c) is satisfied, however, the pole contribution
is the dominant contribution to the asymptotic behavior of the propagated
wave field and the saddle point contribution is asymptotically negligible by
comparison. For example, for the spacetime point depicted in Figure 18,
Equation (115) is satisfied if either c c1 or c c6 and the value of is
then c for either of these two pole locations. Furthermore, for the space
time value depicted in Figure 18, < c and the inequality (SP+n , ) >
(p) is satisfied if c1 < c < c6, and > c and the inequality (SP+n , )
< (c) is satisfied if either 0 c < c1 or c > c6. Consequently, the
pole contribution is asymptotically negligible in comparison to the saddle
point contribution to the propagated wave field at the spacetime value
depicted in Figure 18 if c1 < c < c6, whereas the pole contribution
is the dominant contribution to the asymptotic behavior of the propagated
wave field and the saddle point contribution is asymptotically negligible in
comparison to it if either 0 c < c1 or c > c6.
Based upon these results, it is seen that the signal arrival for a fixed value of
c occurs at the spacetime point c satisfying Equation (115). Notice
that this definition of the signal arrival yields a signal velocity (Oughstun &
Sherman, 1975, 1980, 1988)
c
vc c (116)
c c
that is independent of the initial pulse envelope function and the propaga-
tion distance z, depending only upon the dispersive medium properties and
the value of c. Notice that this pulse velocity measure always satisfies rel-
ativistic causality; that is, vc(c) c for all c. In addition, notice that
Brillouins definition of the signal velocity in Equation (2.38) in Part I is
in terms of s and so is referred to here as vs in order to distinguish it from
the correct definition given in Equation (116).
In summary, the signal arrival in a single-resonance Lorentz model
dielectric separates naturally into two distinct cases dependent upon the
value of the real angular frequency c of the pole in comparison to the crit-
ical angular frequency value SB defined in Equation (63). For values of c
in the angular frequency interval 0 c SB, the signal arrival is due to the
crossing of the isotimic contour () (sp) with the simple pole singu-
larity at c, where sp denotes the location of the upper near saddle
point SP+n for 1 < < 1, the second-order near saddle point SPn at
1, and the near saddle point SP+n for all > 1. For such values of c, the
signal due to the pole contribution at c is preceded by the first and
second precursor fields, and the signal evolves essentially undisturbed as
increases above c. For pole values c > SB, however, the signal arrival first
occurs due to the crossing of the isotimic contour () (SP+d , ) through
the distant saddle point SP+d with the simple pole singularity. This first arrival
occurs at some spacetime point 2 (1, SB) for finite c. At some later
spacetime point 2 (SB, 0), this pole is again crossed, but in the opposite
direction, by the isotimic contour () (SP+n , ) through the upper near
saddle point SP+n , rendering the pole contribution asymptotically less dom-
inant than the second precursor field. Finally, for some still later spacetime
point > 0, the pole is again crossed in the original direction by the isotimic
contour () (SP+n , ) through the near saddle point SP+n so that it
finally becomes asymptotically dominant over all other contributions to
the propagated wave field AH(z,t) for all remaining spacetime values. Con-
sequently, for pulsed sources with c > SB there is the existence of a
so-called prepulse (Oughstun & Sherman, 1988, 1994) due to the interrup-
tion of the signal evolution by the second precursor field which becomes
dominant over the pole contribution for some short spacetime interval.
This prepulse formation is seen to be an integral part of the dynamic evo-
lution of the second precursor field superimposed upon the evolution of
the signal contribution. The spacetime evolution of the signal contribution
when c > SB may then be considered to be separated into three parts: the
so-called prepulse which is preceded by the first precursor and then followed
by the second precursor field superimposed upon the signal contribution,
which is then finally followed by the signal which remains dominant for
all later spacetime points. It is important to keep in mind that the prepulse
is not independent of the signal evolution. Indeed, the prepulse formation is
simply a consequence of the superposition of the signal (or pole) contribu-
tion with the second precursor field which becomes dominant over the sig-
nal for a finite spacetime interval.
The spacetime point c at which the pole contribution is of equal
dominance with the second precursor field is defined by the relation
SPn+ ; c c ; c 0 , (117)
where SP+n () denotes the location of the upper near saddle point SP+n for
c 2 [0, 1), the second-order near saddle point SPn at c 1, and the near
saddle point SP+n for all c > 1. For all spacetime points > c, the pole
contribution is the dominant contribution to the asymptotic behavior of the

propagated wave field AH(z,t) for all c 0.
For angular frequency values c > SB, the pole contribution is the
dominant contribution to the asymptotic behavior of the propagated wave
field AH(z,t) for spacetime points in the interval 2 (c1, c2), where 1 <
c1 < SB and where SB < c2 < c for finite c. The spacetime point
c1 is defined by the relation
SPd+, c1 c ; 1 < c1 < SB , c > SB , (118)
at which point the pole contribution is of equal dominance with the first
precursor field, this pole contribution remaining dominant over the first pre-
cursor for all > c1. However, at the spacetime point c2 defined by
the relation
SPn+,c2 c ; SB < c2 < 0 , c > SB , (119)
the second precursor is of equal dominance with the pole contribution, and
over the subsequent spacetime interval 2 (c2, c), the second precursor
field is dominant over the pole contribution. Finally, at the spacetime point
c, these two contributions are again of equal dominance, and for all
later spacetime points > c, the pole contribution remains as the dom-
inant contribution to the asymptotic behavior of the propagated wave field
AH(z, t).
Physically, the first (or Sommerfeld) precursor field is due to the high-
frequency (above absorption band) energy present in the frequency
spectrum of the initial pulse as filtered by the material dispersion, whereas
the second (or Brillouin) precursor field is due to the low-frequency
(below resonance) energy present in the initial pulse spectrum as filtered
by the material dispersion. The pole contribution is physically due to the
frequency component in the initial pulse spectrum at the angular
frequency c.
The main signal arrival is defined to occur at that spacetime value c
satisfying Equation (115) at which the pole contribution given in
Equation (110) becomes the dominant contribution to the asymptotic
behavior of the propagated wave field AH(z, t). The velocity at which this
spacetime point in the wave field propagates through the dispersive
medium is defined as the main signal velocity, given by
c
vc c , (120)
c c
where c is the vacuum speed of light. Furthermore, for angular frequency

values c > SB, there is the appearance of a so-called prepulse whose front
arrives at the spacetime point c1 satisfying Equation (118) when the
pole contribution becomes the dominant contribution to the asymptotic
behavior of the propagated wave field, and whose back arrives at the
spacetime point c2 satisfying Equation (119) when the second precur-
sor field becomes the dominant contribution to the asymptotic behavior of
the propagated wave field. The velocity at which the front of this prepulse
propagates through the dispersive medium is called the anterior presignal
velocity, given by
c
vc1 c ; c > SB , (121)
c1 c
and the velocity at which the back of this prepulse propagates through the
dispersive medium is called the posterior presignal velocity, given by
c
vc2 c ; c > SB : (122)
c2 c
From the inequalities given in Equations (117)(119), these three pulse
velocities are seen to satisfy the inequality
c c
vc c < vc2 c < < vc1 c < c: (123)
0 SB
Notice that the main signal, anterior presignal, and posterior presignal veloc-
ities depend only upon the value of c and the medium parameters.
The angular frequency dependence of these three signal velocities is pres-
ented in Figure 19 for a single-resonance Lorentz model dielectric charac-
terized by Brillouins choice of the medium parameters (0 4 1016 r/s,
p
p 20 1016 r=s, 0 0.28 1016 r/s). The numerical values for these
signal velocity graphs are obtained (Oughstun, 1978) using Equations (117)
(119) to determine accurate numerical estimates of the values of c(c),
c1(c), and c2(c) through a comparison of the numerically determined
behavior of (, ) at either the near or distant saddle points with the value
of () at the angular frequency value c. As evident from these numerical
results presented in Figure 19, the main signal velocity vc(c) attains a min-
imum value near the resonance frequency of the medium in agreement
with Baerwald (1930). The actual minimum occurs at the value c min
p
0 1 + 20 =0 where (c) attains its minimum value along the positive
real axis. Consequently, the signal velocity does not peak to the vacuum
vc1/c
0.8
vc2/c
0.6
vE/c
v/c
vc/c
0.4
0.2
0
0 w0 5 w1 wSB 10 15
wc (r/s) 1016
Figure 19 Angular frequency dependence of the relative main signal velocity vc(c)/c
1/c(c), relative anterior presignal velocity vc1(c)/c 1/c1(c), and relative poste-
rior presignal velocity vc2(c)/c 1/c2(c) in a single-resonance Lorentz model dielec-
tric characterized by Brillouin's choice of the medium parameters. The dashed curve
describes the frequency dependence of the relative energy transport velocity
vE(c)/c in the dispersive medium. Reprinted with kind permission of Springer Science
+Business Media from Oughstun (2009), Fig. 15.3, p. 518.
speed of light c near resonance, as indicated by Brillouin (1914, 1960), but

rather attains a minimum value near resonance. The dashed curve in
Figure 19 illustrates the frequency dependence of Loudons energy velocity
[see Equation (14)], which is seen to be an upper envelope to the signal
velocity curves. This identification forms part of the basis for the energy
velocity description of the precursor fields given by Sherman and
Oughstun (1981, 1995).
4. EXTENSIONS OF THE SOMMERFELDBRILLOUIN

THEORY TO THE SINGULAR AND WEAK DISPERSIVE
LIMITS
Two special cases of the dispersive signal problem occur in the
so-called singular and weak dispersion limits, as described by Oughstun
et al. (2010). The material absorption in a Lorentz medium, as described
by the amplitude attenuation coefficient () (/c)ni() for real [see
Equation (1.38) in Part I], decreases when either 0 ! 0 or when N ! 0. In

the first case, the material dispersion becomes increasingly localized about
the resonance frequency as 0 ! 0 and is thus referred to as the singular dis-
persion limit. The term singular is used here in the mathematical sense that
the function () fails to be well-behaved at a point in some well-defined
manner, in this case in terms of its differentiability at 0. In the second
case, the absorption vanishes while the material dispersion nr() approaches
unity at all frequencies as N ! 0 and is consequently referred to here as the
weak dispersion limit. In this limit, dn()/d ! 0 as N ! 0. In the group
velocity description, the strength of the material dispersion is typically
measured through the derivative dnr()/d as that is what appears in the
coefficients of the Taylor series expansion of the real propagation factor
n o
Re k in a hypothetical lossless dispersive medium (see
Section 1.4 of Part I.).
The singular and weak dispersion limits have received considerable
attention recently because it is easier to experimentally observe optical pre-
cursors when both limits hold. In that case, the material dispersion is accu-
rately described by the Lorentz model. In particular, Jeong, Dawes, and
Gauthier (2006) observed precursor-like waveforms when a Heaviside
step-function signal propagated through a gas of cold potassium atoms. They
found that the precursors have an amplitude that is comparable to the ampli-
tude of the incident signal and persists for many nanoseconds for the case
when the signal frequency is very close to the resonance frequency of the
Lorentz model dielectric (jc 0 j0 ), making them easier to detect
directly using standard fast detectors and oscilloscopes.
These two limiting cases of singular and weak dispersion are fundamen-
tally different in their effects upon ultra-wideband pulse propagation in gen-
eral and upon dispersive signal propagation in particular and are thus treated
separately in the following two sections.
4.1 Dispersive Signal Propagation in the Singular

Dispersion Limit
From either of Equation (2.21) in Part I or (39), the first-order near saddle
points SP
n of the complex phase function (, ) for a single-resonance
Lorentz medium coalesce into a single second-order saddle point SPn at
2
SPn 1 0 i (124)
3
22 2
when 1 0 + 300 p4 . At the spacetime point 0, the dominant
0
near saddle point SP+n crosses the origin [SP+n (0) 0] so that its contribu-
tion to the asymptotic behavior of the propagated wave field experiences

zero exponential attenuation [ SPn+ 0 , 0 0], the peak amplitude
point in the wave field decaying only as z1/2 with z ! 1. At the subse-
quent spacetime point 1, this contribution to the asymptotic wave
experiences a small (but nonzero) amount of exponential attenuation with
propagation distance as well as a z1/3 algebraic decay as z ! 1, provided
that 0 > 0. In the singular dispersion limit as 0 ! 0, however, the two near
saddle points SPn coalesce into a single second-order saddle point at the ori-
gin when 1 ! 0, resulting in a peak amplitude that experiences zero
exponential attenuation, the amplitude now decaying only as z1/3 as
z ! 1, as described by Brillouin (1960) (see Section 3.3 of Part I).
The numerically determined peak amplitude decay with relative prop-
agation distance z/zd is presented in Figure 20 for an input Heaviside unit
10 0
Peak amplitude of the Brillouin precursor
101
de = 3.02 1013 r/s
d 0 = 0.1 de
d 0 = 0.01 de
102 d 0 = 0.001 d e
103
0 20 40 60 80 100
z/zd
Figure 20 Numerically determined peak amplitude decay of a Heaviside unit step-

function signal with below-resonance carrier frequency c 0.769 0 in a single-
resonance Lorentz model dielectric as a function of the relative propagation distance
z/zd for several decreasing values of the phenomenological damping constant
0 > 0. Reprinted with kind permission of Springer Science+Business Media from
Oughstun (2009), Fig. 15.107, p. 654.
step-function-modulated signal with below-resonance angular carrier fre-

quency c 3.0 1014 r/s in a single-resonance Lorentz-model dielectric
with resonance frequency 0 3.9 1014 r/s (chosen to correspond
to the experimental value in Jeong et al., 2006) and plasma frequency
p 3.05 1014 r/s for several values of the damping constant 0. Here
zd 1 c (125)
denotes the e1 amplitude penetration depth at c. Notice that zd also

depends on the damping constant 0 through Equations (1.30) and (1.38b)
in Part I. Nevertheless, its value is constant for each curve in Figures 20
and 21, varying from zd 7.69 m when 0 e 3.02 1013 r/s to
zd 0.72 cm when 0 0.001e. The dashed line in the figure describes
0.02
z/zd = 10
Brillouin
0.01 precursor
A(z,t)
0
Sommerfeld
precursor
0.01
0.02
2 3 4 5 6
t (1010 s)
Figure 21 Propagated signal with below-resonance angular carrier frequency c
0.769 0 in a single-resonance Lorentz model dielectric in the singular dispersion limit.
The -symbol in the plot marks the time t0 0z/c following the onset of the Brillouin
precursor. Reprinted with kind permission of Springer Science+Business Media from
Oughstun (2009), Fig. 15.09, p. 655.
the pure exponential attenuation described by the BeerLambertBouguer

law5 relation ez=zd for the amplitude decay.
The algebraic power of the measured peak amplitude attenuation pres-
ented in Figure 20 was determined (Oughstun, 2005) by plotting the log-
arithm of the peak amplitude data Apeak B(z/zd)p versus the logarithm
of the relative propagation distance z/zd, showing that the power p increases
from a value approaching 1/2 as z=zd ! 1 to a value approaching 1/3
as z=zd ! 1 when 0 is decreased such that the inequality 0/0 1 is
satisfied.
An example of the numerically computed dynamical wave field evolu-
tion of a unit step-function signal with below resonance angular carrier fre-
quency c 3.0 1014 r/s in the singular dispersion limit is presented in
Figure 21. The propagated wave field was calculated at 10 absorption depths
into a single-resonance Lorentz medium with resonance frequency 0 3.9
1014 r/s, plasma frequency p 3.05 1014 r/s, and phenomenological
damping constant 0 3.02 1010 r/s. Because 0/0 7.74 105, this
case is well within the singular dispersion regime. A distinguishing feature of
this field evolution in the singular dispersion limit is the sharp leading edge of
the Brillouin precursor just preceding the spacetime point ct/z 0 at
which the peak amplitude point occurs, followed by a relatively slow decay
in the field amplitude as time t increases above 0z/c with z > 0. The rate
of this decay with propagation distance for > 0 is determined by the

absorption coefficient Imfk g evaluated at the near saddle point
SP+n , which is found to be proportional to the damping constant
0 (Oughstun, 2009).
4.2 Dispersive Signal Propagation in the Weak Dispersion Limit

In the weak dispersion limit as N ! 0, the material dispersion approaches
that of vacuum at all frequencies, i.e., n() ! 1. This then introduces a
rather curious difficulty into any numerical simulation of signal propagation
in this weak dispersion limit as the number of sample points required to
accurately model the propagated signal behavior rapidly increases as the
number density N goes to zero. In order to circumvent this problem,
5
The BeerLambertBouguer law was originally discovered by Pierre Bouguer, as published in his Essai
dOptique sur la Gradation de la Lumiere (Claude Jombert, Paris, 1729) and subsequently cited by Johann
Heinrich Lambert in Photometri (V. E. Klett, Augsburg, 1760). The result was then extended by August
Beer in Einleitung in die hohere Optik (Friedrich Viewig, Braunschweig, 1853) to include the concen-
tration of solutions in the expression of the absorption coefficient for the intensity of light.
Oughstun et al. (2010) developed an approximate equivalence relation that

allows one to compute the propagated wave field behavior in an equivalent
dispersive medium that is strongly dispersive. This approximate equivalence
relation, which becomes exact in the limit as N ! 0 for each given medium,
follows from the general integral representation of the propagated wave field
given in Equation (1.58) in Part I.
Two different propagation problems for the same initial pulse are iden-

tical, provided that the relation k 1 z1 t1 k 2 z2 t2 is satisfied
for all . Equating real and imaginary parts for real results in the pair of
relations
1 z1 t1 2 z2 t2 , (126)
1 z1 2 z2 , (127)
both of which must be satisfied for all . For the absorptive part, the equiv-
alence relation
1
z2 z1 , 8 (128)
2
results. If the two media differ only through their densities, then
p 1
n 1 + Ng ! 1 + Ng as N ! 0,
2
because () (/c)ni() for real , where g 4q2e =me =
2 20 + 2i0 . In this case, ni 12 Ngi and the above equivalence
relation becomes
N1
z2 z1 , (129)
N2
the accuracy of this result increasing as N1, N2 ! 0 with N2 > N1. The
corresponding equivalence relation for the phase part then results in

N1 z1
t2 t1 + 1 , (130)
N2 c
p
which is simply a uniform displacement in time. Because p 4Nq2e =me ,
the pair of equivalence relations given in Equations (129) and (130) may be
expressed as
2p1
z2 z1 , (131)
2p2
!
2p1 z1
t2 t1 + 1 : (132)
2p2 c
The first part of this equivalence relation is the essential part. For example, if
N1/N2 1 102, then z2 z1 102 and t2 t1 (0.33 108 s/m)z1.
In that case, the propagated wave field structure illustrated in Figure 21 also
applies to the case when the medium plasma frequency p is reduced by
the factor 10 and the propagation distance z is increased by the factor 100
provided that the time origin is adjusted according to Equation (130).
The accuracy of this equivalence relation is illustrated in the sequence of
propagated signal plots presented in Figures 2224. In each case, the initial
signal has below-resonance angular carrier frequency c 3.0 1014 r/s
and is propagating in a single-resonance Lorentz model dielectric with res-
onance frequency 0 3.90 1014 r/s and phenomenological damping
constant 0 3.02 1012 r/s. The plasma frequency in case 1, illustrated
in Figure 22, is p1 3.05 1014 r/s, resulting in an absorption depth
14
0.1 wp1 = 3.05 10 r/s
4
zd = 7.238 10 m
0.05
A(z,t)
0.05
0.1
0 1 2 3 4
12
t z/c ( 10 s)
Figure 22 Propagated wave field at five absorption depths in a single-resonance
Lorentz medium with plasma frequency p1 3.05 1014 r/s. Notice that A(z, t) 0
for all t < z/c.
of zd 1(c) 7.238 105 m. In case 2, illustrated in Figure 23, the

plasma frequency has been reduced by two orders of magnitude (and the
number density by four orders of magnitude) to the value p2 3.05
1012 r/s, resulting in an absorption depth of zd 1(c) 0.458 m.
Finally, in case 3, illustrated in Figure 24, the plasma frequency has been
reduced by an additional two orders of magnitude (and the number density
by eight orders of magnitude from that in case 1) to the value p3 3.05
1010 r/s, resulting in an absorption depth of zd 1(c) 4.582 103 m.
Each of these propagated wave field structures was computed at z 5zd
with a 223 point FFT and the time origin in each was shifted by the amount
z/c in order to align them temporally for the sole purpose of ease of
comparison.
Each of the propagated wave field structures presented in Figures 2224
exhibits the field structure described by the uniform asymptotic theory for a
below-resonance carrier frequency: the high-frequency Sommerfeld pre-
cursor As(z, t) arriving at the speed of light point t z/c, followed by the
evolution of the low-frequency Brillouin precursor Ab(z, t), which is then
12
0.1 wp2 = 3.05 10 r/s
zd = 0.458 m
0.05
A(z,t)
0.05
0.1
0 1 2 3 4
12
t z/c ( 10 s)
Figure 23 Propagated wave field at five absorption depths in a single-resonance
Lorentz medium with plasma frequency p2 3.05 1012 r/s. Notice that A(z, t) 0
for all t < z/c.
10
wp3 = 3.05 10 r/s
0.1 3
zd = 4.582 10 m
0.05
A(z,t)
0.05
0.1
0 1 2 3 4
12
t z/c ( 10 s)
Figure 24 Propagated wave field at five absorption depths in a single-resonance Lorentz
medium with plasma frequency p3 3.05 1010 r/s. Notice that A(z, t) 0 for all t < z/c.
followed by the signal contribution Ac(z, t) that is primarily oscillating at the

angular carrier frequency c of the input signal.
The unique advantage of the modern asymptotic description lies in the
fact that the complete spacetime evolution of each component of the prop-
agated wave field is described separately from the other components. Each
individual field component, as described by the uniform asymptotic theory
presented in Section 5 with numerically determined saddle point locations,
is presented in Figure 25 for case 2 (cf. Figure 23). These results clearly reveal
the full spacetime evolution of the separate Sommerfeld As(z, t) and
Brillouin Ab(z, t) precursors as well as the signal evolution Ac(z, t) that is par-
tially obscured in the total field evolution A(z, t).
5. PULSED PLANE-WAVE TRANSMISSION THROUGH

AN INTERFACE
The classical description of a dispersive signal propagation is restricted
to the special case of propagation through an unbounded, Lorentz-type
dielectric. In practical applications and experiments involving pulsed
AS(z,t)
0
AB(z,t)
0
Ac(z,t)
0
A(z,t)
0 0.5 1 1.5 2 2.5 3 3.5 4

t z/c ( 1012 s)
Figure 25 Separate Sommerfeld precursor As(z,t), Brillouin precursor Ab(z, t), and signal
Ac(z, t) components of the propagated wave field and their superposition to yield the
total propagated field A(z, t) at five absorption depths in a single-resonance Lorentz
medium in the weak dispersion limit with plasma frequency p3 3.05 1012 r/s.
Notice that A(z, t) 0 for all t < z/c.
electromagnetic wave propagation, the initial pulse must be transmitted from

one medium (typically vacuum) into a dispersive half-space. The study of a
linearly polarized electromagnetic plane-wave pulse that is propagating across
a planar interface separating vacuum from a dispersive medium with an arbi-
trary angle of incidence was first considered by Colby (1915). As in the
unbounded case, the transmission problem also has an exact integral represen-
tation in the frequency domain when the medium is homogeneous, isotropic,
linear, and temporally dispersive, as considered here. More specifically,
Colby, inspired by the recently published work of Sommerfeld (1914) and
Brillouin (1914), set out to study the transmission of a step-modulated sine
wave from vacuum into a Lorentz half-space. However, he restricted his anal-
ysis to early times and frequencies removed from the absorption band of the
material so that damping could be assumed negligible. He then expanded the
integrand in a Laurent series centered at the origin which enabled him to
invert each term separately in terms of Bessel functions.
Much later, Gitterman and Gitterman (1976) again considered the trans-
mission of a step-modulated signal into a Lorentz half-space. They used both
low- and high-frequency expansions of the dielectric permittivity to obtain
closed-form expressions for the locations of the saddle points of the phase
function, applying nonuniform asymptotic methods to produce a set of
explicit time-domain descriptions of the transmitted signal, each with its
own region of validity. The combination of their approximations for the
saddle point locations with nonuniform asymptotic theory severely limits
the accuracy of these descriptions, even within each of their regions of valid-
ity (see Cartwright, 2011). The same problem, but limited to the case of nor-
mal incidence, was considered by Mokole and Samaddar (1999). Following
the work of Oughstun and Sherman (1994) and based upon accurate
approximations for the saddle point locations and employing uniform
asymptotic methods, they provided a time-domain asymptotic description
of the early-time response of the transmitted field. In addition, they recast
the integral representation of the reflected field into a form that is
readily amenable to numerical calculation.
A complete uniform asymptotic description of a step-modulated sinusoid
obliquely incident on a Lorentz half-space, valid as the propagation (or pen-
etration) distance into the Lorentz medium increases, was finally given by
Cartwright (2011). The asymptotic approximations are relatively straight-
forward extensions of the asymptotic approximations derived in the study
of linearly polarized electromagnetic plane-wave pulse propagation through
unbounded media. The expressions are completely uniform in time and thus
improve and extend the early-time results of Colby (1915), improve and
unify the limited approximations provided by Gitterman and Gitterman
(1976), and extend the work of Mokole and Samaddar (1999) to all times
and arbitrary angle of incidence. These closed-form solutions provide both
the arrival time and angle of refraction of prominent physical features of the
transmitted pulse.
5.1 Formulation
Consider an incident, linearly polarized electromagnetic plane-wave pulse
propagating in Region 1 (z < 0) that impinges upon the planar boundary
at z 0 with Region 2 (z > 0), where Region 1 is vacuum (E1 E0)
and Region 2 is a single-resonance Lorentz medium, as depicted in
Figure 26. The relative dielectric permittivity of Region 2 is described by
the frequency-dependent expression
x
Hr
Er kR Et kT
Region 1
e1 = e0 , m1 = m0 aR Ht
aT
n z
a
Region 2
Ei kI e2 = e2(w) , m2 = m0
Hi
Figure 26 Geometric illustration of a pulsed TE-polarized plane wave obliquely incident
upon a planar interface at z 0.
b2
E2 1 , (133)
2 20 + 2i0
[see Equation (1.30) in Part I]. In both regions, the magnetic permeability is
that of free space ( 0).
If the incident signal Ei(r, t) is a transverse electric (TE)-polarized plane
wave with angle of incidence , where 0 < /2, then the transmitted
electric and magnetic field components are given exactly by Ex Ez
Hy 0 and the integral representations (see Marozas, 1998 for a thorough
derivation)
0 1 0 1
Ey Z ia + 1 T? ;
B C 1 B C
@ Hx A E i @ T? ;cos N2 ;=0 A
2 ia1 (134)
Hz T? ;sin =0
expikT ; r it
d:
Here, a is greater than the abscissa of absolute convergence of the incident
signal, E~ is the temporal Fourier transform of the electric field of the
i
incident signal, r is the position vector in Region 2, and T?() is the trans-
mission coefficient for the TE mode, given by
2
T? ; : (135)
1 + N2 ;
The wavevector in Region 2 is given by
D E
kT ; sin,0, N2 ;cos , (136)
c c
where the function N2(; ) is defined as
s
b2 =cos
N2 ; 1 2 (137)
20 + 2i0
p
with 0 0 =E0 denoting the impedance of free space.
If the incident signal Ei(r, t) is a transverse magnetic (TM)-polarized
plane wave with angle of incidence , where 0 < /2, then the
transmitted electric and magnetic field components are given exactly by
Ey Hx Hz 0 and the integral representations
0 1 0 1
Ex Z ia + 1 Tjj ;cos N2 ;=n2
B C 1 B C
E~ B C
i
@ Ez A @ Tjj ;sin=n2 A
2 ia1 (138)
Hy Tjj ;n2 =0
expikT ; r it
d:
Here, Tjj(; ) is the transmission coefficient for the TM mode, given by
2n2
Tjj ; (139)
n22 + N2 ;
p
where n2 E2 =E0 is the complex index of refraction in Region 2.
As for the SommerfeldBrillouin problem, the incident electric field at
z 0 is given by the canonical step-modulated sinusoid
E i t utsinc t, (140)
where u(t) denotes the Heaviside step function and c 2 + is the fixed car-
rier frequency of the signal. Substitution of the temporal Fourier transform
of the incident electric field E~ into either Equation (134) or (138), and
i
scaling all angular frequencies by the resonance frequency 0 results in the

expression
Z ia + 1
1 1 d0
H Re ^
g;exp ^ ; d
, ^ , (141)
2 ia1 ^ ^c c
where H denotes one of the six field components given in either

^
Equation (134) or (138) and g; is its corresponding amplitude function
without the spectral factor E~ , which appears separately in the integrand
i
of Equation (141). Here, the scaled variables are indicated by a caret (e.g.,
^ =0 ), the complex phase function ,
^ is defined as
hz x i
^ ; i
, ^ sin ,
^ cos N2 ; (142)
d d
with
d x sin + zcos (143)
the penetration depth into Region 2 along the refracted ray, and ct/d is a
spacetime variable defined for all d > 0. These exact integral representa-
tions of the transmitted components are now in the form of those considered
in the study of a step-modulated sine wave propagating through an
unbounded single-resonance Lorentz material (see Section 3), the only dif-
ferences being in the additional amplitude functions and the functional form
of the phase function. Thus, the same uniform asymptotic methods that are
used in the study of a step-modulated sine wave propagating through an
unbounded, single-resonance Lorentz material may be used in the study
of a step-modulated sine wave impinging on a single-resonance Lorentz
half-space. As described by Cartwright (2011), the similarities and differ-
ences between the two problems are now given along with the final
asymptotic expressions for the transmitted pulse.
5.2 Asymptotic Description

First, as stated by Colby (1915), for values of 1, convergence requires the
contour to be enclosed in the upper-half plane and an application of Jordans
lemma shows that the integral is identically zero; that is, the signal does not
experience superluminal travel. For values of > 1, the contour must be
enclosed in the lower-half plane. However, the phase function has two
branch cuts in the lower half plane, symmetrically located across the imag-
inary axis between the branch points
p
^ 1, 2 i^0 1 ^0 2 ,
(144)
q
^ 3, 4 i^0
1 + b^2 =cos 2 ^0 2 : (145)
Rather than perform integration around these branch cuts, saddle point
methods are the preferred method of analysis when > 1.
When compared to the complex phase function in the unbounded case,

viz., i ^ n2
^
with ct/z, the phase function given in
Equation (1.47) in Part I may be considered as applicable to a denser mate-
rial with index of refraction scaled by z=dcos and with a delayed space
time variable . Thus, the number and the behavior of the saddle points of
will be similar to those in the unbounded case. Specifically, there exist
four saddle points of the complex phase function (142) whose locations
move in the complex -plane ^ for fixed x,z and increasing > 1. The
two distant saddle points ^ SPD ; are symmetrically located about the
imaginary axis and evolve in the region jj ^ > j ^ 3 j. They are
infinite-order saddle points located at 2i0 1 when 1, become
first-order saddle points in the finite complex plane for > 1, and approach
the branch points ^ 3, 4 as ! 1. Their contribution to the integral
(141) is responsible for the early-time, high-frequency behavior of the
field, the Sommerfeld precursor S. The two near saddle points
^ SPN ; reside on the upper and lower imaginary axis when 1+,
travel vertically toward each other with increasing , coalesce at a point
on the negative imaginary axis at the spacetime point 1, and then split
apart and are symmetrically located about the imaginary axis as they
approach the branch points ^ 1, 2 in the lower-half plane. Their contribu-
tion to the integral Equation (141) is responsible for the low-frequency
behavior of the field, the Brillouin precursor B. As the contour of inte-
gration is continuously deformed through the valleys of the accessible sad-
dle points, the contour will encircle the simple pole located at ^ c , which is
responsible for the steady-state portion of the propagated field, the signal
contribution c. The uniform methods of Handelsman and Bleistein
(1969) for infinite-order saddle points located at infinity, Chester et al.
(1957) for two coalescing saddle points, and L. Felsen and Marcuvitz
(1973) for saddle point coalescing with a pole may then be applied to
obtain expressions for the Sommerfeld precursor, the Brillouin precursor,
and the signal contribution, respectively. The asymptotic approximation to
the transmitted field, valid as d ! 1, is then given by the superposition of
these three contributions
H S + B + c : (146)
The first term in each asymptotic expansion, uniformly valid for all 1, is
given in the following (Cartwright, 2011).
5.2.1 Sommerfeld Precursor

^ SPD denote the locations of the distant saddle points in the complex
Let
^
-plane at the spacetime point . The asymptotic approximation of the
Sommerfeld precursor S is given by the expression

d0
S r,ct=d Re exp i B;
c
(147)
d0 d0
0 ;J0 A; 2iA; 1 ;J1 A; ,
c c
as d ! 1, for all 1, where Jn() denotes the Bessel function of the first
kind of integer order n. The coefficients appearing in (147) are given by
n o n o
A; Im ^ SPD+ , B; iRe ^ SPD+ , (148a)
2 !1=2

1 4 ig ^ SPD+ 1 4A3
0 ;
2 ^ SPD+ ^ c 2A i2
^ SP +
D
!1=2 3
(148b)

ig^ SP 1 4A 3
+ D
2 5,
^ SPD
^c 2A i ^ SPD
2 !1=2

1 4 ig ^ SPD+ 1 4A3
1 ;
4A ^ SPD+ ^ c 2A i2
^ SP + D
!1=2 3
(148c)

ig^ SPD 1 4A3
2 5,
^ SPD
^c 2A i ^ SPD
where the arguments , have been omitted in Equation (148a)(148c) for

ease of presentation.
5.2.2 Brillouin Precursor

Let ^ SPN denote the locations of the near saddle points in the complex
^
-plane at the spacetime point . The asymptotic approximation of the
Brillouin precursor B(z, t) is given by
( (
d0 1 c 1=3 i2=3
B r,ct=d Re exp 0 ; e Ai;
c 2 d0
i
^ SPN+ ;h + ;
g
^ SPN+
^c
!
i
+ ^ SPN ;h ;
g
^ SPN
^c

1 c 2=3 i4=3 1
+ 1=2 e Ai ;
21 ; d0
i
g^ SPN+ ;h + ;
^ SPN+
^c
!))
i
g^ SPN ;h ; ,
^ SPN
^c
(149)
where Ai() denotes the Airy function and the coefficients appearing in
Equation (149) are given by
h i1=3
1h i
1=2 3
0 ; ^ SPN+ +
^ SPN , 1 ; ^ SPN+
^ SPN ,
2 4
(150a)
" 2=3 # !1=2
1=2
c 21
; 1 ei2=3 , h ; 2
,
d0 ^ SPN+
(150b)
for 61. At the critical value 1, when the two near saddle points coalesce
into a single second-order saddle point N, these coefficients take on the
limiting values
!1=3
2
lim h ; 3 h1 1 ;, (151a)
!1 N
" #
1 i i i
lim ^ SPN+ h + +
g ^ SPN h
g gN h1 ,
!1 2 ^ SPN+ ^c ^ SPN
^c N ^c
(151b)
" #
1 i i
lim ^ SPN+ h +
g ^ SPN h
g
!1 21=2 ^ SPN+ ^c ^ SPN
^c
1
(151c)
2 d ^
ig
h1 :
d ^ ^ c
^ N
Again, the arguments , have been omitted in Equations (150)(151), for

ease of presentation.
5.2.3 Signal Contribution

^ SPD and
Let ^ SPN denote the locations of the distant and near saddle
^
points, respectively, in the complex -plane at the spacetime point . The
asymptotic approximation of the signal contribution is given by
2c r,ct=d
( " r#
d0
Re i u g ^ SPD+ ; i sgnD erfc i sgnD D ;
c
r
d0 1 c d0
exp ^ c , ; + exp ^ SPD , ;
+
c D ; d0 c
r!
d0
+ i u g^ SPN+ ; i sgnN erfc i sgnN N ;
c
r
d0 1 c d0
exp ^ c , ; + exp ^ SPN , ;
+
c N ; d0 c

d0
+ u s Re ig ^ c ;exp ^ c ,; ,
c
(152)
as d ! 1. Here, u i is the residue of E~

i
^ at
^ c , erfc() is the comple-
mentary error function
Z 1
2
ex dx,
2
erfc p (153)

and the function D,N(; ) is defined as
h i1=2
D;N ; ^ D;N
+
^ c , ;
, ; : (154)
The sign function sgnD,N 1 if Im D;N ; 0, respectively, and

u( s) is the unit step function that turns on at the spacetime points s
when the path of integration encloses the pole located at c, which is defined
by either Im D 0 or Im N 0, depending upon the location of ^c.
5.2.4 The Total Transmitted Field

Equations (147)(154) are uniformly valid for all > 1 and increase in
accuracy as d ! 1. To provide a visual example, Cartwright (2011) consid-
^ c 3=4
ered an incident step-modulated sine wave with carrier frequency
obliquely incident at an angle of 10 on a Lorentz half-space with
the same (uncharacteristically absorbent) material parameters used by
Brillouin (1960), as well as by Mokole and Samaddar (1999),
p
0 4 1016 s1 , b^ b=0 5=2, and ^0 0 =0 0:07. The three
contributions S,B,c and the total transmitted field Ey S + B + c
of the electric field component of the TE-polarized plane-wave pulse at
the observation point (x, z) (0.1 m, 0.2 m) are shown in Figure 27
as a function of . In this case, g; ^ T? ; ^ is used in
Equations (147)(154). The transmitted magnetic field components
Hx,Hz for the TE case are shown in Figure 28a and b, respectively.
Here, the computations of Hx and Hz require g; ^ T? ;^
^
cos N2 ;= 0 ^ T? ;sin
and g; ^ =0 , respectively [see
Equation (134)].
B
Ey (V/m)
c
0.2
0.1
0
0.1
0.2
1 2 3 4 5 6
q
Figure 27 The three contributions S, B, c and the total transmitted field Ey S +
B + c of the electric field component of the TE-polarized plane-wave pulse with
10 at the observation point (x, z) (0.1 m, 0.2 m).
A 104
6
Hx, TE case (A/m) 2
8
1 3 4 5
q
B 105
8
6
Hz, TE case (A/m)
6
1 2 3 4 5
q
Figure 28 The magnetic field components (a) Hx and (b) Hz for the TE-polarized plane-
wave pulse. The same material and pulse parameters and observation point used for
Figure 27 are used here.
5.3 Transmitted Signal Evolution

Notice first that the lower edge of the absorption band of the material, which
corresponds with Re ^ 1 [see Equation (144)], is a function of the material
properties only and hence remains fixed with changes in . Thus, normal-
ized frequencies below 1 ^0 2 1=2 will comprise the low-frequency con-
tribution to the propagated field regardless of the angle of incidence. In
contrast, the upper edge of the absorption band, which corresponds with
Re ^ 3 , is a function of both material properties and the angle of inci-

dence such that
^ 3 1:
lim Re (155)
!=2
Hence, only normalized frequencies above 1 + b^2 =cos 2 ^0 2 1=2 will be

apparent in the Sommerfeld precursor S as d ! 1. The instantaneous
angular frequency of S, approximately given by Re SP+D(; )
(Oughstun, 2009; Oughstun & Sherman, 1988), is infinite at 1 (where
^ 3 as ! 1. As with the
the field amplitude is zero) and approaches Re
unbounded case, the amplitude of S initially increases with to a maxi-
mum value and then decays exponentially (see Figure 27).
The real angle of refraction T, perpendicular to the planes of constant
phase, is determined from the Equation
tan
tan T : (156)
^ g
RefN2 ;
In order to determine the angle of refraction ST of the front of the pulse,
Equation (156) is evaluated at the distant saddle point SP+D(; ), yielding
!
tan
ST arctan : (157)
RefN2 SPD+ ;g
Because Re {N2(SP+D(; ))}! 1 as ! 1+, it follows that ST ! as

! 1+, in keeping with the physical interpretation that the material is trans-
parent to high frequencies. As increases from unity, the distant saddle point
^ 1 , a zero of N2 ;.
SP+D(; ) approaches ^ Thus, the Sommerfeld precur-
sor continuously refracts away from the normal of the plane of incidence
with ST ! /2 as ! 1. However, the amplitude of S becomes negli-
gible as ! 1. These results are in keeping with the results given by
Gitterman and Gitterman (1976), but contradict the conclusion by Colby
that the ray will be bent continuously toward the normal (Colby, 1915).
The low-frequency contribution to the transmitted field is given by the
Brillouin precursor B as d ! 1. In the unbounded case, the Brillouin pre-
cursor of a step-modulated sinusoid is known to have a peak amplitude point
that decays algebraically with propagation distance. This peak amplitude
point occurs when a saddle point of the phase function crosses the origin
at the spacetime point equal to the dc value of the refractive index. In
the case of the step-modulated sinusoid transmitted into a Lorentz
half-space, the near saddle point SP+N also crosses the origin, thereby pre-
serving this special feature of the Brillouin precursor, but now the peak
amplitude point occurs at the spacetime point
z cos x sin
0 N2 0; + (158)
d d
and is refracted at an angle of
!
tan sin
0T arctan arctan p (159)
RefN2 0;g cos 2 + b^2
into Region 2, in agreement with the numerical computations performed by
Blaschak and Franzen (1995). These authors studied precursor propagation
assuming a trapezoidal modulated sinusoid at oblique incidence on a Lorentz
p
half-space with material parameters b 24 1016 s1 , 0 4 1016 s1,
and 0 0.28 1016 s1. With an angle of incidence 80, the authors
found, using purely numerical estimates, the leading edge of the Sommerfeld
precursor to be at an angle of ST 80 and the peak of the Brillouin pre-
cursor to be 0T 40, in agreement with that given by both Equations (157)
and (159). The angle of transmission of the entire Brillouin precursor BT is
found by substitution of SP+N(; ) into Equation (156). As ! 1,
SPN+ ! ^ 1 and BT ! 0 so that the Brillouin precursor is continuously
bent toward the normal with increasing , in agreement with that given
in Gitterman and Gitterman (1976).
The signal contribution, characterized by the smooth turn on of oscilla-
tions at the carrier frequency of the incident pulse, is refracted at the constant
angle

tan
T arctan
c
(160)
RefN2 ^ c ;g
for all . The velocity of the main signal is defined as vc c/c, where c is
the spacetime point at which the signal contribution is of equal dominance
with the Brillouin precursor (Oughstun & Sherman, 1980), given implicitly
by the Equation
^ c ;:
Re SPN+ c , c ; Re (161)
An example of the dependence of the refractive angles ST, BT, cT on is
shown in Figure 29 for the same material and signal parameters, angle of
incidence 10, and observation point (x, z) (0.1 m, 0.2 m) used
30
Transmission angle 25
S
20 a T
B
a T
15 c
a T
10
1 1.5 2 2.5 3 3.5 4 4.5 5

q
Figure 29 The angle of refraction of the Sommerfeld precursor ST, the Brillouin precur-
sor BT , and the signal contribution cT as a function of . The open circle represents the
angle of refraction 0T of the peak amplitude point of B and the asterisk represents the
spacetime point c at which the main signal velocity is defined. The same material and
pulse parameters, angle of incidence, and observation point used for Figure 27 are
used here.
in Figure 27. The open circle represents the angle of refraction 0T of the
peak amplitude point 0 at which B experiences algebraic attenuation.
The asterisk represents the spacetime point c at which the main signal
velocity is defined.
The relative velocity of the main signal vS/c 1/c for various angles of
incidence is shown in Figure 30 using the same material and signal param-
eters and observation point (x, z) (0.1 m, 0.2 m) used in Figure 27.
Here, the angles of incidence are 0, 10, 30, 50, 70. As with the
unbounded case, the signal velocity obtains its minimum approximately
^ 0 and is subluminal for all .
at ^
6. CONCLUDING REMARKS
The fundamental analysis that was begun by Sommerfeld and
Brillouin in 1914 concerning the propagation velocity of a pulse in a Lorentz
model dielectric has now seen its conclusion in 2007 with the publication of
0.7
0.6
0.5
0.4
vs/c
0.3
0.2
0.1
0
0 1 2 3 4 5
w, s-1
Figure 30 The relative velocity of the main signal vS/c 1/c for angles of incidence
0, 10, 30, 50, 70depicted by the solid, dotted, dashed, and dashed-dotted cur-
ves, respectively. The same material and pulse parameters and observation point used
for Figure 27 are used here.
the complete uniform asymptotic description of the dynamical signal evo-

lution (Cartwright & Oughstun, 2006), some 93 years in the making.
Because of the mathematical rigor presented by this solution, this asymptotic
description represents the standard by which all other approximate descrip-
tions, especially the group velocity description, should be compared in order
to quantify their accuracy. In particular, the statement by Brillouin that the
group velocity is only valid when it agrees with the physically correct signal
velocity remains entirely valid as stated. That said, claims, both theoretical
and experimental, regarding so-called superluminal pulse velocities need
to be carefully reexamined in light of this rigorous theory, the results of
which can also be extended into the nonlinear optics domain
(Palombini & Oughstun, 2010).
This asymptotic theory has also been extended to other dispersive
models, particularly the Debye model (Oughstun, 2005) and the Drude
model (Cartwright & Oughstun, 2009), as well as to a variety of other
canonical pulse types (see Oughstun, 2009). More importantly, it has also
led to a physical description (Oughstun & Sherman, 1988; Sherman &
Oughstun, 1995) of dispersive pulse dynamics in dispersive attenuative
media exhibiting resonance dispersion that reduces to the group velocity
description in the limit of vanishing damping (i.e., in the singular dispersion
limit). Recent attempts to simplify this description by using inexact approx-
imations are, in the authors view, completely unnecessary as the asymptotic
expressions presented here can easily be coded on a wide variety of personal

computers.
Finally, the recent extension of this asymptotic theory to the transmission
of a pulse across a dielectric interface (Cartwright, 2011; Palombini &
Oughstun, 2010) has opened up an entirely new avenue of research in which
the individual precursor fields, which are a characteristic of the dispersive
medium, can be used for medical imaging as well as therapy. Security and
surveillance applications are all too obvious.
ACKNOWLEDGMENTS
The research review presented in this chapter by the authors was supported, in part, by the
United States Air Force Office of Scientific Research (AFOSR).
REFERENCES
Baerwald, H. (1930). U ber die fortpflanzung von signalen in disperdierenden medien. Annals
of Physics, 7, 731760.
Blaschak, J. G., & Franzen, J. (1995). Precursor propagation in dispersive media from short-
rise-time pulses at oblique incidence. Journal of the Optical Society of America B: Optical
Physics, 12(7), 15011512.
Bleistein, N. (1966). Uniform asymptotic expansions of integrals with stationary point near
algebraic singularity. Communications on Pure and Applied Mathematics, XIX(4), 353370.
Bleistein, N., & Handelsman, R. (1975). Asymptotic expansions of integrals. New York: Dover.
Brillouin, L. (1914). U ber die fortpflanzung des licht in disperdierenden medien. Annals of
Physics, 44, 204240.
Brillouin, L. (1960). Wave propagation and group velocity. New York: Academic.
Cartwright, N. A. (2004). Uniform asymptotic description of the unit step function modulated sinu-
soidal signal. PhD dissertation: University of Vermont.
Cartwright, N. A. (2011). Electromagnetic plane-wave pulse transmission into a Lorentz
half-space. Journal of the Optical Society of America. A, 28(12), 26472654.
Cartwright, N. A., & Oughstun, K. E. (2006). Uniform asymptotics applied to ultrawideband
pulse propagation. SIAM Review, 49(4), 628648.
Cartwright, N. A., & Oughstun, K. E. (2009). Ultrawideband pulse propagation through a
lossy plasma. Radio Science, 44. http://dx.doi.org/10.1029/2009RS004201.
Chester, C., Friedman, B., & Ursell, F. (1957). An extension of the method of steepest
descents. Proceedings of the Cambridge Philosophical Society, 53, 599611.
Colby, W. (1915). Signal propagation in dispersive media. Physics Review, 5, 253265.
Felsen, L., & Marcuvitz, N. (1973). Radiation and scattering of waves. Englewood Cliffs, NJ:
Prentice-Hall, Inc.
Felsen, L. B. (1963). Radiation from a uniaxially anisotropic plasma half-space. IEEE Trans-
actions on Antennas and Propagation, AP-11, 469484.
Felsen, L. B., & Marcuvitz, N. (1959). Modal analysis and synthesis of electromagnetic fields.
Polytechnic Institute of Brooklyn Microwave Research Institute Report.
Gitterman, E., & Gitterman, M. (1976). Transient processes for incidence of a light signal on a
vacuum-medium interface. Physical Review A, 13(2), 763776.
Handelsman, R. A., & Bleistein, N. (1969). Uniform asymptotic expansions of integrals that
arise in the analysis of precursors. Archive for Rational Mechanics and Analysis, 35, 267283.
Jeong, H., Dawes, M. C., & Gauthier, D. J. (2006). Direct observation of optical precursors
in a region of anomalous dispersion. Physical Review Letters, 96, 143901.
Loudon, R. (1970). The propagation of electromagnetic energy through an absorbing dielec-
tric. Physica A, 3, 233245.
Marozas, J. A. (1998). Angular spectrum representation of ultrawideband electromagnetic pulse
propagation in lossy, dispersive dielectric slab waveguides. PhD dissertation: University of
Vermont.
Mokole, E. L., & Samaddar, S. N. (1999). Transmission and reflection of normal incident,
pulse electromagnetic plane waves upon a Lorentz half-space. Journal of the Optical Society
of America B: Optical Physics, 16(5), 812831.
Neufeld, J. (1970a). Revised formulation of the macroscopic Maxwell theory. I. Fundamen-
tals of the proposed formulation. II Nuovo Cimento, LXV B(1), 3368.
Neufeld, J. (1970b). Revised formulation of the macroscopic Maxwell theory. II. Propaga-
tion of an electromagnetic disturbance in dispersive media. II Nuovo Cimento, LXVI B(1),
5176.
Olver, F. W. J. (1970). Why steepest descents? SIAM Review, 12(2), 228247.
Oughstun, K. E. (1978). Propagation of optical pulses in dispersive media. PhD dissertation: The
Institute of Optics, University of Rochester.
Oughstun, K. E. (2005). Dynamical evolution of the Brillouin precursor in Rocard-Powles-
Debye model dielectrics. IEEE Transactions on Antennas and Propagation, 53(5),
15821590.
Oughstun, K. E. (2006). Electromagnetic and optical pulse propagation 1. New York: Springer.
Oughstun, K. E. (2009). Electromagnetic and optical pulse propagation 2. New York: Springer.
Oughstun, K. E., Cartwright, N. A., Gauthier, D., & Jeong, H. (2010). Optical precursors in
the singular and weak dispersion limits. Journal of the Optical Society of America B: Optical
Physics, 27(8), 16641670.
Oughstun, K. E., & Sherman, G. C. (1975). Optical pulse propagation in temporally disper-
sive Lorentz media. Journal of the Optical Society of America, 65(10), 1224A.
Oughstun, K. E., & Sherman, G. C. (1980). Comparison of the signal velocity of a pulse with
the energy velocity of a time-harmonic field in Lorentz media. In Proceedings of the URSI
symposium on electromagnetic wave theory, Munich.
Oughstun, K. E., & Sherman, G. C. (1988). Propagation of electromagnetic pulses in a linear
dispersive medium with absorption (the Lorentz medium). Journal of the Optical Society of
America B: Optical Physics, 5(4), 817849.
Oughstun, K. E., & Sherman, G. C. (1989). Uniform asymptotic description of electromag-
netic pulse propagation in a linear dispersive medium with absorption (the Lorentz
medium). Journal of the Optical Society of America. A, 6(9), 13941420.
Oughstun, K. E., & Sherman, G. C. (1994). Pulse propagation in causal dielectrics. Berlin:
Springer-Verlag.
Palombini, C. L., & Oughstun, K. E. (2010). Optical precursor fields in nonlinear pulse
dynamics. Optics Express, 18(22), 2310423120.
Sherman, G. C., & Oughstun, K. E. (1981). Description of pulse dynamics in Lorentz media
in terms of the energy velocity and attenuation of time-harmonic waves. Physical Review
Letters, 47, 14511454.
Sherman, G. C., & Oughstun, K. E. (1995). Energy velocity description of pulse propagation
in absorbing, dispersive dielectrics. Journal of the Optical Society of America B: Optical
Physics, 12, 229247.
Sommerfeld, A. (1914). U ber die fortpflanzung des lichtes in disperdierenden medien. Annals
of Physics, 44, 177202.
Stratton, J. A. (1941). Electromagnetic theory. New York: McGraw-Hill.
Wyns, P., Foty, D. P., & Oughstun, K. E. (1988). Numerical analysis of the precursor fields in
linear dispersive pulse propagation. Journal of the Optical Society of America. A, 6(9),
14211429.
CHAPTER FOUR
Quantum Limits in Optical

Interferometry
ski, Marcin Jarzyna, Jan Koodyn
Rafal Demkowicz-Dobrzan ski
Contents
1. Introduction 346
2. Quantum States of Light 351
2.1 Mode Description 351
2.2 Gaussian States 353
2.3 Definite Photon-Number States 358
2.4 Particle Description 358
2.5 Mode Versus Particle Entanglement 359
3. MZ Interferometry 362
3.1 Phase-Sensing Uncertainty 362
3.2 Coherent State Interferometry 364
3.3 Fock State Interferometry 365
3.4 Coherent + Squeezed-Vacuum Interferometry 366
3.5 Definite Photon-Number State Interferometry 367
3.6 Other Interferometers 369
4. Estimation Theory 373
4.1 Classical Parameter Estimation 374
4.2 Quantum Parameter Estimation 382
5. Quantum Limits in Decoherence-Free Interferometry 387
5.1 QFI Approach 388
5.2 Bayesian Approach 390
5.3 Indefinite Photon-Number States 393
5.4 Role of Entanglement 399
5.5 Multipass Protocols 400
6. Quantum Limits in Realistic Interferometry 401
6.1 Decoherence Models 403
6.2 Bounds in the QFI Approach 408
6.3 Bayesian Approach 419
6.4 Practical Schemes Saturating the Bounds 421
7. Conclusions 426
Acknowledgments 427
References 427

346 Rafal Demkowicz-Dobrza
nski et al.
1. INTRODUCTION
Without much exaggeration one may say that optics is basically the sci-
ence of light interference. Light interference effects were behind the final
acceptance of classical wave optics and abandoning the Newtonian corpuscu-
lar theory of light. Conceptual insight into the process of interfering light
waves prompted development of a number of measurement techniques
involving well controlled interference effects and gave birth to the field of
optical interferometry (Hariharan, 2003). At the fundamental level, classical
interferometry is all about observing light intensity variations (intensity
fringes) resulting from a change in relative phases between two (or more)
overlapping light waves, e.g., when a single light beam is split into a number
of paths with tunable optical path-length differences and made to interfere on
the screen. The number of applications is stunning and ranges from basic
length measurements via spectroscopic interferometric techniques to the most
spectacular examples involving stellar interferometry and gravitational-wave
detectors (Pitkin, Reid, Rowan, & Hough, 2011).
Coherent properties of light as well as the degree of overlap between the
interfering beams determine the visibility of the observed intensity fringes
and are crucial for the quality of any interferometric measurement. Still,
when asking for fundamental limitations on precision of estimating, e.g.,
a phase difference between the arms of a MachZehnder (MZ) interferom-
eter, there is no particular answer within a purely classical theory, where
both the light itself as well as the detection process are treated classically.
In classical theory, intensity of light can in principle be measured with arbi-
trary precision and as such allows to detect in principle arbitrary small phase
shifts in an interferometric experiment.
This, however, is no longer true when semiclassical theory is considered
in which the light is still treated classically but the detection process is quan-
tized, so that instead of a continuous intensity parameter the number of
energy quanta (photons) absorbed is being measured. The absorption pro-
cess within the semiclassical theory has a stochastic character and the number
of photons detected obeys Poissonian or super-Poissonian statistics (Fox,
2006). If light intensity fluctuations can be neglected, the number of photons
detected, N, follows the Poissonian statistics with photon-number standard
p
deviation N hN i, where hN i denotes the mean number of photons
detected. This implies that the determination of the relative phase difference
Quantum Limits in Optical Interferometry 347
between the arms of the interferometer, based on the number of photons

detected at the output ports, will be affected by the relative uncertainty
p
/ N =hN i 1= hN i referred to as the shot noise. The shot noise
plays a fundamental role in the semiclassical theory and in many cases it
is indeed the factor limiting achievable interferometric sensitivities. In
modern gravitational-wave detectors, in particular, the shot noise is
the dominant noise term in the noise spectral density for frequencies
above few hundred Hz (LIGO Collaboration, 2011, 2013; Pitkin
et al., 2011).
Yet, the shot noise should not be regarded as a fundamental bound
whenever nonclassical states of light are consideredwe review the essential
aspects of nonclassical light relevant from the point of view of interferometry
in Section 2. The sub-Poissonian statistics characteristic for the so-called
squeezed states of light may offer a precision enhancement in interferometric
scenarios by reducing the photon-number fluctuations at the output ports.
First proposals demonstrated that sending coherent light together with the
squeezed vacuum state into the two separate input ports of a MZ interfer-
ometer offers estimation precision beating the shot noise and attaining the
1/hN i2/3 scaling of the phase estimation precision (Caves, 1981), while con-
sideration of more general two-mode squeezed states showed that even
1/hN i scaling is possible (Bondurant & Shapiro, 1984; Yurke, McCall, &
Klauder, 1986). A number of papers followed, studying in more detail
the phase-measurement probability distribution and proposing various strat-
egies leading to 1/hN i precision scaling (Braunstein, 1992; Dowling, 1998;
Holland & Burnett, 1993; Sanders & Milburn, 1995). Similar observations
have been made in the context of precision spectroscopy, where spin-
squeezed states have been shown to offer a 1/N scaling of atomic transition
frequency estimation precision, where N denotes the number of atoms
employed (Wineland, Bollinger, Itano, & Heinzen, 1994; Wineland,
Bollinger, Itano, Moore, & Heinzen, 1992). Section 3 provides a detailed
framework for deriving the above results. However, already at this point
a basic intuition should be conveyed that one cannot go beyond the shot
noise limit whenever an interferometric experiment may be regarded in a
spirit that each photon interferes only with itself. In fact, the only possibility
of surpassing this bound is to use light sources that exhibit correlations
in between the constituent photons, e.g., squeezed light, so that the
interference process may benefit from the properties of interphotonic
entanglement.
nski et al.
These early results provided a great physical insight into the possibilities
of quantum-enhanced interferometry and the class of states that might be of
practical interest for this purpose. The papers lacked generality, however, by
considering specific measurement estimation strategies, in which the error in
the estimated phase was related via a simple error propagation formula to the
variance of some experimentally accessible observable, e.g., photon-number
difference at the two output ports of the interferometer, or by studying the
width of the peaks in the shape of the phase-measurement probability
distribution.
Given a particular state of light fed into the interferometer, it is a priori
not clear what is the best measurement and estimation strategy yielding the
optimal estimation precision. Luckily, the tools designed to answer these
kinds of questions had already been present in the literature under the name
of quantum estimation theory (Helstrom, 1976; Holevo, 1982). The Quantum
Fisher Information (QFI) as well as the cost of Bayesian inference provide a
systematic way to quantify the ultimate limits on performance of phase-
estimation strategies for a given quantum state, which are already optimized
over all theoretically admissible quantum measurements and estimators. The
concept of the QFI and the Bayesian approach to quantum estimation are
reviewed in Section 4. As a side remark, we should note, that by treating
the phase as an evolution parameter to be estimated and separating explicitly
the measurement operators from the estimator function, quantum estima-
tion theory circumvents some of the mathematical difficulties that arise if
one insists on the standard approach to quantum measurements and attempts
to define the quantum phase operator representing the phase observable being
measured (Barnett & Pegg, 1992; Lynch, 1995; Noh, Fouge`res, & Mandel,
1992; Summy & Pegg, 1990).
The growth of popularity of the QFI in the field of quantum metrology
was triggered by the seminal paper of Braunstein and Caves (1994) advocat-
ing the use of QFI as a natural measure of distance in the space of quan-
tum states. The QFI allows to pin down the optimal probe states that are
the most sensitive to small variations of the estimated parameter by esta-
blishing the fundamental bound on the corresponding parameter sensitivity
valid for arbitrary measurements and estimators. Following these lines of
reasoning the 1/N bound, referred to as the Heisenberg limit, on the phase
estimation precision using N-photon states has been claimed fundamental
and the NOON states were formally proven to saturate it (Bollinger,
Itano, Wineland, & Heinzen, 1996). Due to close mathematical analo-
gies between optical and atomic interferometry (Bollinger et al., 1996;
Lee, Kok, & Dowling, 2002) similar bounds hold for the problem of atomic
transition-frequency estimation and more generally for any arbitrary unitary
parameter estimation problem, i.e., the one in which an N-particle state
evolves under a unitary UN, U expiH^ , with H^ being a general
single-particle evolution generator and the parameter to be estimated
(Giovannetti, Lloyd, & Maccone, 2004, 2006).
A complementary framework allowing to determine the fundamental
bounds in interferometry is the Bayesian approach, in which one assumes
the estimated parameter to be a random variable itself and explicitly defines
its prior distribution to account for the initial knowledge about before per-
forming the estimation. In the case of interferometry, the typical choice is
the flat prior p() 1/2 which reflects the initial ignorance of the phase.
The search for the optimal estimation strategies within the Bayesian
approach is possible thanks to the general theorem on the optimality of
the covariant measurements in estimation problems satisfying certain group
symmetry (Holevo, 1982). In the case of interferometry, a flat prior guaran-
tees the phase-shift, U(1), symmetry and as a result the optimal measurement
operators can be given explicitly and they coincide with the eigenstates of
the Pegg-Barnett phase operator (Barnett & Pegg, 1992). This makes it
possible to optimize the strategy over the input states and for simple cost
functions allows to find the optimal probe states (Berry & Wiseman,
2000; Buzek, Derka, & Massar, 1999; Luis & Perina, 1996). In particular,
for the 4sin 2 =2 cost function which approximates the variance for small
phase deviations , the corresponding minimal estimation uncertainty has
been found to read /N for large N providing again a proof of the
possibility of achieving the Heisenberg scaling, yet with an additional coef-
ficient. It should be noted that the optimal states in the above approach that
have been found independently in Berry and Wiseman (2000); Luis and
Perina (1996); Summy and Pegg (1990) have completely different structure
to the NOON states which are optimal when QFI is considered as the figure
of merit. This is not that surprising taking into account that the NOON
states suffer from the 2/N ambiguity in retrieving the estimated phase,
and hence are designed only to work in the local estimation approach when
phase fluctuations can be considered small. Derivations of the Heisenberg
bounds for phase interferometry using both the QFI and Bayesian
approaches are reviewed in Section 5. We also discuss the problem of
deriving the bounds for states with indefinite photon number in which case
replacing N in the derived bounds with the mean number of photons hN i
is not always legitimate, so that in some cases the naive Heisenberg
nski et al.
bound 1/hN i may in principle be beaten (Anisimov et al., 2010;

Giovannetti & Maccone, 2012; Hofmann, 2009).
Further progress in theoretical quantum metrology stemmed from the
need to incorporate realistic decoherence processes in the analysis of the
optimal estimation strategies. While deteriorating effects of noise on preci-
sion in quantum-enhanced metrological protocols have been realized by
many authors working in the field (Caves, 1981; Datta et al., 2011;
Gilbert, Hamrick, & Weinstein, 2008; Huelga et al., 1997; Huver,
Wildfeuer, & Dowling, 2008; Rubin & Kaushik, 2007; Sarovar &
Milburn, 2006; Shaji & Caves, 2007; Xiao, Wu, & Kimble, 1987), it has
long remained an open question to what extent decoherence effects may
be circumvented by employing either more sophisticated states of light or
more advanced measurements strategies including, e.g., adaptive techniques.
With respect to the most relevant decoherence process in optical appli-
cations, i.e., the photonic losses, strong numerical evidence based on the
QFI (Demkowicz-Dobrza nski et al., 2009; Dorner et al., 2009) indicated
that in the asymptotic limit of large number of photons the precision of
p
the optimal quantum protocols approaches const= N , and hence the gain
over classical strategies is bound to a constant factor. This fact has been first
rigorously proven within the Bayesian approach (Koody nski &
Demkowicz-Dobrza nski, 2010) and then independently using the QFI
(Knysh, Smelyanskiy, & Durkin, 2011). Both approaches yielded the same
fundamental bound on precision in the lossy optical interferometry:
p
1 =N , where is the overall power transmission of an
interferometric setup. This bound is also valid after replacing N with hN i
when dealing with states of indefinite photon number, and moreover can
be easily saturated using the most popular scheme involving a coherent
and a squeezed-vacuum state impinged onto two input ports of the MZ
interferometer (Caves, 1981). This fact also implies that the presently
implemented quantum-enhanced schemes in gravitational wave detection,
based on interfering the squeezed vacuum with coherent light, operate close
to the fundamental bound (Demkowicz-Dobrza nski, Banaszek, & Schnabel,
2013), i.e., they make almost optimal use of nonclassical features of light for
enhanced sensing given light power and loss levels present in the setup.
Based on the mathematical analysis of the geometry of quantum channels
(Fujiwara & Imai, 2008; Matsumoto, 2010) general frameworks have been
developed allowing to find fundamental bounds on quantum precision
enhancement for general decoherence models (Demkowicz-Dobrza nski,
Koody nski, & Guta, 2012; Escher, de Matos Filho, & Davidovich, 2011;
Demkowicz-Dobrza nski & Maccone, 2014). These tools allow to investi-
gate optimality of estimation strategies for basically any decoherence model
and typically provide the maximum allowable constant factor improvements
p
forbidding better than 1= N asymptotic scaling of precision. Detailed
presentation of the above mentioned results is given in Section 6.
Other approaches to derivation of fundamental metrological bounds
have been advocated recently. Making use of the calculus of variations it
was shown in Knysh, Chen, and Durkin (2014) how to obtain exact formu-
las for the achievable asymptotic precision for some decoherence models,
while in Alipour (2014); Alipour, Mehboudi, and Rezakhani (2014) a var-
iants of QFI have been considered in order to obtain easier to calculate, yet
weaker, bounds on precision. While detailed discussion of these approaches
is beyond the scope of the present review, in Section 6.2.3 we make use of
the result from Knysh et al. (2014) to benchmark the precision bounds
derived in the case of phase diffusion noise model.
The paper concludes with Section 7 with a summary and an outlook on
challenging problems in the theory of quantum-enhanced metrology.
2. QUANTUM STATES OF LIGHT

The advent of the laser, light-squeezing and single-photon light
sources triggered developments in interferometry that could benefit from
the nonclassical features of light (Buzek & Knight, 1995; Chekhova,
2011; Torres, Banaszek, & Walmsley, 2011). In this section, we focus on
the quantum-light description of relevance to quantum optical interferom-
etry. We discuss the mode description of light and the most commonly used
states in quantum opticsthe Gaussian states. In the end, we consider states
of definite photon number and study their particle description, in particular,
investigating their relevant entanglement properties.
2.1 Mode Description

Classically, an electromagnetic field can be divided into orthogonal modes
distinguished by their characteristic spatial, temporal, and polarization prop-
erties. This feature survives in the quantum description of light, where for-
mally we may associate a separate quantum subsystem with each of these
modes. Each subsystem is described by its own Hilbert space and, because
nski et al.
photons are bosons, can be occupied by an arbitrary number of particles

(Mandel & Wolf, 1995; Walls & Milburn, 1995). The most general M-mode
state of light may be then written as:
X
n, n0 jnihn0 j, Tr 1, 0 (1)
n, n0
with n fn1 , . .. , nM g and jni jn1 i jnM i representing a Fock state
with exactly ni photons occupying the i-th mode. States jnii may be further
expressed in terms of the respective creation and annihilation operators a^{i , aî
obeying â i , a^{j ij :
a^{n p p
jni i pi j0i, a^jni i ni jni 1i, a^{i jni i ni + 1jni + 1i, (2)
n!
where j0i is the vacuum state with no photons at all.
In the context of optical interferometry, modes are typically taken to be
distinguishable by their spatial separation, corresponding to different arms of
an interferometer, whereas the various optical devices such as mirrors,
beam-splitters or phase-delay elements transform the state on its way
through the interferometer. Eventually, photon numbers are detected in
the output modes allowing to infer the value of the relative phase difference
between the arms of the interferometer.
In many applications, the above standard state representation may not be
convenient and phase-space description is used insteadin particular, the
Wigner function representation (Schleich, 2001; Wigner, 1932). Adopting
the convention in which the quadrature operators read xî a^{i + aî and
a {i aî , the Wigner function may be regarded as a quasi-probability
pî i^
distribution on the quadrature phase space:
Z
1 M 0 M 0 ip0 xxx
^ 0
pp
^
W x,p d x d p Tr e , (3)
2 2 M
where x fx1 , . . ., xM g, x^ fx^1 , . .. , x^M g and similarly for p and p.
^ As a
consequence, the Wigner function is real, integrates to 1 over the whole
phase space and its marginals yield the correct probability densities of each
of the phase space variables. Yet, since it may take negative values it cannot
be regarded as a proper probability distribution. Most importantly, it may be
reconstructed from experimental data either by tomographic methods
(Schleich, 2001) or by direct probing of the phase space (Banaszek,
Radzewicz, Wodkiewicz, & Krasinski, 1999), and hence is an extremely

useful representation both for theoretical and experimental purposes.
2.2 Gaussian States

From the practical point of view, the most interesting class of states are the
Gaussian states (DAriano, Macchiavello, & Paris, 1995; Olivares & Paris,
2007; Pinel et al., 2012; Pinel, Jian, Fabre, Treps, & Braun, 2013). The great
advantage of using them is that they are relatively easy to produce in the lab-
oratory with the help of standard laser sources and nonlinear optical ele-
ments, which allow to introduce nonclassical features such as squeezing
or entanglement. Gaussian states have found numerous application in vari-
ous fields of quantum information processing (Adesso, 2006) and are also
extensively employed in quantum metrological protocols.
Gaussian states of M modes are fully characterized by their first and sec-
ond quadrature moments and are most conveniently represented using the
Wigner function which is then just a multidimensional Gaussian distribution
1
p e zhzî
T 1
W z zhzî
1
2
, (4)
2 det
M
where for a more compact notation we have introduced: the phase space var-
iable z fx1 ,p1 , ..., xM , pM g, the vector containing mean quadrature values
R
hzî fhx^1 i, hp^1 i, ... , hx^M i, hp^M ig, hzî i Trzî d2M z W zzi , and
the 2M dimensional covariance matrix :
1
i, j hzî z^j + z^j zî i hzî ihz^j i: (5)
2
Gaussian states remain Gaussian under arbitrary evolution involving
Hamiltonians at most quadratic in the quadrature operators, which includes
all passive devices such as beam splitters and phase shifters as well as single-
and multimode squeezing operations. Below we focus on a few classes of
Gaussian states highly relevant to interferometry.
2.2.1 Coherent States

Coherent states are the Gaussian states with identity covariance matrix
1, so that the uncertainties are equal for all quadratures saturating
the Heisenberg uncertainty relations 2xi2pi 1 and there are no corre-
lations between the modes. Mean values of quadratures may be arbitrary and
correspond to the coherent state complex amplitude hxî + ihpî=2:
nski et al.
These are the states produced by any phase-stabilized laser, what makes them
almost a fundamental tool in the theoretical description of many quantum
optical experiments. Moreover, coherent sates have properties that resemble
features of classical light, and thus enable to establish a bridge between the
quantum and classical descriptions of light.
In the standard representation, an M-mode coherent state
ji j1 i jM i is a tensor product of single-mode coherent states,
whereas a single-mode coherent state is an eigenstate of the respective anni-
hilation operator:
a^ji ji, (6)

where jjei and jj, are respectively the amplitude and the phase of a
coherent state. Equivalently, we may write
^
ji Dj0i, (7)
^ eâ a^ is the so-called displacement operator or write the

where D
{
coherent state explicitly as a superposition of consecutive Fock states:
2
X1
n
ji ejj =2 p jni: (8)
n0 n!
From the formula above, it is clear that coherent states do not have a definite
photon number and if a photon number n is measured its distribution follows
2n
the Poissonian statistics Pn ejj jjn! with average hni jj2 and stan-
2
dard deviation n jj. Thus, the relative uncertainty n/hni in the mea-
p
sured photon-number scales like 1= hni and hence in the classical limit of
large hni the beam power may be determined up to arbitrary precision.
Moreover, the evolution of the coherent state amplitude is identical to
the evolution of a classical-wave amplitude. In particular, an optical phase
delay transforms the state ji into j eii. These facts justify a common
jargon of calling coherent states the classical states of light, even though
for relatively small amplitudes different coherent states may be hard to dis-
2
tinguish due to their non-orthogonality jhjij2 ejj . More generally,
we call cl a classical state of light if and only if it can be written as a mixture
of coherent states:
Z
cl d2M Pjihj (9)
with P() 0, which is equivalent to the statement that cl admits a non-

negative Glauber P-representation (Glauber, 1963; Walls & Milburn, 1995).
Classical states are often used as a benchmark to test the degree of possible
quantum enhancement which may be obtained by using more general states
outside this class.
2.2.2 Single-Mode Squeezed States
Heisenbergs uncertainty principle imposes that xp 1 for all possible
quantum states. Single-mode states that saturate this inequality are called
the single-mode squeezed states (Walls & Milburn, 1995). As mentioned above,
coherent states fall into such a category serving as a special example for which
x p 1. Yet, as for general squeezed states x 6 p, the noise in one
of the quadratures can be made smaller than in the other. Formally, a single-
mode squeezed state may always be expressed as
j, ri D ^
^ Srj0i, (10)
^ exp1 r a^2 1 r a^{2 is the squeezing operator, r jrjei is a com-
where Sr 2 2
plex number and jrj and are the squeezing factor and the squeezing angle,
respectively. In fact, any pure Gaussian one-mode state may be written in
the above form. For 0, uncertainties in the quadratures x and p read
x er and p erreduction of noise in one quadrature is accompanied
by an added noise in the other one. This may be conveniently visualized in
the phase-space picture by error disks representing uncertainty in quadra-
tures in different directions, see Figure 1. In such a representation squeezed
states correspond to ellipses while coherent states are represented by circles.
Importantly, the fact that the uncertainty of one of the quadratures can be
less relatively to the other makes it possible to design an interferometric
scheme where the measured photon-number fluctuations are below that
of a coherent state and allows for a sub-shot noise phase estimation precision,
Figure 1 Phase-space diagrams denoting uncertainties in different quadratures for

momentum squeezed states (a) and for position squeezed states (b). Dashed circles
denote corresponding uncertainties for coherent states x p.
nski et al.
see Section 3. As squeezed states in general cannot be described as mixtures

of coherent states, they are nonclassical and their features cannot be fully
described by classical electrodynamics. Nevertheless, they can be relatively
easily prepared using nonlinear optical elements in the process of parametric
down conversion (Bachor & Ralph, 2004).
The special type of squeezed states which is most relevant from the met-
rological perspective is the class of the squeezed vacuum states that possess
vanishing mean values of their quadratures, i.e., hzî 0:
^
jri Srj0i: (11)
In the Fock basis, a squeezed vacuum state reads
X
1 n
1 Hn 0 tanh r 2 i n
jri p p e 2 jni, (12)
cosh r n0 n! 2
where Hn(0) denotes values of n-th Hermite polynomial at x 0. As for odd

n, Hn(x) is antisymmetric and thus Hn(0) 0, it follows that squeezed vac-
uum states are superpositions of Fock states with only even photon numbers.
The average number of photons in a squeezed vacuum state is given by
hni sinh 2 r, what means that, despite their name, a squeezed vacuum states
contain photons, possibly a lot of them.
2.2.3 Two-Mode Squeezed States

The simplest nonclassical two-mode Gaussian state is the so-called two-
mode squeezed vacuum state or the twin-beam state (Walls & Milburn,
1995). Mathematically, such a state is generated from the vacuum by a
two-mode squeezing operation, so that
ji2 S^2 j0,0i, (13)

{
where S^2 exp a^b^ ^
a { b^ and jjei, whereas in the Fock basis it
reads
1 X 1
ji2 1n ei n tanh n jn, ni: (14)
cosh n0
A notable feature of the twin-beam state, which may be clearly seen from
Equation (14), is that it is not a product of squeezed states in modes a
and b, but rather it is correlated in between them being a superposition of
terms with the same number of photons in both modes. Its first moments
of all quadratures are zero, hzî 0, whereas in the case of jj its

covariance matrix has a particularly simple form:
0 1
cosh2 0 sinh 2 0
B 0 cosh 2 0 sinh2 C
B @ sinh 2
C:
A (15)
0 cosh2 0
0 sinh 2 0 cosh2
Such a covariance matrix clearly indicates the presence of correlations

between the modes and since the state (13) is pure this implies immediately
the presence of the mode-entanglement. In fact, in the limit of large squeezing
coefficient jj ! 1 such twin-beam state becomes the original famous
EinsteinPodolskyRosen state (Adesso, 2006; Banaszek & Wodkiewicz,
1998) that violates assumptions of any realistic local hidden variable theory.
Twin-beam states may be generated in a laboratory by various nonlinear
processes such as four- and three-wave mixing (Bachor & Ralph, 2004;
Reid & Drummond, 1988). Alternatively, they may be produced by mixing
two single-mode squeezed vacuum states with opposite squeezing angles on
a fiftyfifty beam-splitter.
2.2.4 General Two-Mode Gaussian States

General two-mode Gaussian state is rather difficult to write in the Fock basis,
so it is best characterized by its 4 4 real symmetric covariance matrix

11 12
, (16)
21 22
where ij represent blocks with 2 2 matrices describing correlations

between the i-th and the j-th mode, and the vector of the first moments
hzî fhx^1 i, hp^1 i, hx^2 i, hp^2 ig. This in total gives up to fourteen real param-
eters describing the state: ten covariances, two displacement amplitudes and
two phases of displacement. General Gaussian states are in principle feasible
within current technological state of art, as any pure Gaussian state can the-
oretically be generated from the vacuum by utilizing only a combination of
one-, two-mode squeezing, and displacement operations with help of
beam-splitters and one-mode rotations (Adesso, 2006). Furthermore, mixed
Gaussian states are obtained as a result of tracing out some of the system
degrees of freedom, which is effectively the case in the presence of light
losses, or by adding a Gaussian noise to the state.
nski et al.
2.3 Definite Photon-Number States

Gaussian states are important from the practical point of view due to the rel-
ative ease with which their may be prepared. From a conceptual point of
view, however, when asking fundamental questions on limits to quantum
enhancement in interferometry, states with a definite photon number prove
to be a better choice. The main reason is that photons are typically regarded
as a resource in interferometry and when benchmarking different interfer-
ometric schemes it is natural to restrict the class of states with the same num-
ber of photons, i.e., the same resources consumed. A general M-mode state
consisting of N photons is given by:
X
N n, n0 jnihn0 j, (17)
0
jnjjn jN
P
where jnj i ni , so that the summation is restricted only to terms with
exactly N photons in all the modes. Apart from the vacuum state j0i no
Gaussian state falls into this category.
States with an exact photon number are extensively used in other fields of
quantum information processing, including quantum communication and
quantum computing (Kok et al., 2007; Pan et al., 2012). Most of the quan-
tum computation and communication schemes are designed with such states
in mind, as they provide the most intuitive and clear picture of the role the
quantum features play in these tasks. For large N, however, states with a def-
inite photon numbers are notoriously hard to prepare and states with N of
the order of 10 can only be produced with the present technology pushed to
its limits (Hofheinz et al., 2008; Sayrin et al., 2011; Torres et al., 2011).
When considering states with definite N, it is also possible to easily switch
between the mode and particle description of the states of light, which is a
feature that we discuss in the following section.
2.4 Particle Description

When dealing with states of definite photon number, instead of thinking
about modes as quantum subsystems that possess some number of excitations
(photons), we may equivalently consider the first quantization formalism
and regard photons themselves as elementary subsystems. Fundamentally,
photons are indistinguishable particles and since they are bosons their wave
function should always be permutationally symmetric. Still, it is common in
the literature to use a description in which photons are regarded as distin-
guishable particles and adopt a notation such that
jmi jm1 i1 jm2 i2 jmN iN (18)

denotes a product state of N photons, where the i-th photon occupies
the mode mi. We explicitly add subscripts to the kets above labeling each
constituent photon, in order to distinguish this notation from the mode
description of Equation (1), where kets denoted various modes and not
the distinct particles. The description (18) is legitimate provided there are
some degrees of freedom that ascribe a meaning to the statement the i-th
photon. For example, in the case when photons are prepared in non-
overlapping time-bins, the time-bin degree of freedom plays the role of
the label indicating a particular photon, whereas the spatial characteristics
determine the state of a given photon. Nevertheless, if we assume the
overall wave function describing also the temporal degrees of freedom of
the complete state to be fully symmetric, the notion of the i-th photon
becomes meaningless.
A general pure state of N distinguishable photons has the form:
X
j N i cm jmi (19)
m
P
where m jcm j2 1. If indeed there is no additional degree of freedom that
makes the notion of the i-th photon meaningful, the above state should
posses the symmetry property such that cm c(m), where is an arbitrary
permutation of the N indices.
Consider for example a Fock state jni jn1 i...jnM i of N n1 + + nM
indistinguishable photons in M modes. In the particle description, the state
has the form:
r
n1 ! nM ! X
jni jf1, . .. , 1 , 2. . ., 2 , . .. , M, . . ., M gi, (20)
N!
|{z} |{z} |{z}
n1 n2 nM
where the sum is performed over all nontrivial permutations of the

indices inside the curly brackets (Shankar, 1994). Since all quantum
states may be written in the Fock basis representation, by the above con-
struction one can always translate any quantum state to the particle
description.
2.5 Mode Versus Particle Entanglement

One of the most important features which makes the quantum theory different
from the classical one is the notion of entanglement (Horodecki, Horodecki,
nski et al.
Horodecki, & Horodecki, 2009). This phenomenon plays also an important

role in quantum metrology and is often claimed to be the crucial resource for
the enhancement of the measurement precision (Giovannetti et al., 2006;
Pezze & Smerzi, 2009). Conflicting statements can be found in the literature,
however, as some of the authors claim that entanglement is not indispensable
to get a quantum precision enhancement (Benatti & Braun, 2013). This con-
fusion stems simply from the fact that entanglement is a relative concept depen-
dent on the way we divide the relevant Hilbert space into particular
subsystems. In order to clarify these issues, it is necessary to explicitly study
relation between mode and particle entanglement, i.e., entanglement with
respect to different tensor product structures used in the two descriptions.
Firstly, let us go through basic definitions and notions of entanglement.
The state AB of two parties A and B is called separable if and only if one can
write it as a mixture of product of states of individual subsystems:
X A B
AB pi i i , pi 0: (21)
i
Entangled states are defined as all states that are not separable. A crucial fea-
ture of entanglement is that it depends on the division into subsystems. For
example, consider three qubits A, B and C and their joint quantum state
P
ABC 1i, j0 12 jiih jjjiih jjj0ih0j. This state is separable with respect
to the ABjC cut but is entangled with respect to the AjBC cut.
As a first example, consider a two-mode Fock state j1iaj1ib which rep-
resents one photon in mode a and one photon in mode b. This state written
in the mode formalism of Equation (1) is clearly separable. On the other
hand, photons are indistinguishable bosons and if we would like to write
their state in the particle formalism of Equation (18) we have to symmetrize
over all possible permutations of particles, thus obtaining the state
1
j1ia j1ib p jai1 jbi2 + jbi1 jai2 : (22)
2
In this representation, the state is clearly entangled. We may thus say that the
state contains particle entanglement but not the mode entanglement.
If, however, we perform the HongOuMandel experiment and send
the j1iaj1ib state through a balanced beam-splitter which transforms mode
p p
annihilation operators as a^ ! ^ ^
a + b= 2, b^! ^ ^
a b= 2, the resulting
state reads:
1 1
j1ia j1ib ! p j2ia j0ib j0ia j2ib p jai1 jai2 jbi1 jbi2 , (23)
2 2
which is both mode and particle entangled. Mode entanglement emerges
because the beam-splitter is a joint operation over two modes that intro-
duces correlations between them. On the other hand, it is a local operation
with respect to the particles, i.e., it can be written as U U in the particle
representation, and does not couple photons with each other. Thus, using a
beam-splitter one may change mode entanglement but not the content of
particle entanglement.
As a second example, consider two modes of light, a and b, each of
them in coherent state with the same amplitude , jiajib. This state
clearly has no mode entanglement. Since this state does not have a definite
photon number, in order to ask questions about the particle entanglement
we first need to consider its projection on one of the N-photon
subspacesone can think of a nondemolition total photon-number
measurement yielding result N. After normalizing the projected state we
obtain:
s

1 X
N
N
j N i jia jib N p jnia jN nib , (24)
N
2 n0 n
which in the particle representation reads:
1 O N
j N i p jaii + jbii (25)
2N i1
and is clearly a separable state. The fact that products of coherent states con-
tain no particle entanglement is in agreement with our definition of classical
state given in Equation (9) being a mixture of products of coherent states.
A classical state according to this definition will contain neither mode nor
particle entanglement.
As a last example, consider the case of particular interest for quantum
interferometry, i.e., a coherent state of mode a and a squeezed vacuum state
of mode b: jiajrib. Again this state has no mode entanglement. On the other
hand it is particle entangled. To see this, consider, e.g., the two-photon
sector, which up to irrelevant normalization factor reads:
jijr iN 2 / 2 j2ia j0ib + tanh r j0ia j2ib 2 jai1 jai2 + tanh r jbi1 jbi2 (26)
nski et al.
and contains particle entanglement provided both and r are nonzero.

We argue and give detailed arguments in Section 5 that it is indeed the par-
ticle entanglement and not the mode entanglement that is relevant in
quantum-enhanced interferometry. See also Killoran, Cramer, and Plenio
(2014) for more insight into the relation between mode and particle
entanglement.
3. MZ INTERFEROMETRY
We begin the discussion of quantum-enhancement effects in
optical interferometry by discussing the paradigmatic model of the MZ
interferometer. We analyze the most popular interferometric schemes
involving the use of coherent and squeezed states of light accompanied
by a basic measurement-estimation procedure, in which the phase is esti-
mated based on the value of the photon-number difference between the
two output ports of the interferometer. Such a protocol provides us with
a benchmark that we may use in the following sections when discussing the
optimality of the interferometry schemes both with respect to the states of
light used as well as the measurements and the estimation procedures
employed.
3.1 Phase-Sensing Uncertainty

In the standard MZ configuration, depicted in Figure 2, a coherent state
of light is split on a balanced beam-splitter, the two beams acquire phases
a, b, respectively, interfere on the second beam-splitter and finally the
photon numbers na, nb are measured at the output ports. Let a^, b^ and a^0 , b^0
be the annihilation operators corresponding to the two input and the two
na
a a
a
b b nb
b
Figure 2 The MachZehnder interferometer, with two input light modes a, b and two
output modes a0 , b0 . In a standard configuration, a coherent state of light ji is sent into
mode a. In order to obtain quantum enhacement, one needs to make use of the b input
port as well, sending, e.g., the squeezed vacuum state jri.
output modes, respectively. The combined action of the beam-splitters

and the phase delays results in the effective transformation of the
annihilation operators:
0 i
a^ 1 1 i e a 0 1 i a^

b^0 2 i 1 0 eib i 1 b^
(27)
cos =2 sin=2 a^
eia + b =2 ,
sin=2 cos =2 b^
where b a is the relative phase delay and for convenience we

assume that the beams acquire a /2 or /2 phase when transmitted
through the first or the second beam-splitter, respectively. The common
phase factor ei(a+b)/2 is irrelevant for further discussion in this section
and will be omitted.
In order to get a better insight into the quantum-enhancement effects
in the operation of the MZ interferometer, it is useful to make use of
the so-called JordanSchwinger map (Schwinger, 1965) and analyse
the action of the MZ interferometer in terms of the algebra of the angu-
lar momentum operators (Yurke et al., 1986). Let us define the
operators:
1 { ^ ^{ i { ^ J^ 1 ^ {
a b + b a^, J^y b^ a^ a^{ b,
J^x ^ z a { a^ b^ b,
^ (28)
2 2 2
which fulfill the angular momentum commutation relations Jî , J^j iEijk J^k
while the corresponding square of the total angular momentum reads:

N^ N^ {
J^ + 1 , N^ a^{ a^ + b^ b,
^
2
(29)
2 2
where N^ is the total photon-number operator. The action of linear optical

elements appearing in the MZ interferometer can now be described
as rotations in the abstract spin space: a^0 U aÛ { , b^0 U bU ^ {,
U expiJ^s, where J^ f J^x , J^y , J^z g and , s are the angle and the
axis of the rotation, respectively. In particular, the balanced beam split-
ter is a rotation around the x axis by an angle /2: U expi 2 J^x ,
while the phase delay is a rotation around the z axis: U expiJ^z .
Instead of analysing the transformation of the annihilation operators,
nski et al.
it is more convenient to look at the corresponding transformation of the

Ji operators themselves:
0 1
0 01 0 1 cos sin 0
^
Jx 1 0 0 B C
@ J^0y A @ 0 0 1 A B sin cos 0 C
@ 0 0 1A
J^0z 0 1 0
0 10 1 0 10 1 (30)
1 0 0 J^x cos 0 sin J^x
@ 0 0 1 A @ J^y A @ 0 1 0 A @ J^y A,
0 1 0 J^z sin 0 cos J^z
which makes it clear that the sequence of /2, and /2 rotations around
axes x, z and x, respectively, results in an effective rotation around the
y axis.
Using the above formalism, let us now derive a simple formula for uncer-
tainty of phase-sensing based on the measurement of the photon-number
difference at the output. Note that nâ n^b 2J^z , so the photon-number dif-
ference measurement is equivalent to the J^z measurement. Utilizing
Equation (30) in the Heisenberg picture, the average Jz evaluated on the interfe-
0
rometer output state may be related to the average of Jz of the input state jiin as
h J^z i cosh J^z iin sin h J^x iin : (31)

In order to assess the precision of -estimation, we also calculate the variance
of the J^z operator of the output state of the interferometer:
2 Jz cos 2 2 Jz jin + sin 2 2 Jx jin 2 sin cos cov Jx , Jz jin , (32)

where cov Jx , Jz 12 h J^x J^z + J^z J^x i h J^x ih J^z i is the covariance of the
two observables. The precision of estimating can now be quantified via
a simple error propagation formula:
Jz

:
d J^z (33)

d
3.2 Coherent State Interferometry

Let us now analyze the precision given by Equation (33) for the standard
optical interferometry with the input state jiin jij0i, representing a
coherent state and no light at all being sent into the input modes a and b
of the interferometer in Figure 2. The relevant quantities required for

calculating the precision given in Equation (33) read:
1 1
h J^z iin jj2 , h J^x iin 0, 2 Jz jin 2 Jx jin jj2 , cov Jx , Jz jin 0
2 4
(34)
yielding the precision:
1
jj 1 1
jij0i
2 p , (35)
1 2 jsinj hN ijsin j
jj jsin j
2

where the average photon number hN i N^ jj2 . The above formula
p
represents 1= hN i shot noise scaling of precision characteristic for the clas-
sical interferometry. The shot noise is a consequence of the Jz effectively
representing the Poissonian fluctuations of the photon-number difference
measurements at the output ports. Yet, although such fluctuations are
-independent, the average photon-number difference h Jzi changes with
with speed proportional to jsin j appearing in Equation (35), so that
the optimal operating points are at /2,3/2.
3.3 Fock State Interferometry

We can attempt to reduce the estimation uncertainty using more general
states of light at the input. For example, we can replace the coherent state
with an N-photon Fock state, so that jiin jN ij0i. This is an eigenstate
of J^z and hence 2Jzjin 0, and only the 2Jxjin contributes to the J^z
variance at the output: 2 Jz 14 N sin 2 . Since h J^z i 12 N cos , the
corresponding estimation uncertainty reads:
1
jN ij0i p , (36)
N
being again shot-noise limited. The sole benefit of using the Fock state is the
lack of -dependence of the estimation precision. This, however, is scarcely
of any use in practice, since one may always perform rough interferometric
measurements and bring the setup close to the optimal operating points
before performing more precise measurements there. Moreover, the
-dependence in Equation (35) can be easily removed by taking into
account not only the photon-number difference observable J^z but also
nski et al.
the total photon-number measured. By using their ratio as an effective

observable in the r.h.s. of Equation (33), or in other words by considering
the visibility observable, the formula (35) is replaced by (36).
3.4 Coherent + Squeezed-Vacuum Interferometry

We thus need to use more general input states in order to surpass the shot noise
limit. Firstly, let us note that sending the light solely to one of the input ports
will not provide us with the desired benefit. As in the end, only a photon-
number measurement is assumed, which is not sensitive to any relative phase
differences between various Fock terms of the output state, any scenario
involving a single-beam input may always be translated to the the situation
in which an incoherent mixture of Fock states is sent onto the input port. Since
the variance is a convex function with respect to state density matrices, and
thus increasing under mixing, such strategies are of no use for our purposes.
Let us now consider a scheme were apart from the coherent light we
additionally send a squeezed-vacuum state into the other input port
(Caves, 1981): jiin jijri, see Figure 2. This kind of strategy is being
implemented in current most advanced interferometers designed to detect
gravitational waves like LIGO or GEO600 (LIGO Collaboration, 2011,
2013; Pitkin et al., 2011). Assuming for simplicity that r is real, the relevant
quantities required to calculate the estimation precision read:
1
hN^ i jj2 + sinh 2 r, h Jz iin jj2 sinh 2 r, h Jx iin 0
2
1 1
Jz jin jj + sinh 2r , cov Jx , Jz jin 0,
2 2 2
(37)
4 2
1 2
Jx jin jj cosh2r Re2 sinh 2r + sinh 2 r :
2
4
Hence, the usage of squeezed-vacuum as a second input allows to reduce the
variance 2Jxjin thanks to the Re2 sinh 2r term above, which is then max-
imized by choosing the phase of the coherent state such that Re(). This
corresponds to the situation, when the coherent state is displaced in phase
space in the direction in which the squeezed vacuum possesses its lowest var-
iance. With such an optimal choice of phase, substituting the above formulas
into Equation (33), we obtain the final expression for the phase-estimation
precision:
r
1
cot 2 jj2 + sinh 2 2r + jj2 e2r + sinh 2 r
2 (38)
jijri 2 :
jj sinh 2 r
The optimal operation points are again clearly /2, 3/2, since at them
the first term under the square root, which is nonnegative, vanishes. For a
fair comparison with other strategies, we should fix the total average number
of photons hN i, which is regarded as a resource, and optimize the split of
energy between the coherent and the squeezed vacuum beams in order
to minimize . This optimization can only be done numerically, but
the solution can be well approximated analytically in the regime of hN i
1.
p
In this regime, the squeezed vacuum should carry approximately hN i=2
p
of photons, so the squeezing factor obeys sinh 2 r 14 e2r hN i=2 while
the majority of photons belongs to the coherent beam. The resulting
precision reads
q
p p
hN i
1 hN i=2 hNi + hN i=2 hN i
1 1
jijri p (39)
hNi hNi=2 hN i3=4
and proves that indeed this strategy offers better than shot noise scaling of
precision.
While the above example shows that indeed quantum states of light
may lead to an improved sensitivity, the issue of optimality of the proposed
scheme has not been addressed. In fact, keeping the measurement-
estimation scheme unchanged, it is possible to further reduce the estima-
tion uncertainty by sending a more general two-mode Gaussian states of
light with squeezing present in both input ports and reach the / 1=hN i
scaling of precision (Olivares & Paris, 2007; Yurke et al., 1986). We skip
the details here, as the optimization of the general Gaussian two-mode
input state minimizing the estimation uncertainty of the scheme considered
is cumbersome. More importantly, using the tools of estimation theory
introduced in Section 4, we will later show in Section 5 that even with
jijri class of input states it is possible to reach the / 1=hN i scaling of
precision, but this requires a significantly modified measurement-
estimation scheme (Pezze & Smerzi, 2008) and different energy partition
between the two input modes.
3.5 Definite Photon-Number State Interferometry

Even though definite photon-number states are technically difficult to
prepare, they are conceptually appealing and we make use of them to
demonstrate explicitly the possibility of achieving the 1/N Heisenberg
scaling of estimation precision in MZ interferometry. We have already
shown that sending an N-photon state into a single input port of the
nski et al.
interferometer does not lead to an improved precision compared with a

coherent-state-based strategy. Therefore, we need to consider states with
photons being simultaneously sent into both input ports. A general
N-photon two-mode input state can be written down using the angular
momentum notation as
X
j
jiin cm j j, mi, (40)
mj
where j N/2 and j j, mi j j + mij j mi in the standard mode-occupation

notation. In particular, j j, ji jN ij0i corresponds to a state with all the
photons being sent into the upper input port. One can easily check that
angular momentum operators introduced in Equation (28) act in a standard
way on the j j, mi states. For concreteness, assume N is even and consider the
state (Yurke et al., 1986):

1 1 N N N

N

j iin p j j,0i + j j, 1i p + +1 1 ,
2 2 2 2 2 2
(41)
for which:
1 1 p 1
h J^z iin , h J^x iin j j + 1, 2 Jz jin ,
2 2 2 (42)
1 1
2 Jx jin j j + 1 , cov Jx , Jz jin 0:
2 4
Plugging the above expressions into Equation (33), we get:
p
cos 2 + sin 2 j j + 1 1
p : (43)
jsin + cos j j + 1j
The optimal operation point corresponds to sin 0, where we benefit
from large h J^x iin and low 2Jzjin making the state very sensitive to rotations
around the y axis. The resulting precision reads:
1 N
1 2
p , (44)
j j + 1 N
indicating the possibility of achieving the Heisenberg scaling of precision.

We should note here, that a simpler state jiin j j, 0i jN/2ijN/2i called
the twin-Fock state where N/2 photons are simultaneously sent into each of
the input ports, is also capable of providing the Heisenberg scaling of pre-
cision (Holland & Burnett, 1993), but requires a different measurement-
estimation scheme which goes beyond the analysis of the average
photon-number difference at the output, see Section 5.3.3.
One should bear in mind that the expressions for attainable precision
presented in this section are based on a simple error propagation formula cal-
culated at a particular operating point. Therefore, in order to approach any
of the precisions claimed, one needs to first lock the interferometer to oper-
ate close to an optimal point, which also requires some of the resources to be
consumed. Rigorous quantification of the total resources needed to attain a
given estimation precision starting with a completely unknown phase may
be difficult in general. We return to this issue in Section 5, where we are able
to resolve this problem by approaching it with the language of Bayesian
inference.
3.6 Other Interferometers

Even though we have focused on the MZ interferometer setup, analogous
results could be obtained for other optical interferometric configurations,
such as the Michelson interferometer, the FabryPerot interferometer, as
well as the atomic interferometry setups utilized in: atomic clocks operation
(Diddams, Bergquist, Jefferts, & Oates, 2004), spectroscopy (Leibfried et al.,
2004), magnetometry (Budker & Romalis, 2007) or the BEC interferometry
(Cronin, Schmiedmayer, & Pritchard, 2009). We briefly show below that
despite physical differences the mathematical framework is common to all
these cases and as such the results presented in this review, even though
derived with the simple optical interferometry in mind, have a much
broader scope of applicability.
The Michelson interferometer is depicted in Figure 3a. Provided
the output modes a0 , b0 can be separated from the input ones a, b via an
optical circulator, the Michelson interferometer is formally equivalent to
the MZ interferometer. The input output relations are identical as in
Equation (27) with 2(b a), as the light acquires the relative phase
twicetraveling both to and from the end mirrors.
Consider now the FabryPerot interferometer depicted in Figure 3b.
We assume for simplicity that both mirrors have the same power trans-
mission coefficient T and the phase is acquired while the light travels
nski et al.
Figure 3 Other popular two-input/two-output mode interferometers: (a) Michelson

interferometer with one-way phase delays a, b in the respective arms and
(b) FabryProt interferometer with one-way phase delay and power transmission
T of the mirrors.
from one mirror to the other inside the interferometer. The resulting
inputoutput relation reads:
0
a^ 1
^0
b 2 T i sin T cos
p
2i 1 T sin p
T a^
(45)
T 2i 1 T sin b^
and up to an irrelevant global phase may be rewritten as:
0
a^ cos =2 i sin =2 a^
,
b^0 i sin =2 cos =2 b^ ! (46)
T
2 arcsin p :
T 2 + 41 T sin 2
In terms of the angular momentum operators, the above transformation is

simply a -rotation around the x axis. Thus, up to a change of the rotation
axis, the action of the FabryPerot interferometer with phase delay is
equivalent to the one of the MZ interferometer with phase delay . Know-
ing the formulas for the estimation precision of in the MZ interferometer,
we can easily calculate the corresponding estimation precision of via the
error propagation formula obtaining:
T 2 + 41 T sin 2
p :
@ 4T 1 T cos (47)

@
As a consequence, the above expression allows to translate all the results
derived for the MZ interferometer to the FabryPerot case.
Ramsey interferometry is a popular technique for performing precise

spectroscopic measurements of atoms. It is widely used in atomic clock
setups, where it allows to lock the frequency of an external source of radi-
ation, , to a selected atomic transition frequency, 0, between the single-
atom excited and ground states, jei and jgi (Diddams et al., 2004). In a typical
Ramsey interferometric experiment, N atoms are initially prepared in the
ground state and are subsequently subjected to a /2 Rabi pulse, which
transforms each of them into an equally weighted superposition of ground
and excited states. Afterwards, they evolve freely for time t, before finally
being subjected to a second /2 pulse, which in the ideal case of 0
would put all the atoms in the exited state. In case of any frequency
mismatch, the probability for an atom to be measured in an excited state
is cos 2 =2, where (0 )t. Hence, by measuring the number of
atoms in the excited state, one can estimate and consequently knowing
t the frequency difference 0 . Treating two-level atoms as spin-1/2
particles with their two levels corresponding to up and down projec-
tions of the spin z component, we may introduce the total spin operators
P k k
Jî 1 N
2 k0
î , i x, y, z, where î are standard Pauli sigma matrices
acting on the k-th particle. Evolution of a general input state can be
written as:
j i eiJ^x =2 eiJ^z eiJ^x =2 jiin : (48)
which is completely analogous to the MZ transformation (30) with /2

pulses playing the role of the beam-splitters, see Figure 4. The total spin
z operator can be written as J^z 2n^g nê , where n^g , nê denote the
ground and excited state atom number operators, respectively. Therefore,
measurement of J^z is equivalent to the measurement of the difference of
excited and nonexcited atoms analogously to the optical case where it
corresponded to the measurement of photon-number difference at the
two output ports of the interferometer. Fluctuation of the number of
atoms measured limits the estimation precision and in case of uncorre-
lated atoms is referred to as the projection noise, which may be regarded as
an analog of the optical shot noise. An important difference from the optical
case, though, is that when dealing with atoms we are restricted to consider
states of definite particle-number. Thus, there is no exact analogue of
coherent or squeezed states that we consider in the photonic case. We
can therefore regard atomic Ramsey interferometry as a special case
of the MZ interferometry with inputs restricted to states of definite
nski et al.
Figure 4 Formal equivalence of Ramsey and MZ interferometry. The two atomic levels
which are used in the Ramsey interferometry play analogous roles as the two arms of
the interferometer, while the /2 pulses equivalent from a mathematical point of view
to the action of the beam-splitters. Quantum-enhancement in Ramsey interferometry
may be obtained by preparing the atoms in a spin-squeezed input state reducing
the variances of the relevant total angular momentum operators, in a similar fashion
as using squeezed states of light leads to an improved sensitivity in the MZ
interferometry.
particle-number, discussed in Section 3.5, and further relate the precision of

estimating the frequency difference to the precision of phase estimation via

: (49)
t
Beating the projection noise requires the input state of the atoms to share
some particle entanglement. From an experimental point of view, the most
promising class of states are the so-called one-axis or two-axis spin-squeezed
states, which may be realized in BEC and atomic systems interacting with
light (Kitagawa & Ueda, 1993; Ma, Wang, Sun, & Nori, 2011). In fact, these
states may be regarded as a definite particle-number analogues of optical
squeezed states. In particular, starting with atoms in a ground state the
two-axis spin-squeezed states may be obtained via:
^2 ^2
j i e 2 J + J jgiN , (50)
where J^ J^x iJ^y . The above formula resembles the definition of an opti-
cal squeezed state given in Equation (10), where a^, a^{ operators are replaced
by J^ , J^+ . With an appropriate choice of squeezing strength as a function
of N it is possible to achieve the Heisenberg scaling of precision / 1=N

(Ma et al., 2011; Wineland et al., 1994).
Analogous schemes may also be implemented in BEC (Cronin et al.,
2009; Gross, Zibold, Nicklas, Esteve, & Oberthaler, 2010). In particular,
BEC opens a way of realizing a specially appealing matter-wave interferom-
etry, in which, similarly to the optical interferometers, the matter-wave is
split into two spatial modes, evolves and finally interferes resulting in a spatial
fringe pattern that may be used to estimate the relative phase acquired by the
atoms (Shin et al., 2004). Such a scheme may potentially find applications in
precise measurements of the gravitational field (Anderson & Kasevich,
1998). Yet, in this case, the detection involves measurements of positions
of the atoms forming the interference fringes, what makes the estimation
procedure more involved than in the simple MZ scheme (Chwede nczuk,
Hyllus, Piazza, & Smerzi, 2012; Chwede nczuk, Piazza, & Smerzi, 2011),
but eventually the precisions for the optimal estimation schemes should
coincide with the ones obtained for the MZ interferometry.
Finally, we should also mention that atomic ensembles interacting with
light are excellent candidates for ultra-precise magnetometers (Budker &
Romalis, 2007). Collective magnetic moment of atoms rotates in the pres-
ence of magnetic field to be measured, what again can be seen as an analogue
of the MZ transformation on Jî in Equation (30). The angle of the atomic
magnetic-moment rotation is determined by sending polarized light which
due to the Faraday effect is rotated proportionally to the atomic magnetic-
moment component in the direction of the light propagation. In standard
scenarios, the ultimate precision will be affected by both the atomic projec-
tion noise, due to characteristic uncertainties of the collective magnetic-
moment operator for uncorrelated atoms, and the light shot noise. The
quantum enhancement of precision may again be achieved by squeezing
the atomic states (Sewell et al., 2012; Wasilewski et al., 2010) as well as
by using the nonclassical states of light (Horrom, Singh, Dowling, &
Mikhailov, 2012), what in both cases allows to go beyond the projection
and shot-noise limits.
4. ESTIMATION THEORY
In this section, we review the basics of both classical and quantum
estimation theory. We present Fisher Information (FI) and Bayesian
nski et al.
approaches to determining the optimal estimation strategies and

discuss tools particularly useful for analysis of optical interferometric
setups.
4.1 Classical Parameter Estimation

The essential question that has been addressed by statisticians long before the
invention of quantum mechanics is how to most efficiently extract informa-
tion from a given data set, which is determined by some nondeterministic
process (Kay, 1993; Lehmann & Casella, 1998).
In a typical scenario, we are given an N-point data set
x fx1 , x2 , . .. ,xN g which is a realization of N independent identically dis-
tributed random variables, XN, each distributed according to a common
Probability Density Function (PDF), p(X), that depends on an unknown
parameter we wish to determine. Our goal is to construct an estimator

N x that should be interpreted as a function which outputs the most
accurate estimate of the parameter based on a given data set. Importantly,

as the estimator N is build on a sample of random data, it is a random
variable itself and the smaller are its fluctuations around the true value
the better it is.
Typically, two approaches to the problem of the choice of the
optimal estimator are undertaken. In the so-called frequentist or classi-
cal approach, is assumed to be a deterministic variable with an
unknown value that, if known, could in principle be stated to any pre-
cision. In this case, one of the basic tools in studying optimal estimation
strategies is the FI, and hence we will refer to this approach as the
FI approach. In contrast, when following the Bayesian paradigm, the
estimated parameter is a random variable itself that introduces some
intrinsic error that accounts for the lack of knowledge about we pos-
sess prior to performing the estimation. We describe both approaches in
detail below.
4.1.1 FI Approach
In this approach, p(X) is regarded as a family of PDFs parametrized by
the parameter to be estimated based on the registered data x. The per-

formance of a given estimator N x is quantified by the Mean Square Error
(MSE) deviation from the true value :
D Z
2 2 E 2
N N x dN x p x N x : (51)
A desired property for an estimator is that it is unbiased:

Z

N dN x p x N x , (52)
so that on average it yields the true parameter value. The optimal unbiased
estimator is the one that minimizes 2N for all . Looking for the optimal
estimator may be difficult and it may even be the case that there is no single
estimator that minimizes the MSE for all .
Still, one may always construct the so-called Cramer-Rao Bound (CRB)

that lower-bounds the MSE of any unbiased estimator N (see, e.g., Kay,
1993 for a review):
1
2 N , (53)
N F p
where F is the FI, and can be expressed using one of the formulas below:
Z * 2 + 2
1 @ p x 2 @ @
F p dx ln p ln p : (54)
p x @ @ @2
The basic intuition is that the bigger the FI is the higher estimation precision
may be expected. The FI is nonnegative and additive for uncorrelated events,
h i h i h i
so that F p1

, 2
F p 1
+ F p2
, for p
1, 2
x1 , x2 p1 2
x1 p x2
h i
and in particular: F pN N F p , which can be easily verified using
the last expression in definition (54). The FI is straightforward to calculate
and once an estimator is found that saturates the CRB it is guaranteed to
be optimal. In general, estimators saturating the CRB are called efficient.
The sufficient and necessary condition for efficiency is the following condition
on the PDF and the estimator (Kay, 1993):
@
ln p x N F p k x : (55)
@

An estimator satisfying the above equality exist only for a special class of
PDFs belonging to the so called exponential family of PDFs, for which:
a0
lnp x a + bx + cdx, , (56)
c 0
where a(), c() and b(x), d(x) are arbitrary functions and primes denote dif-
ferentiation over . In general, however, the saturability condition cannot
be met.
nski et al.
Note that in general FI is a function of , so that depending on the

true value of the parameter, the CRB puts weaker or stronger constraints
on the minimal MSE. Actually, one is not always interested in the
optimal estimation strategy that is valid globallyfor any potential value
of but may want to design a protocol that works optimally for
confined to some small parameter range. In this case, one can take a local
approach and analyze the CRB at a given point 0. Formally, der-
ivation of the CRB at a given point requires only a weaker local unbiased-
ness condition:

@

1 (57)
@ N 0
at a given parameter value

0. FI at 0 is a local quantity that depends only on
@ p X
p X and @ , as explicitly stated in Equation (54). As a
0 0
result, the FI is sensitive to changes of the PDF of the first order in
0. Looking for the optimal locally unbiased estimator at a given point
0 makes sense provided one has a substantial prior knowledge that the true
value of is close to 0. This may be the case if the data is obtained from a
well controlled physical system subjected to small external fluctuations or if
some part of the data had been used for preliminary estimation narrowing
the range of compatible to a small region around 0. In this case, even
if the condition for saturability of the CRB cannot be met, it still may be
possible to find a locally unbiased estimator which will saturate the CRB
at least at a given point 0. The explicit form of the estimator can easily
be derived from Equation (55) by substituting 0:

@ lnp x
N0 x 0 + NFp1 @ : (58)
0 0
Fortunately, difficulties in saturating the CRB are only present in the

finite-N regime. In the asymptotic limit of infinitely many repetitions of
an experiment, or equivalently, for an infinitely large sample, N ! 1, a
particular estimator called the Maximum Likelihood (ML) estimator saturates
the CRB (Kay, 1993; Lehmann & Casella, 1998). The ML estimator
formally defined as
ML
N x argmax p x (59)

is a function that for a given instance of outcomes, x, outputs the

value of parameter for which this data sample is the most probable.
For finite N, the ML estimator is in general biased, but becomes un-
biased asymptotically lim N !1 hML
N xi and saturates the CRB
ML
lim N !1 N 2 x Fp .
4.1.2 Bayesian Approach

In this approach, the parameter to be estimated, , is assumed to be a random
variable that is distributed according to a prior PDF, p(), representing the
knowledge about one possesses before performing the estimation, while
p(xj) denotes the conditional probability of obtaining result x for parameter
value . Notice, a subtle change in notation from p(x) in the FI approach to
p(xj) in the Bayesian approach reflecting the change in the role of which
is a parameter in the FI approach and a random variable in the Bayesian
approach. If we stick to the MSE as a cost function, we say that the estimator

N x is optimal if it minimizes the average MSE
ZZ

2 2
N d dx pxjp N x , (60)
which, in contrast to Equation (51), is also averaged over all the values of the
parameter with the Bayesian prior p(). Making use of the Bayes theorem
we can rewrite the above expression in the form
Z Z

2 2
N dx px d pjx N x : (61)
From the above formula, it is clear that the optimal estimator is the one that
minimizes terms in square bracket for each x. Hence, we can explicitly
derive the form of the Minimum Mean Squared Error (MMSE) estimator
Z
@ 2
d pjx N x 0
@ N Z
MMSE
) N x d pjx hipjx (62)
which simply corresponds to the average value of the parameter computed

with respect to the posterior PDF, p(jx). The posterior PDF represents
nski et al.
the knowledge we possess about the parameter after inferring the informa-
tion about it from the sampled data x. The corresponding MMSE reads:
Z Z 2

2
N dx px d pjx hipjx
Z (63)
2

dx px pjx ,
so that it may be interpreted as the variance of the parameter computed

again with respect to the posterior PDF p(jx) that is averaged over all
the possible outcomes.
The optimal estimation strategy within the Bayesian approach depends
explicitly on the prior PDF assumed. If either the prior PDF will be very
sensitive to variations of or the physical model will predict the data to
be weakly affected by any parameter changes, so that p(xj)p() p(),

the minimal 2 N will be predominantly determined by the prior distribu-
tion p() and the sampled data will have limited effect on the estimation pro-
cess. Therefore, it is really important in the Bayesian approach to choose an
appropriate prior PDF that, on one hand, should adequately represent our
knowledge about the parameter before the estimation, but, on the other, its
choice should not dominate the information obtained from the data.
In principle, nothing prevents us to consider more general cost functions,

C, , that in some situations may be more suitable than the squared error.
The corresponding optimal estimator will be the one that minimizes the
average cost function
ZZ

hCi d dx p pxj C N x, : (64)
In the context of optical interferometry, the estimated quantity of interest

is the phase which is a circular parameter, i.e., may be identified with a point
on a circle or more formally as an element of the circle group U(1) and in
particular + 2n. Following Holevo (1982), the cost function should
respect the parameter topology and the squared error is clearly not the proper

choice. We require the cost function to be symmetric, C, C, ,

group invariant, i.e., 82U1 : C + , + C, , and periodic,

C + 2n, C, . This restricts the class of cost functions to:

X
1

C, C cn cos n with : (65)
n0
Furthermore, we require C() must rise monotonically from C(0) 0 at

0 to some C() Cmax at , so that C0 () 0 and the coef-
ficients cn must fulfill following constraints:
X
1 X
1
cn 0, 1n cn Cmax ,
n0 n0
X
1 X1 (66)
n cn 0,
2
n 1 cn 0,
2 n
n1 n1
which may be satisfied by imposing 8n>0 : cn 0 and taking cn to decay

at least quadratically with n. Lastly, for the sake of compatibility we
would like the cost function to approach the standard variance for
small so that C 2 + O 4 , which is equivalent to
P1 2
n1 n cn 2.
In all the Bayesian estimation problems considered in this work, we will
consider the simplest cost function that satisfies all above-mentioned condi-
tions with c0 c1 2, 8n>1 : cn 0 which reads explicitly:

2
C, 4 sin : (67)
2
Following the same argumentation as described in Equation (62) when
minimizing the averaged MSE, one may prove that for the above chosen
C
cost function the average cost hCi is minimized if an estimator, N x,
can be found that for any possible data sample x collected satisfies the
condition
Z C

d pjx sin N x 0: (68)
4.1.3 Example: Transmission Coefficient Estimation

In order to illustrate the introduced concepts, let us consider a simple
model of parameter estimation, where a single photon impinges on the
beamsplitter with power transmission and reflectivity equal, respectively,
p and q 1 p. The experiment is repeated N times and based on the data
obtainednumber of photons transmitted and the number of photons
reflectedthe goal is to estimate the transmission coefficient p. This
problem is equivalent to a coin-tossing experiment, where we assume an
nski et al.
unfair coin, which heads and tails occurring with probabilities p and
q 1 p, respectively.
Probability that n out of N photons get transmitted is governed by the
binomial distribution

N
pp n
N
pn 1 pN n : (69)
n
p N=p1 p and hence the CRB imposes a lower

The FI equals FpN
bound on the achievable estimation variance:
p1 p
2 p N : (70)
N
Luckily the binomial probability distribution belongs to the exponential
family of PDFs specified in Equation (56), and by inspecting saturability con-
dition Equation (55) it may be easily checked, that the simple estimator

p N n n=N saturates the CRB. It is also worth mentioning that the optimal
estimator also coincides with the ML estimator, hence in this case the ML
estimator is optimal also for finite N and not only in the asymptotic regime.
In the context of optical interferometry, we will deal with an analogous
situation, where photons are sent into one input port of an interferometer
and p, q correspond to the probabilities of detecting a photon in one of the
two output ports. For a MZ interferometer, see Section 3, the probabilities
depend on the relative phase delay difference between the arms of an inter-
ferometer : p sin 2 =2, q cos 2 =2, which is the actual parameter
of interest. Probability distribution as a function of then reads:

N
p n
N
sin=22n cos=22N n (71)
n
and the corresponding FI and the CRB take the form
h i 1
2
F pN N , N : (72)
N
Interestingly, FI does not depend on the actual value , what suggests that
the achievable estimation precision may be independent of the actual param-
eter value. However, for such a parametrization, the CRB saturability
condition Equation (55) does not hold and there is no unbiased estimator
saturating the bound. Nevertheless, using Equation (58), we may still

write a locally unbiased estimator saturating CRB at 0: N n
0 tan 0 =2 + 2n=N sin 0 , which is possible provided sin0 6 0.
Since the CRB given in Equation (72) can only be saturated locally it is
worth looking at the ML estimator which we know will perform optimally
in the asymptotic regime N ! 1. Solving Equation (59) we obtain
r
ML n
N n argmax lnp n 2 arctan
N
(73)
N n
and in general we notice that there are two equivalent maxima. This ambi-
guity is simply the result of invariance of the PDF p(n) with respect to the
change ! and might have been expected. Hence in practice we would
need some additional information, possibly coming from prior knowledge or
other observations, in order to distinguish between this two cases and be enti-
tled to claim that the ML estimator saturates the CRB asymptotically for all .
We now analyze the same estimation problem employing the Bayesian
approach. Let us first consider the case where p is the parameter to be esti-
mated and the relevant conditional probability pN njp pN p n is given by
Equation (69). Choosing a flat prior distribution p(p) 1 and the mean
square as a cost function, we find using Equations (62) and (63) that the
MMSE estimator and the corresponding minimal averaged MSE equal:
n + 1 D 2 E 1
p N n , pN , (74)
N +2 6N + 2
which may be compared with the previously discussed FI approach where

the optimal estimator was p n n=N and the resulting variance when
R1
averaged over all p would yield h2 p N i 0 dp pp1 N 1=6N . Hence,
in the limit N ! 1 these results converge to the ones obtained previously
in the FI approach. This is a typical situation that in the case of large amount
of data the two approaches yield equivalent results (van der Vaart, 2000).
We now switch to parametrization and consider pN nj pN n as in
Equation (71). Assuming flat prior distribution p() 1/(2) and the pre-
viously introduced natural cost function in the case of circular parameter

C , 4sin 2 =2, due to estimation ambiguity we realize
that the condition for the optimal estimator given in Equation (68) is satisfied
C
for a trivial estimator N n 0 which does not take into account the
measurement results at all. This can be understood once we realize that
the ambiguity in the sign of estimated phase and the possibility of esti-
mating the phase with the wrong sign is worse than not taking into account
the measured data at all. In order to obtain a more interesting result, we
need to consider a subset of possible values of over which the
nski et al.
reconstruction is not ambiguous. If we choose 2 [0, ), and the

corresponding prior p() 1/, Equation (68) yields the following optimal
form of the estimator and its corresponding minimal cost calculated
according to Equation (64)

C 2
N n arctan (75)
f N, n
!
1 X q2
N
hCi 2 1 4 + f N , n , (76)
N + 1 n0
where f(N, n) (N 2n)(n 1/2)!(N n 1/2)!/n!(N n)!. Despite its

complicated form, the above formulas simplify in the limit N ! 1.
C
The optimal estimator approaches the ML estimator N n 2 arctan
p
n=N n while the average cost function approaches hCi 1/N indicat-
ing saturation of the CRB and confirming again that in the regime of many
experimental repetitions the two discussed approaches coincide.
4.2 Quantum Parameter Estimation

In a quantum estimation scenario, the parameter is encoded in a quantum
state which is subject to a quantum measurement Mx yielding measure-

ment result x with probability p x Tr Mx . The estimation strategy

is complete once an estimator function x is given ascribing estimated
parameters to particular measurement results. The quantum measurement
may be a standard projective von-Neumann measurement,
Mx Mx0 Mx x, x0 , or a generalized measurement where the measurement
operators form a Positive Operator
R Valued Measure (POVM) with the only
constraints being Mx 0, d x Mx 1 (Bengtsson & Zyczkowski,
2006; Nielsen & Chuang, 2000). Establishing the optimal estimation strategy
corresponds then not only to the most accurate inference of the parameter
value from the data, but also to a nontrivial optimization over the class of all
POVMs to find the measurement scheme maximizing the precision. How-
ever, as soon as we decide on a particular measurement scheme, we obtain a
model that determines probabilities describing the sampled data p(x) and
their quantum mechanical origin becomes irrelevant. Hence, the estimation
problem becomes then fully classical and all the techniques developed in
Section 4.1 apply.
4.2.1 QFI Approach

The problem of determining the optimal measurement scheme for a partic-
ular estimation scenario is nontrivial. Fortunately, analogously as in the clas-
sical estimation it is relatively easy to obtain useful lower bounds on the
minimal MSE. The Quantum Cramer-Rao Bound (QCRB) (Braunstein &
Caves, 1994; Helstrom, 1976; Holevo, 1982) is a generalization of the
classical CRB (53), which lower-bounds the variance of estimation for all
possible locally unbiased estimators and the most general POVM
measurements1:
1 n o
2 N with FQ Tr L 2 , (77)
N FQ
where FQ is the QFI while the Hermitian operator L[] is the so-called
Symmetric Logarithmic Derivative (SLD), which can be unambiguously defined

for any state via the relation _ 12 L + L . Crucially,
QFI is solely determined by the dependence of on the estimated param-
eter, and hence allows to analyze parameter sensitivity of given probe states
without considering any particular measurements norP estimators. Explicitly,
the SLD when written in the eigenbasis of i i jei i hei j
reads:

X 2hei j_ ej

L jei ej , (78)
i, j
i + j
where the sum is taken over the terms with nonvanishing denominator.
Analogously to the FI (54), the QFI is an additive quantity when calculated
h i
on product states and in particular FQ N N FQ . Thus the N
term in the denominator of Equation (77) may be equivalently interpreted
as the number of independent repetitions of an experiment with a state to
form the data sample x of size N, or a single shot experiment with a multi-
party state N
.
Crucially, as proven by Braunstein and Caves (1994) and Nagaoka
(2005), there always exist a measurement strategya projection measure-
ment in the eigenbasis of the SLDfor which the FI calculated for the
resulting probability distribution equals the QFI, and consequently the
1
For clarity of notation, in what follows we drop the tilde symbol when writing estimator variance.
nski et al.
bounds (53) and (77) coincide. Hence the issue of saturability of the QCRB
amounts to the problem of saturability of the corresponding classical CRB.
As discussed in detail in Section 4.1.1, the bound is therefore globally satu-
rable for a special class of probability distributions belonging to the so called
exponential family, and if this is not the case the saturability is achievable
either in the asymptotic limit of many independent experiments N ! 1
or in the local approach when one estimates small fluctuation of the param-
eter in the vicinity of a known

value 0.
For pure states, , the QFI in Equation (77) simplifies to

2 dj i
FQ 4 _ j_ _ j , j_ i : (79)
d
Yet, for general mixed states, calculation of the QFI involves diagonalization
of the quantum state , in order to calculate the SLD, and becomes tedious
for probe states living in highly dimensional Hilbert spaces.
Interestingly, the QFI may be alternatively calculated by considering
purifications ji of a given family of mixed states on an extended Hilbert

space TrE , where by E we denote an ancillary space
needed for the purification. It has been proven by Escher et al. (2011) that
the QFI of any is equal to the smallest QFI of its purifications
n

2 o
FQ min FQ 4 min _ j_ _ j : (80)

Even though minimization over all purification may still be challenging, the
above formulation may easily be employed in deriving upper bounds on
QFI by considering some class of purifications. Since upper bounds on
QFI translate to lower bounds on estimation uncertainty this approach
proved useful in deriving bounds in quantum metrology in presence of
decoherence (Escher, Davidovich, Zagury, & de Matos Filho, 2012;
Escher et al., 2011).
Independently, in Fujiwara and Imai (2008) another purification-based
QFI definition has been constructed:

FQ 4 min _ j_ : (81)

Despite apparent difference, Equations (80) and (81) are equivalent and one
can prove that any purification minimizing one of them is likewise optimal
for the other causing the second term of Equation (80) to vanish. Although
for any suboptimal jiEquation (80) must provide a strictly tighter bound
on QFI than Equation (81), the latter definition, owing to its elegant form
allows for a direct and efficient procedure for derivation of the precision
bounds in quantum metrology (Demkowicz-Dobrza nski et al., 2012;
Koody nski & Demkowicz-Dobrza nski, 2013). Derivations of the precision
bounds using the above two techniques in the context of optical interferom-
etry are discussed in Section 6.
For completeness, we list below some other important properties of
the QFI. QFI does not increase under a parameter-independent quantum
channel
FQ FQ , (82)
where is an arbitrary completely positive map. QFI appears in the lowest
order expansion of the measure of fidelity F between two quantum states
p2
p p
F 1 ,2 Tr 1 2 1 , (83)
1
F , + d 1 FQ d2 + Od4 : (84)
4
QFI is convex
!
X X X
FQ pi i
pi FQ i
, pi 1, pi 0, (85)
i i i
which reflects the fact that mixing quantum states can only reduce achiev-
able estimation sensitivity. In a commonly encountered case, specifically in
the context of optical interferometry, when the estimated parameter is
encoded on the state by a unitary
U U{ , U eiH^ , (86)
where H^ is the generating Hamiltonian, the general formula for QFI

reads:
X 2jhei jH^ jej ij2 i j 2
FQ (87)
i, j
i + j
where jeii, i form the eigendecomposition of . Note that in this case the
QFI does not depend on the actual value of . For the pure state estimation
case, jihj, the QFI is proportional to the variance of H^ :
FQ j i 42 H 4hjH^ ji hjH^ ji
2
(88)
nski et al.
and the QCRB (77) takes a particular appealing form resembling the form of
the energy-time uncertainty relation:
1
2 2 H : (89)
4N
We conclude the discussion of QFI properties by mentioning a very recent
and elegant result proving that for unitary parameter encodings, QFI is pro-
portional to the convex roof of the variance of H^ (Toth & Petz, 2013; Yu,
2013):
X
FQ 4 min pi 2 H i , (90)
fpi , j i ig
i
where the minimum is performed over all decompositions of

P
i pi j i ih i j, and 2H(i ) denotes the variance of H^ calculated
on j (i )i.
4.2.2 Bayesian Approach

As the quantum mechanical estimation scenario, with a particular measure-
ment scheme chosen, represents nothing but a probabilistic model with out-
come probabilities p(x), we may also apply the Bayesian techniques
described in Section 4.1.2. The quantum element of the problem, however,
i.e., minimization of the average cost function over the choice of measure-
ments is in general highly nontrivial. Fortunately, provided the problem
possesses a particular kind of symmetry it may be solved using the concept
of covariant measurements (Holevo, 1982).
In the context of optical interferometry, it is sufficient to consider the
unitary parameter encoding case as defined in (86), where the estimated
parameter will correspond to the phase difference in an interferometer,
see Figure 5 in Section 5. Let us denote a general POVM measurement

as Mx and the corresponding estimator as . Since we need to minimize
the average cost over both the measurements and the estimators it is
convenient to combine these two elements into one by labeling the POVM
R
elements with the estimated values themselves: M dxMx x.
The expression for the average cost function, (64), takes the form:
ZZ
d d
hCi pTrU U{ M C , : (91)
2 2
and leaves us with the problem of minimization hCi over a general POVM
R
M with standard constraints M 0, d
2 M 1. Note that for clarity we

use the normalized measure d 2 .

Provided the problem enjoys the shift symmetry, so that p( + 0)

p(), C + 0 , + 0 C,, it may be shown that one does not loose
optimality by restricting the class of POVM measurements to
M U U { , (92)

which is a special case of the so-called covariant measurements (Bartlett,

Rudolph, & Spekkens, 2007; Chiribella, DAriano, & Sacchi, 2005;
Holevo, 1982). If we take flat prior distribution p() 1, and the natural

cost function C, 4 sin 2 =2 introduced in Section 4.1.2, sym-
metry conditions are fulfilled and substituting (92) into (91) we get a simple
expression:
n
o
hCi Tr C , (93)

R {

2
where C 4 d 2 U U sin 2 is the final quantum state averaged
with the cost function. Looking for the optimal Bayesian strategy now
amounts to minimization of the above quantity over with the POVM
R {
constraints requiring that 0, d2 U U
1. As shown in Sections 5
and 6, this minimization is indeed possible for optical interferometric esti-

mation models and allows to find the optimal Bayesian strategy and the
corresponding minimal average cost.
5. QUANTUM LIMITS IN DECOHERENCE-FREE

INTERFEROMETRY
In order to analyze the ultimate precision bounds in interferometry,
one needs to employ the quantum estimation theory introduced in
Section 4.2. In this section, we review the most important results of
decoherence-free interferometry leaving the analysis of the impact of
decoherence for Section 6. This will provide us with precision benchmarks
to which we will be able to compare realistic estimation schemes and analyze
the reasons for the departure of practically achievable precisions from ideal-
ized scenarios.
nski et al.
Figure 5 Instead of a particular MachZehnder interferometric strategy, a general

quantum interferometric scheme involves a general input probe state jiin that is sub-
ject to a unitary phase delay operation U followed by a general quantum measurement

(POVM) Mx. Finally, an estimator x is used to obtain the estimated value of the
phase delay.
Formally, interferometry may be regarded as a channel estimation prob-

lem where a known state jiin is sent through a quantum channel
U eiJ^z , with an unknown parameter , where J^z 12 a{ a b{ b is
the z component of angular momentum operator introduced in
Section 3.1. Quantum measurement M^ x is performed at the output state
and the value of is estimated based on the obtained outcome x through

an estimator x, see Figure 5. Pursuing either QFI or Bayesian approach
it is possible to derive bounds on achievable precision that are valid
irrespectively of how sophisticated is the measurement-estimation strategy
employed and how exotic the input states of light are. We start by consid-
ering definite photon-number states using both QFI and Bayesian
approaches and then move on to discuss issues that arise when discussing
fundamental bounds taking into account states of light with indefinite
photon number.
5.1 QFI Approach

As discussed in detail in Section 4, the QFI approach is particularly well
suited to analyze problems where one wants to estimate small deviations
of around a known value 0, as in this local regime the QCRB, (77),
is known to be saturable. This is, for example, the case of gravitational-wave
interferometry in which one sets the interferometer at the dark fringe and
wants to estimate small changes in the interference pattern induced by
the passing wave (LIGO Collaboration, 2011, 2013; Pitkin et al., 2011).
Since the state of light at the output, j i eiJ^z jiin , is pure, the QFI
may be calculated using the simple formula given in Equation (79).
Realizing that j_ i iJ^z j i we get the QFI and the corresponding

QCRB on the estimation precision:
1
FQ 4 h j J^z j i jh j J^z j ij2 42 Jz ,
2
: (94)
2Jz
Note that the form of the QCRB above may be regarded as an analogue of
the Heisenberg uncertainty relation for phase and angular momentum.
Clearly, according to the above bound, the optimal probe states for inter-
ferometry are the ones that maximize Jz. We fix the total number of pho-
tons and look for N-photon states maximizing Jz. A General N-photon
input state and the explicit form of the corresponding QFI read:
X
N
jiin cn jnijN ni, (95)
n0
" !2 #
XN X
N
FQ 4 jcn j n
2 2
jcn j n
2
: (96)
n0 n0
Let us first consider the situation in which jiin is a state resulting from
sending N photons on the balanced beam-splitter, as discussed in
Section 3.3. This time, however, we do not insist on a particular measure-
ment nor estimation scheme, but just want to calculate the corresponding
QCRB on the sensitivity. Written in the photon-number basis the state
takes the form:
s

XN
1 N
jiin jnijN ni, (97)
n0
2N n
for which FQ N and this results in the shot noise bound on precision:
1
jN ij0i p : (98)
N
Recall, that this bound is saturated with the simple MZ interferometric
scheme discussed in Equation (36), which is a proof that for the considered
probe state this measurement-estimation scheme is optimal. From a particle-
description point of view, see Section 2.4, the above considered state is a
pure product state with no entanglement between the photons. More gen-
erally, the shot noise limit, sometimes referred to as the standard quantum
limit, is valid for all N-photon separable states (Giovannetti, Lloyd, &
nski et al.
Maccone, 2011; Pezze & Smerzi, 2009), and going beyond this limit
requires making use of inter-photon entanglement, see Section 5.4.
Let us now investigate general input states, which possibly may be
entangled. Consider the state of the form jNOONi p12 jN ij0i + j0ijN i
which is commonly referred to as the NOON state (Bollinger et al.,
1996; Lee et al., 2002). QFI for such state is given by FQ N2 and
consequently
1
, (99)
N
which is referred to as the Heisenberg limit. In fact, NOON state gives the best
possible precision as it has the biggest variance of J^z among the states with a
given photon number N (Bollinger et al., 1996; Giovannetti et al., 2006).
Still, the practical usefulness of the NOON states is doubtful. The difficulty
in their preparation increases dramatically with increasing N, and with pre-
sent technology the experiments are limited to relatively small N, e.g., N 4
(Nagata, Okamoto, OBrien, Sasaki, & Takeuchi, 2007) or N 5 (Afek,
Ambar, & Silberberg, 2010). Moreover, even if prepared, their extreme sus-
ceptibility to decoherence with increasing N, see Section 6, makes them
hardly useful in any realistic scenario unless N is restricted to small values.
Taking into account that experimentally accessible squeezed states of light
offer a comparable performance in the decoherence-free scenario, see
Section 5.3, and basically optimal asymptotic performance in the presence
of decoherence, see Section 6, there is not much in favor of the NOON
states apart from their conceptual appeal.
5.2 Bayesian Approach

Let us now look for the fundamental precision bounds in the
Bayesian approach (Berry & Wiseman, 2000; Hradil et al., 1996). We
assume the flat prior distribution p() 1/2 reflecting our complete initial
ignorance on the true phase value, and the natural cost function

C , 4sin 2 =2, see Section 4.1.2. Thanks to the phase shift
symmetry of the problem, see Section 4.2.2, we can restrict the class of mea-
surements to covariant measurements M U U{ , where is the seed
measurement operator, to be optimized below. Using Equation (93), and
R 2
noting that 0 d 2 4sin =2e 2nm n, m1 + n, m + 1 , the aver-
2 inm
P
aged cost for a general N-photon input state ji N n0 cn jn, N ni reads:
!
X
N
hC i 2 2Re cn cn1 n, n1 : (100)
n1
R
Because is Hermitian, the completeness condition d ^ ^{
2 U U 1
implies that nn 1, while due to the positive semidefiniteness condi-
p
tion 0, jnm j nn mm 1. Therefore, the real part in the subtracted
P
term in Equation (100) can at most be N n1 jcn jjcn1 j, which will be the
case for n,m ei(nm), where n arg cn . This is a legitimate positive
semidefinite operator, as it can be written as jeNiheNj, with
P
jeN i N n0 e jn,N ni (Chiribella et al., 2005; Holevo, 1982). Thus,
in
for a given input state the optimal Bayesian measurement-estimation

strategy yields
!
X
N
hC i 2 1 jcn cn1 j : (101)
n1
For the uncorrelated input state (97), the average cost reads:
s
!
X 1
1 N N N N !1 1
hC i 2 1 N : (102)
2 n0 n n+1 N
Since in the limit of small estimation uncertainty the considered cost func-
tion approaches the MSE, we may conclude that in the limit of large N:
N !1 1
p , (103)
N
which coincides with the standard shot-noise limit derived within the QFI
approach.
Note a subtle difference between the above solution and the solution of
the optimal Bayesian transmission coefficient estimation problem discussed
in Section 4.1.3 with the parametrization employed. The formulas for
probabilities in Section 4.1.3 can be regarded as arising from measuring each
of uncorrelated photons leaving the interferometer independently, while in
the present considerations we have allowed for arbitrary quantum measure-
ments, which are in general collective. Importantly, we account for the
adaptive protocols in which a measurement on a subsequent photon depends
on the results obtained previously (Koody nski & Demkowicz-Dobrza nski,
2010)practically these are usually additional controlled phase shifts
nski et al.
allowing to keep the setup at the optimal operation point (Higgins, Berry,
Bartlett, Wiseman, & Pryde, 2007). This approach is therefore more general
and in particular, does not suffer from a ambiguity that forced us to
restrict the estimated region to [0, ) in order to obtain nontrivial results
given in Equation (75).
Let us now look for the optimal input states. From Equation (102) it is
clear that we may restrict ourselves to real cn. Denoting by c the vector con-
taining the state coefficients cn, we rewrite the formula for hC i in a more
appealing form
hCi 2 cT Ac, An, n1 An1, n 1, (104)

from which it is clear that minimizing the cost function is equivalent to find-
ing the eigenvector with maximal eigenvalue of the matrix A, which has all
its entries zero except for its first off-diagonals. This can be done analytically
(Berry & Wiseman, 2000; Luis & Perina, 1996; Summy & Pegg, 1990) and
the optimal state, which we will refer to as the sine state, together with the
resulting cost read
r
XN
2 n+1
jiin sin jnijN ni, (105)
n0
2+N N +2
h iN !1 2
hC i 2 1 cos : (106)
N +2 N2
Again, in the large N limit we may identify the average cost with the average
MSE, so that the asymptotic precision reads:
N !1
: (107)
N
Analogously, as in the QFI approach, we arrive at the 1/N Heisenberg scaling
of precision, but with an additional constant factor , reflecting the fact
that Bayesian approach is more demanding as it requires the strategy to
work well under complete prior ignorance of the value of the estimated
phase. Even more interestingly, the factor remains in the asymptotic
Bayesian formula even if one considers narrow priors, provided they are reg-
ular and are not changed while approaching the asymptotic limit (Jarzyna &
Demkowicz-Dobrza nski, 2015). Note also that the structure of the optimal
states is completely different from the NOON states. In fact, the NOON
states are useless in absence of any prior knowledge on the phase, since they
are invariant under 2/N phase shift, and hence cannot unambiguously
resolve phases differing by this amount.
5.3 Indefinite Photon-Number States

We now consider a more general class of states with indefinite photon num-
bers and look for optimal probe states treating the average photon number
hN i as a fixed resource. A state with an indefinite number of photons
may posses in general coherences between sectors with different total num-
bers of particles. These coherences may in principle improve estimation pre-
cision. However, a photon-number measurement performed at the output
ports projects the state on one of the sectors and necessarily all coherences
between different total photon-number sectors are destroyed. In order to
benefit from these coherences, one needs to make use of a more general
scheme such as, e.g., homodyne detection, where an additional phase refer-
ence beam is needed, typically called the local oscillator, which is being
mixed with the signal light at the detection stage. Usually, the local oscillator
is assumed to be strong, classical field with a well defined phase. In other
words, it provides one with reference frame with respect to which phase
of the signal beams can be measured (Bartlett et al., 2007; Mlmer,
1997). Thus, it is crucial to explicitly state whether the reference beam
is included in the overall energy budget or is it treated as a free resource.
Otherwise one may arrive at conflicting statements on the achievable fun-
damental bounds ( Jarzyna & Demkowicz-Dobrza nski, 2012).
5.3.1 Role of the Reference Beam

As an illustrative example, consider an artificial one mode scheme with input
in coherent state ji ji which passes through the phase delay . Strictly
speaking, this is not an interferometer and one may wonder how one can
possibly get any information on the phase by measuring the output state. Still
ji evolves into a formally different state j i Uji jeii, where
U ein^, and since the corresponding QFI is nonzero

FQ 4 hjn^2 ji jhjnjij
^ 2 4jj2 , (108)
it is in principle possible to draw some information on the phase by measur-
ing j i. Clearly, the measurement required cannot be a direct photon-
number measurement, and an additional phase reference beam needs to
be mixed with the state before sending the light to the detectors. In a fair
approach, one should include the reference beam into the setup and assume
that whole state of signal + reference beam is averaged over a global
undefined phase. This formalizes the notion of the relative phase delay
it is defined with respect to reference beam and there is no such thing like
absolute phase delay.
nski et al.
More formally, let jiar jiajir, be the original coherent state used for
sensing the phase, accompanied by a coherent reference beam ji. The
corresponding output state reads j iar jeiiajir. Now, the phase
plays the role of the relative phase shift between the two modes, with a clear
physical interpretation. The combined phase shift in the two modes, i.e.,
an operation Ua Ur einâ + n^r , has no physical significance as it is not
detectable without an additional reference beam. Hence, before calculating
the QFI or any other quantity determining fundamental precision bounds,
one should first average the state jiar over the combined phase shift and
treat the resulting density matrix as the input probe state
Z 2 X 1
d a r
U U jiar hjjihjUa{ Ur{ pN N : (109)
0 2 N 0
This operation destroys all the coherences between sectors with different
total photon number, and the resulting state is a mixture of states N with
different total photon numbers N, appearing with probabilities pN.
In the absence of a reference beam, i.e., when 0, the above averaging
kills all the coherences between terms with different photon numbers in the
mode a:
Z X1
d i i jj2 jj2n
je ihe jj0ih0j e jnihnjj0ih0j (110)
2 n0
n!
which results in a state insensitive to phase delays and gives F 0, restoring

our physical intuition. On the other hand, F 4jj2 obtained previously is
recovered in the limit jj2 ! 1, meaning that reference beam is classical
consists of many more photons than the signal beams.
The above averaging prescription, is valid in general also when label a
refers to more than one mode. Considering the standard MZ interferom-
eter fed with a state jiab with an indefinite photon number and no addi-
tional reference beam available one again needs to perform the averaging
over a common phase shift. If this is not done, one may obtain conflicting
results on, e.g., QFI for seemingly equivalent phase shift operations such as
0
U einâ or U einâ n^b =2 . The reason is that, without the common
phase averaging, one implicitly assumes the existence of a strong external
classical phase reference to with respect to which the phase shifts are
defined. In particular, U assumes that the second mode is perfectly locked
with the external reference beam and is not affected by the phase shift,
0
whereas U assumes that there are exactly opposite phase shifts in the
two modes with respect to the reference. Different choices of phase-shift

distribution may lead to a factor of 2 or even factor of 4 discrepancies in
the reported QFIs in apparently equivalent optical phase estimation
schemessee Jarzyna and Demkowicz-Dobrza nski (2012) for further dis-
cussion and compare with some of the results that were obtained without
the averaging performed ( Joo, Munro, & Spiller, 2011; Spagnolo
et al., 2012).
One can also understand why ignoring the need for a reference beam
may result in underestimating the required energy resources. Consider a
P
singe mode state with an indefinite photon number ji N n0 cn jni evolv-
in^
ing under e and note that, from the phase sensing point of view, this
situation is formally equivalent to a two-mode state with a definite photon
P inâ
number N: jiab N n0 cn jnijN ni, evolving under e . Still, the aver-
P
age photon-number consumed in the one mode case is hN i N 2
n0 jcn j n
and is in general smaller than N.
5.3.2 Optimal Indefinite Photon-Number Strategies

Looking for the optimal states with fixed average photon number is in general
more difficult than in the definite photon-number case. Still, if we agree
with the above-advocated approach to average all the input states over a
common phase-shift transformation as in Equation (109), then the resulting
state is a probabilistic mixture of definite photon-number states. Intuitively,
it is then clear that, instead of sending the considered averaged state, it is
more advantageous to have information which particular component N
of the mixture is being sent. This would allow to adjust the
measurement-estimation procedure to a given component and improve
the overall performance. This intuition is reflected by the properties of both
the QFI and the Bayesian cost, which are, respectively, convex and concave
quantities (Helstrom, 1976):
!
X X
FQ pN N pN FQ N , (111)
N N
* !+
X X
C pN N pN hCN i: (112)
N N
This, however, implies that knowing the solution for the optimal definite
photon-number probe states, by adjusting the probabilities pN with which
different optimal N are being sent, we may determine the optimal strategies
with the average photon-number fixed.
nski et al.
Taking the QFI approach for a moment, we recall that the optimal
N-photon state, the NOON state, yields FQ(N) N2. Let us consider a
strategy where a vacuum state and the NOON state are sent with probabil-
ities 1 p and p, respectively, with the constraint on the average photon
number pN hN i. The corresponding QFI reads FQ (1 p) 0 + pN2.
Substituting p hN i/N we get FQ NhN i. Therefore, while keeping hN i
fixed we may increase N arbitrarily and in principle reach FQ 1,
suggesting the possibility of arbitrary good sensing precisions (Rivas &
Luis, 2012; Zhang, Jin, Cao, Liu, & Fan, 2013). Note in particular, that
a naive generalization of the Heisenberg limit to hN1 i, does not hold,
and the strategies beating this bound are referred to as sub-Heisenberg
strategies (Anisimov et al., 2010). A universally valid bound may be
q
written as 1= hN^ i (Hofmann, 2009), but the question remains,
2
whether the bound is saturable, and in particular does quantum mechanics

indeed allows for practically useful estimation protocols leading to the sub-
Heisenberg scaling of precision. Closer investigations of that problem
proves such hypothesis to be false (Berry, Hall, Zwierz, & Wiseman,
2012; Giovannetti & Maccone, 2012; Jarzyna & Demkowicz-
Dobrza nski, 2015; Tsang, 2012; Zwierz, Perez-Delgado, & Kok, 2010).
In principle, we may achieve the sub-Heisenberg precision in the local
estimation regime but in order for the local strategy to be valid, we should
know the value of estimated parameter with prior precision of the same
order as the one we want to obtain, what makes the utility of the proce-
dure questionable. Actually, if no such assumption on the priori knowl-
edge is made, the Heisenberg scaling in the form const/hN i is
the best possible scaling of precision.
This claim can also be confirmed within the Bayesian approach with flat
prior phase distribution. For large N, the minimal Bayesian cost behaves like
hCN i 2 =N 2 , see Equation (105). Since this function is convex, taking
convex combinations of the cost for two different total photon numbers N1,
N2, such that p1N1 + p2N2 hN i will yield the cost higher than 2/hN i2,
and the corresponding uncertainty /hN i, indicating the universal
validity of the Heisenberg scaling of precision (Jarzyna & Demkowicz-
Dobrza nski, 2015).
5.3.3 Gaussian States

From a practical point of view, rather than looking for the optimal indefinite
photon-number states for interferometry it is more important to analyze
experimentally accessible Gaussian states. The paradigmatic example of a
Gaussian state applied in quantum-enhanced interferometry is the two mode

state jijricoherent state in mode a and squeezed vacuum in mode b. We
have already discussed this example in Section 3.4, and calculated the pre-
cision for a simple measurement-estimation scheme. For such states, sent
through a fiftyfifty beam splitter, QFI can been calculated explicitly
( Jarzyna & Demkowicz-Dobrza nski, 2012; Ono & Hofmann, 2010;
Pezze & Smerzi, 2008):
FQ jj2 e2r + sinh 2 r: (113)
For the extreme cases jj2 0, sinh 2 r hN i and jj2 hN i, sinh 2 r 0,

this formula gives FQ hN i, implying the shot noise scaling. Most impor-
tantly, optimization of Equation (113) over and r with constraint
jj2 + sinh 2 r hNi gives asymptotically the Heisenberg limit 1/hN i,
making this strategy as good as the NOON-one for large number of pho-
tons. Moreover, this bound on precision can be saturated by estimation strat-
egies based on photon-number (Pezze & Smerzi, 2008; Seshadreesan,
Anisimov, Lee, & Dowling, 2011) or homodyne (DAriano et al., 1995)
measurements. This also proves that a simple measurement-estimation strat-
egy discussed in Section 3.4 which yielded 1/hN i3/4 scaling of precision is
not optimal. Unlike the simple interferometric scheme where it was optimal
p
to dedicate approximately hN i photons to the squeezed beam, from the
QFI point of view it is optimal to equally divide the number of photons
between the coherent and the squeezed beam.
More generally, finding the fundamental limit on precision achievable
with general Gaussian states, requires optimization of the QFI or the
Bayesian average cost function over general two-mode Gaussian input
states, specified by the covariance matrix and the vector of first moments,
see Section 2.2. For the decoherence-free case this was done in Pinel
et al. (2012, 2013). Crucial observation is that for pure states, the overlap
between two M-mode Gaussian states j i and j + di is given by
(up to the second order in d)
Z 2 !
2 d2 dW z
jh j + d ij 1 242 d 2M
z (114)
4 d
where W(x) is the Wigner function (5) of state j i. Thus, because

jh j + d ij2 1 14 FQ d2 we may write that
nski et al.
Z ! 1
dW z 2 2M 2
242 d z : (115)
d
In terms of the covariance matrix and the first moments hzi, the formula
takes an explicit form:
2 !! 12
dhziT 1 dhzi 1 d 1
+ Tr : (116)
d d 4 d
Formal optimization of the above equation was done by Pinel et al. (2012),
however, the result was a one-mode squeezed-vacuum state, which in order
to carry phase information needs to be assisted by a reference beam. Unfortu-
nately, performing a common phase-averaging procedure described in
Section 5.3.1 in order to calculate the precision in the absence of additional
phase reference destroys the Gaussian structure of the state and makes the opti-
mization intractable. Luckily, for path symmetric states, i.e., the states invariant
under the exchange of interferometer arms, the phase averaging procedure does
not affect the QFI ( Jarzyna & Demkowicz-Dobrza nski, 2012). Hence, assum-
ing the path-symmetry the optimal Gaussian state is given by jrijri with
sinh 2 r hNi=2two squeezed vacuums send into the input ports of the
interferometerand its corresponding QFI leads to the QCRB
1 1
p (117)
hN ihN i + 2 hN i
The state also achieves the Heisenberg limit for a large number of photons in
the setup but it does not require any external phase reference. It is also worth
noting that this state, while being mode-separable is particle entangled and is
feasible to prepare with current technology for moderate squeezing
strengths. However, the enhancement over optimal squeezedcoherent
strategy is rather small and vanish for large number of photons. Precisions
obtained in squeezedcoherent and squeezedsqueezed scenarios are
depicted in Figure 6.
One can also study Gaussian states within the Bayesian framework.
Optimal seed operator can be easily generalized from the definite
P
photon-number case to 1 N 0 jeN iheN j. Conceptually, the whole treat-
ment is the same as in the definite photon-number case. However, the
expressions and calculations are very involved and will not be
presented here.
1

N
1

N
N
Figure 6 Limits on precision obtained within QFI approach when using two optimally
squeezed states in both modes jrijri (black, solid), coherent and squeezed-vacuum
states jijri (black, dashed). For comparison, precision achievable with simple coherent
and squeezed vacuum MZ interferometric scheme discussed in Section 3.4 is also
depicted (gray, dashed).
5.4 Role of Entanglement

The issue of entanglement is crucial in quantum interferometry as it is
known that only entangled states can beat the shot noise scaling
(Pezze & Smerzi, 2009). This statement is sometimes questioned, pointing
out the example of the squeezed + coherent light strategy, where the inter-
ferometer is fed with seemingly unentangled jijri input state. The reason
of confusion is the conflict of notions of mode and particle entanglement.
As discussed in detail in Section 2.5, the two notions are not compatible,
and there are states which are particle entangled, while having no mode
entanglement and vice versa. In the context of interferometry, it is the
particle entanglement that is the source of quantum enhancement of
precision. In order to avoid criticism based on the ground of fundamental
indistinguishability of particles and, therefore, a questionable physical
content of the distinguishable particle-based entanglement picture on
the fundamental level (Benatti, Floreanini, & Marzolino, 2010), we should
stress that when considering models involving indistinguishable particles
one should regard this statement as a formal (but still a meaningful and
useful) criterion where the particles are treated as formally distinguishable
as described in Section 2.4.
To see this, let us consider first a separable input state of N photons of the
form 1 N , where i denotes the state of the i-th photon. Since
nski et al.
the phase shift evolution affects each of the photons independently

UNUN and the QFI is additive on product states we may write:
X
N
FQ FQ i NFQ max (118)
i1
where max denotes state from the set fi gi1, ..., N for which QFI takes the
largest value. But for a one photon state, the maximum value of QFI is equal
to 1, so
1
FQ N, p : (119)
N
P i i
For general separable states i pi 1 N , it is sufficient to use
the convexity of QFI, together with Equation (119) to obtain the same con-
clusion. Above results imply that QFI, or precision, can be interpreted as a
particle-entanglement witness, i.e., all states that give precision scaling better
than the shot noise must be particle entangled (Hyllus et al., 2012; Toth, 2012).
The seemingly unentangled state jijri when projected on the definite
total photon-number sector, indeed contains particle entanglement as was
demonstrated in Section 2.5. This fact should be viewed as the fundamental
source of its ability for performing quantum-enhanced sensing. It is also
worth stressing, that unlike mode entanglement, particle entanglement is
invariant under passive optical transformation like beam splitters, delay lines
and mirrors, which makes it a sensible quantity to be treated as a resource for
quantum-enhanced interferometry.
5.5 Multipass Protocols

A common method, used in, e.g., gravitational-wave detectors, to improve
interferometric precision is to let the light bounce back and forth through
the phase delay element many times so that the phase delay signal is enhanced
as shown in Figure 7. This method is used in GEO600 experiment (LIGO
Collaboration, 2011), where the light bounces twice in each of the sensing
arms, making the detector as sensitive as the one with arms twice as long.
Up to some approximation, one can also view the FabryPerot cavities placed
on top of the MZ design as devices forcing each of the photon to pass
multiple-times through the arms of the interferometer and acquire a multiple
of the phase delay (Berry et al., 2009; Demkowicz-Dobrza nski et al., 2013).
Consider a single photon in the state (after the first beam splitter)
jiin p12 j01i + j10i. After passing through the phase shift N times, it
Figure 7 A multipass interferometric protocol. A standard phase delay element is rep-

laced by a setup which makes the beam to pass through the phase delay multiple num-
ber of times.
evolves into j i p12 j01i + eiN j10i. The phase is acquired N times
faster compared with a single pass case, mimicking the behavior of a single
pass experiment with a NOON state. Hence, the precision may in prin-
ciple be improved by a factor of N. Treating the number of single photon
passes as a resource, it has been demonstrated experimentally (Higgins
et al., 2007) that in the absence of noise such a device can indeed achieve
the Heisenberg scaling without resorting to entanglement and efficiency of
various multipass protocols has been analyzed in detail in Berry et al.
(2009). This is not to say, that all quantum strategies are formally equivalent
to single-photon multipass strategies. As will be discussed in the next
section, the NOON states are extremely susceptible to decoherence, in
particular loss, and this property is shared by the multipass strategies. Other
quantum strategies prove more advantageous in this case, and they do not
have a simple multipass equivalent (Demkowicz-Dobrza nski, 2010;
Kaftal & Demkowicz-Dobrza nski, 2014).
6. QUANTUM LIMITS IN REALISTIC INTERFEROMETRY

In this section, we revisit the ultimate limits on precision derived in
Section 5 taking into account realistic noise effects. We study three
decoherence processes that are typically taken into account when discussing
imperfections in interferometric setups. We consider the effects of phase dif-
fusion, photonic losses and the impact of imperfect visibility, see Figure 8. In order
to establish the ultimate limits on the estimation performance, we first ana-
lyze the above three decoherence models using the QFI perspective and sec-
ondly compare the bounds obtained with the ones derived within the
Bayesian approach.
For the most part of this section, we will consider input states with
definite number of photons, N, so that
nski et al.
Figure 8 Schematic description of the decoherence processes discussed that affect the
performance of an optical interferometer: (i) phase diffusion representing stochastic fluc-
tuations of the estimated phase delay, (ii) losses in the respective a/b arms represented
by fictitious beam splitters with 0 a/b 1 transmission coefficients, and (iii) imperfect
visibility indicated by a mode mismatch of the beams interfering at the output beam
splitter.
X
N
in jiin hj, jiin cn jnijN ni: (120)
n0
Similarly as in Section 5, this will again be sufficient to draw conclusions also

on the performance of indefinite photon-number states, which will be dis-
cussed in detail in Section 6.4.1.
In what follows, it will sometimes prove useful to switch from mode to
particle description, see Section 2.4, and treat photons formally as distin-
guishable particles but prepared in a symmetrized state. This approach is
illustrated in Figure 9 where each photon is represented by a different hor-
izontal line, and travels through a phase encoding transformation
i
Ui ei^z =2 , where î
z is a z Pauli matrix acting on the i-th qubit.
The combined phase encoding operation is a simple tensor product
P i
UN ei i ^z =2 eiJ^z , recovering the familiar formula but with J^z
being now interpreted as the z component of the total angular momentum
which is the sum of individual angular momenta. The photons are then
subject to decoherence that acts in either correlated or uncorrelated man-
ner. In the case of phase diffusion, (i) the decoherence has a collective char-
acter since each of the photons experiences the same fluctuation of the
phase being sensed, while in the case of loss (ii) and imperfect visibility
(iii) the decoherence map has a tensor structure N reflecting the fact that
it affects each photon independently. In the latter case of independent
decoherence models, the overall state evolution is uncorrelated and may
be written as:
Figure 9 General metrological scheme in case of photons being treated as formally dis-
tinguishable particles. Each photon travels through a phase encoding transformation
U. Apart from that all photons are subject to either correlated (i) (phase diffusion),
or uncorrelated (ii), (iii) (loss, imperfect visibility) decoherence process.
N {
in , U U : (121)
6.1 Decoherence Models

In general, decoherence is a consequence of the uncontrolled interactions of
a quantum system with the environment. Provided the system is initially
decoupled from the environment, the general evolution of a quantum sys-
tem interacting with the environment mathematically corresponds to a
completely positive trace preserving map . Every can be written using the
Kraus representation (Nielsen & Chuang, 2000):
X X {
out in Ki in Ki{ , Ki Ki 1, (122)
i i
where Ki are called the Kraus operators.

Effects of decoherence inside an interferometer are taken into account by
replacing the unitary transformation U describing the action of the ideal
interferometer, see Section 5, with its noisy variant :
!
X
in Ki U in U{ Ki{ : (123)
i
The formal structure of the above formula corresponds to a situation in

which decoherence happens after the unitary phase encoding. This of course
might not be true in general. Still, for all the models considered in this review
the decoherence part and the unitary part commute and therefore the order
in which they are written is a matter of convenience.
nski et al.
6.1.1 Phase Diffusion

Phase diffusion, also termed as the collective dephasing or the phase noise repre-
sents the effect of fluctuation of the estimated phase delay . Such effect may
be caused by any process that stochastically varies the effective optical lengths
traveled by the photons, such as thermal deformations or the micro-motions
of the optical elements. We model the optical interferometry in the presence
of phase diffusion process by the following map:
Z1
in d p U in U{ , (124)
1
where the phase delay is a random variable distributed according to prob-

ability distribution p(). Note that the above form is actually the Kraus rep-
p
resentation of the map with Kraus operators K p U . In case
p() is a Gaussian distribution with variance and the mean equal to
2
the estimated parameter , p p
1
e 2 , the output state reads
2
explicitly (Genoni, Olivares, & Paris, 2011):
X
N 2
cn cm? e 2 n m einm jn,N ni hm, N mj, (125)
n, m0
where cn are parameters of the input state given as in (120). The above equa-
tion indicates that due to the phase diffusion the off-diagonal elements of
are exponentially suppressed at a rate increasing in the antidiagonal
directions.
6.1.2 Photonic Losses

In the lossy interferometer scenario, the fictitious beam-splitters introduced in
the interferometer arms with respective power transmission coefficient a/b
account for the probability of photons to leak out. Such a loss model is
relatively general, as due to the commutativity of the noise with the phase
accumulation (Demkowicz-Dobrza nski et al., 2009), it accounts for the
photonic losses happening at any stage of the phase sensing process.
Moreover, losses at the detection as well as the preparation stages can
be moved inside the interferometer provided they are equal in both
arms. This makes the model applicable in typical experimental realization
of quantum-enhanced interferometry (Kacprowicz, Demkowicz-
Dobrza nski, Wasilewski, & Banaszek, 2010; Spagnolo et al., 2012; Vitelli,
Spagnolo, Toffoli, Sciarrino, & De Martini, 2010), and most notably, when
analyzing bounds on quantum enhancement in gravitational-wave detectors
(Demkowicz-Dobrza nski et al., 2013).
Loss decoherence map may be formally described using the following
set of Kraus operators (Dorner et al., 2009):
s s
1 a a^{ a^ la 1 b lb b^{ b^ ^lb
la
Kla , lb a a^ b b (126)
la ! lb !
where the values of index la/b corresponds to the number of photons lost in
mode a/b, respectively. For a general N-photon input state of the form
(120), the density matrix representing the output state of the lossy interfer-
ometer reads
M
N NX
N 0

U in U{ pla , lb la , lb la , lb ,
(127)
N 0 0 la 0
lb N N 0 la
P 2 la , lb
where pla , lb N n0 jcn j bn is the binomially distributed probability of
losing la and lb photons in arms a and b, respectively, with

la , lb n nla la N n N nlb
bn 1 a b 1 b lb , (128)
la a lb
while the corresponding conditional pure states read:
q
NXlb
cn ein
l , l p
l , l
bn a b jn la ,N n lb i: (129)
a b
nla
pla , lb
The direct sum in Equation (127) indicates that the output states of different
total number of surviving photons, N 0 , belong to orthogonal subspaces,
which in principle could be distinguished by a nondemolition, photon-
number counting measurement.
In the particle-approach when photons are considered as formally distin-
guishable particles, the loss process acts on each of the photons indepen-
dently, see Figure 9, so that the overall decoherence process has a tensor
product structure N, with being a single particle loss transformation.
At the input stage, each photon occupies a two-dimensional Hilbert space
spanned by vectors jai, jbi representing the photon traveling in the mode
a/b, respectively. In order to describe loss, however, and formally keep
nski et al.
the number of particles constant, one needs to introduce a third photonic

state at the output, jvaci, representing the state of the photon being lost.
Then formally, maps states from the input two-dimensional Hilbert space
to the output three-dimensional Hilbert space, and can be fully specified by
P
means of the Kraus representation, 3i1 Ki Ki{ , where K1, K2, K3
are given, respectively, by the following matrices:
0 p 1 0 1 0 1
a 0 0 0 0 0
@ 0 p b A , @ p 0A , @0 A
p :
0 0 (130)

0 0 1 a 0 0 1 b
Intuitively, the above Kraus operators account for no photon loss, photon
loss in mode a and photon loss in mode b, respectively. When applied to
symmetrized input states, this loss model yields output states
N UN in U{N N
in , (131)
equivalent to the ones given in Equation (127), where U is a single photon
phase-shift operation U ei^z =2 .
6.1.3 Imperfect Visibility

In real-life optical interferometric experiments, it is always the case that the
light beams employed do not contribute completely to the interference pat-
tern. Due to spatiotemporal or polarization mode-mismatch, caused for
example by imperfect wave-packet preparation or misalignment in the optical
elements, the visibility of the interference pattern is diminished (Leonhardt,
1997; Loudon, 2000). This effect may be formally described as an effective
loss of coherence between the two arms a and b of an interferometer.
Consider a single photon in a superposition state of being in modes a
and b, respectively: ji jai + jbi. If other degrees of freedom such
as, e.g., polarization, temporal profile, etc., were identical for the two
states jai, jbi, we could formally write (jai + jbi)j0iX, where j0iX
denotes the common state of additional degrees of freedom. Loss of coher-
ence may be formally described as the transformation of the state jij0iX
into (Banaszek, Horodecki, Karpinski, & Radzewicz, 2013)
p p
ji jaij0iX + 1 jaij + iX
p p
+ jbij0iX + 1 jbij iX ,
(132)
where j+iX, jiX are states orthogonal to j0iX, corresponding to photon

traveling in, e.g., orthogonal transversal spacial modes as depicted in
Figure 8 (iii), in which case parameter can be interpreted as transmission
of fictitious beam splitters that split the light into two orthogonal modes.
Assuming we do not control the additional degrees of freedom the effective
state of the photon is obtained by tracing out the above state over X,
obtaining the effective single-photon decoherence map:

jj2
jihj TrX jihj , (133)
jj2
where the off-diagonal terms responsible for coherence, are attenuated by

coefficient , what corresponds to the standard dephasing map (Nielsen &
Chuang, 2000). Written using the Kraus representation, the above map
reads:
r r
X
2
1+ 1
Ki Ki{ , K1 1, K2 z: (134)
i1
2 2
Note that similarly to the loss model we have modeled the noise with the
use of fictitious beam-splitters to visualize the effects of decoherence. Now,
as we know that a beam-splitter acts on the photons contained in its two-
mode input state in an uncorrelated manner, the effective map on the full
N-photon input state is N. In case of atomic systems, this would be a typ-
ical local dephasing model describing uncorrelated loss of coherence
between the two relevant atomic levels (Huelga et al., 1997). Still, there
is a substantial difference from the loss models as the dephased photons
are assumed to remain within the spatially confined beams of the
interferometer arms.
We can relate the two models by a simple observation, namely that if
the photons lost in the loss model with a b were incoherently
injected back into the arms of the interferometer, we would recover the
local dephasing model with the corresponding parameter . It should
therefore come as no surprise, when we derive bounds on precisions for
the two models in Sections 6.2.1 and 6.2.2, that for the same the
local dephasing (imperfect visibility) model implies more stringent bounds
on achievable precision than the loss model. Intuitively, it is better to get
rid of the photons that lost their coherence and do not carry information
about the phase, rather than to inject them back into the setup.
nski et al.
The structure of the output state is more complex than in the case of
phase diffusion and loss models. This is because the local dephasing noise not
only transforms the input state into a mixed state, but due to tracing out
some degrees of freedom, the output state
X
2
N
in Ki1 KiN UN in UN { Ki{1 Ki{N (135)
i1 , ..., iN 1
is no longer supported on the bosonic space spanned by the fully symmetric

states jnijN ni. This makes it impossible to use the mode-description in
characterization of the process. Even though it is possible to write down the
explicit form of the above state (Fr owis, Skotiniotis, Kraus, & D ur, 2014;
Jarzyna & Demkowicz-Dobrza nski, 2015) decomposing the state into
SU(2) irreducible subspaces, we will not present it here for the sake of con-
ciseness, especially that it will not be needed in derivation of the fundamental
bounds.
6.2 Bounds in the QFI Approach

Once we have the formula for the output states , given a particular
decoherence model, we may use Equation (77) to calculate QFI, FQ(),
which sets the limit on practically achievable precision of estimation of .
In order to obtain the fundamental precision bound for a given number N
of photons used, we need to maximize the resulting FQ over input states
jiin, which will in general be very different from the NOON states
which maximize the QFI in the decoherence-free case. This is due to
the fact that the NOON states are extremely susceptible to decoherence,
e.g., loss of a single photon makes the state completely useless for phase
sensing. Unfortunately, for mixed states, the computation of the QFI
requires in general performing the eigenvalue decomposition of and
such a minimization ceases to be effective for large N. Therefore, while
it is relatively easy to obtain numerical bounds on precision and the form
of optimal states for moderate N (Demkowicz-Dobrza nski, 2010; Dorner
et al., 2009; Huelga et al., 1997), going to the large N regime poses a huge
numerical challenge, making determination of the asymptotic bounds for
N ! 1 with brute force optimization methods infeasible.
Over the past few years, elegant methods have been proposed that allow
to circumvent the above mentioned difficulties and obtain explicit bounds
on precision based on QFI for arbitrary N, and in particular grasp the
asymptotic precision scaling (Demkowicz-Dobrza nski et al., 2012;

Demkowicz-Dobrza nski & Maccone, 2014; Escher et al., 2011; Knysh
et al., 2014). These methods include: the minimization over channel purifications
method (Escher et al., 2011) which is applicable in general but requires some
educated guess to obtain a useful bound, as well as classical and quantum sim-
ulation methods (Demkowicz-Dobrza nski et al., 2012) which are applicable
when the noise acts in an uncorrelated manner on the probes, but have an
advantage of being explicit and convey some additional physical intuition on
the bounds derived. Description of the newly published method (Knysh
et al., 2014) which is based on continuous approximation of the probe states
and the calculus of variations is beyond the scope of this review.
We will present the methods by applying them directly to interferometry
with each of the decoherence models introduced above. We invert the order
of presentation of the bounds for the decoherence models compared with
the order in Section 6.1, as this will allow us to discuss the methods in
the order of increasing complexity. The simplest of the methods, the classical
simulation (CS), will be applied to the imperfect visibility model, while the
quantum simulation (QS) will be discussed in the context of loss. Finally,
minimization over channel purification method will be described in the
context of the phase diffusion model, to which classical and QS methods
are not applicable due to noise correlations. We should emphasize that
the methods of Escher et al. (2011) and Knysh et al. (2014) can also be suc-
cessfully applied to uncorrelated noise models. Still, CS and QS approaches
are more intuitive and that is why we present derivations based on them
even though the other techniques yield equivalent bounds.
In order to appreciate the significance of the derived bounds, we will
always compare them with the precision achievable with a state of N
uncorrelated photons as given in Equation (97). The ratio between this
quantities bounds the amount of quantum-precision enhancement that
can be achieved with the help of quantum correlations present in the input
state of N photons.
6.2.1 Imperfect Visibility

The fundamental QFI bound on precision in case of imperfect visibility or
equivalently the local dephasing model has been derived in Escher et al.
(2011), Demkowicz-Dobrza nski et al. (2012), and Knysh et al. (2014)
and reads:
nski et al.
s
1 2 1
p , (136)
2 N
where is the dephasing parameter, see Section 6.1.3. For the optimal
p p
uncorrelated input state, j in i jai + jbi= 2N we get 1= 2 N ,
and hence the quantum precision enhancement which is the ratio of the
bound on precision achievable for the optimal strategy and the precision
p
for the uncorrelated strategy is bounded by a constant factor of 1 2 .
CS Method
The derivation of the formula (136) presented below makes use of the CS
method (Demkowicz-Dobrza nski et al., 2012), which requires viewing the
quantum channel representing the action of the interferometer from a geo-
metrical perspective. The set of all physical quantum channels,
: LHin ! LHout , that map between density matrices defined on
the input/ouput Hilbert spaces (Hin=out ) constitutes a convex set
(Bengtsson & Zyczkowski, 2006). This is to say that given any two quantum
channels 1, 2, their convex combination p1 + (1 p)2, 0 p 1
is also a legitimate quantum channel. Physically corresponds then to a
quantum evolution that is equivalent to a random application of 1, 2
transformations with probabilities p, 1 p, respectively.
As derived in Section 6.1.3, within the imperfect visibility (local
dephasing) decoherence model: N(in), and hence the relevant
quantum channel, has a simple tensor structure. Consider a single-photon
channel , which -dependence we may depict as a trajectory within
the set of all single-photon quantum maps, see Figure 10. The question
of sensing the parameter can now be translated to the question of deter-
mining where on the trajectory a given quantum channel lies.
Consider a local CS of a quantum channel trajectory in the vicinity of
a given point 0, 0 + (Demkowicz-Dobrza nski et al., 2012;
Matsumoto, 2010),
X
p x x + O 2 , (137)
x
which represents the variation of the channel up to the first order in as a

classical mixture of some -independent channels fx gx , where the
dependence is present only in the mixing probabilities p(x). Under such
a construction the random variable X distributed according to p(x) specifies
Figure 10 The space of all quantum channels, , which map between density matrices
specified on two given Hilbert spaces, : LHin ! LHout , represented as a con-
vexset. The estimated parameter specifies a trajectory, (black curve), in such
a space. From the point of view of the QFI, any two channel trajectories, e.g., and

(gray curve), are equivalent at a given 0 as long as they and their first derivatives
with respect to coincide there. Moreover, they can be optimally classically simulated at
0 by mixing two channels lying on the intersection of the tangent to the trajectory and
the boundary of the set: .
probabilistic choice of channels x that reproduces the local action of in

the vicinity of 0. Crucially, the QFI is a local quantitysee discussions in
Section 4.2.1and at a given point 0 is a function only of the quantum
state considered and its first derivative with respect to the estimated param-
eter. Consequently, when considering the parameter being encoded in a
quantum channel, all the channel trajectories at a given point 0 are equiv-
alent from the point of view of QFI if they lead to density matrices that are
identical up to the first order in . In other words, we can replace with

any and obtain the same QFI at a given point 0 provided 0 0 and

d d
d d , see Figure 10. This means that when constructing a local CS
0 P
of the quantum channel at 0, we need only to satisfy x p0 xx 0 ,
P dp x d
as well as x d j0 x d j0 .
Crucially, as the maps in Equation (135) act independently on each
photon, we can simulate the overall N with N independent random vari-
N
ables, X , associated with each channel. The estimation procedure can now
be described as

! X N ! N N
! j in i ! , : (138)
where N classical random variables are employed to generate the desired

quantum map N . It is clear that a strategy in which we could infer the
nski et al.

parameter directly from XN, i.e., ! X N ! , can perform only better
than the scheme where the information about is firstly encoded into the
quantum channel which acts on the probe state and afterwards decoded from
the measurement results performed on the output state. This way, we may
always construct a classically scaling lower bound on the precision, or equiv-
alently an upper bound on the QFI of (135):
h i
FQ Fcl pN N Fcl p , (139)
which is determined by the classical FI (54) evaluated for the probability dis-
tribution p(X). Importantly, Demkowicz-Dobrza nski et al. (2012) have
shown that in this particular case thanks to the parameter being unitary
encoded, it is always optimal to choose a CS depicted in Figure 10, which
employs for each only two channels lying at the points of intersection
of the tangent to the trajectory with the boundary of the quantum maps set.
Such an optimal CS leads to the tightest upper bound specified in

Equation (139): FQ N = + , where are the distances to
d
the boundary marked in Figure 10, 0 d j0 .
Looking for parameters amounts to a search of the distances one can
go along the tangent line to the trajectory of such that the corresponding
map is still a physical quantum channel, i.e., a completely positive trace
preserving map. This is easiest to do making use of the ChoiJamiolkowski
isomorphism (Choi, 1975; Jamiokowski, 1972) which states that with
each quantum channel, : LHin ! LHout , we can Nassociate a positive
operator 2 L(Hout Hin ), so that ( I )(jIihIj), where
PdimHin
jIi i1 jiijii is a maximally entangled state on Hin Hin , while I is
the identity map on L(Hin ). Since the complete positivity of is equivalent
d
to positivity of the operator, one needs to analyze 0 d j0 and
find maximum so that the above operator is still positive-semidefinite.
Taking the explicit form of the for the case of optical interferometry
with imperfect visibility, see Section 6.1.3, one can show that
p
1 2 = (Demkowicz-Dobrza nski et al., 2012), which yields the
ultimate quantum limit on precision given by Equation (136).

The expression for the QFI of the output state (127) in the asymptotic limit
of large N has been first derived by Knysh et al. (2011). Yet, the general
frameworks proposed by Escher et al. (2011) and Demkowicz-Dobrza nski

et al. (2012) for generic decoherence allowed to reconstruct this bound on
precision with the following result:
s s!
1 1 a 1 b 1
+ p , : (140)
2 a b N
where a, b are transmission in the two arms of the interferometer, respec-

tively, see 6.1.2. This bound simplifies to
r
1
(141)
N
in the case of equal losses, and since the precision achievable with
p
uncorrelated states is given by 1= N , the maximal quantum-enhancement
p
factor is 1 . In the following, we derive the above bounds using the QS
approach of Demkowicz-Dobrza nski et al. (2012) and Koodynski and
Demkowicz-Dobrza nski (2013).
QS Method
Unfortunately, in the case of loss the simple CS method yields a trivial bound
0, since the tangent distances to the boundary of the set of quantum
channels are 0 in this case. It is possible, however, to derive a useful
bound via the QS method which is a natural generalization of the CS
method. The QS method has been described in detail and developed for
general metrological schemes with uncorrelated noise by Koody nski
and Demkowicz-Dobrza nski (2013) stemming from the works of
Demkowicz-Dobrza nski et al. (2012) and Matsumoto (2010).
As shown in Figure 11, local QS amounts to reexpressing the action of
for 0 + by a larger -independent map that also acts on the
auxiliary -dependent input , up to the first order in :

TrE + O2 : (142)
P
Note that for x p xjxi hxj, and jxihxj x, QS becomes
equivalent to the CS of Equation (137), so that CS is indeed a specific
instance of the more general QS. An analogous reasoning as in the case
of CS leads to the conclusion that we may upper-bound the QFI of the over-
all output state, here (127) for the case of losses, as
nski et al.

Figure 11 The quantum simulation (QS) of a channel. The action of the channel is
simulated up to the first order in the vicinity of a given point 0 using a -independent
channel and an auxiliary state that contains the full information about the esti-
mated parameter .
h i h n h ioi h i
FQ N
in F Q Tr E N
N
in F Q N
N FQ ,
since TrE N[] is just a parameter independent map, under which the over-
all QFI may only decreasesee Equation (82). Last equality follows from the
additivity property of the QFI, which, similarly to Equation (139), constrains
h i
FQ N to scale at most linearly for large N. Similarly to the case of CS, in
order to get the tightest bound one should find QS that yields the smallest
FQ[ ], which in principle is a nontrivial task.
Fortunately, Koody nski and Demkowicz-Dobrza nski (2013) have
demonstrated that the search for the optimal channel QS corresponds to
the optimization over the Kraus representation of a given channel.
Without loss of generality we may assume that jihj is a pure
-dependent state while [] U U is unitary. For a given QS, we
may write the corresponding Kraus representation of the channel by
choosing a particular basis jiiE in the E space: Ki() EhijUjiE. In order
for the QS to be valid, these Kraus operators should P correspond to a legit-
imate Kraus representation of the channel i Ki Ki . Two
Kraus representation of a given quantum channel are equivalent if and only
if they are related by a unitary matrix u:

X
K i uij Kj , (143)
j
which may in principle be also dependent. Since we require QS to be only

locally valid in the vicinity of 0, the above equation as well as its first deriv-
ative needs to be fulfilled only at 0. Because of that, the search for the opti-

mal Kraus representation K i (or equivalently the optimal QS) may be
restricted to the class of transformations where u() ei(0)h with h being
any Hermitian matrix that shifts the relevant derivative of Ki() at 0, so that
_ P
K i 0 Ki 0 and K i 0 K_ i 0 + i j hij Kj 0 . As shown by
Koodynski and Demkowicz-Dobrza nski (2013), the problem of finding the

optimal QS, i.e., the minimal FQ jihj which we term as the FQS, can be
formally rewritten as
FQS min s s:t:
h
X _ { _ s X _ (144)
K i 0 K i 0 12 , K i 0 { Ki 0 0,
i
4 i
where the parameter s has the interpretation of FQ jihj for the partic-
ular QS at 0 and the constraints imposed in Equation (143) are necessary
and sufficient for the QS required transformation U and the state ji
to exist.
The above optimization problem may not always be easy to solve. Still,
its relaxed version:
X _ X _
Ki 0 { K_i 0 k ,

min k
h
Ki 0 { K i 0 0, (145)
i i
where kk is the operator norm, can always be cast in the form of an explicit
semidefinite program, which can be easily solved numerically (Demkowicz-
Dobrza nski et al., 2012). Numerical solution of the semidefinite program
provides a form of the optimal h which may then be taken as an ansatz
for further analytical optimization.
Plugging in the Kraus operators Ki() KiU representing the lossy
interferometer, see Equation (130), and following the above described
procedure one obtains
4
FQS q q2
1a 1b (146)
+a b
for the optimal h given by

1 a 4 b 4
hopt diag , , + , (147)
8 1 a a 1 b a
where FQS a b
. This indeed reproduces the bounds given in (140).
a b
In order to provide the reader with a simple intuition concerning the QS
method, we shall present an elementary construction of the QS for lossy
interferometer in the special case of a b 1/2. In this case, the bound
nski et al.
p
(140) yields 1= N which implies that using optimal entangled probe
state at the input under 50% losses cannot beat the precision which can be
obtained by uncorrelated probes in the ideal lossless scenario.
Consider the action of the single photon lossy channel on the pure
input state ji jai + jbi:
1 1
jihj j ih j + jvacihvacj (148)
2 2
with j i eijai + jbi, which represents 1/2 probability of photon sens-

ing the phase undisturbed and the 1/2 probability of the photon being lost.
p
Let the auxiliary state for QS be ji ei j0i + j1i= 2. The joined input +
auxiliary state reads:
1 1
jiji p ei j0ijai + j1ijbi + p ei j0ijbi + j1ijai : (149)
2 2
The map realizing the QS consists now of two steps. First the controlled
NOT operation is performed with the auxiliary system being the target
qubit, this transform the above state to: p12 j0iei jai + jbi + p12 j1i
ei jbi + jai. The second step is the measurement of the auxiliary system.
If the result j0i is measured (probability 1/2), the system is left in the correct
state j i and the map leaves it unchanged, if the j1i is measured the state of
the photon is not the desired one, in which case the map returns the jvaci
state. This map is therefore a proper QS of the desired lossy interferometer
transformation for a b 1/2. Since the auxiliary state employed in this
p
construction was ji ei j0i + j1i= 2 for which FQ(jihj N) N this
p
leads to the anticipated result 1= N .

Since the phase diffusion model, see Section 6.1.1, is an example of a cor-
related noise model, it cannot be approached with the CS and QS methods.
The study of the behavior of the QFI within the phase-diffusion model was
for the first time carried out by Genoni et al. (2011) considering indefinite-
photon-number Gaussian input states and studied numerically the achiev-
able precision and the structure of optimal input states. Yet, the fundamental
analytical bounds on precision have not been verified until Escher et al.
(2012), where the phase noise has been approached using the minimization
over purifications method of Escher et al. (2011) and most recently using the
calculus of variations approach of Knysh et al. (2014).
Minimization Over Purification Method

The minimization over purification method of Escher et al. (2011) is
based on the observation, already mentioned in Section 4.2.1, that QFI
for a given mixed quantum state, here (127), is not only upper bounded
by the QFI of any of its purifications, but there always exists an optimal
E h Ei
opt
purification, opt
, for which F Q FQ , where
n ED o

TrE opt opt . As such statement does not rely on the form
of the transformation jiin ! but rather on the properties of the output
state itself, the framework of Escher et al. (2011) in principle does not put
any constraints on the noise-model considered. Note that, even if the opti-
mal purification itself is difficult to find, any suboptimal purification yields a
legitimate upper bound on the QFI and hence may provide a nontrivial
precision bound.
In order to get a physical intuition regarding the purification method,
consider a physical model of the phase diffusion where light is being reflected
from a mirror which position fluctuations are randomly changing the effec-
tive optical length. Formally, the model amounts to coupling the phase delay

generator J^z to the mirror position quadrature xÊ p1 aÊ + a^{E (Escher
2
et al., 2012). Assuming the mirror, serving as the environment E, to reside in
the ground state of a quantum oscillator j0iE before interaction with the light
beam, the pure output state reads:
p
eiJ^z ei 2J^z xÊ ji j0i : (150)
in E
2 p
Thanks to the fact that jhxj0ij2 ex = , the reduced state
Z1

dx E xj jx E (151)
1
indeed coincides with the correct output phase diffused state (125). There-
fore, this is a legitimate purification of the interferometer output state in
presence of phase diffusion.
nski et al.
E

Consider now another purification eiH^ E generated by a
local ( 0 + ) rotation of the mirror modes, i.e., a unitary operation on
the system E. We look for a transformation of the above form which hope-
fully erases as much information
E on the estimated phase as possible, so that

QFI for the purified state will be minimized leading to the best bound
on the QFI of . Choosing H^ E ^
p E we obtain the following upper
bound on the QFI

FQ min FQ ei^p E ji

p 2 4 2 Jz
min 22 + 4 1 2 2 Jz ,
1 + 4 2 Jz
and thus a lower limit on the precision
r r
1 1
+ 2 + 2, (152)
4 Jz N
where we plugged in 2Jz N2/4 corresponding to the NOON state
which maximized the variance for N-photon states. See also an alternative
derivation of the above result that has been proposed recently in
Macieszczak (2014). Crucially, the above result proves that the phase diffu-
p
sion constrains the error to approach a constant value as N ! 1, which
p
does not vanish in the asymptotic limit, what contrasts the 1= N behavior
characteristic for uncorrelated noise models. Note also that due to the cor-
related character of the noise, the bound (152) predicts that it may be more
beneficial to perform the estimation procedure on a group of k particles and
p
then repeat the procedure independently times obtaining 1= reduction
in estimation error, rather than employing N k in a single experimental
shot (Knysh et al., 2014).
Only very recently, the exact ultimate quantum limit for the N-photon
input states has been derived in (Knysh et al., 2014)
r
2
+ 2 , (153)
N
showing that the previous bound was not tight, with the second term fol-
lowing the HL-like asymptotic scaling of the noiseless decoherence-free
Bayesian scenario stated in Equation (107). In fact, as proven by Knysh
et al. (2014), the optimal states of the noiseless Bayesian scenario, i.e., the
sine states (105), attain the above correct quantum limit. In Section 6.3.2, we
show that within the Bayesian approach, with the phase-diffusion effects
incorporated, the sine states are always the optimal inputs.
6.3 Bayesian Approach

Minimizing the average Bayesian cost, as given by Equation (93), over input
probe states jiin is in general more demanding than minimization of the
QFI due to the fact that it is not sufficient to work within the local regime
and analyze only the action of a channel and its first derivative at a given
estimation point as in the QFI approach. For this reason, we do not apply
the Bayesian approach to the imperfect visibility model as obtaining the
bounds requires a significant numerical and analytical effort (Jarzyna &
Demkowicz-Dobrza nski, 2015), and constrain ourselves to loss and phase
diffusion models.

The optimal Bayesian performance of N-photon states has been studied by
Koodynski and Demkowicz-Dobrza nski (2010). Assuming the natural cost
function (67) and the flat prior phase distribution, the average cost (93) reads:
n
o
hC i Tr C (154)

R
2
where C 4 d 2 sin 2 and is given by (127). The optimal
measurement seed operator can be found analogously as in the
decoherence-free case. The block diagonal form of , implies that without
L
losing optimality one can assume N N 0 0 jeN 0 i heN 0 j with
PN 0 0
jeN 0 i n0 jn, N ni. Physically, the block-diagonal structure of
indicates that the optimal covariant measurement requires a non-
demolition photon-number measurement to be performed before
carrying out any phase measurements, so that the orthogonal subspaces,
labeled by the number of surviving photons N 0 , may be firstly distinguished,
and subsequently the measurement which is optimal in the lossless case is
performed (Koody nski & Demkowicz-Dobrza nski, 2010). Plugging in
the explicit form of together with the explicit form of the output state
, we arrive at
n,X N nq
l , l la , lb
hCi 2 c Ac,T
An, n1 An1, n bn a b bn1 , (155)
la , lb 0
nski et al.
where A is a symmetric N + 1 N + 1 matrix that is nonzero only on

its first off-diagonals, bln a , lb are the binomial coefficients previously defined
in Equation (128), while c is a state vector containing coefficients cn of the
N-photon input state (120).
The minimal average cost (154) for the lossy interferometer then equals
hC i min 2 max , where max is the maximal eigenvalue of the matrix A
and the corresponding eigenvector cmax provides the optimal input state
coefficients. hC i min quantifies the maximal achievable precision and in
the N ! 1 limit may be interpreted as the average MSE (60) due to

the convergence of the cost function (67) to the squared distance as ! .
The procedure described above allows only to obtain numerical values of
the achievable precision, and ceases to be feasible for N ! 1. The main
result of Koody nski and Demkowicz-Dobrza nski (2010) was to construct
a valid analytical lower bound on the minimal average cost (154):
h i
hC imin 2 1 Amax cos , (156)
N +2
where Amax max 1nN fAn, n1 g is the largest element of the matrix A,
contained within its off-diagonal entries (155). The bound yields exactly the
same formula as the QFI bound (140), proving that in this case the Bayesian
and QFI approaches are equivalent:
s s!
p p 1 1 a 1 b 1
hC i hC imin + p , (157)
2 a b N
where represents the fact that Bayesian cost approximated the variance
only in the limit of large N. The fact that both approaches lead to the same
ultimate bounds on precision suggests that the optimal input states may be
approximated for N ! 1 up to an arbitrary good precision with states
manifesting only local finite-number of particle correlations and may in
particular be efficiently simulated with the concept of matrix-product
states ( Jarzyna & Demkowicz-Dobrza nski, 2013, 2015).

Similarly to the case of losses discussed in the previous section, we study the
estimation precision achieved within the Bayesian approach but in the pres-
ence of phase diffusion. The analysis follows exactly in the same way, so that
likewise assuming no prior knowledge and the natural cost function introduced
in Equation (93) the formula for the average cost reads
n
o
hC i Tr C 2 cT Ac, (158)
where this time one may think of the effective state, as of the input state in
which is firstly averaged over the Gaussian distribution dictated by the evo-
lution (124) and then over the cost function in accordance with
Equation (93). The optimal seed element of the covariant POVM is identical
as in the decoherence-free case jeN i heN j and the matrix A possesses
again only nonzero entries on its first off-diagonals, but this time all of them

are equal to e 2 . As a result, the minimal average cost (158) may be evaluated
analytically following exactly the calculation of Berry and Wiseman (2000)

for the noiseless scenario, which leads then to max 2e 2 cos and
N +2
hence
N !1
2
hC imin 2 1 e 2 cos 2 1e 2 + e 2 2 : (159)
N +2 N
The optimal input states are the same as in the decoherence-free case, i.e.,
they are the N-photon sine states of Equation (105). Note that in contrast to
the photonic loss which is an example of an uncorrelated noise, the minimal
average cost (159) does not asymptotically coincide with the QFI-based
precision limit (153) unless 1.
6.4 Practical Schemes Saturating the Bounds

Deriving the fundamental bounds on quantum-enhanced precision in pres-
ence of decoherence is interesting in itself from a theoretical a point of view.
Still, a practical question remains whether the bounds derived are saturable
in practice. Note that NOON states and the sine states that are optimal in
case of QFI and Bayesian approaches in the decoherence-free case are noto-
riously hard to prepare apart from regime of very small N. For large photon
numbers, the only practically accessible states of light are squeezed Gaussian
states and one of the most popular strategies in performing quantum-
enhanced interferometry amounts to mixing a coherent beam with a
squeezed vacuum state on the input beam splitter of the MZ interferometer,
see Section 3.4. We demonstrate below that in presence of uncorrelated
decoherence, such as loss or imperfect visibility, this strategy is indeed
optimal in the asymptotic regime of large N and allows to saturate
nski et al.
the fundamental bounds derived above. We will not discuss the phase-
diffusion noise, since the estimation uncertainty is finite in the asympto-
tic limit, and the issue of saturating the asymptotic bound becomes
trivial as practically all states lead to the same asymptotic precision value,
while saturating the bound for finite N requires the use of experimentally
inaccessible sine states.
6.4.1 Bounds for Indefinite Photon-Number States

Derivation of the bounds presented in this section both in the QFI
and Bayesian approaches assumed definite-photon-number states at the
input. We have already discussed the issue of translating the bounds
from a definite-photon-number input state case to a general indefinite-
photon-number state case in Section 5.3 in the case of decoherence-free
metrology, where we have observed that due to quadratic dependence of
QFI on number of photons used, maximization of QFI over states with fixed
averaged photon number hN i is ill defined and arbitrary high QFI are in prin-
ciple achievable. Controversies related to this observation, discussed in
Section 5.3.2, are fortunately not present in the noisy metrology scenario.
For the decoherence models, analyzed in this paper, the QFI scales at
most linearly with N. Following the reasoning presented P in Section 5.3,
consider a mixture of different photon-number states N pN N . Since in
the presence of decoherence FQ(N) cN, where c is a constant coefficient
that depends on the type and strength of the noise considered, thanks to the
convexity of the QFI we can write:
!
X X X
FQ pN N pN FQ N pN cN chN i: (160)
N N
Hence the bounds on precision derived in Sections 6.2.2 (losses) and 6.2.1
(imperfect visibility) are valid also under replacement of N by hN i. Still, one
may come across claims of precisions going beyond the above mentioned
bounds typically by a factor of two (Aspachs, Calsamiglia, Munoz
Tapia, & Bagan, 2009; Joo et al., 2011). This is only possible, however, if
classical reference beam required to perform, e.g., the homodyne detection
is not treated as a resource. As discussed in detail in Section 5.3.1, we take the
position that such reference beams should be treated in the same way as the
light traveling through the interferometer and as such also counted as a
resource.
6.4.2 Coherent + Squeezed Vacuum Strategy

In section Section 3.1, we have derived an error propagation formula for the
phase-estimation uncertainty (33) for the standard MZ interferometry in
absence of decoherence. For this purpose, we have adopted the Heisenberg
picture and expressed the precision in terms of expectation values, variances
and covariances of the respective angular momentum observables calculated
for the input state. Here, we follow the same procedure but take additionally
into account the effect of imperfect visibility (local dephasing) and loss. For
simplicity, in the case of loss we restrict ourselves to equal losses in both arms.
The Heiseberg picture transformation of an observable O^ corresponding to
a general map (123) reads
X
^ i U O,
U{ Ki{ OK ^
(161)
i
where * is called the conjugated map.

For a more direct comparison with the decoherence-free formulas of
Section 3.1, we explicitly include the action of the MZ input and output bal-
anced beam splitters in the description of the state transformationin terms of
Figure 8 this corresponds to moving jiin to the left and to the right of the
figure. In case of loss, the decoherence map has the same form as given in

Section 6.1.2, but with U ei 2 y , while in the case of imperfect visibility
q q
the Kraus operators (134) will be modified to K1 1 +2 1, K2 1 2 y , so
that the local dephasing is defined with respect to the y rather than the z axis.
For the two decoherence models, the resulting Heisenberg picture transfor-
mation of the Jz observable yields (Ma et al., 2011):
h J^z i
cosh J^z iin sinh J^x iin ,

2 Jz hN i (162)
f + cos 2 2 Jz jin + sin 2 2 Jx jin +
2 4
2 sincos cov Jx , Jz jin :
where f() (1 )/ for the loss model and f() (1 2)/2 in the case of
local dephasing model. The above expressions have a clear intuitive inter-
pretation. The signal h J^z i is rescaled by a factor compared with the
decoherence-free case, while the variance apart from the analogous rescaling
is enlarged by an additional noise contribution f()hN i/4 due to lost or
dephased photons.
nski et al.
In order to calculate the precision achievable with coherent + squeezed-

vacuum strategy, we may use the already obtained quantities presented in
Equation (37). After substituting the input variances and averages into
Equation (162) and optimally setting Re() as before, we arrive
at a modified version of the formula (38) for the phase estimation precision:
s

1 jj2 + sinh 2 r
cot 2 jj + sinh 2r + jj2 e2r + sinh 2 r + f
2 2
2 sin 2
jijr i 2 :
jj sinh 2 r
(163)
The optimal operation points are again /2, 3/2. Considering the
asymptotic limit hN i jj2 + sinh 2 r ! 1 and assuming the coherent beam
to carry the dominant part of the energy jj
sinh 2 r, the formula for
precision at the optimal operation point reads:
p p
jijr i hN ie2r + f hN i e2r + f
p : (164)
hN i hN i
Clearly, even for relatively small squeezing strength r the e2r term becomes
negligible, and hence we can effectively approach arbitrary close precision
given by:
p
jijr i f
p , (165)
hN i
which recalling the definition of f() for the two decoherence models con-
sidered coincides exactly with the fundamental bounds (136), (141) derived
before. This proves that the fundamental bounds can be asymptotically
saturated with a practical interferometric scheme. One should note that
this contrasts the noiseless case and the suboptimal performance of simple
estimation scheme based on the photon-number difference measurements,
see Equation (39).
To summarize the results obtained in this section, in Figure 12, we pre-
sent a plot of the maximal achievable precision for the lossy interferometer in
q
the equal-losses scenario with 0.9, i.e., 1= F Q with optimal
(127) as a function of N compared with the NOON state-based strategy
as well as the asymptotic bound (140). On the one hand, the NOON states
remain optimal for relatively small N( 10), for which the effects of losses
Figure 12 The phase estimation precision of an interferometer with equal losses in both
arms ( 0.9). The performance of the optimal N-photon input states (120) is shown
(solid black) that indeed saturate the asymptotic quantum limit (140) (dotted):
p
1 =N. The NOON states (solid grey) achieve nearly optimal precision only
for low N ( 10) and rapidly diverge becoming out-performed by classical strategies.
For comparison, the precision attained for an indefinite photon-number scheme is pres-
ented, i.e., a coherent state and squeezed vacuum optimally mixed on a beam-splitter
(Caves, 1981) (dashed), which in the presence of loss also saturates the asymptotic quan-
tum limit (140).
may be disregarded. This fact supports the choice of NOON-like states in

the quantum-enhanced experiments with small number of particles
(Krischek et al., 2011; Mitchell, Lundeen, & Steinberg, 2004; Nagata
et al., 2007; Okamoto et al., 2008; Resch et al., 2007; Xiang, Higgins,
Berry, Wiseman, & Pryde, 2010). However, one should note that in the
presence of even infinitesimal losses, the precision achieved by the NOON
states quickly diverges with N, because their corresponding output state
QFI, FNOON
Q NN2, decays exponentially for any < 1.
Most importantly, it should be stressed that the coherent + squeezed
vacuum strategy discussed above has been implemented in recent
gravitational-wave interferometry experiments (LIGO Collaboration, 2011,
2013). The main factor limiting the quantum enhancement of precision in
this experiments is loss, which taking into account detection efficiency, optical
instruments imperfections and imperfect coupling was estimated at the level of
38% (LIGO Collaboration, 2011). In Demkowicz-Dobrza nski et al. (2013), it
has been demonstrated that the sensing precision achieved in LIGO
nski et al.
Collaboration (2011) using the 10dB squeezed vacuum (corresponding to the

squeezing factor e2r 0.1) was strikingly close to the fundamental bound and
only 8% further reduction in estimation uncertainty would be possible if more
advanced input states of light were used.
7. CONCLUSIONS
In this review, we have showed how the tools of quantum estimation
theory can be applied in order to derive fundamental bounds on achievable pre-
cision in quantum-enhanced optical interferometric experiments. The main
message to be conveyed is the fact that while the power of quantum enhance-
ment is seriously reduced by the presence of decoherence, and, in general, the
Heisenberg scaling cannot be reached, nonclassical states of light offer a notice-
able improvement in interferometric precision and simple experimental
schemes may approach arbitrary close the fundamental quantum bounds. It
is also worth noting that in the presence of uncorrelated decoherence the
Bayesian approaches coincide asymptotically with the QFI approaches easing
the tension between this two often competing ways of statistical analysis.
We would also like to mention an inspiring alternative approach to the
derivation of limits on precision of phase estimation, where the results are
derived making use of information theoretic concepts such as rate-distortion
theory (Nair, 2012) or entropic uncertainty relations (Hall & Wiseman,
2012). Even though the bounds derived in this way are weaker than the bou-
nds presented in this review and obtained via Bayesian or QFI approaches,
they carry a conceptual appeal encouraging to look for deeper connections
between quantum estimation and communication theories.
Let us also point out, that while we have focused our discussion on opti-
cal interferometry using the paradigmatic MZ model, the same methods can
be applied to address the problems of fundamental precision bounds in
atomic interferometry (Cronin et al., 2009), magnetometry (Budker &
Romalis, 2007), frequency stabilization in atomic clocks (Diddams et al.,
2004) as well as the limits on resolution of quantum-enhanced lithographic
protocols (Boto et al., 2000). All these setups can be cast into a common
mathematical framework, see Section 3.6, but the resulting bounds will
depend strongly on the nature of dominant decoherence effects and the rel-
evant resource limitations such as: total experimental time, light power,
number of atoms, etc., as well as on the chosen figure of merit. In particular,
it is not excluded that in some atomic metrological scenarios one may still
p
obtain a better than 1= N of precision if decoherence is of a special form
allowing for use of the decoherence-free subspaces (Dorner, 2012; Jeske,
Cole, & Huelga, 2013) or when its impact may be significantly reduced
by considering short evolution times, in which the SQL-like scaling bounds
may in principle be circumvented by: adjusting decoherence geometry
(Arrad, Vinkler, Aharonov, & Retzker, 2014; Chaves, Brask,
Markiewicz, Koody nski, & Acin, 2013; D ur, Skotiniotis, Fr
owis, &
Kraus, 2014; Kessler, Lovchinsky, Sushkov, & Lukin, 2013) or by consid-
ering non-Markovian short-time behavior (Chin, Huelga, & Plenio, 2012;
Matsuzaki, Benjamin, & Fitzsimons, 2011).
We should also note that application of the tools presented in this review
to a proper analysis of fundamental limits to the operation of quantum-
enhanced atomic clocks (Andre`, Srensen, & Lukin, 2004; Leibfried
et al., 2004) is not that direct as it requires taking into account precise fre-
quency noise characteristic of the local oscillator, allowing to determine the
optimal stationary operation regime of the clock (Macieszczak,
Demkowicz-Dobrza nski, & Fraas, 2013) ideally in terms of the Allan vari-
ance taken as a figure of merit (Fraas, 2013). A deeper theoretical insight into
this problem is still required to yield computable fundamental bounds.
The applicability of the tools presented has also been restricted to single
phase parameter estimation. A more general approach may be taken, were
multiple-phases (Humphreys, Barbieri, Datta, & Walmsley, 2013) or the
phase as well as the decoherence strength itself are the quantities to be esti-
mated (Crowley, Datta, Barbieri, & Walmsley, 2014; Knysh & Durkin,
2013). This poses an additional theoretical challenge as then the multi-
parameter quantum estimation theory needs to be applied, while most of
the tools discussed in this review are applicable only to single-parameter
estimation. Developing nontrivial multiparameter fundamental bounds for
quantum metrology is therefore still an open field for research.
ACKNOWLEDGMENTS
This research work supported by the FP7 IP project SIQS co-financed by the Polish Ministry
of Science and Higher Education, Polish NCBiR under the ERA-NET CHIST-ERA
project QUASAR, and Foundation for Polish Science TEAM project.
REFERENCES
Adesso, G. (2006). Entanglelement of Gaussian states.. Universita degli Studi di Salerno,
Unpublished doctoral dissertation.
Afek, I., Ambar, O., & Silberberg, Y. (2010). High-NOON states by mixing quantum and
classical light. Science, 328(5980), 879881. http://dx.doi.org/10.1126/science.1188172.
nski et al.
Alipour, S. (2014). Dynamical and kinematic bounds for quantum metrology in open systems
(ArXiv e-prints arXiv:1403.8033).
Alipour, S., Mehboudi, M., & Rezakhani, A. T. (2014). Quantum metrology in open sys-
tems: Dissipative Cramer-Rao bound. Physical Review Letters, 112, 120405. http://dx.
doi.org/10.1103/PhysRevLett.112.120405.
Anderson, B. P., & Kasevich, M. A. (1998). Macroscopic quantum interference from atomic
tunnel arrays. Science, 282(5394), 16861689. http://dx.doi.org/10.1126/
science.282.5394.1686.
Andre`, A., Srensen, A. S., & Lukin, M. D. (2004). Stability of atomic clocks based on
entangled atoms. Physical Review Letters, 92, 230801.
Anisimov, P. M., Raterman, G. M., Chiruvelli, A., Plick, W. N., Huver, S. D., Lee, H., &
Dowling, J. P. (2010). Quantum metrology with two-mode squeezed vacuum: Parity
detection beats the Heisenberg limit. Physical Review Letters, 104, 103602. http://dx.
Arrad, G., Vinkler, Y., Aharonov, D., & Retzker, A. (2014). Increasing sensing resolution
with error correction. Physical Review Letters, 112, 150801. http://dx.doi.org/10.1103/
PhysRevLett.112.150801.
Aspachs, M., Calsamiglia, J., Munoz Tapia, R., & Bagan, E. (2009). Phase estimation for
thermal Gaussian states. Physical Review A, 79, 033834. http://dx.doi.org/10.1103/
PhysRevA.79.033834.
Bachor, H.-A., & Ralph, T. (2004). A guide to experiments in quantum optics. Weinheim:
Wiley-VCH.
Banaszek, K., Horodecki, P., Karpinski, M., & Radzewicz, C. (2013). Quantum mechanical
which-way experiment with an internal degree of freedom. Nature Communications, 4, 2594.
Banaszek, K., Radzewicz, C., Wodkiewicz, K., & Krasinski, J. S. (1999). Direct measure-
ment of the Wigner function by photon counting. Physical Review A, 60, 674677.
http://dx.doi.org/10.1103/PhysRevA.60.674.
Banaszek, K., & Wodkiewicz, K. (1998). Nonlocality of the Einstein-Podolsky-Rosen state
in the Wigner representation. Physical Review A, 58, 43454347. http://dx.doi.org/
10.1103/PhysRevA.58.4345.
Barnett, S. M., & Pegg, D. T. (1992). Limiting procedures for the optical phase operator.
Journal of Modern Optics, 39(10), 21212129. http://dx.doi.org/10.1080/
09500349214552141.
Bartlett, S. D., Rudolph, T., & Spekkens, R. W. (2007). Reference frames, superselection
rules, and quantum information. Reviews of Modern Physics, 79(2), 555. http://dx.doi.org/
10.1103/RevModPhys.79.555.
Benatti, F., & Braun, D. (2013). Sub-shot-noise sensitivities without entanglement. Physical
Review A, 87, 012340. http://dx.doi.org/10.1103/PhysRevA.87.012340.
Benatti, F., Floreanini, R., & Marzolino, U. (2010). Sub-shot-noise quantum metrology
with entangled identical particles. Annals of Physics, 325(4), 924935. http://dx.doi.
org/10.1016/j.aop.2010.01.005.
Bengtsson, I., & Zyczkowski, K. (2006). Geometry of quantum states: An introduction to quantum
entanglement. Cambridge: Cambridge Univeristy Press.
Berry, D. W., Hall, M. J. W., Zwierz, M., & Wiseman, H. M. (2012). Optimal Heisenberg-
style bounds for the average performance of arbitrary phase estimates. Physical Review A,
86, 053813. http://dx.doi.org/10.1103/PhysRevA.86.053813.
Berry, D. W., Higgins, B. L., Bartlett, S. D., Mitchell, M. W., Pryde, G. J., & Wiseman, H. M.
(2009). How to perform the most accurate possible phase measurements. Physical Review A,
80(5), 052114. http://dx.doi.org/10.1103/PhysRevA.80.052114.
Berry, D. W., & Wiseman, H. M. (2000). Optimal states and almost optimal adaptive mea-
surements for quantum interferometry. Physical Review Letters, 85(24), 50985101.
http://dx.doi.org/10.1103/PhysRevLett.85.5098.
Bollinger, J. J., Itano, W. M., Wineland, D. J., & Heinzen, D. J. (1996). Optimal frequency
measurements with maximally correlated states. Physical Review A, 54(6),
R4649R4652. http://dx.doi.org/10.1103/PhysRevA.54.R4649.
Bondurant, R. S., & Shapiro, J. H. (1984). Squeezed states in phase-sensing inter-
ferometers. Physical Review D, 30, 25482556. http://dx.doi.org/10.1103/
PhysRevD.30.2548.
Boto, A. N., Kok, P., Abrams, D. S., Braunstein, S. L., Williams, C. P., & Dowling, J. P.
(2000). Quantum interferometric optical lithography: Exploiting entanglement to beat
the diffraction limit. Physical Review Letters, 85, 27332736. http://dx.doi.org/10.1103/
Braunstein, S. L. (1992). Quantum limits on precision measurements of phase. Physical Review
Letters, 69, 35983601. http://dx.doi.org/10.1103/PhysRevLett.69.3598.
Braunstein, S. L., & Caves, C. M. (1994). Statistical distance and the geometry of quantum
states. Physical Review Letters, 72(22), 34393443. http://dx.doi.org/10.1103/
Budker, D., & Romalis, M. (2007). Optical magnetometry. Nature Physics, 3, 227234.
http://dx.doi.org/10.1038/nphys566.
Buzek, V., Derka, R., & Massar, S. (1999). Optimal quantum clocks. Physical Review Letters,
82, 22072210. http://dx.doi.org/10.1103/PhysRevLett.82.2207.
Buzek, V., & Knight, P. L. (1995). I: Quantum interference, superposition states of light, and
nonclassical effects. In E. Wolf (Ed.), Annals of Physics: Vol. 34 (pp. 1158). Amsterdam:
Elsevier, http://dx.doi.org/10.1016/S0079-6638(08)70324-X.
Caves, C. M. (1981). Quantum-mechanical noise in an interferometer. Physical Review D,
23(8), 16931708. http://dx.doi.org/10.1103/PhysRevD.23.1693.
Chaves, R., Brask, J. B., Markiewicz, M., Koody nski, J., & Acin, A. (2013). Noisy metrol-
ogy beyond the standard quantum limit. Physical Review Letters, 111, 120401. http://dx.
Chekhova, M. (2011). Chapter 4Polarization and spectral properties of biphotons.
In E. Wolf (Ed.), Progress in optics: Vol. 56 (pp. 187226). Amsterdam: Elsevier.
http://dx.doi.org/10.1016/B978-0-444-53886-4.00004-6.
Chin, A. W., Huelga, S. F., & Plenio, M. B. (2012). Quantum metrology in non-Markovian
environments. Physical Review Letters, 109(23), 233601. http://dx.doi.org/10.1103/
Chiribella, G., DAriano, G. M., & Sacchi, M. F. (2005). Optimal estimation of group trans-
formations using entanglement. Physical Review A, 72, 042338. http://dx.doi.org/
10.1103/PhysRevA.72.042338.
Choi, M.-D. (1975). Completely positive linear maps on complex matrices. Linear Algebra and
its Applications, 10, 285290. http://dx.doi.org/10.1016/0024-3795(75)90075-0.
Chwede nczuk, J., Hyllus, P., Piazza, F., & Smerzi, A. (2012). Sub-shot-noise interferometry
from measurements of the one-body density. New Journal of Physics, 14(9), 093001.
http://dx.doi.org/10.1088/1367-2630/14/9/093001.
Chwede nczuk, J., Piazza, F., & Smerzi, A. (2011). Phase estimation from atom position mea-
surements. New Journal of Physics, 13(6), 065023. http://dx.doi.org/10.1088/1367-
2630/13/6/065023.
Cronin, A. D., Schmiedmayer, J., & Pritchard, D. E. (2009). Optics and interferometry with
atoms and molecules. Reviews of Modern Physics, 81, 10511129. http://dx.doi.org/
10.1103/RevModPhys.81.1051.
Crowley, P. J. D., Datta, A., Barbieri, M., & Walmsley, I. A. (2014). Tradeoff in simulta-
neous quantum-limited phase and loss estimation in interferometry. Physical Review A,
DAriano, G. M., Macchiavello, C., & Paris, M. G. (1995). Optimized phase detection. Phys-
ics Letters A, 198(4), 286294. http://dx.doi.org/10.1016/0375-9601(95)00009-R.
nski et al.
Datta, A., Zhang, L., Thomas-Peter, N., Dorner, U., Smith, B. J., & Walmsley, I. A. (2011).
Quantum metrology with imperfect states and detectors. Physical Review A, 83, 063836.
Demkowicz-Dobrza nski, R. (2010). Multi-pass classical vs. quantum strategies in lossy phase
estimation. Laser Physics, 20, 11971202. http://dx.doi.org/10.1134/S1054660X10090306.
Demkowicz-Dobrza nski, R., Banaszek, K., & Schnabel, R. (2013). Fundamental quantum
interferometry bound for the squeezed-light-enhanced gravitational wave detector GEO
600. Physical Review A 88. http://dx.doi.org/10.1103/PhysRevA.88.041802, 041802(R).
Demkowicz-Dobrza nski, R., Dorner, U., Smith, B. J., Lundeen, J. S., Wasilewski, W.,
Banaszek, K., et al. (2009). Quantum phase estimation with lossy interferometers. Phys-
ical Review A, 80, 013825. http://dx.doi.org/10.1103/PhysRevA.80.013825.
Demkowicz-Dobrza nski, R., Koody nski, J., & Guta, M. (2012). The elusive Heisenberg
limit in quantum enhanced metrology. Nature Communications, 3, 1063. http://dx.doi.
org/10.1038/ncomms2067.
Demkowicz-Dobrza nski, R., & Maccone, L. (2014). Using Entanglement Against Noise in
Quantum Metrology. Physical Review Letters, 113, 250801.
Diddams, S. A., Bergquist, J. C., Jefferts, S. R., & Oates, C. W. (2004). Standards of time and
frequency at the outset of the 21st century. Science, 306(5700), 13181324. http://dx.doi.
org/10.1126/science.1102330.
Dorner, U. (2012). Quantum frequency estimation with trapped ions and atoms. New Journal
of Physics, 14(4), 043011. http://dx.doi.org/10.1088/1367-2630/14/4/043011.
Dorner, U., Demkowicz-Dobrza nski, R., Smith, B. J., Lundeen, J. S., Wasilewski, W.,
Banaszek, K., et al. (2009). Optimal quantum phase estimation. Physical Review Letters,
102, 040403. http://dx.doi.org/10.1103/PhysRevLett.102.040403.
Dowling, J. P. (1998). Correlated input-port, matter-wave interferometer: Quantum-noise
limits to the atom-laser gyroscope. Physical Review A, 57, 4736. http://dx.doi.org/
10.1103/PhysRevA.57.4736.
Dur, W., Skotiniotis, M., Fr owis, F., & Kraus, B. (2014). Improved quantum metrology
using quantum error correction. Physical Review Letters, 112, 080801. http://dx.doi.
org/10.1103/PhysRevLett.112.080801.
Escher, B. M., Davidovich, L., Zagury, N., & de Matos Filho, R. L. (2012). Quantum met-
rological limits via a variational approach. Physical Review Letters, 109, 190404. http://dx.
Escher, B. M., de Matos Filho, R. L., & Davidovich, L. (2011). General framework for
estimating the ultimate precision limit in noisy quantum-enhanced metrology. Nature
Physics, 7, 406411. http://dx.doi.org/10.1038/nphys1958.
Fox, M. (2006). Quantum optics. Oxford: Oxford University Press.
Fraas, M. (2013). An analysis of the stationary operation of atomic clocks (ArXiv e-prints
arXiv:1303.6083).
Fr
owis, F., Skotiniotis, M., Kraus, B., and D ur, W. (2014). Optimal quantum states for
frequency estimation (ArXiv e-prints arXiv:1402.6946).
Fujiwara, A., & Imai, H. (2008). A fibre bundle over manifolds of quantum channels and its
application to quantum statistics. Journal of Physics A: Mathematical and Theoretical, 41,
255304. http://dx.doi.org/10.1088/1751-8113/41/25/255304.
Genoni, M. G., Olivares, S., & Paris, M. G. A. (2011). Optical phase estimation in the pres-
ence of phase diffusion. Physical Review Letters, 106, 153603. http://dx.doi.org/10.1103/
Gilbert, G., Hamrick, M., & Weinstein, Y. S. (2008). Use of maximally entangled N-photon
states for practical quantum interferometry. Journal of the Optical Society of America B:
Optical Physics, 25, 1336. http://dx.doi.org/10.1364/JOSAB.25.001336.
Giovannetti, V., Lloyd, S., & Maccone, L. (2004). Quantum-enhanced measurements:
Beating the standard quantum limit. Science, 306, 1330. http://dx.doi.org/10.1126/
science.1104149.
Giovannetti, V., Lloyd, S., & Maccone, L. (2006). Quantum metrology. Physical Review
Giovannetti, V., Lloyd, S., & Maccone, L. (2011). Advances in quantum metrology. Nature
Photonics, 5, 222229. http://dx.doi.org/10.1038/nphoton.2011.35.
Giovannetti, V., & Maccone, L. (2012). Sub-Heisenberg estimation strategies are ineffective.
Physical Review Letters, 108, 210404. http://dx.doi.org/10.1103/PhysRevLett.108.210404.
Glauber, R. J. (1963). Coherent and incoherent states of the radiation field. Physics Review,
131, 2766. http://dx.doi.org/10.1103/PhysRev.131.2766.
Gross, C., Zibold, T., Nicklas, E., Esteve, J., & Oberthaler, M. K. (2010). Nonlinear atom
interferometer surpasses classical precision limit. Nature, 464, 11651169. http://dx.doi.
org/10.1038/nature08919.
Hall, M. J. W., & Wiseman, H. M. (2012). Heisenberg-style bounds for arbitrary estimates of
shift parameters including prior information. New Journal of Physics, 14, 033040. http://
dx.doi.org/10.1088/1367-2630/14/3/033040.
Hariharan, P. (2003). Optical interferometry. Amsterdam: Elsevier.
Helstrom, C. W. (1976). Quantum detection and estimation theory. London: Academic press.
Higgins, B. L., Berry, D. W., Bartlett, S. D., Wiseman, H. M., & Pryde, G. J. (2007).
Entanglement-free Heisenberg-limited phase estimation. Nature, 450, 393396.
http://dx.doi.org/10.1038/nature06257.
Hofheinz, M., Weig, E. M., Ansmann, M., Bialczak, R. C., Lucero, E., Neeley, M., et al.
(2008). Generation of Fock states in a superconducting quantum circuit. Nature, 454,
310314. http://dx.doi.org/10.1038/nature07136.
Hofmann, H. F. (2009). All path-symmetric pure states achieve their maximal phase sensi-
tivity in conventional two-path interferometry. Physical Review A, 79(3), 033822. http://
dx.doi.org/10.1103/PhysRevA.79.033822.
Holevo, A. S. (1982). Probabilistic and statistical aspects of quantum theory. Amsterdam: North
Holland.
Holland, M. J., & Burnett, K. (1993). Interferometric detection of optical phase shifts at the
Heisenberg limit. Physical Review Letters, 71(9), 13551358. http://dx.doi.org/10.1103/
Horodecki, R., Horodecki, P., Horodecki, M., & Horodecki, K. (2009). Quantum entanglement.
Reviews of Modern Physics, 81, 865942. http://dx.doi.org/10.1103/RevModPhys.81.865.
Horrom, T., Singh, R., Dowling, J. P., & Mikhailov, E. E. (2012). Quantum-enhanced
magnetometer with low-frequency squeezing. Physical Review A, 86, 023803. http://
dx.doi.org/10.1103/PhysRevA.86.023803.
Hradil, Z., Myska, R., Perina, J., Zawisky, M., Hasegawa, Y., & Rauch, H. (1996). Quan-
tum phase in interferometry. Physical Review Letters, 76, 42954298. http://dx.doi.org/
10.1103/PhysRevLett.76.4295.
Huelga, S. F., Macchiavello, C., Pellizzari, T., Ekert, A. K., Plenio, M. B., & Cirac, J. I.
(1997). Improvement of frequency standards with quantum entanglement. Physical
Review Letters, 79(20), 38653868. http://dx.doi.org/10.1103/PhysRevLett.79.3865.
Humphreys, P. C., Barbieri, M., Datta, A., & Walmsley, I. A. (2013). Quantum enhanced
multiple phase estimation. Physical Review Letters, 111, 070403. http://dx.doi.org/
10.1103/PhysRevLett.111.070403.
Huver, S. D., Wildfeuer, C. F., & Dowling, J. P. (2008). Entangled Fock states for robust
quantum optical metrology, imaging, and sensing. Physical Review A, 78, 063828.
Hyllus, P., Laskowski, W., Krischek, R., Schwemmer, C., Wieczorek, W., Weinfurter, H.,
et al. (2012). Fisher information and multiparticle entanglement. Physical Review A, 85,
022321. http://dx.doi.org/10.1103/PhysRevA.85.022321.
Jamiokowski, A. (1972). Linear transformations which preserve trace and positive semi-
definiteness of operators. Reports on Mathematical Physics, 3, 275278. http://dx.doi.
org/10.1016/0034-4877(72)90011-0.
nski et al.
Jarzyna, M., & Demkowicz-Dobrza nski, R. (2012). Quantum interferometry with and
without an external phase reference. Physical Review A, 85. http://dx.doi.org/
10.1103/PhysRevA.85.011801, 011801(R).
Jarzyna, M., & Demkowicz-Dobrza nski, R. (2013). Matrix product states for quantum
metrology. Physical Review Letters, 110, 240405. http://dx.doi.org/10.1103/
Jarzyna, M., & Demkowicz-Dobrzanski, R. (2015). True precision limits in quantum
metrology. New Journal of Physics, 17, 013010.
Jeske, J., Cole, J. H., & Huelga, S. (2013). Quantum metrology in the presence of spatially
correlated noise: Restoring Heisenberg scaling (ArXiv e-prints arXiv:1307.6301).
Joo, J., Munro, W. J., & Spiller, T. P. (2011). Quantum metrology with entangled coherent
states. Physical Review Letters, 107, 083601. http://dx.doi.org/10.1103/
Kacprowicz, M., Demkowicz-Dobrza nski, R., Wasilewski, W., & Banaszek, K. (2010).
Experimental quantum enhanced phase-estimation in the presence of loss. Nature
Kaftal, T., & Demkowicz-Dobrza nski, R. (2014). Usefulness of an enhanced Kitaev phase-
estimation algorithm in quantum metrology and computation. Physical Review A, 90,
062313.
Kay, S. M. (1993). Fundamentals of statistical signal processing: Estimation theory (S. E. Alan V.
Oppenheim, Ed.). Prentice Hall: Upper Saddle River.
Kessler, E. M., Lovchinsky, I., Sushkov, A. O., & Lukin, M. D. (2013). Quantum Error
Correction for Metrology. Physical Review Letters, 112, 150802.
Killoran, N., Cramer, M., & Plenio, M. B. (2014). Extracting entanglement from identical
particles. Physical Review Letters, 112, 150501. http://dx.doi.org/10.1103/
Kitagawa, M., & Ueda, M. (1993). Squeezed spin states. Physical Review A, 47, 51385143.
Knysh, S., Smelyanskiy, V. N., & Durkin, G. A. (2011). Scaling laws for precision in quan-
tum interferometry and the bifurcation landscape of the optimal state. Physical Review A,
Knysh, S. I., Chen, E. H., & Durkin, G. A. (2014). True limits to precision via unique quan-
tum probe (ArXiv e-prints arXiv:1402.0495).
Knysh, S. I., & Durkin, G. A. (2013). Estimation of phase and diffusion: Combining quantum
statistics and classical noise (ArXiv e-prints arXiv:1307.0470).
Kok, P., Munro, W. J., Nemoto, K., Ralph, T. C., Dowling, J. P., & Milburn, G. J. (2007).
Linear optical quantum computing with photonic qubits. Reviews of Modern Physics, 79,
135174. http://dx.doi.org/10.1103/RevModPhys.79.135.
Koody nski, J., & Demkowicz-Dobrza nski, R. (2010). Phase estimation without a priori
phase knowledge in the presence of loss. Physical Review A, 82, 053804. http://dx.
doi.org/10.1103/PhysRevA.82.053804.
Koody nski, J., & Demkowicz-Dobrza nski, R. (2013). Efficient tools for quantum metrology
with uncorrelated noise. New Journal of Physics, 15, 073043. http://dx.doi.org/
10.1088/1367-2630/15/7/073043.
Krischek, R., Schwemmer, C., Wieczorek, W., Weinfurter, H., Hyllus, P., Pezze, L., et al.
(2011). Useful multiparticle entanglement and sub-shot-noise sensitivity in experimental
phase estimation. Physical Review Letters, 107, 080504. http://dx.doi.org/10.1103/
Lee, H., Kok, P., & Dowling, J. P. (2002). A quantum Rosetta stone for interferometry.
Journal of Modern Optics, 49(1415), 23252338. http://dx.doi.org/10.1080/
0950034021000011536.
Lehmann, E. L., & Casella, G. (1998). Theory of point estimation (Vol. 31). New York:
Springer.
Leibfried, D., Barrett, M. D., Schaetz, T., Britton, J., Chiaverini, J., Itano, W. M., et al.
(2004). Toward Heisenberg-limited spectroscopy with multiparticle entangled states.
Science, 304, 14761478.
Leonhardt, U. (1997). Measuring the quantum state of light (Vol. 22). Cambridge:
Cambridge Univeristy Press.
LIGO Collaboration. (2011). A gravitational wave observatory operating beyond the quantum
shot-noise limit. Nature Physics, 7, 962965. http://dx.doi.org/10.1038/nphys2083.
LIGO Collaboration. (2013). Enhanced sensitivity of the LIGO gravitational wave detector
by using squeezed states of light. Nature Photonics, 7(8), 613619. http://dx.doi.org/
10.1038/nphoton.2013.177.
Loudon, R. (2000). The quantum theory of light. Oxford: Oxford University Press.
Luis, A., & Perina, J. (1996). Optimum phase-shift estimation and the quantum description of
the phase difference. Physical Review A, 54, 45644570. http://dx.doi.org/10.1103/
PhysRevA.54.4564.
Lynch, R. (1995). The quantum phase problem: A critical review. Physics Reports, 256(6),
367436. http://dx.doi.org/10.1016/0370-1573(94)00095-K.
Ma, J., Wang, X., Sun, C., & Nori, F. (2011). Quantum spin squeezing. Physics Reports,
509(2), 89165. http://dx.doi.org/10.1016/j.physrep.2011.08.003.
Macieszczak, K. (2014). Upper bounds on the quantum Fisher information in the presence of
general dephasing (ArXiv e-prints arXiv:1403.0955).
Macieszczak, K., Demkowicz-Dobrza nski, R., & Fraas, M. (2013). Bayesian quantum
frequency estimation in presence of collective dephasing. New Journal of Physics, 16,
113002.
Mandel, L., & Wolf, E. (2000). Optical coherence and quantum optics. Cambridge: Cambridge
Univeristy Press.
Matsumoto, K. (2010). On metric of quantum channel spaces., 1006.0300v1 (ArXiv
e-prints).
Matsuzaki, Y., Benjamin, S. C., & Fitzsimons, J. (2011). Magnetic field sensing beyond the
standard quantum limit under the effect of decoherence. Physical Review A, 84, 012103.
Mitchell, M. W., Lundeen, J. S., & Steinberg, A. M. (2004). Super-resolving phase measure-
ments with a multi-photon entangled state. Nature, 429, 161164. http://dx.doi.org/
10.1038/nature02493.
Mlmer, K. (1997). Optical coherence: A convenient fiction. Physical Review A, 55,
Nagaoka, H. (2005). On Fisher information of quantum statistical models. In M. Hayashi
(Ed.), (pp. 113125). Singapore: World Scientific, (Chapter 9).
Nagata, T., Okamoto, R., OBrien, J. L., Sasaki, K., & Takeuchi, S. (2007). Beating the stan-
dard quantum limit with four entangled photons. Science, 316, 726729. http://dx.doi.
org/10.1126/science.1138007.
Nair, R. (2012). Fundamental limits on the accuracy of optical phase estimation from rate-
distortion theory (ArXiv e-prints arXiv:1204.3761).
Nielsen, M. A., & Chuang, I. L. (2000). Quantum computing and quantum information.
Cambridge: Cambridge University Press.
Noh, J. W., Fouge`res, A., & Mandel, L. (1992). Operational approach to the phase of a
quantum field. Physical Review A, 45, 424442. http://dx.doi.org/10.1103/
PhysRevA.45.424.
Okamoto, R., Hofmann, H. F., Nagata, T., OBrien, J. L., Sasaki, K., & Takeuchi, S. (2008).
Beating the standard quantum limit: Phase super-sensitivity of N-photon interferome-
ters. New Journal of Physics, 10(7), 073033. http://dx.doi.org/10.1088/1367-2630/10/
7/073033.
Olivares, S., & Paris, M. G. A. (2007). Optimized interferometry with Gaussian states. Optics
and Spectroscopy, 103, 231236. http://dx.doi.org/10.1134/S0030400X07080103.
nski et al.
Ono, T., & Hofmann, H. F. (2010). Effects of photon losses on phase estimation near the
Heisenberg limit using coherent light and squeezed vacuum. Physical Review A, 81,
Pan, J.-W., Chen, Z.-B., Lu, C.-Y., Weinfurter, H., Zeilinger, A., & zukowski, M. (2012).
Multiphoton entanglement and interferometry. Reviews of Modern Physics, 84, 777838.
http://dx.doi.org/10.1103/RevModPhys.84.777.
Pezze, L., & Smerzi, A. (2008). Mach-Zehnder interferometry at the Heisenberg limit with
coherent and squeezed-vacuum light. Physical Review Letters, 100(7), 073601. http://dx.
Pezze, L., & Smerzi, A. (2009). Entanglement, nonlinear dynamics, and the Heisenberg limit.
Physical Review Letters, 102, 100401. http://dx.doi.org/10.1103/PhysRevLett.
102.100401.
Pinel, O., Fade, J., Braun, D., Jian, P., Treps, N., & Fabre, C. (2012). Ultimate sensitivity of
precision measurements with intense Gaussian quantum light: A multimodal approach.
Physical Review A, 85. http://dx.doi.org/10.1103/PhysRevA.85.010101, 010101(R).
Pinel, O., Jian, P., Fabre, C., Treps, N., & Braun, D. (2013). Quantum parameter estimation
using general single-mode Gaussian states. Physical Review A, 88. http://dx.doi.org/
10.1103/PhysRevA.88.040102, 040102(R).
Pitkin, M., Reid, S., Rowan, S., & Hough, J. (2011). Gravitational wave detection by inter-
ferometry (ground and space). Living Reviews in Relativity. 14(5). http://dx.doi.org/
10.12942/lrr-2011-5.
Reid, M. D., & Drummond, P. D. (1988, Jun). Quantum correlations of phase in non-
degenerate parametric oscillation. Physical Review Letters, 60, 27312733. http://dx.
Resch, K. J., Pregnell, K. L., Prevedel, R., Gilchrist, A., Pryde, G. J., OBrien, J. L., et al.
(2007). Time-reversal and super-resolving phase measurements. Physical Review Letters,
98(22), 223601. http://dx.doi.org/10.1103/PhysRevLett.98.223601.
Rivas, A., & Luis, A. (2012). Sub-Heisenberg estimation of non-random phase shifts. New
Journal of Physics, 14(9), 093052. http://dx.doi.org/10.1088/1367-2630/14/9/093052.
Rubin, M. A., & Kaushik, S. (2007). Loss-induced limits to phase measurement precision
with maximally entangled states. Physical Review A, 75, 053805. http://dx.doi.org/
10.1103/PhysRevA.75.053805.
Sanders, B. C., & Milburn, G. J. (1995). Optimal quantum measurements for phase estima-
tion. Physical Review Letters, 75, 29442947. http://dx.doi.org/10.1103/
Sarovar, M., & Milburn, G. J. (2006). Optimal estimation of one parameter quantum chan-
nels. Journal of Physics A: Mathematical and General, 39, 8487. http://dx.doi.org/
10.1088/0305-4470/39/26/015.
Sayrin, C., Dotsenko, I., Zhou, X., Peaudecerf, B., Rybarczyk, T., Gleyzes, S., et al. (2011).
Real-time quantum feedback prepares and stabilizes photon number states. Nature, 477,
7377. http://dx.doi.org/10.1038/nature10376.
Schleich, W. P. (2001). Quantum optics in phase space. Berlin: Wiley-VCH.
Schwinger, J. (1965). Quantum theory of angular momentum. New York: Academic Press.
Seshadreesan, K. P., Anisimov, P. M., Lee, H., & Dowling, J. P. (2011). Parity detection
achieves the Heisenberg limit in interferometry with coherent mixed with squeezed
vacuum light. New Journal of Physics, 13, 083026. http://dx.doi.org/10.1088/1367-
2630/13/8/083026.
Sewell, R. J., Koschorreck, M., Napolitano, M., Dubost, B., Behbood, N., & Mitchell, M. W.
(2012). Magnetic sensitivity beyond the projection noise limit by spin squeezing. Physical
Review Letters, 109, 253605. http://dx.doi.org/10.1103/PhysRevLett.109.253605.
Shaji, A., & Caves, C. M. (2007). Qubit metrology and decoherence. Physical Review A, 76,
Shankar, R. (1994). Principles of quantum mechanics. New York: Plenum.
Shin, Y., Saba, M., Pasquini, T. A., Ketterle, W., Pritchard, D. E., & Leanhardt, A. E. (2004).
Atom interferometry with Bose-Einstein condensates in a double-well potential. Physical
Spagnolo, N., Vitelli, C., Lucivero, V. G., Giovannetti, V., Maccone, L., & Sciarrino, F.
(2012). Phase estimation via quantum interferometry for noisy detectors. Physical Review
Letters, 108(23), 233602. http://dx.doi.org/10.1103/PhysRevLett.108.233602.
Summy, G. S., & Pegg, D. T. (1990). Phase optimized quantum states of light. Optics
Communication, 77(1), 7579. http://dx.doi.org/10.1016/0030-4018(90)90464-5.
Torres, J. P., Banaszek, K., & Walmsley, I. (2011). Chapter 5Engineering nonlinear optic
sources of photonic entanglement. In E. Wolf (Ed.), Progress in optics: Vol. 56,
(pp. 227331). Amsterdam: Elsevier. http://dx.doi.org/10.1016/B978-0-444-53886-
4.00005-8.
Toth, G. (2012). Multipartite entanglement and high-precision metrology. Physical Review A,
Toth, G., & Petz, D. (2013). Extremal properties of the variance and the quantum Fisher
information. Physical Review A, 87, 032324. http://dx.doi.org/10.1103/PhysRevA.87.
032324.
Tsang, M. (2012). Ziv-Zakai error bounds for quantum parameter estimation. Physical Review
van der Vaart, A. W. (2000). Asymptotic statistics. Cambridge: Cambridge Univeristy Press.
Vitelli, C., Spagnolo, N., Toffoli, L., Sciarrino, F., & De Martini, F. (2010). Enhanced
resolution of lossy interferometry by coherent amplification of single photons.
Physical Review Letters, 105(11), 113602. http://dx.doi.org/10.1103/PhysRevLett.105.
113602.
Walls, D. F., & Milburn, G. J. (1995). Quantum optics. Berlin: Springer Verlag.
Wasilewski, W., Jensen, K., Krauter, H., Renema, J. J., Balabas, M. V., & Polzik, E. S.
(2010). Quantum noise limited and entanglement-assisted magnetometry. Physical
Wigner, E. P. (1932). On the quantum correction for thermodynamic equilibrium. Physics
Review, 40, 749759. http://dx.doi.org/10.1103/PhysRev.40.749.
Wineland, D. J., Bollinger, J. J., Itano, W. M., & Heinzen, D. J. (1994). Squeezed atomic
states and projection noise in spectroscopy. Phys. Rev. A, 50, 6788. http://dx.doi.
org/10.1103/PhysRevA.50.67.
Wineland, D. J., Bollinger, J. J., Itano, W. M., Moore, F. L., & Heinzen, D. J. (1992). Spin
squeezing and reduced quantum noise in spectroscopy. Phys. Rev. A, 46,
R6797R6800. http://dx.doi.org/10.1103/PhysRevA.46.R6797.
Xiang, G.-Y., Higgins, B. L., Berry, D. W., Wiseman, H. M., & Pryde, G. J. (2010).
Entanglement-enhanced measurement of a completely unknown optical phase. Nature
Xiao, M., Wu, L.-A., & Kimble, H. J. (1987). Precision measurement beyond the shot-noise
limit. Physical Review Letters, 59(3), 278281. http://dx.doi.org/10.1103/PhysRevLett.
59.278.
Yu, S. (2013). Quantum Fisher information as the convex roof of variance (ArXiv e-prints
arXiv:1302.5311).
Yurke, B., McCall, S. L., & Klauder, J. R. (1986). Su(2) and SU(1,1) interferometers. Physical
Review A, 33(6), 40334054. http://dx.doi.org/10.1103/PhysRevA.33.4033.
Zhang, Y. R., Jin, G. R., Cao, J. P., Liu, W. M., & Fan, H. (2013). Unbounded quantum
Fisher information in two-path interferometry with finite photon number. Journal of
Physics A: Mathematical and Theoretical, 46, 035302. http://dx.doi.org/10.1088/1751-
8113/46/3/035302.
Zwierz, M., Perez-Delgado, C. A., & Kok, P. (2010). General optimality of the Heisenberg
limit for quantum metrology. Physical Review Letters, 105, 180402. http://dx.doi.org/
10.1103/PhysRevLett.105.180402.
CHAPTER FIVE
Macroscopic Theory of Optical

Momentum
Brandon A. Kemp*
*College of Engineering, Arkansas State University, Jonesboro, Arkansas, USA
Contents
1. Introduction 437
2. Classical Electrodynamics 440
2.1 Electrodynamics of Moving Media 441
2.2 The Quasi-Stationary Approximation 446
2.3 Electrodynamics of Quasi-Stationary Media 449
3. Optical Force, Momentum, and Stress 452
3.1 Electromagnetic Force 452
3.2 Electromagnetic Momentum 455
3.3 Electromagnetic Stress 460
4. Momentum and Stress in Dispersive Media 467
4.1 Field and Material Contributions 468
4.2 Time-Averaged Energy and Momentum Propagation 470
5. Discussion 472
5.1 Conclusions 473
5.2 Application to Optical Manipulation 475
Appendix A. Scattering by a Cylinder 478
Appendix B. Scattering by a Sphere 481
Acknowledgments 485
References 485
1. INTRODUCTION
Light possesses energy and momentum. This statement should be obvi-
ous any time we feel the warmth of the sun or look at the tail of a comet. Light
leaves the sun, travels long distances through space, and does work on pon-
derable media at remote distances. Conservation principles ensure that the
energy and momentum are not lost only to be regained after the elapsed time

438 Brandon A. Kemp
required for light to reach its destination. Therefore, energy and momentum
must be accounted for within light itself. Maxwell (1891) first applied elec-
tromagnetic wave theory to predict and calculate the radiation pressure of
light. This prediction was soon verified by the observation of light pressure
on reflectors in vacuum by Lebedew (1901) and Nichols and Hull (1903).
Poynting (1905) quantitatively validated the theory with detailed measure-
ments of radiation pressure. Such experiments were relegated to scientific val-
idation of the electromagnetic wave theory until the invention of the laser.
Ashkin (1970) first demonstrated the manipulation of small dielectric
particles by laser light when particles suspended in water were drawn into
the axis of a laser beam and accelerated in the direction of propagation.
Rough calculations for the radiation pressure were used for designing the
experiments which involved a laser focused upon spheres a few wavelengths
in diameter. In addition to radiation pressure, the particles experienced a
trapping force that pulled the particles into the laser beam. This force was
reversed when the relative index of refraction was inverted. Air bubbles,
with dielectric permittivity less than that of the surrounding water, were
observed to be pushed out of the beam, while the radiation pressure
remained in the beam propagation direction. The trapping force and radi-
ation pressure are commonly known as the gradient force and the scattering
force, respectively, due to the separation based on the ray optics approxima-
tion for large particles or the Rayleigh approximation for small particles.
Optical levitation (Ashkin & Dziedzic, 1971, 1976) and radiation pres-
sure on a liquid surface (Ashkin & Dziedzic, 1973) were among the many
experiments that followed. Lasers were also proposed as a means to trap
atoms (Ashkin, 1978; Ashkin & Gordon, 1979), and single atoms were even-
tually trapped and cooled by optical tweezers (Chu, Bjorkholm, Ashkin, &
Cable, 1986; Chu, Bjorkholm, Ashkin, Gordon, & Holberg, 1986; Chu,
Holberg, Bjorkholm, Cable, & Ashkin, 1985). Colloidal dielectric particles
were also trapped in three dimensions using a single highly focused laser
beam (Ashkin, Dziedzic, Bjorkholm, & Chu, 1986). Such optical tweezers
result when a the gradient force is larger than the scattering force. Optical
manipulation has since found many application in physics, biology, chem-
istry, and medicine (Ashkin, 1997, 2000), and optical tweezers have become
a necessary tool for many researchers in microbiology (Ashkin &
Dziedzic, 1987).
The description of optical manipulation was further expanded by Burns,
Fournier, and Golovchenko (1989) when optical binding was first observed.
Optical binding can be used to create optical matter, a lattice of dielectric
Macroscopic Theory of Optical Momentum 439
particles held together by multiple particle interactions within an optical

field (Burns, Fournier, & Golovchenko, 1990). The optical scattering, gra-
dient, and binding forces provide a generic description for the optical
manipulation of colloidal particles. In the past few years, there has been a
renewed interest in the theory of optical momentum transfer to matter.
Much of this can be attributed to the growth in nanotechnology, where
optical forces can play a significant role in the dynamics of material systems
(Dholakia & Zemnek, 2010; Grier, 2003).
In spite of the simple language used to describe optical manipulation,
detailed modeling of optical forces can be a daunting task. There are two
steps involved when employing a macroscopic model of classical electro-
magnetics. First, the total electromagnetic fields must be determined. Sec-
ond, the electromagnetic force is calculated from the total fields. The first
task is accomplished by a variety of analytical and numerical solutions
(e.g., FEM, FDTD), which are generally well known. The second task is
shrouded in controversy. A quick review of the literature reveals a number
of electromagnetic force density equations. The most commonly used are
listed in Table 1 along with the momentum continuity formulation to which
they are inherently tied (Kemp, 2011). Although the list in Table 1 is not
exhaustive, it illustrates the source of confusion surrounding optical forces
as each listed force density may predict different force distributions in matter
under certain conditions (Mansuripur, Zakharian, & Wright, 2013). The
myriad of force densities are related to the dispute over the momentum
of light in media. This will become evident in Section 2. For now, it is suf-
ficient to say that we may calculate the total force on an object by either
integrating a force density over the volume or by considering the total
change in optical momentum.
Table 1 Leading Electromagnetic Force Densities Applied to Linear Media

Formulation Force Density (N/m3)
Minkowski 2 2 r + E + J B
12 E rE 12 H

Chu r PE r 0 M H 0 @ M E0 E
+ J + @ P 0 H
@t @t

Einstein-Laub + P rE + M r H + J + @ P 0 H 0 @ M E0 E
0 @t @t
Abraham 2 2 r + E + J B + @ fG
12 E rE 12 H Ag
M G
@t
r PE + @@tP B + r M
B + E + J B
Amperian
440 Brandon A. Kemp
The confusion over applicability and interpretation of the various elec-

tromagnetic force equations is related to a dispute regarding the momentum
of electromagnetic waves known as the AbrahamMinkowski debate. The
debate is more than a century old and relates to the correct form of optical
momentum in media (Abraham, 1909, 1910; Minkowski, 1908). Since, it
has been suggested that the debate is linked to the wave-particle duality
of a photon (Leonhardt, 2006): the Minkowski momentum

pM k n (1)
c
is related to the reduced wavelength in matter and the Abraham momentum
1
pA mv (2)
n c
is related to Einstiens mass equivalent m =c 2 , where is the reduced
Planks constant, k is the wavenumber, n is the refractive index, is the angu-
lar frequency, and c is the speed of light in vacuum. Both experiment and the-
ory have been applied in attempt to resolve the debate over the past century,
which have been reviewed by Brevik (1979), Pfeifer, Nieminen, Heckenberg,
and Rubinsztein-Dunlop (2007), Barnett and Loudon (2010), Baxter and
Loudon (2010), Milonni and Boyd (2010), and Kemp (2011).
This correspondence presents a macroscopic view of optical momentum
and forces with three fundamental aims. First, a description of how media are
modeled within classical electrodynamics illuminates the differences
between the various optical force and momentum densities in Section 2.
Second, interpretation of these differences allows the identification of the
kinetic momentum of light and how materials contribute to the electro-
magnetic stress in Sections 3 and 4. Finally, using these formulations and
interpretations, it is shown how optical manipulation experiments can be
modeled based on the developed conclusions in Section 5. The overall goal
is to present a classical framework for predicting experiments involving
optical momentum transfer to matter, such as involved with the growing
applications in optical manipulation.
2. CLASSICAL ELECTRODYNAMICS
To address energy, momentum, and charge conservation in electro-
dynamics, we use the subsystem mathematical framework. This is necessary
since there are a number of formulations of classical electrodynamics (e.g.,

Minkowski, Chu, and Amperian) commonly used today which differ in
how matter is modeled. Since each formulation separates the field and
material responses differently, there are energy, momentum, and charge
continuity equations presented for the various formulations used.
The energy and momentum continuity are expressed for any given sub-
system j by the equations (Penfield & Haus, 1967)
@Wj r , t
j r ,t r Sj r , t (3a)
@t

@ G j r , t
fj r , t r Tj r , t (3b)
@t
where fj is the force density, j is the power density, T j is the momentum
flux or stress tensor, Sj is the power flux, G j is the momentum density, and
Wj is the energy density. This indicates that each j may represent any sub-
system such as electromagnetic, hydrostatic, thermodynamic, kinetic, etc.
Closing the overall system, such that
X
j r , t 0 (4a)
j
X
fj r ,t 0, (4b)
j
is is a statement of energy and momentum conservation. Additionally,

we find that charge is generally a conserved quantity in any formulation
of electrodynamics and is written as
@j r , t
0 r Jj r ,t , (5)
@t
where Jj r ,t and j r , t denote a current density and charge density for a
given formulation, respectively.
2.1 Electrodynamics of Moving Media

Multiple formulations of Maxwells classical field theory have been advanced
over the years. Each differ in how ponderable media are modeled. Here,
we review the three most commonly used today; the Chu, Amperian,
and Minkowski formulations. The reader is directed to Kong (2005),
442 Brandon A. Kemp
Kemp (2011), and Penfield and Haus (1967) for additional information
regarding these and other, less used, formulations.
2.1.1 Minkowski Formulation

The MaxwellMinkowski equations
r ,t @
r H Dr , t J r ,t (6a)
@t
r ,t + @ B
r E r ,t 0 (6b)
@t
r ,t r , t
r D (6c)
r , t 0,
r B (6d)
combine the field and material contributions within the time and space
dependent field vectors E r ,t, H
r , t, D r ,t, which contain
r , t, and B
the response of the material via macroscopic constitutive relations. The free
charge density r ,t and free current density J r ,t are the only field-free
quantities present when considering the Minkowski formulation. In a linear,
isotropic medium, which is effectively dispersionless at frequencies of interest,
r , t EE
the constitutive relations D r , t and B
r ,t H r , t yield the
terms for the momentum continuity equation in the Minkowski formulation
1 r ,trE 1 H r ,t H
r ,tr
fM r , t E r ,t E
2 2 (7a)
r , t + J r , t B
+ r ,tE r , t
1 r ,t + B
r ,t H

r , t I
TM r ,t D r ,t E
2 (7b)
D r ,t B
r ,tE r , tH r ,t
G r ,t B
M r , t D r , t, (7c)
r ,t E
where 12 E r ,trE is the well-known Helmholtz force. The
corresponding energy continuity equations are defined by
M r , t J r ,t E r ,t (8a)
SM r ,t E r , t H r ,t (8b)
1
WM r , t D r , t Er , t + B
r ,t H
r ,t : (8c)
2
In Section 4, the consequences of dispersion on the momentum of the wave
are explored. A statement of charge conservation results by taking the
divergence of Equation (6a) and utilizing Equation (6c). Charge conserva-

tion in the Minkowski formulation is in terms of the free charge and currents
@r , t
0 r J r , t : (9)
@t
Equations (7) and (8) are derived using the predefined constitutive rela-
tions with Equation (6). This indicates that the desired momentum and
energy terms, along with the f and terms, depend on the constitutive rela-
tions used in modeling the media. Thus, when using the Minkowski formu-
lation, the combination of both field and material contributions are present
in the description of the momentum and energy of the electromagnetic
subsystem.
2.1.2 Chu Formulation

The MaxwellChu equations
rH C r ,t E0 @ EC r , t J e r , t (10a)

@t
@
r EC r , t + 0 H C r , t J h r ,t (10b)
@t
E0 r EC r ,t e r ,t (10c)
0 r H C r , t h r , t (10d)
separate the electric field EC r , t and magnetic field H C r ,t from the mate-
rial response and are, therefore, often referred to as the EH representation.
The Chu formulation models the material responses by equivalent electric
current density Je r , t, magnetic current density Jh r ,t, electric charge
density e r ,t, and magnetic charge density h r , t. These quantities for
moving media with local velocity field v r , t are defined as
@
Je r ,t PC r ,t + r PC r , t v r , t + JC r , t (11a)
@t
@ C r , t v r ,t
Jh r ,t 0 M C
r ,t + 0 r M (11b)
@t
e r ,t r PC r ,t + C r , t (11c)
h r , t 0 r M C r , t, (11d)
where M C r , t is the magnetization, PC r , t is the polarization, JC r , t is
the free current density, and C r ,t is the free charge density of the given
444 Brandon A. Kemp
medium. Here, the subscript C indicates that the values involved within the
Chu formulation differ from similar terms in other common formulations.
The momentum continuity equation in the EH representation is defined by
the terms
feh r ,t e EC + h H C Jh E0 EC
C + Je o H (12a)
1
Teh r ,t E0 EC EC + 0 H C H C I E0 EC EC 0 H
C HC (12b)
2
eh r ,t E0 0 EC H
G C , (12c)
and the corresponding terms in the energy continuity equation are
eh r , t Je EC + Jh H
C (13a)
Seh r ,t EC HC (13b)
1
Weh r , t E0 EC EC + 0 H C H
C , (13c)
2
where the subscript eh denotes quantities resulting from the EH representa-
tion which are quadratic in the fields. In these equations, the exact response
of the material has been left open. That is, the material could be nonlinear,
anisotropic, or dispersive. The Chu formulation combines the bound and
free electric and magnetic charges to conserve the charge within the system.
Charge conservation can be expressed in terms of the total charge and
current densities
@e r , t
0 r Je r ,t (14a)
@t
@ r ,t
0 r Jh r ,t h : (14b)
@t
2.1.3 The Amperian Formulation

The MaxwellAmperian equations
1 @
r BA r , t E0 EA r ,t Jeb r ,t (15a)
0 @t
@
r EA r , t + BA r , t 0 (15b)
@t
E0 r EA r , t eb r , t (15c)
r BA r ,t 0 (15d)
separate the electric EA r , t and magnetic BA r , t fields from the material
response and is, therefore, often referred to as the EB representation. The
Amperian formulation models the material responses by equivalent electric
current density Jeb r ,t and electric charge density eb r ,t. These quantities
for moving media are defined as
@ A r ,t v r , t

Jeb r , t P A r ,t 0 E0 M
@t (16a)
+ r M A r , t + PA r ,t v r , t + J A r , t

eb r ,t r PA r ,t 0 E0 M A r , t v r , t + A r , t, (16b)
where M A r , t is the magnetization, PA r , t is the polarization, JA r , t is

the free current density, and A r ,t is the free charge density of the given
medium. Here, the subscript A indicates that the values involved within the
Amperian formulation differ from similar terms in other common formula-
tions. The momentum continuity equation in the EB representation is
defined by the terms
feb r , t eb EA + Jeb BA (17a)

1
Teb r ,t E0 EA EA + 1 1
0 B A B A I E0 E A EA 0 B A B A (17b)
2
G eb r ,t E0 EA BA , (17c)
and the corresponding terms in the energy continuity equation are
eb r , t Jeb EA (18a)

Seb r , t 1
0 E A BA (18b)
1
Web r ,t E0 EA EA + 1
0 BA BA , (18c)
2
where the subscript eb denotes quantities resulting from the EB representa-
tion which are quadratic in the fields. In these equations, the exact response
of the material has been left open. That is, the material could be nonlinear,
anisotropic, or dispersive. Charge conservation is expressed in terms of the
total charge and current density
@ r , t
0 r Jeb r ,t eb : (19)
@t
446 Brandon A. Kemp
2.2 The Quasi-Stationary Approximation

While the AbrahamMinkowski debate originated out of relativistic consid-
erations, the primary differences between the theories can be studied inde-
pendent of material motion (Nelson, 1991). For simplicity, this paper
presents a quasi-stationary analysis of electrodynamics. That is, we apply a
limiting process such that v ! 0 and m ! 1 simultaneously such that the
kinetic energy 12 mv2 of the material approaches zero while the momentum
m v remains, in general, a nonzero vector. This stationary approximation is
commonly applied in the literature and in text book calculations (Kong,
2005) where optical momentum is imparted to material while the optical
energy is conserved. This is illustrated in the following two examples of opti-
cal energy incident from vacuum onto a perfect reflector. First, an example is
given within the framework of classical electrodynamics (Daly &
Gruenberg, 1967; Sheppard & Kemp, 2014). Second, a simple argument
of a Doppler shifted photon is given (Kemp & Grzegorczyk, 2011). Both
illustrate that energy is not conserved in the quasi-stationary approximation.
However, the approximation is useful for modeling many experimental
configuration where the kinetic energy of the material is much smaller than
the energy of the electromagnetic waves.
2.2.1 Plane Wave Example

Consider the electromagnetic fields resulting from a plane wave incident in
the z^ direction upon a perfect electrical conductor at z vt. Complex
notation is used to represent time-harmonic fields such that, for example,
the time-harmonic electric field E is related to the complex field E by
Er , t fE
r eit g, (20)
where {} is the real-part operator. Average values of derived quantities,
which are quadratic in the fields, are computed directly from the complex
fields. For example, the time-averaged vacuum Poynting power is given by
1 fE H
hSi g, (21)
2
where * denotes complex conjugation. The incident and reflected electric
and magnetic fields are

E i xE
^ 0 eiki z + i t + Reikr zr t (22a)
ik z + t
i ^
H yH0 e i i
Reikr zr t , (22b)
p
where H0 E0 E0 =0 , ki i/c, and kr r/c. The boundary condition
z^ E + 0 at z vt yields the phase matching condition
v B
kiv i krv r and the reflection coefficient
+1
R , (23)
1
where v/c. Note that the Doppler effect is a consequence of phase
matching at a moving boundary

1+
r i : (24)
1
Since the incident region is vacuum, there is no ambiguity as to the appli-
cable energy and momentum equations. The incident and reflected power
flow are
1
i z^1 E0 H0 zhS
hSi i Ei H ^ ii (25a)
2 2
1
i z^1 E0 H0 jRj2 zhS
hSr i Ei H ^ r i: (25b)
2 2
The total electromagnetic power flow into a closed volume at the reflector
interface is
hSi i
hPelec i hSi i hSr i 4 : (26)
c 1 2
The change in stored electromagnetic energy is
v nE0 0 o 1 + 2
hPstored i E E + H H 2hSi i : (27)
2 2 2 1 2
Therefore, the mechanical work that must be done to the system to keep the
reflector at constant velocity is
1 +
hPmech i hPstored i hPelec i 2hSi i : (28)
1
Likewise, the force on the reflector is
hF elec i z^hTi i + hTr i + vhGi i + hGr i, (29)
where the z^ components of the momentum densities and the z^z^ compo-
nents of the stress tensors are
448 Brandon A. Kemp
1
hGi i i hS i i
E0 0 Ei H (30a)
2 c

1
hGr i E0 0 Er H r hS r i (30b)
2 c
1 nE0 0 o hSi i
hTi i Ei Ei + H i H i (30c)
2 2 2 c
1 nE0 0 o hSr i
hTr i Er Er + H r H r : (30d)
2 2 2 c
The mechanical force required to keep the reflector from accelerating is
hSi i 1 +
hF mech i hF elec i z2
^ : (31)
c 1
Obviously, v hF mech i hPmech i so that the mechanical force and work are
required to keep the reflector at constant velocity. The case of v ! 0
produces the well known result hF elec i z2hS ^ i i=c, as is obtained from
the stationary reflector case treated in most electromagnetic textbooks
(Kong, 2005).
Since hPmechi is proportional to v, it will go to zero under a stationary
material approximation, but lim v!0 hF mech i z2hS^ i i=c. If the mechanical
balance force is removed, the reflector will accelerate from v 0 and the
kinetic energy of the mirror will be increased. However, this problem is typ-
ically treated using a stationary media approximation such that the Doppler
shift is ignored. It is equivalent to taking a quasi-stationary approximation
where v ! 0 while m ! 1 such that the kinetic momentum of the material
mv remains a nonzero number and the kinetic energy of the material 12 mv v
goes to zero.
2.2.2 Doppler Shifted Photon Example

Next, consider a photon with energy normally incident upon a mirror
initially at rest as in Figure 1. The reflected photon is red shifted due to the
Doppler effect described in Equation (24). For simplicity, assume that jvj c
so that
v
0 1 + , (32)
c
where 0 is the Doppler-shifted frequency of the reflected photon and m is
the mass of the mirror, and v zv
^ is the the velocity of the mirror after
e0 , m0

Ei

ki
x
n
z
y
z=0
Figure 1 A plane wave incident upon a perfect electrical conductor. The Fresnel reflec-
tion coefficient of the reflector in the stationary frame is R 1.
reflecting the photon. The nonrelativistic Doppler shift is applied since

assuming that the mass of the mirror is much greater than the effective mass
of the photon (i.e., m =c 2 ). Energy conservation states that the initial
energy of the photon is equal to the sum of the final kinetic energy of the
photon plus the final kinetic energy of the mirror
1 v
mv2 + 1 : (33)
2 c
Solving Equation (33) yields solutions for the mirror energy and momentum
1 2 v
mv (34a)
2 c

mv 2 : (34b)
c
If, however, we let v ! 0 a priori as in the quasi-stationary approximation,
the calculated kinetic energy of the mirror will be zero while the momen-
tum after reflection will be unchanged.
2.3 Electrodynamics of Quasi-Stationary Media

Under the quasi-stationary approximation, the velocity terms in Maxwells
equations go to zero. That is, v ! 0 in all of the equations of Section 2.1.
Also, the field vectors in each formulation are the same. That is
450 Brandon A. Kemp
EC EA E, and likewise for all other fields and sources. Because of this,
we may simply state the polarization and magnetization in terms of the field
vectors such that
P r , t E0 Er , t
r ,t D (35a)
r ,t B
0 M r , t H
r , t, (35b)
0
which are general enough to include causal (i.e., lossy and dispersive) and
anisotropic media.
Obviously, the Minkowski, Chu, and Amperian force densities given in
Table 1 correspond to the corresponding formulations for stationary media.
Therefore, the stationary force densities in Table 1 may be used only when
the velocity of the material can be safely ignored, such as when relativistic
effects are insignificant. Two other equations, the Einstein-Laub and the
Abraham force densities, are also given in Table 1. Both have been discussed
extensively (see for example reviews by Pfeifer et al. (2007) and Milonni and
Boyd (2010). Here, Equation (35) are used to rearrange terms in the formu-
lations given in Section 2.1 to arrive at the Einstein-Laub and the Abraham
force densities. These mathematical rearrangements are exact and consistent
with Maxwells equations. However, such excercises shouldnt be taken as
basis for proof or interpretation (Grzegorczyk & Kemp, 2008).
Einstein and Laub (1908) proposed a formulation which models both
dielectric and magnetic response as effectively bound dipoles. This differs
from the Chu formulation, which models media as effectively bound electric
and magnetic charges, and the Amperian formulation, which models media
as bound electric charges and infinitesimal current loops. However, at the
macroscopic level, these microscopic viewpoints are effectively lost, and
the real difference is in which field terms give W, S, and T defining
G,
the electromagnetic subsystem for the formulation. Using vector calculus
identities to write the force on bound electric charges and bound magnetic
charges
E P rE r PE
r P (36a)
H
0 r M 0 M rH H
0 r M (36b)
allows for substitution into the Chu force given in Table 1. The tensor terms
are then moved to the other side of the momentum continuity equation
such that

@ P 0 @ M E0 E
+ P rE + 0 M r H + J + 0 H
@t @t

@
1

E0 0 E H r E0 E E + 0 H H I D E B H ,
@t 2
(37)
where the relations in Equation (35) have been used. Equation (37) is the
momentum continuity equation for the Einstein-Laub formulation. The
energy continuity equation is reported as identical to Chu (Mansuripur
et al., 2013).
Likewise, terms can be rearranged from the Minkowski formulation to
arrive at the Abraham formulation. This is accomplished by simply adding
the term known as the Abraham force @G M G A =@t to the Minkowski
force density and subtracting the same term from the momentum density
to arrive at the momentum continuity equation
1 2 1 2 @
E rE H r + E + J B + G
M GA
2 2 @t (38)
@
1
E0 0 E H r D E + B H I DE B H ,
@t 2
where G M is the Minkowski momentum density and G A is the Abraham

momentum density given in Table 2. It should be pointed out that a symmet-
ric form of the Abraham tensor has been perpetuated in the literature (Brevik,
1979; Kemp, 2011). However, the asymmetric stress tensor given in
Equation (38) was originally presented by Abraham (1909). The energy con-
tinuity equation is reported as identical to Minkowski (Pfeifer et al., 2007).
The leading electromagnetic momentum continuity equations in sta-
tionary media are defined by the force densities in Table 1 along with the
Table 2 Leading Electromagnetic Momentum Densities and Stress Tensors

Formulation Momentum Density Stress Tensor

Minkowski D B 1
2
I DE BH
D E + B H

Chu E0 0 E H

2 E0 E E + 0 H H I E0 E E 0 H H
1

Einstein-Laub E0 0 E H
1
2 E0 E E + 0 H H I D E B H

Abraham E0 0 E H
1
2 D E + B H I DE B H

Amperian E0 E B 1
2 E0 E E + 1 B B I E0 EE 1 BB
0 0
452 Brandon A. Kemp
momentum densities and stress tensors in Table 2. The form of the energy
and momentum continuity equations are given in Equations (3a) and (3b).
The Minkowski, Chu, and Amperian energy terms have been given in
Section 2.1. The Einstein-Laub energy continuity terms are identical to
the Chu formulations, and the Abraham energy continuity terms are
identical to the Minkowski formulation. In general, torque density can be
calculated as r f, with the exception of the Einstein-Laub torque den-
sity which is supplemented by the terms P E + 0 M H (Einstein &
Laub, 1908; Mansuripur et al., 2013).
3. OPTICAL FORCE, MOMENTUM, AND STRESS

The electromagnetic force may be computed directly from the force
density or by considering spatial variations of the stress tensor and temporal
variations of the momentum density. In the following sections, consideration
is given to these contributions to the total force. In certain cases, different
force densities give identical predictions for the total force on a material
object, and these cases are defined. Additionally, there are situations where
either the momentum density G or the stress tensor T may be isolated.
These situations are treated individually to explore the nature of G and
T for nondispersive media. This general nature will be used later to model
optical manipulation experiments. Finally, this section is concluded with
a presentation of optical momentum in causal media, which must include
dispersion and loss in the material model contribution.
3.1 Electromagnetic Force

The force distribution inside an object depends upon the equation used to
compute the force density. In general, the equation takes the form of
Equation (3b) where the force densities fr , t are given in Table 1 and
r ,t and T r , t are given in Table 2. The
the corresponding values for G
total electromagnetic force F t on an object results from integration of
an electromagnetic force density fr , t over the volume V. Equivalently,
one may choose to apply the divergence theorem to reduce the contribution
of an electromagnetic stress tensor T to an integral over the surface A with
outward pointing area element dA enclosing the volume V so that the total
force is calculated equivalently by
Z I
@ T r , t:
F t dV Gr , t dA (39)
V @t A
In the following sections, this equation is studied in regards to situations

when the various formulations give equivalent results.
3.1.1 Equivalence of Total Force

The equivalence of total force for different formulations and calculation
methods has been demonstrated or proven by a number of researchers.
For example, see reports by Kemp, Grzegorczyk, and Kong (2005),
Barnett and Loudon (2006), Loudon and Barnett (2006), and Mansuripur
(2008). To show which total force equations are equivalent at any point
in time, refer to the diagram in Figure 2a depicting an object surrounded
by vacuum. If we choose to integrate the force density fr , t over the vol-
ume completely enclosing the object as depicted by the dashed line, the total
force is given by Equation (39), where the tensor reduces to the vacuum
Maxwell stress tensor
1
Tvac E0 E E + 0 H E0 EE 0 H
H H:
(40)
2
This results because the surface of integration is outside the material.
The momentum density G must still be chosen based on the formulation.
Therefore, any formulations which share a common momentum density
(a) (b)
e 0, m0 e b, mb
e,m e,m
Figure 2 Diagram for illustration of total force calculation. (a) The total force on an
object in vacuum is computed by volume integration of a force density within a region
which completely encloses the material object. The volume is enclosed by the surface
depicted by the dashed line. (b) The total force on an object embedded is computed in a
similar way by considering a thin vacuum region between the two materials.
454 Brandon A. Kemp
will produce identical results for the total force on an object at all points in
time. For example, the Chu, Einstein-Laub, and Abraham force densities are
equivalent in terms of total force. Since the Amperian formulation differs
only in the modeling of magnetic media, it will also produce identical results
for total force for dielectric materials.
As a second case, consider an object which is submerged or embedded
inside another medium as depicted in Figure 2b. In this case, the total force
will still be the same for two formulations if their momentum densities are
identical, but it is important to clarify how the force is computed. If the force
is computed by integrating the force density fr , t over a volume, there will
generally be both volume and surface forces. The surface forces arise due to
material discontinuities at the boundaries (Kemp, Grzegorczyk, & Kong,
2006a; Mansuripur, 2004). Therefore, there may be initial ambiguity as
to which material, the embedded object or the embedding medium, to
which the surface forces should be assigned. To alleviate this ambiguity, ima-
gine there is a thin layer of vacuum in the boundary. A volume integration
procedure can be applied, as before, to the embedded object. Since the sur-
face of the integration is in the vacuum region, the tensor reduces to the
vacuum stress tensor and the force separation for the two media is unambig-
uous. Allowing the vacuum region to vanish, or mathematically taking the
limit of the gap to approach zero, yields the original field problem with the
forces correctly separated (Kemp & Grzegorczyk, 2011). In this case, how-
ever, there will generally be forces on both the embedded object and the
surrounding medium. The conclusion remains the same; any formulations
which share a common momentum density will produce identical results for
the total electromagnetic force on a material object at any point in time.
3.1.2 Equivalence of Total Time-Average Force

Systems are often modeled under the influence of time-harmonic excitation.
In these situations, it is the time-average, or cycle-average, force that is the
observable quantity. The total average force on an object due to time-
harmonic excitation is
I
hF i dA hT i: (41)
A
The momentum density does not contribute to the time-average force of

time-harmonic fields. This is because the the @ G=@t term is comprised of
terms such as @=@tcos t, @=@tsin t, and @=@tcos tsint,
2 2
which all vanish under time-averaging.

The total average force can be computed by integrating an average force

density over a volume that includes the entire object. All such computations
will yield identical results regardless of the force density applied, assuming it
is consistent with Maxwells equations. To see why, consider again the dia-
gram in Figure 2a. Choosing a volume which completely encloses the mate-
rial object allows for the force to be computed by a surface integration along
a contour which lies completely in the vacuum region as depicted by the
dashed lines. In this case, the momentum density G does not contribute
and all formulations reduce to the divergence of the time-averaged vacuum
Maxwell stress tensor, which is given by

1 1
hTvac i E0 E E + 0 H H E0 EE 0 H H : (42)
2 2
A similar argument holds for a material object which is embedded inside
another medium. It is only necessary to consider separation of the surface
force contributions between the embedded material and the embedding
medium as described in the previous section and depicted in Figure 2b.
The total time-average force on a material object is unambiguous. Any
formulation will produce an identical value for the total time-average elec-
tromagnetic force on a material object. Because of this, it is impossible to
determine the time-domain force equation from a time-average force equa-
tion, and such assumptions lead to serious errors (Kemp, 2013).
3.2 Electromagnetic Momentum

In this section, contributions of the electromagnetic momentum density
r , t are isolated. This is accomplished by considering pulses of finite
G
extent. When the volume V of integration is extended such that the material
and the pulse are completely contained, the fields on the surface A are zero.
Therefore, the stress tensor is also zero on A and Equation (39) reduces to
Z
@
F t dV G r ,t: (43)
V @t
3.2.1 Comparison of Momentum Densities

It has been established that the different predictions in total force on a mate-
rial object arises from the variety of momentum density terms for G r , t, not
the stress tensor terms T r , t, and the leading momentum densities have
been given in Table 2. Only three unique momentum densities result from
456 Brandon A. Kemp
the five leading formulations reviewed. First, the Abraham momentum den-
sity, G A E0 0 E H
shows up in the Chu, Einstein-Laub, and Abraham
formulations. Second, the Livens momentum density G L E0 E B (see
references by Livens (1918), Nelson (1991), and Scalora et al. (2006)) is
included in the Amperian subsystem. Third, the Minkowski momentum
density G M D B completes the Minkowski momentum continuity
equation. For each form, the momentum of an electromagnetic pulse trans-
mitted into a medium can be written in terms of vacuum quantities. In this
way, the various forms can be compared and contrasted (Kemp, 2011).
For comparison, consider a plane wave pulse normally incident upon a lin-
p
ear, transparent material with index of refraction n c E. For simplicity, the
material is considered to be nondispersive in the frequency range of interest and
the quasi-stationary approximation is applied. The incident pulse is described
by the electric field E xE
^ 0 z, tcosk0 z t. Here, E 0(z, t) is an envelope
function for the harmonic wave cos k0 z t and k0 is the wavenumber in the
incident vacuum region. The incident energy density Wi z E0 E 20 z, t and
momentum density G i z, t E0 0 Ei z, t H
i z, t zW
^ i z, t=c are
unambiguous since the wave is incident from vacuum. The incident optical
energy is defined as the volume integral over the entire incident pulse. The
one-dimensional pulse varies only in z so that the incident energy is
ZZZ Z
Ei dVWi z dzWi z (44)
V
2
and has units (J/m ). Similarly, the unambiguous incident momentum is
determined by integrating the incident momentum density at an instant
prior to the field interaction with the material. It is given by
ZZZ Z
Wi z Ei
pi dV G i z z^ dz z^ , (45)
V c c
and has units (N s/m2).
We may also arrive at quantities for the reflected energy density Wr(z)
R Wi(z) and reflected momentum density Gr(z) R2Wi(z)/c, where R2 is
2
the reflectivity (Kong, 2005). The reflected energy in terms of the incident
energy is
Er R2 Ei , (46)
and the reflected momentum in terms of the incident momentum is
pr R2 pi : (47)
In order to express the energy and momentum in the material in terms of

unambiguous vacuum quantities, the excitation energy and momentum
are defined as
Eexc Ei Er Ei 1 R2 (48)
pexc pi pr pi 1 + R ,
2 (49)
which represent the difference between the incident and reflected pulses.
No assumptions are made as to the nature of momentum propagation or
transfer in the material. The intent, rather, is to study the difference between
Abraham, Livens, and Minkowski momenta in media by developing precise
relations for the different momenta in terms of the vacuum quantities.
The transmitted Abraham momentum density is
t z^0 T 2 G
A z E0 0 Et H
G i z z^Wi z 1 R2 , (50)
c
where 0 is the wave impedance of free space, is the wave impedance of the
material, T is the transmission coefficient, and the relation (0/)T2 1 R2
resulting from the boundary conditions has been used. The Livens momen-
tum density is given by
L z E0 Et Bt E0 0 Et H
G t + E0 0 Et M
t
Wi z t Et , (51)
z^ 1 R2 E0 0 M
c
where the last term is the hidden momentum as coined by Shockley and
James (1967). Similarly, the transmitted Minkowski momentum density is
M z EEt H
t zn i z zn Wi z
G ^ 2 0 T 2G ^2 1 R2 : (52)
c
The spatial length of the transmitted pulse is decreased proportional to
the factor n1 due to the change in velocity. The Abraham momentum
of the transmitted pulse is
Z

pA dzG A z, t z^1 Ei 1 R2 z^ 1 Eexc : (53)
z nc nc
The Livens momentum of the transmitted pulse reduces to
Z Z
1 t Et ,
pL dzGL z,t z^ Eexc dzE0 0 M (54)
z nc z
which differs from the Abraham form by the hidden momentum. Finally,
the Minkowski momentum density yields the momentum
458 Brandon A. Kemp
Z
M z,t z^ 1 n2 Ei n
pM dzG 1 R2 z^ Eexc (55)
z n c c
transmitted into the medium. Consequently, both the direction and magni-
tude of the momentum transferred to the surface depend upon this choice
according to Loudon (2004). In more exact mathematical terms, this means
that the subsystem required to close the overall system depends upon which
formulation of electromagnetics is applied.
3.2.2 Field Momentum

The various forms for the momentum of light differ in how the material
response is included. The total system may be subdivided in a number of
ways such that portions of the material response are included in the electro-
magnetic subsystem. It is desirable to determine the kinetic momentum of
light, which is the momentum which contains only field contributions with-
out contributions from the mass of the material. The thought experiment by
Balazs (1953) allows for the determination of the kinetic momentum by
studying the center-of-mass displacement of a material slab as an electromag-
netic pulse passes through. A number of authors have presented versions of
this thought experiment. For example, see the works of Loudon (2004),
Scullion and Barnett (2008), Barnett (2010), Kemp (2011), and Griffiths
(2012).
Here, the analysis is simplified by considering a nondispersive magne-
todielectric material that is impedance matched to the surrounding vacuum.
p
The slab is comprised of a material with index of refraction n c E and
p p
wave impedance =E 0 =E0 . An electromagnetic wave pulse
depicted in Figure 3 has an initial free space momentum Ei /c. The pulse
is delayed with respect to the free space path by the distance L (n 1)d
by the slab of thickness d since the velocity in the material is v c/n. In order
to maintain uniform motion of the center-of-mass energy, the required
kinetic momentum of the material while the pulse overlaps spatially with
the slab is

Ei 1
pm 1 : (56)
c n
Momentum conservation requires that the electromagnetic contribution to
the total momentum be the difference between the total momentum of the
incident pulse and the material momentum given by Equation (56).
(a)
n = c me
d
(b)
Figure 3 Diagram of Balazs thought experiment for determining the kinetic momen-
tum of light. An electromagnetic wave (a) traveling through an impedance-matched
magnetodielectric slab is delayed by a length L (n 1)d in comparison to (b) an iden-
tical, unimpeded pulse traveling through vacuum.
Therefore, the electromagnetic momentum of the pulse is the Abraham

momentum
1 Ei
pA : (57)
nc
This analysis is actually very similar to the reasoning applied to arrive at
Equation (2). The effective mass of the pulse remains constant as it traverses
through a medium requiring that the momentum changes in proportion to
the velocity.
The conclusion of the Balazs thought experiment is that the Abraham
momentum density gives momentum contained within the electromagnetic
fields. As previously stated, the Abraham momentum density appears in a
number of subsystems (e.g., Chu and Einstein-Laub). This analysis excludes
other forms, such as the Livens momentum and the Minkowski momentum,
as being the kinetic momentum of light. Of course, the Livens momentum
appears in the Amperian formulation, and it differs from the Abraham
momentum by the hidden momentum defined in Section 3.2.1. The topic
of hidden momentum continues to receive significant attention (Griffiths &
Hnizdo, 2013; Mansuripur, 2012; Saldanha, 2013). The fact that a hidden
momentum term must be added to the Livens momentum in the Balazs
460 Brandon A. Kemp
thought experiment does not imply that the Amperian formulation violates
momentum conservation. It does mean, however, that the Amperian for-
mulation is not considered as a possible choice for the kinetic subsystem
of light. It must be concluded that the macroscopic formulation of electro-
dynamics presented as the Amperian formulation (i.e., EB representation)
does not provide the field only contribution to energy and momentum.
Although, microscopic reasoning can be applied to justify the inclusion of
the hidden momentum model (Griffiths & Hnizdo, 2013), the hidden
momentum term cannot be simply added to the momentum density in
the Amperian formulation without being accounted for equally in either
the force density term or stress tensor term of the formulation. This math-
ematical rigour is necessary to ensure that the macroscopic momentum con-
tinuity equation is consistent with Maxwells equations. As previously
demonstrated, such mathematical rearrangements defines a new subsystem,
but this author is unaware of such a formally proposed formulation and
interpretation. Such a subsystem would effectively include the Abraham
momentum density with a modified version of the Amperian force density
and/or stress tensor. It should also be reiterated that the results of this section
make no claims regarding the microscopic nature of matter response.
Although the different formulations provide conceptual views of material
response (e.g., effective dipoles, bound charges, or current loops), at the
macroscopic level these details are lost (Kong, 2005; Penfield & Haus,
1967), and the important features of the macroscopic theories pertain to
which macroscopic fields appear in the continuity equations.
3.3 Electromagnetic Stress

In this section, contributions of the electromagnetic stress tensor are isolated.
This is accomplished by computing the time-average force of time-
harmonic waves. As previously shown, such computations nullify contribu-
tions from the momentum density term.
3.3.1 Momentum Flow Into Media

As a first example, consider an electromagnetic wave normally incident from
vacuum upon a dielectric boundary at z 0. It is desirable to study the
momentum flow from the Chu and Einstein-Laub formulations as leading
contenders for the kinetic subsystem of electrodynamics (Kemp, 2011).
However, in this analysis, no distinction is made between the two. The fields
in the incident region are
E0 xE
^ i eikz + Reikz (58a)
0 y^Hi eikz Reikz :
H (58b)
The excitation power flow is the power available to the medium
1
0 1 Ei Hi 1 jRj2 1 jRj2 hSi i,
hSexc i n^ E0 H (59)
2 2
where hSi i 12 fEi Hi g is the power flow of the incident wave. The reflec-
tion coefficient is R (n 1)/(n + 1). Therefore, the excitation power flow
can be written as
" #
n + 12 n 12 4n
hSexc i hSi i < Si > njT j2 hSi i, (60)
n + 1 n + 1 n + 12
where njTj2 4n/(n+1)2 is the transmissivity (Kong, 2005). The excitation

energy is completely defined in terms of the vacuum fields. Therefore,
the excitation power flow does not include material contributions. Like-
wise, the excitation momentum flow is the momentum available to the
medium

1 nE0 2 0 2 o 1 E i Hi 2
hTexc i jE0 j + jH 0 j 1 + jRj , (61)
2 2 2 2 c
and it is unambiguously defined in terms of the Tzz component of the free
space Maxwell stress tensor in the incident region. Substituting for jRj2 gives
" #
n + 12 n 12 hSi i 2n2 + 2 hSi i
hTexc i + : (62)
n + 12 n + 12 c n + 12 c
Finally, the excitation momentum flux reduces to

1 1 hSexc i
hTexc i n + : (63)
2 n c
Inside the dielectric, both the Chu and Einstein-Laub formulations yield the
electromagnetic stress

1 nE0 2 0 2 o 1 1 1 E t Ht
hTeh i jEt j + jH t j n+
2 2
2 2 2 n c
(64)
1 1 hSexc i
n+ :
2 n c
462 Brandon A. Kemp
The electromagnetic stress appears as the average of n/c and 1/nc times the
excitation power flow. In this form, the transmitted momentum flow may
be written as the average of the transmitted Abraham and Minkowski
momenta (Mansuripur, 2004). However, it is noted that the transmitted
momentum flow given in Equation (64) is equal to the excitation momen-
tum in Equation (63). Therefore, the EH representation of the stress tensor
in media yields a momentum flow which is equal to the excitation momen-
tum flow at normal incidence. Both the Chu and Einstein-Laub formula-
tions yield zero surface pressure in this case as the net momentum flow is
conserved.
As a second example, consider an impedance-matched material such that
p p p
n c E and =E 0 =E0 . The reflectivity is zero in this case. The
excitation momentum flow is simply given by the incident momentum flow
1 nE0 2 0 2 o 1 Ei Hi hSi i
hTexc i jE0 j + jH 0 j : (65)
2 2 2 2 c c
The transmitted momentum flow given by either the Chu or Einstein-Laub
formulations is
1 nE0 2 0 2 o 1 Et Ht hSi i
hTeh i jE t j + jH t j : (66)
2 2 2 2 c c
In this case, the EH representation yields zero force on the surface since the
momentum flow is conserved.
3.3.2 Stresses in Dielectric Fluids

At normal incidence upon a planar boundary of dielectric and magnetic
media, the EH representation provides for conserved momentum flow at
the boundary. However, this is only a special case. In general, oblique inci-
dence or curved surfaces give rise to surface forces. Furthermore, wave inter-
ference produces volume forces due to field variations (Kemp et al., 2006a).
These forces, in turn, produce material stresses in addition to the electro-
magnetic stresses. Before considering the stresses due to interference fringes,
it is illustrative to consider the electromagnetic and material stress contained
within a plane wave. A time-harmonic, unbounded plane electromagnetic
wave exerts zero time-average force on a dielectric fluid. However, a time-
average material stress exists within a propagating plane wave.
Consider a wave with a planar front which is ramped up from zero to
electric and magnetic field amplitudes E0 and H0. The velocity of the wave
p
in the dielectric fluid is v c/n, where n c 0 E. The electric and magnetic
fields are
Ez,t E0 z, tcoskz t (67a)

Hz,t H 0 z, tcos kz t, (67b)
where E0 z, t xE 0 z,t y^H0 gz vt. The function g

^ 0 gz vt and H
(z vt) is the ramp function which is slowly varying in comparison to and
approximately linear over the spatial length L so that its temporal duration is
L/v. For z > vt + L, the electric and magnetic field amplitudes are zero, and
for z < vt, the electric and magnetic field amplitudes are E0 and H0. Before
the planar front arrives at a region of the material, the electromagnetic and
material stresses are zero. The average electromagnetic stress in the region
z < vt is given by the hTzzi component of the Chu or Einstein-Laub stress
tensors

1 nE0 2 0 2 o 1 1 hS0 i
he zi E + H n+ , (68)
2 2 0 2 0 2 n c
where hS0 i 12 E0 H0 .
The material stress is determined by closing the system. Inside the fluid,
the electromagnetic force is balanced by an equal and opposite force of the
material fe + fm 0. The material stress is written as fm fe rm . Con-
sidering the one-dimensional problem, the material stress at t 0 is
Z L Z L
@P
m z dzf e z^ dz 0 H z: ^ (69)
0 0 @t
Because the ramp is approximately linear, the @ P=@t term is constant over
the span [0, L] and can be approximated as P0v/L, where P0 E0(n2 1)E0.
The 0 H term is linear, so the integration yields 1 0 H
L. The material stress
2
reduces to

1
P0 0 1 S0
m z v H0 L n : (70)
L 2 2 n c
Taking the time-average in the region z < vt (i.e., once the time-harmonic
plane wave has been well established) yields

1 1 hS0 i
hm zi n : (71)
2 n c
464 Brandon A. Kemp
This gives the material response contribution to the average stress of a plane
wave. Adding the material and electromagnetic components gives the total
stress or momentum flow associated with the plane wave h(z)i nhS0i/c,
which is consistent with the Minkowski momentum. For comparison, eval-
uation of the Minkowski stress tensor in the region z < vt yields the total
stress associated with the fields and material response
1 nE o hS0 i
hMin zi E02 + 0 H02 n : (72)
2 2 2 c
3.3.3 Radiation Pressure

Jones and Richards (1954) and later Jones and Leslie (1978) measured the
pressure on submerged mirrors due to optical pressure as a function of refrac-
p
tive index n E=E0 . The conclusion of the JRL experiments is that the
observed pressure on the submerged mirror is proportional to n. The gen-
erality of this conclusion has since been questioned by Mansuripur (2007),
who correctly pointed out that the electromagnetic force on the mirror
depends upon the type of mirror used. The force on a submerged perfect
reflector indeed depends upon the phase of the reflection coefficient
Rmirror ei. For example, a perfect electrical conductor (PEC) reflects
with a phase and experiences an electromagnetic pressure of twice
the Minkowski momentum, while a perfect magnetic conductor (PMC)
reflects with a phase 0 and experiences an electromagnetic pressure of
twice the Abraham momentum. In this section, the stress of the material
response is added to the electromagnetic stress to deduce the observable
pressure of light on a submerged reflector with respect to a submerging fluid
(Kemp, 2012; Kemp & Grzegorczyk, 2011).
Consider a time-harmonic plane wave E xE ^ 0 eikz incident normally
from a dielectric liquid onto a perfect reflector at z 0. A small vacuum
gap of thickness d is introduced so that the reflector is at z d and the spatial
frequency inside the gap is k0 /c. For the case of a submerged mirror,
the air gap vanishes (d ! 0). Inside the dielectric (z < d), the fields form a
standing wave pattern
E1 r xE
^ i eikz + Reikz (73a)

1 r yÊi 1 eikz Reikz :
H (73b)

Application of the boundary conditions yield a unique solution for the fields.
The expression for R is (Kemp & Grzegorczyk, 2011)
2.5 |E1| = {1 + R }
|H1| = {1 R }
2
Fields
1.5
0.5
0
0 20 40 60 80 100 120 140 160 180
f ()
Figure 4 Normalized fields in a dielectric fluid at the surface of a perfect reflector as a
function of the phase of the mirror reflection coefficient (Rmirror ei. The electric field
(solid line) and magnetic field (dashed lines) are out of phase and are plotted as
jE1j {1 + R} and jH1j {1 R}, respectively.
n cos kd + =2 + i sin kd + =2
R , (74)
n cos kd + =2 i sin kd + =2
where d 0 for the submerged mirror case. The electric and magnetic fields
form out of phase interference patterns. Figure 4 shows that the electric field
is zero at the surface of a PEC ( ) while the magnetic field is zero at the
surface of a PMC ( 0).
The electromagnetic stress is determined from the Tzz component of the
time-averaged Chu or Einstein-Laub stress tensor. Substitution of the fields
in the dielectric region z < 0 yields an equation for the time-averaged elec-
tromagnetic stress in the dielectric fluid in terms of the index of refraction

1 1 i2kz hS0 i
he zi n + n Re : (75)
n n c
Here, hS0 i 12 jE0 j2 = 12 fE0 H0 g is the average incident power.

To close the system, the material response provides an equal and opposite
force to the electromagnetic force. The two subsystems can be written as
h fe zi hr Te zi (76a)

h fm zi hr Tm zi (76b)
466 Brandon A. Kemp
where T m represents the local material stress, and the momentum densities
vanish under time averaging. Since the fluid is nonmagnetic and the wave is
normally incident, the force reduces to a single term
1 1
h fe zi iP1 0 H
1 E E0 0 iE1 H
1 : (77)
2 2
The electromagnetic force acts only only in the
z^ direction, allowing the
use of scalar quantities. The material stress is the integral of the electromag-
netic force
Z
hm zi h fe zidz + h0 i: (78)
Here, the first term is the stress due to the interference of the standing wave
pattern and h0i is the stress due to the individual plane waves. The constant
h0i is twice the value given in Equation (71) since there are counter prop-
agating plane waves present, each contributing equally to the material
response stress. Therefore,

1 hS0 i
h0 i n , (79)
n c
and the time-average material stress reduces to

1 1 i2kz hS0 i
hm zi n + n Re : (80)
n n c
The total stress in the fluid is the sum of the electromagnetic stress in
Equation (75) and the material response stress in Equation (80). The sum,
hS0 i
htotal i he zi + hm zi 2n , (81)
c
gives twice the Minkowski momentum as the observed pressure on the mir-
ror with respect to the submerging fluid regardless of the phase of the reflec-
tion coefficient used. The contributions are plotted in Figure 5 for n 2. In
the case of the PEC reflector ( ) as used in the JRL experiment, the
material stress does not contribute. This is because the fields are zero at
the surface as shown in Figure 4, yielding zero material energy and momen-
tum contributions at z 0. Therefore, any formulation will directly yield
identical results for the pressure on the mirror, which means the JRL exper-
iment does not actually provide a true test for the momentum of light.
5
PMC (f = 0) PEC (f = 180)
4
Stress (N/m2)
3
EM
Material
2 Total
0
0 20 40 60 80 100 120 140 160 180
f ()
Figure 5 Stress in a dielectric fluid at the surface of a perfect reflector as a function of
the phase of the mirror reflection coefficient (Rmirror ei). The electromagnetic pres-
sure (triangles) and material pressure (circles) sum to give a constant value 2nhS0i/c for
the total stress regardless of reflector phase. The incident momentum is normalized
hS0i/c 1 and the index of refraction is n 2.
The PMC case ( 0) is different. Figure 4 shows that the electric field is
nonzero at the mirror surface, and the material contributes to both the local
energy and momentum of the wave. Adding these contributions restores
the Minkowski pressure to the reflector (Kemp, 2012; Kemp &
Grzegorczyk, 2011).
4. MOMENTUM AND STRESS IN DISPERSIVE MEDIA

Real materials exhibit frequency dispersion and losses. This is a
necessary condition of causality (Landau, Lifshitz, & Pitaevskii, 1984).
The inclusion of losses requires that a specific model for P and M be
applied. To arrive at a causal medium model for a dielectric, the equation
of motion for a bounded electron under the action action of an electric field
is applied such that
2
@ r @r
qE m 2 + e + 0r ,2
(82)
@t @t
where q is the electron charge, m is the mass of the electron, e is a damping
constant, 0 is the resonant frequency, and r is the local displacement.
468 Brandon A. Kemp
By defining the polarization P Nqr as the number density N of such oscil-

lators times the dipole moments qr , constitutive relations can be derived for a
Lorentz dielectric in the usual way (Kong, 2005). It has been shown in
Section 3 that the Chu formulation is a reasonable choice for the electro-
magnetic field subsystem. For the derivation of the wave energy in materials
exhibiting both dielectric and magnetic response (e.g., metamaterials), the
stationary approximation of the Chu formulation is usually chosen as the
starting point (Cui & Kong, 2004; Kong, 2005; Ramakrishna &
Grzegorczyk, 2008). Here, the same approach is taken for energy and
momentum. In this section, the dielectric and magnetic material response
are derived using a Lorentz medium model (Kemp, Kong, &
Grzegorczyk, 2007; Loudon, Allen, & Nelson, 1997).
4.1 Field and Material Contributions

The material response to the electromagnetic fields is described by the
differential equations for a Lorentz medium
2
@ @
+ e + e0 P E0 2ep E
2
(83a)
@t2 @t
2
@ @ F2 H,
+ m + m0 M
2
(83b)
@t2 @t mp
where the parameters of the equations have their usual meanings (Cui &
Kong, 2004). To derive the energy of the electromagnetic wave, the mate-
rial Equations (83a) and (83b) are dot multiplied by Je and Jh , respectively.
The resulting equations

1 @ @ P @ P 2 e @ P @ P

Je E + P P + (84a)
2E0 2ep @t @t @t e0
E0 2ep @t @t

@ @ M @ M M + m 0 @ M @ M
Jh H
0
+ m0 M
2
2Fmp @t @t
2 @t Fmp @t
2 @t
(84b)
are then added to the energy conservation equation of the electromagnetic
subsystem given by the Chu formulation. The resulting energy
conservation equation for the electromagnetic wave is in the form of
energy density W, and energy dissi-
Equation (3a) with the energy flow S,
pation given by
S E H (85a)

E0 0 1 @ P @ P 2
W E E + HH+ + e0 P P
2 2 2E0 2ep @t @t
(85b)
0 @M @M 2
+ + m0 M M
2F2mp @t @t
e @ P @ P m 0 @ M @M
+ : (85c)
E0 ep @t @t Fmp @t
2 2 @t
The energy flow S in Equation (85) retains its free-space form even in the
presence of a lossy, dispersive material. Also, the energy density W contains
contributions from the potential energy and kinetic energy of the electric
and magnetic dipoles. Furthermore, the form of Equation (85b) has been
regarded as significant since the energy density remains positive in negative
index of refraction media (Cui & Kong, 2004). Furthermore, the energy dis-
sipation term depends upon the damping factors e and m in
Equation (83). Therefore, 0 in the limiting case of a lossless material,
which indicates the energy of the electromagnetic wave is conserved.
The wave momentum can be derived by a similar method. The material
dispersion Equation (83) are dot multiplied by the dyads rP and 0 rM,
respectively. The resulting vector equations are then added to the electro-
magnetic continuity equation given by the Chu formulation to yield
eh
@G
r Teh + + f + E rP + 0 H M
@t eh
2
@ 2 P 1 @ M 2 0
rP + 2
P r M + M
@t2 e0
E0 2ep 0
@t2 m0
F2mp
@ P e
rP + r M @ M m 0 :
@t E0 2ep @t F2mp
(86)
Vector calculus allows terms to be rewritten as
2
"
#
@ P 1 1 @ P @
P
rP + e0 P
2
r e0 P P
2
@t2 E0 2ep 2E0 2ep @t @t
" #
@ 1 @
P
rP
@t E0 2ep @t
(87a)
470 Brandon A. Kemp

"
#
2
@ M @ M @ M M

rM + m0 M
2 0
r 0
m0 M
2
@t 2 F2mp 2F2mp @t @t
" #
@ 0 @ M
rM
@t F2mp @t
(87b)

f eh + E rP + 0 H M r P E + 0 M H I P E 0 M H
@
+ D B E0 0 E H :
@t
(87c)
By combining Equation (86) and Equation (87), the momentum continuity
equation for the wave can be expressed in the form of Equation (3b) with the
stress tensor, momentum density, and force density
"
1 1 1 @ P @ P 2
T D E + B HI D E B H + P P
2 2 E0 2ep @t @t e0

@M @M M + P E + M H
I (88a)
+ 02 2m0 M
Fmp @t @t 0

D B 1 rP @ P 0 rM
G @M (88b)
E0 ep
2 @t Fmp 2 @t
@ P m 0
f e 2 rP 2 rM @M : (88c)
E0 ep @t Fmp @t
The expressions in Equation (88) define the quantities for the momentum
continuity equation, which complete the wave subsystem along with the
energy expressiongs in Equation (85). The momentum density G contains
the Minkowski momentum D B plus material dispersion terms. Likewise,
the momentum flow T is the Minkowski stress tensor plus dispersive terms.
Note that the momentum dissipation term f depends upon the damping
factors e and m in Equation (83). Therefore, f 0 in the limiting case
of a lossless, unbounded material, which indicates that the momentum of
the electromagnetic wave is conserved.
4.2 Time-Averaged Energy and Momentum Propagation

Consider the propagation of time-harmonic electromagnetic waves or
electromagnetic fields which are contained within a narrow frequency band.
The constitutive parameters follow directly from Equation (83) given eit
dependence (Kong, 2005). The complex permittivity and permeability,
!
2ep
E E0 1 2 (89a)
2e0 + i e
!
F2mp
0 1 2 , (89b)
2m0 + i m
are functions of the frequency and consist of real and imaginary parts den-
oted by E ER + iEI and R + iI. Likewise, the time-average of the
squared polarization and magnetization are
E20 4ep
2
jPj 2
jEj (90a)
2
e0 + e
2 2 2 2
F 2 4mp
j2
jM jH j2 : (90b)
2
m0 + m
2 2 2 2
Similarly, j@ P=@tj2 2 jPj

2 and j@ M =@tj2 2 jM j2 , which can be
applied to determine the time-average values relating to energy and
momentum conservation given in the previous section.
The time average energy density found from Equation (85b) is
" #
E0 2ep 2 + 2e0
hW i 1+ 2 jEj2
2 e0 + e
2 2 2 2
" # (91)
0 F2mp 2 + 2m0
+ 1+ 2
jH j :
2
2 2 2m0 + 2m 2
In the lossless case, e 0 and m 0 implies that both EI 0 and I 0, and
the energy density satisfies the well-known relation (Brillouin, 1960;
Jackson, 1999; Kong, 2005; Landau et al., 1984)
1 @E 2 1 @ 2
hW i jE j + jH j : (92)
4 @ 4 @
The extension of Equation (92) to lossy materials has been criticized due to
the possibility of negative values for negative refractive index metamaterials
(i.e., materials with n < 0). In contrast, the average energy density in
Equation (91) remains positive for all . However, it is the rate of change
in energy that appears in the energy continuity equation, which tends to
472 Brandon A. Kemp
zero upon cycle averaging. That is, h@W/@ti 0, and the resulting conser-
vation equation
1 2 + I jH
j2

hr Si EI jEj (93)
2
is generally regarded as the complex Poyntings theorem, where
g is the time average Poynting power.
1 fE H
hSi 2
Similarly, the average momentum density,
(
1 E0 2ep

hGi D B + k 2
jEj
2 e0 + e
2 2 2 2 2
) (94)
0 Fmp 2
+ k j2 ,
jH
2 2m0 2 + 2m 2
is obtained from Equation (88b). It is simple to show using Equation (89)
that the average momentum given in Equation (94) satisfies (Veselago, 1968)
@E 2 @

1
EE H
+ k + j
hGi jEj jH 2
(95)
2 2 @ @
when the medium is lossless. The time average stress tensor is

1 1
hT i D E + B H I D E B H (96)
2 2
since the dispersive terms in Equation (88a) tend to zero upon cycle aver-
aging. Since the average rate of change in momentum density is zero (i.e.,

h@ G=@ti 0), the momentum conservation theorem for a monochromatic
wave reduces to
1
hr T i fEI E B I H g
D (97)
2
where the tensor is the complex Minkowski tensor. The right-hand side of
Equation (97) is defined as the force density on free currents by Loudon,
Barnett, and Baxter (2005) and Kemp, Grzegorczyk, and Kong (2006b).
5. DISCUSSION
Several conclusions can be made from the preceding sections. Each
conclusion adds to the logical deduction of how optical manipulation
experiments can be modeled within the macroscopic theory of electrody-

namics. These conclusions are listed in the following section followed by
a brief discussion of their applicability to optical manipulation.
5.1 Conclusions
The conclusions are ordered by section number so that a logical flow is pres-
ented and reference to the supporting analysis is maintained.
Conclusions 1: A number of electromagnetic force density equations
persist in the optics literature. Each is tied to a particular formulation of elec-
trodynamics. The most commonly used force equations are considered
herein with the goal of modeling modern experiments involving the manip-
ulation of macroscopic media with light.
Conclusions 2.1: Unique formulations of electrodynamics differ in
how matter is modeled with each resulting in various forms for the electro-
magnetic energy and momentum continuity equations. Within media, the
force density, momentum density, and stress tensor are determined by which
representation of Maxwells theory is applied. When a medium is in motion,
the field vectors are also unique to a particular formulation. In vacuum, the
fields as well as the energy and momentum continuity equations are
unambiguous.
Conclusions 2.2: A quasi-stationary approximation can be applied that
allows for the study of optical momentum transfer to media while optical
energy is conserved. This approximation is common and greatly reduces
the complexity involved with modeling optical manipulation experiments.
Essentially, the Doppler shift of the electromagnetic wave due to material
motion is ignored such that calculations predict momentum exchange
due to reflection or scattering, but optical energy is conserved.
Conclusions 2.3: The most commonly used electromagnetic force
equations assume stationary media, and are each associated with a momen-
tum continuity equation. The sum of the force density, divergence of the
stress tensor, and rate of change of the momentum density must be zero.
Under the quasi-stationary approximation, the field variables are equivalent
among the formulations in vacuum and in media. Terms may be rearranged
between the force density, momentum density, and stress tensor. Such exer-
cises define new electromagnetic subsystems, but the resulting energy and
momentum continuity equations are left for interpretation.
Conclusions 3.1: Any formulations which share a common momen-
tum density will produce identical results for the total electromagnetic force
474 Brandon A. Kemp
on a material object at any point in time, and all formulations produce iden-
tical values for the total time-average electromagnetic force on a material
object. This statement is true as long as the momentum continuity equations
are consistent with Maxwells equations. Additionally, it is impossible to
infer a time-domain force from a time-averaged force.
Conclusions 3.2: The total calculated electromagnetic force on an
object due to an optical pulse of finite extent is due to the time rate of change
of the momentum contained within the pulse. The momentum contained
within a pulse depends upon the momentum density that appears in the
momentum continuity equation. The Abraham momentum gives the
kinetic momentum of light, which is responsible for center of mass transla-
tions of media. Any subsystem that is to be considered as consisting of field
only quantities must include the Abraham momentum density.
Conclusions 3.3: The total average electromagnetic force on an object
due to time-harmonic fields is due to the divergence of the average stress
contained within the fields. The Chu and Einstein-Laub formulations
include the Abraham momentum density in the momentum continuity
equations. Both predict conservation of momentum flow due to normal
incident transmission and reflection at a boundary, both reveal that a plane
electromagnetic wave includes a nonzero average material response stress in
addition to the electromagnetic stress, and both predict an additional mate-
rial contribution to radiation pressure when an object is submerged in a
dielectric fluid. In regards to the last point, the observed optical force on
a submerged object is due to the electromagnetic force on the object and
the electromagnetic force on the submerging fluid. The pressure difference
at the boundary yields the Minkwoski momentum imparted to the sub-
merged object with respect to the submerging fluid, which is key to reduc-
ing the complexity of modeling optical manipulation experiments.
Conclusions 4.1: Causality requirements for media to be lossy and dis-
persive necessitate the inclusion of material dynamics in the wave energy and
momentum continuity equations. Starting with the Chu formulation
and modeling the material response as harmonic oscillators, the energy
and momentum response of the material can be added to the electromagnetic
energy and momentum continuity equation. The resulting wave energy den-
sity, power flux, momentum density, and stress tensor take the Minkowski
forms plus additional material dynamics terms.
Conclusions 4.2: Under time-harmonic excitation, the time-averaged
stress tensor for the electromagnetic wave, which includes both electromag-
netic and material response, reduces to the time-average Minkowski stress
tensor. Therefore, Conclusions 3.2 are valid for dispersive media when the
optical excitation has a narrow bandwidth. Otherwise, the conclusions hold,
but the Minkwoski stress tensor must be supplemented with additional terms
due to material dispersion.
5.2 Application to Optical Manipulation

The conclusions are applied to model the forces exerted on particles by an
optical field. The modeling process requires two steps. First, the total fields
must be found, which can be done by numerical or analytical techniques.
Second, the force on the particle is calculated. Here, cylinders and spheres
are used to represent particles in two-dimensions and three-dimensions,
respectively. The analytical field solutions for cylinders and spheres are
provided in Appendix A and Appendix B, respectively.
A cylindrical particle is used as a computationally efficient way to
model physical phenomena. A laser can be modeled in two dimensions
by a Gaussian beam (Shen & Kong, 1987)
Z1 w 2 k2x
dkx e
w
Einc zE
^ 0 p 4 eikx x + iky y , (98)
2
1
where k2y k2b k2x , k2b 2 b Eb , and w is the beam waist. The electric field
in Equation (98) satisfies the wave equation (Kong, 2005) and is
implemented as a discrete sum of plane waves. The spectrum is actually lim-
ited, and the sum is implemented here within the domain jkx j NA c ,
where NA is the numerical aperture.
Figure 6 illustrates a model of the original experiments by Ashkin (1970).
The arrows depict the observed force on either a polystyrene particle or air
bubble in water due to a Gaussian beam described in Equation (98). The
divergence of the Minkowski stress tensor in the background water is
applied to calculate the time-average force on the 2D particle (i.e., cylinder)
at different points. This approach is consistent with the results of Sections 3.3
and 4.2. The forces in Figure 6 agree with the observed forces by Ashkin
(1970); radiation pressure on both particles is accompanied by a gradient
force that attracts the higher index of refraction particles (polystyrene) into
the center of the beam and repels the lower refractive index air bubbles
out of the center of the beam. Since the air bubbles have a polarization
of P 0, it is actually the kinetic force that pulls the water into the high field
intensity region of the beam. However, the pressure gradient that is
476 Brandon A. Kemp
Figure 6 Force per unit length (represented by the arrows) on a single infinite cylinder
due to an incident Gaussian beam with NA 1, E0 1, w 0.50, and wavelength
0 1064 nm. The background medium is water Eb 1.69E0 and the radius of the par-
ticle is a 0.50. The particle is (a) polystyrene and (b) air.
Figure 7 Classical optical tweezers modeled as a 2D dielectric particle in water. The

arrows represent the force on a silica particle with index of refraction np 1.46 and
radius a 50 nm in water (nb 1.33). The incident laser is modeled as a Gaussian beam
with waist w 0.2 0, 0 514.5 nm, and spectrum defined by NA 1.25.
responsible for manipulating the air bubble can be equivalently computed

using Minkowskis tensor, as demonstrated here.
Figure 7 illustrates classical optical tweezers in two dimensions as origi-
nally demonstrated by Ashkin et al. (1986). The Gaussian beam is focused
such that the so-called gradient force overcomes the scattering force. The
parameters match closely the experimental conditions used in the original
experiments to trap silica beads. Therefore, the modeling approach again
matches the experimental outcome as the optical force is seen to produce
a stable trap near the beam focus.
Optical manipulation of colloidal particles is typically achieved by crea-
tion of an optical intensity gradient as seen with optical tweezers. One such
configuration consists of three lasers with 2/3 incident angle separation
(Fournier et al., 2004). Such a configuration results in a zero net incident
momentum since the vector sum of the incident momentum vectors is zero,
which allows for isolation of the gradient force. This interference is repre-
sented by the background of Figure 8, where white denotes regions of
high field intensity. The force on a 2D polystyrene particle in water is cal-
culated via the divergence of the Minkowski stress tensor as prescribed in
Sections 3.3 and 4.2. Each arrow represents the direction and relative
magnitude of force on a dielectric cylinder placed at the tail of the arrow.
The force on the small particle with radius a 0.150 is seen to closely follow
the high intensity gradient of the incident field with stable optical traps
occurring in the high intensity regions. However, larger particles with radius
a 0.300 are repelled from the high intensity regions and find stable trap-
ping positions in the dark regions of the incident optical interference pattern.
Obviously, this result contrasts the simple description given by the gradient
force obtained in the Rayleigh approximation. The example demonstrates
that such an approximation yields results which are both quantitatively and
qualitatively incorrect for particles that are on the order of a wavelength. It is
also interesting to study the intermediate particle size regime. For example,
the force on a particle with radius a 0.2250 is also plotted over the back-
ground of the incident three plane wave interference pattern in Figure 8. An
interesting feature of this plot is that the force field appears to be non-
conservative, and a prediction of stable optical trapping in these vortices
has been reported by Grzegorczyk and Kong (2007).
A concluding example demonstrates the dependence of radiation pressure
on the background fluid. Consider a spherical particle surrounded by either
vacuum or a negative refractive index medium with negative vacuum param-
eters (E0, 0). The radiation pressure is reversed according to the theory of
Kemp et al. (2007), which is reproduced in Section 4. Figure 9 shows the
force on a sphere as a function of radius a, which is varied from the Rayleigh
regime to the Mie regime. The fields are calculated by the analytical approach
detailed in Appendix B. The results show that the force on the sphere
478 Brandon A. Kemp
Figure 8 Force per unit length (represented by the arrows) on a single infinite cylinder
due to the interference of three plane waves (represented by the background pattern)
of equal amplitude Ei 1 V/m and wavelength 0 532 nm. The incident angles of
the plane waves are {/2, 7/6, 11/6}. The background medium is water Eb 1.69E0,
and the cylinder is polystyrene Ep 2.56E0 with radius (a) a 0.150, (b) a 0.300,
and (c) a 0.2250.
depends upon the refractive index of the background medium. The particle
in vacuum is pushed, while the radiation pressure is reversed when the back-
ground refractive index is negative as first predicted by Veselago (1968).
APPENDIX A. SCATTERING BY A CYLINDER

The geometry of the problem consists of an electromagnetic wave
incident upon an infinite cylinder of radius a aligned in the z direction.
1012
2
1.5
0.5 (eb,mb) = (e0,m0)

Force (N)
0 (eb,mb) = (e0,m0)
0.5
1.5
2
50 100 150 200 250 300 350 400 450 500
a (nm)
Figure 9 Radiation force on a spherical particle with respect to a submerging back-
ground medium as a function of sphere radius a. The particle has is nonmagnetic with
E 4E0. The solid line gives the results for a background medium of vacuum (E0, 0), and
the dashed line depicts the force for a hypothetical inverted vacuum (E0, 0). The
incident field is a plane wave with amplitude E0 1 V/m and vacuum wavelength
0 1064 nm.
The cylinder is characterized by (p, Ep) and the background by (b, Eb).
The incident wave is assumed to be a plane wave with eit dependence.
Many other field distributions, such as a Gaussian beam, can be described
by a sum of plane waves. Therefore, the total solution for such incident fields
can be described as a superposition of solutions resulting from a plane-wave.
The incident, scattered, and internal fields are expanded in cylindrical waves
given by N n, M
n , RgN
n , and RgM
n . The solution is given by Tsang, Kong,
and Ding (2000).
The incident electric field is polarized in the z-direction^ and
propagates in the plane (kz 0). The magnetic fields are obtained from
Faradays law
r E
iH (A.1)
using the identities
rM n
n kN (A.2a)
n kM
rN n: (A.2b)
480 Brandon A. Kemp
The incident fields are regular at the origin and are given by

X
N
Einc zE
^ 0 eik i n kb ,
an RgN (A.3a)
nN
kb X N
inc
H n kb ,,
an RgM (A.3b)
ib nN
where the wavenumber in the background medium is given by the disper-

sion relation k2b 2 b Eb . The scattered fields are
X
N
Escat n kb ,
asn N (A.4a)
nN
kb X N
scat
H n kb ,:
as M (A.4b)
ib nN n
The internal fields are also regular at the origin and can be computed from
X
N
Eint n kp ,
cn RgN (A.5a)
nN
kp X N
int
H n kp , ,
cn RgM (A.5b)
ip nN
where the wavenumber in the cylinder ( < a) is given by k2p 2 p Ep . The

unknown coefficients asn ,cn are determined by matching the boundary
conditions and the coefficients (an) are known from the incident plane wave
expansion Tsang et al. (2000). The expressions for incident, scattered,
and internal fields become exact for N ! 1. The vector cylindrical wave
functions for this particular incidence are given by Stratton (1941)
n k, zkH
N ^ n1 kein (A.6a)

n 1
M n k, ^ i Hn k ein
(A.6b)
n
+ ^ kHn + 1 k Hn1 k ein
1

RgN n k, zkJ
^ n kein (A.6c)

n
RgM n k, ^ i Jn k ein
(A.6d)
n
+ ^ kJn + 1 k Jn k ein :

Here, Hn1 is the Hankel function of the first kind and Jn() is the Bessel
function. The coordinates (, , ) represent the point for field evaluation
(i.e., the observer position). The angle i is used to represent the incident
direction of the illuminating wave.
For a general dielectric and magnetic medium, the boundary conditions
give a 2 2 system of equations that must be solved for the unknown
coefficients,
eini
an in E0 (A.7a)
kb
b1 m22 b2 m12
asn (A.7b)
m11 m22 m21 m12
b1 m21 b2 m11
cn : (A.7c)
m12 m21 m22 m11
The matrix elements and right-hand side for the linear system are
m11 kb Hn1 kb a (A.8a)

m12 kp Jn kp a (A.8b)
1 n
m21 kb Hn + 1 kb a + Hn1 kb a (A.8c)
a
b kp h n i
m22 kp Jn + 1 kp a Jn kp a (A.8d)
p kb a
b1 an kb Jn kb a (A.8e)
h n i
b2 an kb Jn + 1 kb a Jn kb a : (A.8f)
a
APPENDIX B. SCATTERING BY A SPHERE

The solution for a plane wave scattering from a Sphere is referred to as
Mie theory and the solution is given by Kong (2005). The incident electric
^
field is polarized in the x-direction ^
and propagates in the z-direction.
The background is characterized by (Eb, b) with a wavenumber kb.
482 Brandon A. Kemp
The particle is composed of (Ep, p) with wavenumber kp. The incident

fields are
Einc xE
^ 0 eikb z xE
^ 0 eikb rcos (B.1a)
inc yÊ0 eikb z yÊ0 eikb rcos :

H (B.1b)
b b
The solution is found by expanding the incident, scattered, and internal
fields as a sum of spherical waves. The unknown coefficients for the scattered
and internal fields are found from the boundary conditions.
The incident wave is expanded in spherical modes using the identity
X
N
eikrcos in 2n + 1jn krPn cos , (B.2)
n0
where jn(kr) is a spherical Bessel function and Pn cos is the Legendre func-
tion. The approximation is exact as N ! 1.
The incident, scattered, and internal waves are decomposed into TM to
r^ and TE to r^ waves by using scalar potentials e and m respectively.
The potentials are defined by
A r e (B.3a)
H ^ 1 @ e ^ @ e
rA (B.3b)
sin @ @
for TM waves and
Z r m (B.4a)
1 @ @
E r Z ^ m ^ m (B.4b)
sin @ @
for TE waves. The potentials satisfy the Helmholtz equation in spherical
coordinates
r2 + k2 0: (B.5)
The scaler potentials are
X
N
e An zn krPn1 cos cos (B.6a)
n1
X
N
m Bn zn krPn1 cos sin , (B.6b)
n1
where zn(kr) represents solutions to the Bessel equation. The solutions differ
in the three regions by the following.
Incident ( e, m)
zn kr jn kb r
kb in 2n + 1
An E0
b nn + 1
b
Bn An
kb
Scattered ( se , sm )
zn kr hn kb r
kb
A n E0 an
b
Bn E0 bn
hn kb r h1
n kb r is the spherical Hankel function of the first kind.
internal ( ie , im )
zn kr jn kp r
kp
An E0 cn
p
Bn E0 dn
The unknown coefficients an, bn, cn, dn are found using the boundary con-
ditions for the tangential electric field and tangential magnetic field.
The total fields are:
k XN
Er i An Pn1 cos krzn kr + 2z0n kr + krz00n kr cos (B.10a)
E n1
XN
k @ 1 1 0
E iAn P cos zn kr + zn kr (B.10b)
n1
E @ n kr

P 1 cos
+ Bn n zn kr cos (B.10c)
sin
484 Brandon A. Kemp
N
X
k Pn1 cos 1
E iAn zn kr + z0n kr (B.10d)
n1
E sin kr

@ 1
+ Bn Pn cos zn kr sin (B.10e)
@
k XN
Hr i Bn Pn1 cos krzn kr + 2z0n kr + krz00n kr sin (B.10f)
n1
XN
k @ 1 1 0
H iBn P cos zn kr + zn kr (B.10g)
n1
@ n kr

Pn1 cos
An zn kr sin (B.10h)
sin
XN
k Pn1 cos 1 0
H iBn zn kr + zn kr (B.10i)
n1
sin kr

@
+ An Pn1 cos zn kr cos: (B.10j)
@
The spherical Bessel and Hankel functions zn() are given in terms of the
Bessel and Hankel functions Zn() by
r

zn Z 1 : (B.11)
2 n + 2
To evaluate the derivatives, it is helpful to define l n + 2, m n + 1,

p n 1, and q n 2 and use the relation Stratton (1941)
n m
z0n zp zm : (B.12)
2n + 1 2n + 1
The second derivative with respect to the argument is
n
z00n pzq nzn
2n + 12p + 1
m (B.13)
mzn + lzl :
2n + 12m + 1
The derivative of the Legendre function with respect to the argument is, in
general, known. Therefore, a useful relation is
@ m @
Pn cos sin P m cos : (B.14)
@ @cos n
The coefficients as determined from the boundary conditions are

0 0
kb Ep J^n kp aJ^n kb a kp Eb J^n kb aJ^n kp a
an gn 0 0 (B.15a)
kp Eb H^ n kb aJ^n kp a kb Ep J^n kp aH^ n kb a
0 0
kb p J^n kp aJ^n kb a kp b J^n kb aJ^n kp a
bn gn 0 0 (B.15b)
kp b H^ n kb aJ^n kp a kb p J^n kp aH^ n kb a
ikb Ep p
cn gn 0 0 (B.15c)
kb Ep b J^n kp aH^ n kb a kp Eb p H^ n kb aJ^n kp a
ikp p
n gn 0 0 (B.15d)
kb p J^n kp aH^ n kb a kp b H^ n kb aJ^n kp a
where
in 2n + 1
gn : (B.16)
nn + 1
1
The RicattiBessel functions J^n and H^ n H^ n are
Z^n zn (B.17a)
0
Z^n zn + z0n : (B.17b)
ACKNOWLEDGMENTS
The author is grateful to Masud Mansuripur helpful discussions and, in particular, for pointing
out discrepancies in the literature regarding the Abraham stress tensor. The author
acknowledges Cheyenne J. Sheppard for proofreading parts of the manuscript.
This work was sponsored in part by the National Science Foundation EECS Division of
Electrical, Communications, and Cyber Systems (award number ECCS-1150514). The
authors would like to thank the Arkansas Science and Technology Authority for funding
provided to the Center for Efficient and Sustainable Use of Resources (CESUR).
REFERENCES
Abraham, M. (1909). Zur Elektrodynamik der bewegter K orper. Rendiconti del Circolo Mate-
matico di Palermo, 28, 128.
Abraham, M. (1910). SullElettrodinamica di Minkowski. Rendiconti del Circolo Matematico di
Palermo, 30, 3346.
Ashkin, A. (1970). Acceleration and trapping of particles by radiation pressure. Physical
Ashkin, A. (1978). Trapping of atoms by resonance radiation pressure. Physical Review Letters,
40, 729732.
Ashkin, A. (1997). Optical trapping and manipulation of neutral particles using lasers.
Proceedings of the National Academy of Sciences, 94, 48534860.
486 Brandon A. Kemp
Ashkin, A. (2000). History of optical trapping and manipulation of small-neutral particle,

atoms, and molecules. IEEE Journal of Selected Topics in Quantum Electronics, 6, 841856.
Ashkin, A., & Dziedzic, J. M. (1971). Optical levitation by radiation pressure. Applied Physics
Letters, 19, 283285.
Ashkin, A., & Dziedzic, J. M. (1973). Radiation pressure on a free liquid surface. Physical
Ashkin, A., & Dziedzic, J. M. (1976). Optical levitation in high vacuum. Applied Physics
Letters, 28, 333335.
Ashkin, A., & Dziedzic, J. M. (1987). Optical trapping and manipulation of viruses and
bacteria. Science, 235, 15171520.
Ashkin, A., Dziedzic, J. M., Bjorkholm, J. E., & Chu, S. (1986). Observation of a single-
beam gradient force optical trap for dielectric particles. Optics Letters, 11, 288290.
Ashkin, A., & Gordon, J. P. (1979). Cooling and trapping of atoms by resonance radiation
pressure. Optics Letters, 4, 161163.
Balazs, N. (1953). The energy-momentum tensor of the electromagnetic field inside matter.
Physical Review, 91, 408411.
Barnett, S. M. (2010). Resolution of the Abraham-Minkowski dilemma. Physical Review
Letters, 104, 070401.
Barnett, S. M., & Loudon, R. (2006). On the electromagnetic force on a dielectric medium.
Journal of Physics B: Atomic, Molecular and Optical Physics, 39, S671S684.
Barnett, S. M., & Loudon, R. (2010). The enigma of optical momentum in a medium.
Philosophical Transactions of the Royal Society A, 368, 927939.
Baxter, C., & Loudon, R. (2010). Radiation pressure and the photon momentum in dielec-
trics. Journal of Modern Optics, 57, 830842.
Brevik, I. (1979). Experiments in phenomenological electrodynamics and the electromag-
netic energy-momentum tensor. Physics Reports, 52, 133201.
Brillouin, L. (1960). Wave propagation and group velocity. New York, NY: Academic Press.
Burns, M. M., Fournier, J. M., & Golovchenko, J. (1989). Optical binding. Physical Review
Letters, 63, 12331236.
Burns, M. M., Fournier, J. M., & Golovchenko, J. (1990). Optical matter: Crystallization and
binding in intense optical fields. Science, 249, 749754.
Chu, S., Bjorkholm, J. E., Ashkin, A., & Cable, A. (1986). Experimental observation of
optically trapped atoms. Physical Review Letters, 57, 314318.
Chu, S., Bjorkholm, J. E., Ashkin, A., Gordon, J. P., & Holberg, L. (1986). Proposal for
optically cooling atoms to temperatures of the order of 106 K. Optics Letters, 11, 7375.
Chu, S., Holberg, L., Bjorkholm, J. E., Cable, A., & Ashkin, A. (1985). Three-dimensional
viscous confinement and cooling of atoms by resonance radiation pressure. Physical
Cui, T. J., & Kong, J. A. (2004). Time-domain electromagnetic energy in a frequency-
dispersive left-handed medium. Physical Review E, 70, 205106.
Daly, P., & Gruenberg, H. (1967). Energy relations for plane waves reflected from moving
media. Journal of Applied Physics, 38, 44864489.
Dholakia, K., & Zemnek, P. (2010). Colloquium: Gripped by light: Optical binding. Reviews
of Modern Physics, 82(2), 17671791.
Einstein, A., & Laub, J. (1908). Annalen der Physik, 331, 541550.
Fournier, J. M., Boer, G., Delacetaz, G., Jacquot, P., Rohner, J., & Salathe, R. P. (2004).
Building optical matter with binding and trapping forces. Proceedings of SPIE, 5514, 309.
Grier, D. G. (2003). A revolution in optical manipulation. Nature, 424, 810816.
Griffiths, D. J. (2012). Resource letter EM-1: Electromagnetic momentum. American Journal
of Physics, 80, 718.
Griffiths, D. J., & Hnizd, V. (2013). Mansuripurs paradox. American Journal of Physics, 81,
570574.
Grzegorczyk, T. M., & Kemp, B. A. (2008). Transfer of optical momentum: Reconciliations

of the Abraham and Minkowski formulations. Proceedings of SPIE, 7038, 70381S.
Grzegorczyk, T. M., & Kong, J. A. (2007). Analytical prediction of stable optical trapping in
optical vortices created by three TE and TM plane wave incidences. Optics Express, 15,
80108020.
Jackson, J. D. (1999). Classical electrodynamics. New York, NY: John Wiley and Sons.
Jones, R. V., & Leslie, B. (1978). The measurement of optical radiation pressure in dispersive
media. Proceedings of the Royal Society of London A, 360, 347363.
Jones, R. V., & Richards, J. C. S. (1954). The pressure of radiation in a refracting medium.
Proceedings of the Royal Society of London A, 221, 480498.
Kemp, B. A. (2011). Resolution of the Abraham-Minkowski debate: Implications for the
electromagnetic wave theory of light in matter. Journal of Applied Physics, 109, 111101.
Kemp, B. A. (2012). Subsystem approach to the electrodynamics in dielectric fluids. Proceed-
ings of SPIE, 8458, 845803.
Kemp, B. A. (2013). Comment on Revisiting the Balazs thought experiment in the pres-
ence of loss: Electromagnetic-pulse-induced displacement of a positive-index slab having
arbitrary complex permittivity and permeability. Applied Physics A: Materials Science and
Processing, 110(2), 517518.
Kemp, B. A., & Grzegorczyk, T. M. (2011). The observable pressure of light in dielectric
fluids. Optics Letters, 36, 493495.
Kemp, B. A., Grzegorczyk, T. M., & Kong, J. A. (2005). Ab initio study of the radiation
pressure on dielectric and magnetic media. Optics Express, 13, 92809291.
Kemp, B. A., Grzegorczyk, T. M., & Kong, J. A. (2006). Lorentz force on dielectric and
magnetic particles. Journal of Electromagnetic Waves and Applications, 20, 827839.
Kemp, B. A., Grzegorczyk, T. M., & Kong, J. A. (2006). Optical momentum transfer to
absorbing Mie particles. Physical Review Letters, 97, 133902.
Kemp, B. A., Kong, J. A., & Grzegorczyk, T. M. (2007). Reversal of wave momentum in
isotropic left-handed media. Physical Review A, 75, 053810.
Kong, J. A. (2005). Electromagnetic wave theory. Cambridge, MA: EMW Publishing.
Landau, L. D., Lifshitz, E. M., & Pitaevskii, L. P. (1984). Electrodynamics of continuous media
(2nd ed.). New York, NY: Pergamon.
Lebedew, P. (1901). Pressure of light. Annals of Physics, 6, 433458.
Leonhardt, U. (2006). Momentum in an uncertain light. Nature, 444, 823824.
Livens, G. H. (1918). The theory of electricity. Cambridge: Cambridge University Press.
Loudon, R. (2004). Radiation pressure and momentum in dielectrics. Fortschritte der Physik,
52, 11341140.
Loudon, R., Allen, L., & Nelson, D. F. (1997). Propagation of electromagnetic energy and
momentum through an absorbing dielectric. Physical Review E, 55, 10711985.
Loudon, R., & Barnett, S. M. (2006). Theory of radiation pressure on dielectric slabs, prisms,
and single surfaces. Optics Express, 14, 1185511869.
Loudon, R., Barnett, S. M., & Baxter, C. (2005). Radiation pressure and momentum transfer
in dielectrics: The photon drag effect. Physical Review A, 71, 063802.
Mansuripur, M. (2004). Radiation pressure and the linear momentum of the electromagnetic
field. Optics Express, 12, 53755401.
Mansuripur, M. (2007). Radiation pressure on submerged mirrors: Implications for the
momentum of light in dielectric media. Optics Express, 15, 26772682.
Mansuripur, M. (2008). Electromagnetic force and torque in ponderable media. Optics
Express, 16, 1482114835.
Mansuripur, M. (2012). Trouble with the Lorentz law of force: Incompatibility with special
relativity and momentum conservation. Physical Review Letters, 108, 193901.
Mansuripur, M., Zakharian, A. R., & Wright, E. M. (2013). Electromagnetic-force distri-
bution inside matter. Physical Review A, 88, 023826.
488 Brandon A. Kemp
Maxwell, J. C. (1891). A treatise on electricity and magnetism. London: Constable.

Milonni, P. W., & Boyd, R. W. (2010). Momentum of light in a dielectric medium. Advances
in Optics and Photonics, 2, 519553.
Minkowski, H. (1908). Die Grundgleichungen f ur die elektromagnetischen Vorgange in
bewegten K orpern. Nachrichten von der Gesellschaft der Wissenschaften zu G
ottingen, 1, 53111.
Nelson, D. F. (1991). Momentum, pseudomomentum, and wave momentum: Toward
resolving the Minkowski-Abraham controversy. Physical Review A, 44, 39853996.
Nichols, E. F., & Hull, G. F. (1903). The pressure due to radiation. Physics Review, 17, 2650.
Penfield, P., & Haus, H. A. (1967). Electrodynamics of moving media. Cambridge, MA: M.I.T.
Press.
Pfeifer, R. N. C., Nieminen, T. A., Heckenberg, N. R., & Rubinsztein-Dunlop, H. (2007).
Momentum of an electromagnetic wave in dielectric media. Reviews of Modern Physics,
79, 11971216.
Poynting, J. H. (1905). Tangential stress of light obliquely incident on absorbing surface. Phil-
osophical Magazine, 24, 169171.
Ramakrishna, S. A., & Grzegorczyk, T. M. (2008). Physics and applications of negative refractive
index materials. Boca Raton, FL: CRC Press.
Saldanha, P. L. (2013). Comment on Trouble with the Lorentz law of force: Incompatibility
with special relativity and momentum conservation Physical Review Letters, 110,
089403.
Scalora, M., Aguanno, G. D., Mattiucci, N., Bloemer, M. J., Centini, M., Sibilia, C., et al.
(2006). Radiation pressure of light pulses and conservation of linear momentum in dis-
persive media. Physical Review E, 73, 056604.
Scullion, M. G., & Barnett, S. M. (2008). Optical momentum in negative-index media. Jour-
nal of Modern Optics, 55, 23012309.
Shen, L. C., & Kong, J. A. (1987). Applied electromagnetism (3rd ed.). Boston, MA: PWS
Publishing.
Sheppard, C. J., & Kemp, B. A. (2014). Optical pressure deduced from energy relations
within relativistic formulations of electrodynamics. Physical Review A, 89(1), 013825.
Shockley, W., & James, R. P. (1967). Try simplest cases discovery of hidden
momentum forces on magnetic currents. Physical Review Letters, 18, 876879.
Stratton, J. A. (1941). Electromagnetic theory. New York, NY: McGraw-Hill.
Tsang, L., Kong, J. A., & Ding, K. H. (2000). Scattering of electromagnetic waves: Theory and
applications. New York, NY: John Wiley and Sons.
Veselago, V. G. (1968). The electrodynamics of substances with simultaneously negative
values of E and . Soviet Physics Uspekhi, 10, 509514.
CHAPTER SIX
Tunneling of Electromagnetic
Waves in All-Dielectric Gradient
Metamaterials
Alexander B. Shvartsburg1,2,4, Yuri A. Obod3, Oleg D. Volpian3
1
Joint Institute for High Temperatures Russian Academy of Sciences, Moscow, Russia
2
Institute for Space Researches Russian Academy of Sciences, Moscow, Russia
3
4
Far Eastern Federal University, Vladivostok, Russia
Contents
1. Introduction: Nonlocal Dispersion of Waves in Transparent Gradient Dielectrics 490
2. Frustrated TIR from Gradient Dielectric Layers 495
2.1 Evanescent Waves in Uniform Photonic Barriers 496
2.2 TIR of Light from a Gradient Half-Space 498
2.3 Tunneling Through a Transparent Gradient Layer with >0 501
2.4 Propagation of Light in a Gradient Dielectric with <0 507
2.5 Evanescent Modes Near the Transition Point in Gradient Media 509
3. Interplay of Artificial and Natural Dispersion in Tunneling Phenomena 511
3.1 TIR from a Transition Plasma Layer 512
3.2 Evanescent Modes Beyond the Polariton Gap in a Transparent Dielectric 516
3.3 Resonant Magnetic Modes in Dielectric Inclusions 520
3.4 Tunneling Phenomena in Negative and ENZ Metamaterials 523
3.5 Energy Transfer by Reactive Modes in the Near Zone of the Source 527
4. Tunneling of Microwaves Through Gradient TLs 529
4.1 Microwave Analogy of a Gradient Photonic Barrier 531
4.2 Evanescent Voltage and Current Waves in a Gradient TL 536
4.3 Brewster Effect for the TE10 Mode Tunneling Through a Gradient Diaphragm
in a Waveguide 538
4.4 GHz Resonance in Periodic Arrays of Plasmonic Nanowires 543
4.5 Propagation of Microwaves Through a Subwavelength Narrowing in a
Waveguide 545
5. Conclusions: The Mystery of Photon TunnelingParadoxes and Perspectives 551
5.1 Tunneling and Nonlocality 551
5.2 Tunneling of Light as a Stochastic Process 554
5.3 Tunneling of Wave Pulses as a Quasistatic Process 556
Acknowledgments 558
References 558

490 Alexander B. Shvartsburg et al.
1. INTRODUCTION: NONLOCAL DISPERSION OF WAVES

IN TRANSPARENT GRADIENT DIELECTRICS
Electromagnetic waves are known to traverse opaque photonic bar-
riers in the form of evanescent modes, characterized by purely imaginary
wave numbers. Being considered initially as a mathematical concept without
physical meaning, these modes were converted later into a powerful tool for
research and design in several branches of electromagnetics and cross-
disciplinary physics. Tunneling phenomena, initially shrouded in mystery,
have a time-honored history. The primary interest of researchers and engi-
neers in these phenomena was threefold.
First, tunneling was seen as one of the fundamental processes in the
dynamics of waves of different physical nature. The first step in studies of
these phenomena were taken in optics more than a century ago in the the-
oretical analysis of total internal reflection (TIR) of light, performed by
Eikhenwald (2009). Using Maxwells equations, Eikhenwald showed that
a light wave when incident on the boundary between two transparent
dielectrics under an angle exceeding the critical angle for TIR does not van-
ish, but penetrates partially into the subsurface layer of the reflecting
medium. This penetration is accompanied by an exponential attenuation
of the wave; the characteristic spatial scale of attenuation being about one
wavelength. Unlike this effect of TIR of a wave from a nondispersive trans-
parent half-space, the partial reflection from a nontransparent barrier of finite
thickness, named frustrated total internal reflection (FTIR), began to direct
attention to wave transport through dispersive media, e.g., through plasmas.
However, the burst of interest in these intriguing phenomena of wave pen-
etration into the forbidden area (tunneling effect) arose after Gamows
(1928) work on nuclear alpha decay, connected with the penetration of
an alpha particle with energy E through a potential barrier with maximum
U0, surrounding the nucleus, under the paradoxical condition E < U0. In the
framework of Gamows approach, the alpha decay of an atomic nucleus was
explained by tunneling of de Broglie waves, describing the alpha particle,
through this potential barrier. The exponentially small probability of this
effect was calculated by him using the formal analogy between equations
governing electromagnetic and de Broglie waves.
Later on, this approach catalyzed the applications of the tunneling con-
cept to numerous EM wave phenomena in different spectral rangesfrom
TIR in optics (Born & Wolf, 1999) up to the percolation of radio waves
Tunneling of Electromagnetic Waves 491
through the ionosphere (Ginzburg, 1970). These applications were based on

the formal similarity between the stationary Schr odinger equation in quan-
tum mechanics and the Helmholtz equation in classical wave theory; the
tunneling of quantum particles through the forbidden zone was compared
to the propagation of an EM wave through a dispersive medium, the wave
frequency being smaller than the cutoff frequency of the medium. Based on
this equivalence, the electronic tunnel effects and the frustrated optical trans-
mission phenomena were shown to be related, forming a new illustration of
the particle-wave duality.
Second, while resolving the paradox of alpha decay, the tunneling theory
gave rise to a new riddlethe apparent superluminality of tunneling. This
result, called the Hartman paradox, formulated in quantum mechanics more
than half a century ago (Hartman, 1962) was connected with an attempt to
find the tunneling time of nonrelativistic particle through a box-like opaque
potential barrier. Hartman showed that for the case of a thick barrier
! !
j p jd , where p is the particle momentum, d is the barriers width,
and 2 is Planks constant, the tunneling time becomes independent of
d, and, thus, for a sufficiently thick barrier the tunneling velocity vt could reach
superluminal values vt > c. However, the direct measurement of tunneling
time through a quantum barrier proved to be an intricate task, and the idea
to verify Hartmans conclusion by means of the classical tunneling effect of
an EM wave through a macroscopic wave barrier was proposed. Its basis
was the formal similarity between the stationary Schr odinger equation and
the Helmholtz equation, describing both propagating and evanescent EM
modes in continuous media. This similarity opened up the opportunity to per-
form experiments with light beams, easier to perform and interpret than those
with electron waves, since optical tunneling requires a micrometer or even
hundreds of nanometers-sized barrier. Moreover, tunneling processes are
often easier to study with light because the coherence time of an optical wave
packet is usually much longer than that of an electronic wave packet.
The physical community was greatly shocked by the perspectives of
superluminal tunneling the more so, as this perspective was deduced from
standard formulae given in many textbooks without any additional suppo-
sitions. The interpretation of measurement data of tunneling times in optical
and microwave ranges stimulated the appearance of different definitions of
tunneling speeds and times. The debates concerning the possibilities to vio-
late the tenets of Einsteins limit became especially intense on the threshold
of the centenary of the creation of the special theory of relativity in 2005
(see, e.g., Olkhovsky, Recami, & Jakiel, 2004).
Third, despite the absence of a universally adopted theory of tunneling of

electromagnetic waves, these effects are widely used in the design of inno-
vative optoelectronic devices, such as, e.g., enhanced light transmission
through subwavelength periodic hole arrays (Yang & Sambles, 2002), struc-
tures for coherent emission of radiation with controlled directivity (Fu,
Zhang, & Tanner, 2005), and the tunneling of optical similaritons in
nonlinear waveguides (Wang, Li, & Jia, 2008).
The theoretical basis of these researches was presented mainly by the
model of a rectangular box-like photonic barrier, presented, e.g., by the
metallic film; the dielectric permittivity inside this barrier is coordinate
independent. The tunneling problems in these structures are connected with
the local dispersion of material. The aforesaid problems, covered by a series
of detailed monographs (see, e.g., de Fornel, 2001; Sarid & Challener, 2010),
lie outside of this review.
In contrast, this review is focused on new phenomena, connected with
the tunneling of electromagnetic waves in all-dielectric heterogeneous
metamaterials with a prescribed smooth spatial distribution of (gradient
dielectrics). These distributions are provided by elements that are electrically
small in terms of the free-space wavelength. The advent of nanotechnologies
opened the avenue to the creation of gradient dielectric films with control-
lable spatial distributions of , providing, in particular, a series of unusual
tunneling phenomena for EM waves (Shvartsburg & Petite, 2005). The past
decade has witnessed an explosive development of metamaterials, offering an
unprecedented flexibility for manipulating the optical properties of matter
and challenging our traditional understanding of wave physics. Research
of waves tunneling through all-dielectric gradient photonic barriers opens
up a wealth of new possibilities.
The following key concepts, based on nonlocality-assisted tunneling
effects in gradient dielectrics, appear throughout this review:
a. Strong heterogeneity-induced nonlocal dispersion of gradient dielectric
subwavelength layers, both normal and abnormal, which can be formed
in an arbitrary spectral range by means of an appropriate geometry of the
refractive index profile, in a given host material.
The physical fundamentals of these tunneling effects are stipulated by
a peculiar mechanism of wave dispersion in heterogeneous dielectrics.
Note the difference in essence of this mechanism upon both the local
material dispersion, determined by the parameter @ 2 n=@2 , and the spa-
tial dispersion of homogeneous media, resulting in small corrections to
the refractive index of the order of magnitude a= 1, where a is the
period of the crystal lattice or the free path-length of the particles in a

plasma, being the radiation wavelength (Born & Wolf, 1999). Unlike
this, the characteristic frequency, separating the spectral ranges of
tunneling and traveling waves in the gradient dielectric with nonlocal
dispersion, depends, next to the local value of the refractive index n,
upon the gradient and curvature of the spatial distribution of n.
b. The formation of plasma-like dispersion in arbitrary spectral ranges,
accompanied by the appearance of a cutoff frequency in gradient barriers
fabricated from nondispersive dielectrics. Subject to the size of the bar-
rier and the shape of the spatial profile of the refractive index n(z) inside
the barrier of the cutoff frequency can be created in the visible, IR,
THz, or GHz range, i.e., the value can be controlled by technology of
barrier fabrication.
c. Effective energy transfer by evanescent EM modes, tunneling without
attenuation through periodic all-dielectric gradient structures (general-
ized Fresnel formulae). These structures are mostly manufactured as thin
layers on transparent substrates (Smith, Vier, Koschny, & Soukolis, 2005).
d. Circuitry with light at nanoscales based on metamaterials-inspired
nanoelectronics, so-called metatronics (Engheta, 2007). This constitutes
the growing trend in research and design of all-dielectric nanostructures.
The new horizons for circuit elements with light are provided by the
electromagnetics of displacement currents in the structured dielectric
nanosamples without free carriers.
e. Scalability of results, obtained by means of exactly solvable models of
gradient photonic barriers, between optical and microwave spectral
ranges, stimulates parallel researches in gradient radio optics. The use
of microwave upscaling represents a powerful method to investigate
optical tunneling in a very accessible time range. The problem is simpli-
fied due to the possibility of performing microwave experiments with
time scales up to nanoseconds (Shvartsburg, Marklund, Brodin, &
Stenflo, 2008).
This chapter is organized as follows:
Section 2 is devoted to the manifestations of nonlocality in reflectance
and transmittance of light, assisted by tunneling modes in transparent
all-dielectric gradient structures. The drastic influence of heterogeneity-
induced nonlocal dispersion on optical features of these structures is
illustrated by the formation of critical frequencies, determined by the
smoothly varying spatial distribution of the dielectric permittivity; these
frequencies separate the spectral ranges of traveling and tunneling modes
in the nondispersive dielectric without free carriers. The nonlocal dispersion

comes into play, giving rise to surprising phenomena such as tunneling
regimes of wave propagation through dielectric layers with > 0 and, vice
versa, traveling regimes when < 0. The spectral dependence of the critical
angle of TIR from the transition layer and the non-Fresnel transmittance
spectra of periodical gradient nanostructures are shown to arise namely
due to these nonlocality effects.
Generalization of this approach for heterogeneous dielectrics with nat-
ural dispersion, stipulated by the plasmon and polariton resonances, is
described in Section 3. The shifts of spectral stop bands in the IR range, aris-
ing due to the interplay between natural and nonlocal dispersion, are shown.
Owing to these distributions, the formation of evanescent modes proves to
be possible in the spectral range, which is transparent for the same spatially
homogeneous host dielectric. Other examples of energy transport by eva-
nescent modes, arising, in particular, in the far zone of a single-negative
magnetic metamaterial and in the near zones of electric and magnetic
dipoles, are considered.
The further generalization of evanescent modes is illustrated in Section 4
by means of a microwave GHz transmission line (TL) with a continuously
distributed impedance. These distributions are presented by smoothly
shaped plastic porous diaphragms, mimicking the gradient photonic barriers,
considered above. The nonlocal dispersion of microwaves is shown to pro-
vide the tunneling of voltage and current waves in the coaxial as well as in
the nonreflection tunneling of the fundamental mode in a rectangular wave-
guide. Effective energy transport by microwaves, penetrating through the
subwavelength narrows in the waveguide, is described and, the scalability
of the results, obtained for the optical range, to the microwave range is
illustrated.
A brief summary of tunneling concepts for electromagnetic waves and
some related unsolved problems of nonlocality in classical electromagnetics
are given in Section 5.
The evanescent modes in smoothly structured dielectrics are presented
by means of several flexible distributions of the dielectric permittivity,
admitting exact analytical solutions of Maxwell equations, expressed via
simple functions. The experimental verifications of tuning of radiation flows
in tunneling regimes demonstrate the growing potential for applications
of gradient metamaterials in electromagnetics. These results, related to 1D
and 2D problems, are intended to take up the past achievements and future
challenges in the fast development of 3D photonic metamaterials.
The goal of this review is to put in order the scientific results, obtained
sometimes on a case-by-case basis, and to stress the underlying physical
principles of tunneling in transparent heterogeneous media. The tunnel
phenomena, sharing a common stage with quantum mechanics and classical
electrodynamics, present intriguing features that stimulate the ongoing
interest.
Glancing back on this race of researches one can recall the picturesque
and inspired phrase of Prof. T.W. Hansch from his Nobel Lecture
(2005): The most surprising result of these researches would be, if we find
nothing surprising.
2. FRUSTRATED TIR FROM GRADIENT DIELECTRIC

LAYERS
This section is devoted to the tunneling regimes of waves through
gradient photonic barriers, which are viewed below as plane nonpolar
dielectric layers, characterized by some prescribed smooth continuous dis-
tributions of the dielectric permittivity (z) across these layers. These
regimes are shown to be determined by the nonlocal heterogeneity-induced
dispersion (HID). To stress these effects, the material of layers is supposed
to be nonmagnetic, lossless, and isotropic; to visualize the peculiarities of
tunneling phenomena in gradient dielectrics, it is convenient to present
(z) in a form: z n20 U 2 z. Here, n0 is the value of the refractive
index of the material on the layer interface z 0, U(z) is some dimension-
less continuous function, describing the spatial profile of refractive index
inside the layer. To visualize the effects of HID, let us consider the reflec-
tance and transmittance problems for normal incidence of linearly polarized
EM wave, propagating in the z-direction, on the interface z 0 of a
photonic barrier. Expressing the wave components Ex and Hy in terms of
!
the vector potential A
!
! 1@ A ! !
E , H rot A , (1)
c @t
and putting Ax , Ay Az 0, one can reduce the system of Maxwell
equations, related to this configuration, to one equation:
@ 2 @ 2 n20 U 2 z @ 2
+ 2 0: (2)
@z2 @y c2 @t2
To compare and contrast the tunneling phenomena in uniform and gra-

dient layers, the reflectance/transmittance spectra of waves, incident on a
homogeneous dielectric layer under the condition of FTIR (the traditional
rectangular model of an opaque barrier, where U 1) are recalled in
Section 2.1. The changes of critical angle of TIR between two dielectric
half-spaces stipulated by the intermediate gradient transition layer between
them is described in Section 2.2; this TIR process is shown to become
frequency-dependent owing to the HID of the transition layer. The decisive
role of HID is illustrated in Section 2.3, which is devoted to FTIR in the
gradient photonic barrier, fabricated from a transparent dielectric ( > 0);
unlike the exponential decay of tunneling modes in homogeneous layers,
the almost total transmission of energy flows through the gradient periodical
multilayer nanostructure, supported by evanescent and anti-evanescent
modes, is visualized. In contrast to Section 2.3, Section 2.4 is devoted to
the propagation of waves through a stratified medium, containing layers
with < 0, without formation of evanescent modes. The slowly varying
power-law tails in the evanescent fields of some modes of dielectric wave-
guides, distinguished from the usual exponential decay of these tails, are
examined in Section 2.5. The theoretical analysis of the abovementioned
FTIR phenomena in gradient dielectric layers is based on the series of
flexible exactly solvable models of these layers.
2.1 Evanescent Waves in Uniform Photonic Barriers

FTIR and optical tunneling provide a peculiar mechanism of electro-
magnetic wave energy transport through a dielectric layer by means of
so-called evanescent modes. To examine the simplest manifestation of this
mechanism, let us consider a uniform layer (medium 1) of thickness d and
low-refractive index n1 that is embedded in a bulk medium (0) of high
refractive index n0. If a monochromatic wave with wavelength is incident
on the layer from medium 0 at an angle greater than the critical angle
c arcsin n1 =n0 , the tunneling of some part of incident energy flow
across the layer is ensured by FTIR. The TIR is restored for the thick layers
d .
Let us denote the plane of incidence as the (y, z) plane, where the axes z
and y are oriented along the normal to the layers and parallel to their inter-
faces, respectively. Supposing for simplicity the refractive indices n0 and n1 to
be frequency-independent one can write the generating functions, describ-
ing s- and p-polarized evanescent modes inside the embedded layer
s, p expk1 z + Q expk1 z expik0 y sin ; (3)

n1 n0 p n0
k1 , k0 , K 2 sin 2 1, K > 1: (4)
c c n1
The complex reflection/transmission coefficients of the embedded uni-
form layer for s- and p-polarized radiation are given by the formulae (Azzam,
2006)

1 exp2x exp is, p
R s, p ,
1 exp 2 is, p x
(5)
1 exp 2is, p expx
T s, p , x k1 d:
1 exp 2 is, p x
Here, s,p are the interface reflection phase shifts:

K
s 2 arctg , p 2 arctg : (6)
Kcos cos
The phase shifts in reflection r arg R and transmission t arg T

for s- and p-polarized waves are:
tg s, p th x
tg r s, p , tg t s, p : (7)
th x tg s, p
Note that under the layer no phases of evanescent modes accumulate,

and the phases t come from the boundaries. Inspection of formulae (5)(7)
reveals the following salient features of tunneling through the rectangular
barrier:
i. The difference between the reflection and transmission phase shifts for
both s- and p-polarized waves is independent upon the barrier param-
eters (Efimov, Ivanov, & Sementsov, 2001):

r s, p t s, p , (8)
2
ii. An increase of the thickness of the barrier d d , x 1 results in
exponential damping of transmittance: jT j2 exp2x,
iii. The first of the expressions (7) shows that the phases of waves r
reflected upon the thick barriers x 1, become independent upon
the barrier thickness, tending to the interface reflection phase shifts
s,p (6); in this limit the phases of transmitted waves t become inde-
pendent upon the thickness of barrier too.
Phenomena (ii) and (iii) characterize FTIR in uniform nondispersive layers
with sharp boundaries, where the tunneling regime can arise only for
oblique incidence above the critical angle. The drastic changes of this pic-
ture, stipulated by the gradual distribution of dielectric permittivity, are
examined below.
2.2 TIR of Light from a Gradient Half-Space

To illustrate the influence of diffused dielectric boundaries on the TIR phe-
nomena, let us consider the configuration, where the denser medium n0,
occupying the half-space z 0, is contiguous with the gradient medium
n1, whose dielectric permittivity in the area z 0 is described by the
distribution
1 W 2 z
z n21 U 2 z, U 2 z 1 + , g > 1: (9)
g g
To describe the saturation of function U2(z) in the depth of medium, it is

expedient to examine two different decreasing functions W1(z) and W2(z):
z
1 z
W1 z 1 + , W2 z exp : (10)
L L
The dimensionless quantity g and spatial scale L are the free parameters of
distributions (9) and (10). Supposing for simplicity, the refractive indices of

both media to be equal at the boundary z 0 n1 jz0 n0 and taking into
account the vanishing of spatial dependence of refractive index n1 in the
depth of medium z L , we find from (9)
s
1
n1 jzL n0 1 : (11)
g
To exemplify the TIR effects, stipulated namely by the gradient of

refractive index, it makes sense to restrict the analysis by s-polarized waves.
Consider firstly the effects, stipulated by the distribution W1(z) (10). Intro-
ducing the new unknown function f(z), connected with the generating
function , and the new variable
f expik0 ysin
z p , 1 + z=L; (12)
W1 z
we obtain the Bessel equation governing the function f():

d2 f 1 df s2
+ + p 2 f 0;
2
(13)
d2 d
2
Ln0 2
1 n1 2 1 1 Ln0 2
p
2
1 2 sin , s
2
: (14)
c g n0 4 g c
Equation (13) is known to have the solution, damping at the depth of

medium ( 1), under the condition p2 < 0. This condition brings the
value of critical angle c for the gradient layer (9); in a special case

n1 jz0 n0 and g > 1, we have
1
cos 2 c : (15)
g
Solution of Equation (13) under the condition p2 < 0 is given by the

modified Bessel function of the second kind f Ks jpj; finally, the
solution describing the field percolating into the gradient subsurface layer
in a case > c may be written as:
r h
z z
i
1 + Ks jpj 1 + expik0 ysin; (16)
L L
here the values of index s can be both real and imaginary in the frequency
intervals > s and < s ,respectively, where
p
c g
s : (17)
2Ln0
This heterogeneity-induced dispersion of gradient layer results in the
difference in spatial structure of tunneling field : the analysis of func-
tions Ks (Dunster, 1990) shows that the spatial weakening of is monotonic
in the low frequency interval ( < s ), meanwhile in the high frequency
interval ( > s ) the weakening of is accompanied by its spatial
oscillations.
Now let us turn out to the consideration of the same model (9) with
another function W z W2 z (10). Proceeding in a similar fashion, we
can introduce, by analogy with (12), the generating function and the new
variable :
z
x f expik0 ysin , exp : (18)

L
The function f() is determined again by Equation (13); however, the

values of parameters p and s are distinguished from (14):

1 Ln0 2 2 Ln0 2 n21 2 1
p2 , s sin 1 + : (19)
g c c n20 g
The field described by this function f is distinguished from the field (16):
unlike the profile W1(z), the solution of master Equation (13) for W2(z) in a
case g > 0 is given by the Hankel functions of real argument
Hs1 p expz=L , the argument being independent upon the angle of
incidence ; however, the index of this function s can be either real or imag-
inary, subject to this angle. The TIR regime arises for the angles of incidence
> c , the value of critical angle c, determined in Equation (15), is valid for
both models W1(z) and W2(z). Comparing the distributions U2(z) (9), con-
nected with damping functions W1(z) and W2(z), one can see that the initial
values of n2(z) at the interface z 0, as well as their gradients at the interface
z 0, are equal; moreover, the values of n2(z) in the depth of media are
equal too (11). However, the curve W2(z) is decreasing more rapidly
than W1(z); owing to this steepness, the model U2(z) (9) with W2(z)
is approaching to the step-like variation of refractive index at the interface
z 0 and, similar to the step-like variation, the field in this model is
decaying monotonically. To the contrary, the model U2(z) with W1(z)
is sloping near by the interface z 0 more gently, and the spatial structure
of tunneling field ( > c ) in this case can be both monotonic and oscillating
subject to the correlation between the frequencies and s (17); thus the
models W1(z) and W2(z) relate to the dispersive and dispersiveless regimes of
TIR in the gradient dielectric layers, respectively. Moreover, the structure
of the fields for the waves incident under the critical angle c , deter-
mined in the models W1(z) and W2(z) by the Equation (13) under the con-
ditions p 0 and s 0, respectively, proves to be different. These examples
illustrate the dependence of evanescent fields, penetrating inside the gradient
media under the TIR conditions, upon the fine structure of refractive index
distribution n(z) in the gradient layer.
2.3 Tunneling Through a Transparent Gradient Layer

with >0
Unlike Section 2.1, this section is focused on FTIR phenomena for normal
incidence of radiation on a photonic barrier ( 0). In a case of a dispersive
barrier (e.g., a uniform plasma layer, containing free carriers and character-
ized by plasma frequency p), FTIR phenomena are known to arise for the
low frequencies of radiation < p . Resonant tunneling in FTIR was dis-
cussed for periodically curved interfaces (Longhi, 2005). However, similar
effects are inherent to the nonuniform dielectric layer without free carriers
too, if its refractive index is distributed across the plane layer according to
some prescribed law nz n0 U z, which can provide the HID; here
the values n0 and U 1 relate to the plane z 0. Consider the concave pro-
file U(z), characterized by two free parameters L1 and L2 having the dimen-
sion of length
1
z z2
U z 1 + 2 : (20)
L1 L2
1
Note, that the widely used Rayleigh profile U z 1 + z=L
(Rayleigh, 1937) can be viewed as the limiting case of the more flexible
distribution (20), corresponding to the limit L2 ! 1. Distribution (20)
is shown by curve 1 at Figure 1, the spatial scales L1 and L2 are linked with-
the layers thickness d and the minimum value Um via the gradient
parameter y:
1 L2 d d
Um 2
, y , L1 2 , L2 : (21)
1+y 2L1 4y 2y
Equation (2) governing the wave field inside the layer is reduced in
this geometry (@y 0) to an ordinary differential equation. Solutions of
this simplified equation with distribution U(z) (20) describe waves with
both real and imaginary wave numbers, corresponding to propagating
and tunneling regimes of propagation, respectively. The spectral ranges
related to these regimes, separated by some characteristic frequency
, dependent upon the shape and size of profile U(z) (Shvartsburg &
Petite, 2005):
p
2y 1 + y2
: (22)
no d
1.0
0.9
0.8
U2
0.7 1
0.6
2
0.5
0.0 0.2 0.4 0.6 0.8 1.0
z/d
Figure 1 Profiles of dielectric permittivity U2(z) versus the normalized width of the
gradient barrier; curves 1 and 2 relate, respectively, to profile (20) with the value of
parameter y 0.577 and profile (109) with the values of parameters M 3.96, g 2.1.
Note the difference between the characteristic frequencies (22) and (15),
connected with the difference in distributions (20) and (9). Thus, the profile
of refractive index (20) provides the waveguide-like dispersion of a dielectric
layer without free carriers. The frequencies > ( < ) fall in the prop-
agating (tunneling) spectral ranges. Introducing the variable
z
U z1 dz1 , (23)
0
one can present the propagating wave field inside the layer as a result of
interference of forward and backward waves:
Aexpiq + Q expiq ne
p p , q ,
U z c (24)
p
ne n0 1 u2 , u :

Formula (24) describes the high frequency wave ( > , u < 1); in an
opposite case ( < , u > 1) the tunneling field inside the gradient layer
is presented by low frequency evanescent and anti-evanescent waves:
Aexpp + Q expp ne p
t p , p , ne n0 u2 1: (25)
U z c
The parameter Q describing the contribution of the backward wave

to the entire field inside the layer has to be found from the continuity con-
ditions on the layers boundaries. It is essential that in a gradient dielectric,
the waves with imaginary wave numbers arise in the transparent medium
with > 0.
The generating functions (24) and (25) are in need for analysis of field
components inside the gradient layers (20). The components of a tunneling
wave inside the gradient layer possess a complicated spatial structure and
coordinate-dependent mutual phase shift. To examine the spatial structure
of low frequency (u < 1), Ex and Hy components of tunneling fields repre-
sented inside the layer (0 z d) by evanescent modes, one has to substitute
the generating function t (25) into Equations (1):
E0 M 1+R z
Ex p expp + Q expp, M , x , (26)
U z 1 + Q d
n
p h i
Hy iE0 M U z 1 2x ne expp
h 2i o
+ O 1 2x + ne expp ,
2
R is the complex reflection coefficient for single-gradient layer; for simplic-
ity the temporal factor expit is omitted in Equation (22). It is essential
that the phase shift between Ex and Hy inside the layer differs from /2 and,
thus, there is a nonzero energy flux Pz through the layer:
c
Pz Re Ex Hy* : (27)
4
Note that this energy flux is formed by evanescent and anti-
evanescent modes.
Let us consider the tunneling regime ( < , u > 1) of wave travers-
ing the nanostructure, containing m 1 similar adjacent nanofilms with
thickness d, deposited on a homogeneous half-space with refractive index n.
Attributing the number m 1 to the first layer at the far side of this
structure, one can find the explicit formulae for its complex reflection Rm
and transmission Tm coefficients using the standard continuity conditions
for the components (26) on the boundaries of each layer (Shvartsburg &
Maradudin, 2013):

i + n e m
Rm 2 ; (28)
i + n e m
2
p !
p
1 + y2 + y
ne n0 u 1 , t th l 1 u2 , l ln p
2 2 2
,
1 + y2 y
ph p
im (29)
2un0 y 2i n ch l 1 u 2
Ym
p , Tm 1 tm :
1 + y2 i + ne m m1
2
The parameters m in (28) and (29) are linked by the chain of recursive
relations, obtained from the continuity conditions on the interfaces between
the m-th and m 1-th layers (m > 1):
ne m1 + t
m : (30)
ne 1 + tm1 t
The first term in this chain 1, related to the far layer in the structure, is:

n i ne t
1 2t
: (31)
nt i ne
2
The expressions for Rm and Tm coefficients for the case of propagating
waves ( > , u < 1) follows directly from (28) to (29) due to the
replacements
p p p
u2 1 ! i 1 u2 , t ! i tg l u2 1 ,
p
p
(32)
ch l 1 u2 ! cos l u2 1 :
Reflectance/transmittance spectra for a periodic structure, containing

several similar adjacent gradient barriers with concave profiles n(z) (20),
are shown in Figures 2 and 3. Some formal similarity between the tunneling
of waves in a plasma with free carriers in a case < p and in a nonpolar
gradient dielectric without free carriers for the case < was noticed
above; however, the drastic differences between these phenomena have
to be emphasized:
i. The dispersive photonic barriers with cutoff frequencies , restricting
the spectral range of tunneling regime, can be formed in transparent
(a)
2.2
2
1.8
1.6
1.4
1.2
n
1
0.8
0.6
0.4
0.2
0
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400 420
z (nm)
(b)
0.5
0.45
0.4
0.35 1
0.3
|R|2 0.25
2
0.2
0.15
0.1
0.05
0
400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
l (mm)
Figure 2 Reflectance spectra jRj2 of the periodical nanostructure, containing three gra-
dient films (20). (a) Geometry of nanostructure (thickness d 140 nm, the values of
refractive indices n0 2 and nmin 1.5). (b) Curves 1 and 2 present, respectively, the
spectra jRj2 calculated for the three-layer gradient structure, shown on (a), and for
the homogeneous layer with thickness D 3d, refractive index n0 2 and n 1.52.
1
0.95
2
0.9
0.85
0.8 1
0.75
0.7
0.65
0.6
0.55
|T|2
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100
l (nm)
Figure 3 Experimental (1) and theoretical (2) transmittance spectra for periodical nano-
structure, containing 11 layers (20) with width d 140 nm; the values of other param-
eters of this structure are: y 0.577, n 1.52, and n0 1.5.
gradient dielectric structures without free carriers (Re > 0). The
wave energy is transmitted through these structures by means of
evanescent modes.
ii. Interference of evanescent and anti-evanescent modes in a stack of
gradient layers results in the formation of new reflection/transmission
spectra having nothing in common neither with the almost total
reflection nor with the exponentially small transmission of a plasma
layer in the FTIR regime ( < p ). Thus, considering the three-layer
gradient nanostructure, depicted on the Figure 2a, characterized by
the total thickness 3d and cutoff frequency (22), corresponding
to the wavelength c 2c= 1:32 m, one can compare its reflec-
tance with the reflectance of a homogeneous layer with the same
thickness 3d, the other parameters n0 and n being equal too
(Figure 2b). Unlike the calculated oscillating reflectance of the
homogeneous layer (curve 2), the experimentally measured reflec-
tance of the three-layer gradient nanostructure is characterized by
the narrow maximum nearby the blue end of the visible spectrum,
stipulated by HID, accompanied by horizontal plateau, presenting
almost constant small reflectance (jRj2 0:05) in a broad spectral range,
covering both short ( < c ) and long ( > c ) wavelengths. This
plateau corresponds to the broadband range of high transmittance
jT j2 1 jRj2 0:95, including, in particular, the tunneling waves
( > c ).
iii. Increasing the number of gradient nanofilms does not change essentially
the transmittance spectra. Thus, the transmittance spectra of a multi-
layer, containing 11 gradient nanofilms with profile of refractive index
(20), are depicted on Figure 3; curves 1 and 2 present, respectively, the
experimental results and theoretical values, calculated by means of for-
mulae (29)(31). The theoretical and experimental results prove to be
in a good agreement. The narrow minimum at the wavelengths near
450 nm, characterized a decrease of the transmission coefficient
by 1819
, arises as well, as at the Figure 2b, due to HID. The growth
of wavelengths in the interval 600 nm < < 2100 nm, brings the small
oscillations of transparency around the value jT j2 0:9; respectively,
the reflection coefficient keeps a weakly varying value jRj2 0:1 in this
visible and near IR range, resembling the plateau at the same range are
in Figure 2a.
2.4 Propagation of Light in a Gradient Dielectric with <0

While considering the FTIR effects in the homogeneous lossless opaque
layer < 0 in Section 2.1, we had described the evanescent modes,
characterized by the purely imaginary wave numbers k Re k 0. The
conditions of existence of evanescent modes Re k 0 were diversified
in the Section 2.3 due to examples of formation of such modes, in trans-
parent gradient media > 0 too. In contrast, this Section is concerned
with an opposite effectthe propagation of a wave with a real value of its
wave number Re k 6 0 through an opaque layer < 0, occurring in a
dielectric, filling the half-space z 0. To visualize this possibility, one can
consider the wave in a heterogeneous dielectric medium, characterized by
some profile of refraction index n(z/L), where L is the spatial scale of
heterogeneity. Describing this wave by means of a one-dimensional equa-
tion, written by means of normalized coordinate and dimensionless
parameter
d2 z L
2
+ 2 n2 0, , , (33)
d L c
we may seek its solution in the interval z 0 in a form (Erokhin &

Zakharov, 2011)
r z
q0 iz
A exp , q0 q0, z qz1 dz1 , (34)
qz c 0
here A is the wave amplitude. Substitution of solution (34) to Equation (33)

brings the general expression for profile of refractive index n(), supporting
this solution:
( 2 )
2
1 1 d q 3 dq
n2 q2 + 2 2
2 : (35)
2 q d 2q d
Till now the dependence q() remains arbitrary. Let us choose this
dependence, e.g., in a form:
q0 2 2
q 2 , a + b exp : (36)

Figure 4 Square of refractive index in gradient layer n2 (35)(36) is plotted versus the
normalized coordinate ; the values of parameters in Equations (35)(37) are: a 1,
b 5, 2 4, q0 2, and 2 2.
Here, the real quantities a, b, and are the dimensionless free parameters
of spatial dependence of refractive index n(). The spatial distribution of
refractive index (35) inside the interval 0 can be written by substitution
of (36) to (35) in an explicit form, depicted in Figure 4b.

1 q20 2 2 1 2 2 2 2 2
n 2 2 +
2
, b exp : (37)
2
We note that the wave traveling in the z-direction is passing through

the opaque segment with < 0; however, its wave number q (36)
remains real and positive in this segment as well as in the entire interval
z 0. Thus, an evanescent wave does not arise in this opaque layer.
The reflection coefficient, characterizing the normal incidence of wave
with frequency on the interface of dielectric (37), does not contain
the imaginary part:
a + b2 q0
R :
a + b2 + q0
Deep in the medium z L= , the refractive index tends to a constant

value, and the wave (34) is transformed into a harmonic wave with con-
stant amplitude and constant wave number:
q20 h q z
i
0
n2 z L= , z L= Aa exp i 2 t : (38)
a4 ac
Thus, the examples, examined in Sections 2.3 and 2.4, illustrate the flex-
ibility of radiation spectra in the gradient dielectrics, including the cases, which
initially seem to be counterintuitive: a tunneling regime Re k 0 in a trans-
parent metamaterial ( > 0) and a propagation regime Re k 6 0 in an
opaque medium ( < 0).
2.5 Evanescent Modes Near the Transition Point

in Gradient Media
The propagation of EM waves in a gradient dielectric in a direction orthog-
!
onal to r can be viewed as propagation in a planar graded-index wave-
guide. The modes, traveling in this structure, are known to contain
evanescent tails; the decay of these tails at the periphery of the waveguide,
controlled by the spatial distribution of the dielectric permittivity , can dif-
fer substantially from the traditional exponential decay. Thus, the evanescent
tail of a surface mode, propagating along the interface of a gradient dielectric
without free carriers, described by an asymmetric distribution of (e.g., by
the function U2, which is given by (9) within the dielectric, with outside
the dielectric U 2 1) is characterized by nonexponential behavior; in a
dielectric area, the growth of the electric component and the appearance
of zero-crossing points for one of the magnetic components preceed their
exponential decay deep in the medium (Shvartsburg & Maradudin,
2013). On the other hand, there are some distributions of , which provide
monotonic evanescent tails, which decay according to a power law
(Hayata & Koshiba, 1995). To exemplify the former case, let us consider
wave propagation in a dielectric with refractive index n, presented in a form
n2 n20 + 2 z: (39)
Here, is some dimensionless constant, characterizing the maximum
perturbation of the background value of refractive index n0, (z) is an
unknown function, describing the spatial distribution of this perturbation.
Considering wave propagation in the y-direction, orthogonal to the
z-direction, substituting the distribution (39) into the wave equation (2) and
seeking the generating function in a form
n0
f vexpik0 y, k0 , v k0 z, (40)
c
we find the equation governing the function f:
@ 2 f 2 v
+ f 0: (41)
@v2 n20
Unlike the approach, used in the previous sections, based on the search of
solutions determined by the given distribution of , we will seek the distribution
(z), related to the prescribed solution f(v). Thus, for the bell-like solutions
m
f 1 + a2 v 2 , m > 0, (42)
we obtain
2ma2 n20 1 1 + 2ma2 v2
v : (43)
2 1 + a2 v2 2
In the simple case m 1, the solution (42) has a shape of the well-known
Cauchy curve. The profile (z) resembles the symmetrical Mexican hat
with a maximum at v 0 and zero-crossing points v v0 and minima
at v vm
2
an0 ma2 n20 1 + 2m2
0 2m , v0 0, vm : (44)
42 1 + m
Thus, for each wavelength and field structure (42) one can design the
distribution (z), providing the prescribed structure f(v). Herein, the pertur-
bation of refractive index is not small and can exceed at some points the
background value n0. Far from the maximum at v 0 the perturbation
(z) is vanishing: 1 0. The bound mode fields have been found
to be more weakly localized than those of guided modes supported by
the normal waveguide because the evanescent tails of (42) obey, instead
of an exponential law, the power law:

f jvj jaj1 ! jzj2m : (45)
Substitution of the function (40) as the Ax component of the vector
potential to (1) yields the values of the transverse components Ex, Hz and the
longitudinal component of the modal field Hy. The former component is
decaying at the periphery of the localized wave faster than Ex and Hz:
Hy jvj jaj1 ! jzj2m1 . This delayed damping of evanescent tails may
be useful for deep penetration of fields, e.g., for the excitation of electronic
states in the subsurface layers of heterogeneous dielectrics (Zayatz,
Smolyaninov, & Maradudin, 2005).
3. INTERPLAY OF ARTIFICIAL AND NATURAL

DISPERSION IN TUNNELING PHENOMENA
The natural dispersion of materials was ignored in the above analysis.
Many transparent nonpolar dielectric materials do not possess cutoff fre-
quencies in the visible and near IR spectral ranges, while the dependence
of the refractive index n on the wavelength is smooth, and its variations
do not exceed several percents. For example, the values of n for fused silica
are known to decrease monotonically in the visible and near infrared ranges
from n 400 nm 1:47012 to n 2010 nm 1:43794, so that the
relative variation of n in this spectral range does not exceed 22.5%
(Adachi, 1988). Thus, the natural dispersion of optical materials in these
ranges is usually weak and negative. In contrast, the artificial nonlocal dis-
persion in gradient media was shown to provide drastic changes in the func-
tion n() as well as the controlled formation of both positive or negative
dispersion in the host material. This way to strengthen and control disper-
sion seems to be especially useful for the visible spectral range, where the
dispersion of a natural material proves often to be weak.
However, the interplay between natural (local) and artificial (nonlocal)
dispersive effects may become important in some media for definite spectral
ranges in the vicinity of resonances, caused, e.g., by plasmon or polariton
excitations. Simultaneous action of natural and artificial dispersion in these
media is illustrated by the use of generalizations of dielectric functions,
widely used for analysis of both plasmon (46) and polariton (47) resonances
(Maier, 2007):
2p 4e2 Ne
1 , 2p : (46)
2 me
2
2L
1 2 : (47)
2T
Here, 1 is the frequency-independent background dielectric constant,

the plasma frequency p in (46) depends upon the electron density Ne and
electron charge e and effective mass me; T and L in (47) are the frequencies
of transverse and longitudinal vibration modes, forming the edges of the
stop. The interplay of plasmonpolariton resonances in metallic nanofilms
with the nonlocal dispersive effects, stipulated by the arrays of holes in these
nanofilms, has been shown to provide enhanced tunneling of light through
such structures (Darmanyan, Neviere, & Zayats, 2004). The similar effects in
optics of sculptured thin films were examined by Lakhtakia and Messier
(2005). Microwave energy transfer in plasma-filled metallic waveguides,
influenced by waveguide spectra and plasma oscillations, has been the sub-
ject of researches during several decades; see, e.g., the book of Kuzelev and
Rukhadze (2009) and the literature therein.
Below, we will turn our attention to a new trend in the optics of
metamaterials, connected with the tunneling effects in all-dielectric gradient
composite media. Propagation of light in these media, dependent simulta-
neously upon both local and nonlocal dispersion, is considered in
Sections 3.1 and 3.2 in the framework of exactly solvable models of (z).
These sections illustrate the formation of new tunneling spectral ranges in
gradient dielectrics with natural local dispersion, characterized both by a
half-infinite stop band (Section 3.1) and a finite stop band (Section 3.2).
In contrast, the formation of tunneling spectral ranges in media with
man-made local dispersion forms the content of Section 3.3, aimed at the
optically induced magnetization of composite materials. This local,
coordinate-independent, dispersion is provided by the eigenoscillations of
dielectric nanocylinders and nanospheres, embedded in the host material.
Some peculiarities of tunneling of EM radiation in metamaterials with
simultaneously negative values of the permittivity and the permeability
are described in Section 3.4. Unlike the Sections 3.13.4, in which we con-
sider the energy transfer by evanescent modes in the far zone of the source,
Section 3.5 presents an example of energy transfer by evanescent modes
close to the source.
3.1 TIR from a Transition Plasma Layer

Free carriers in semiconductors, belonging to the AIIIBV groups, can be
viewed as a plasma medium, determining the optical properties of these
semiconductors. Thus, the homogeneous n-type semiconductors with con-
stant electron density N0 are characterized by some critical frequency cr,
separating the spectral ranges of high frequency propagating and low
frequency evanescent electromagnetic waves. Neglecting, for simplicity, the

electron collisions in a semiconductor plasma, one can find the frequency
cr for the waves impinging on the semiconductor boundary under the
0:5
angle : cr p 1 sin 2 (Maier, 2007). The waves with frequen-
cies < cr form the evanescent modes, damping exponentially in the sub-
surface layer; the rates of spatial damping are equal for both electric and
magnetic components of evanescent modes.
However, this pattern can vary drastically due to the spatial variations of
free carriers density in the doped polar semiconductors which can stipulate
the variety of structures of evanescent electric and magnetic modes, pene-
trating to the subsurface layers of these media. Thus, the gradual growth
of plasma density N(z) away from the semiconductor surface, tending to
the constant value of N deep in the plasma, results in the continuous decrease
of values of (z), inherent, e.g., to the doped semiconductors due to forma-
tion of a carrier-depletion layer (Meier, 2007). Presenting the distribution of
electron density in such transition layer by a model, similar to (9),
z
1 W 2 z
N z N0 1 + , g > 0, W z exp ,
g g L (48)
z=L,
we can examine the spatial structure of evanescent modes, formed, e.g., by
the low frequency TE-polarized wave, incident on the interface under an
angle . Substitution of distribution (48) to (46) brings the equation for
the generating function in a form of a Bessel equation

d2 1 d s2
+ + p 2 0,
2
(49)
d2 d
herein the values of variable and parameters p2 and s2 are:
z

1 p L 2
exp , p 2
,
L g c
2 (50)
p L 2 1 L
s
2
1+ 1 sin 2 :
c g c
The solution of Equation (49) in the case s2 > 0 is given by Bessel func-
tions Js(p).
Consider first the case of normal incidence on a plasma ( 0). To visu-
alize the salient features of these solutions, it is worthwhile to recall that the
Bessel functions with odd half-integer indices are expressed via the
elementary functions (Abramowitz & Stegun, 1968); thus, in a simplest case

s 0:5, the normalized distributions of electric e(z) and magnetic h(z) com-
ponents of evanescent mode in the subsurface layer are:

Ez sin p exp exp

ez 2 , (51)
Ez 0 sin p
H z p sin p
hz , p cos p , (52)
H z 0 2p
p
1
p exp cos p exp exp sin p exp:

2 2p 2
To exemplify the case of inclined incidence of the same wave on the

same interface, avoiding the massive algebra, one has to choose the angle
, satisfying to condition Lc 1 sin 2; in this case the index s 1:5,
2
the distribution of electric field is:
Y sin p
ez , l cos p, (53)
l p

exp sin p exp
Y exp cos p exp :
2 p
The magnetic components can be calculated by means of substitution of

the generating function J1.5(p) to Equation (1).
The evanescent fields (51)(53) are shown in Figure 5. Inspection of cur-
ves 13 on this figure reveals the properties, distinguishing these fields from
the evanescent fields in a homogeneous medium:
i. The decay of evanescent modes e(z) is not monotonous; the values of
e(z) in the subsurface layers reaches the maximum at the point
p
z0 L ln0 , determined by equation tg p0 2p p0 ; the values
e(z0) exceeding the values of e on the interface z 0;
ii. The distributions h(z) have a zero-crossing point z0 L ln0 and
reach the minimum at the area z > z0 , where hz < 0;
iii. Similar propertiesthe nonmonotonous structure of e(z) and change of
sign of h(z) deep in the plasmaremain valid for the inclined incidence
of TE waves too.
The penetration of waves in semiconductor plasmas, described by another
distributions of electron density, e.g., W z 1 + z=L , 0:5, 1, 1:5,
was considered by Kim (2008).
Figure 5 Penetration of electromagnetic field into the plasma transition layer with var-
iable electron density (48). Normalized values of electric e and magnetic h components
are plotted versus the normalized distance from the plasma boundary in a case
p 2.35. Curves 1 and 2 show the spatial distributions of e and h for the normal inci-
dence and curve 3 illustrates the decrease of e for the inclined incidence (50).
The influence of the boundaries of the confined plasma volume may lead
to plasma depolarization inside the volume. Such effects are important for
three-dimensionally bound particles of a crystalline material, containing
the conduction electrons, e.g., for n-type semiconductors with free carriers.
Thus, for the spheroid particle the electron plasma frequency inside the par-
ticle p is coordinate-independent; however, owing to the depolarization
factor D, is dependent upon the spheroid axial ratio, the frequency p is dis-
tinguished upon the value p, characterized an unbounded plasma
(Dresselhaus, 1955):
2p D 4e2 N0
2p , 2p : (54)
1 + D m
Here, is the dielectric permittivity per unit volume. Thus, the plasma fre-
quency values for two samples of the same material, distinguished only by
their shapes, may be considerably shifted.
The spatial variations of refractive indices in the tailored semiconductors
with free carriers can be formed by the differences in their plasma frequen-
cies, dependent, in their turn, from the effective masses of carriers me: thus,
the difference of plasma frequency p for GaAsme 0:0665m0 and
Alx Ga1x Asme 0:15m0 results in the disparity of the related refractive
indices: n 3:20 (GaAs) and n 3:50 (Alx Ga1x As) (Wilson, Glytsis, &
Gaylord, 1993). The specially designed periodical structures, fabricated from
ultrathin layers of these semiconductor species, can act as optical superlattices
(Nelin, 2007).
3.2 Evanescent Modes Beyond the Polariton Gap

in a Transparent Dielectric
Now let us turn to the tunneling effects in an another spectral struc-
ture, related to a stop band of finite width. It is worthwhile to illustrate
this model by means of a dielectric with a polariton gap, e.g., MgO or
ZnSe (Deych, Livdan, & Lisyansky, 1998), bounded by the resonant
frequencies T and L (48). Recalling the relation between these fre-
quencies 2L 2T 1 + =1 (Bulgakov, Bulgakov, & Nieto-Vesperinas,
1998), where is the so-called oscillator strength, that characterizes
the polarization of molecules, one can model the distribution of 2L(z),
produced by a continuously varying content of polarized molecules in
the subsurface layer of the medium by means of some real dimensionless
function U(z):
2L z 2L0 U 2 z: (55)
Using the distribution (55) one can represent the generalized profile (z)
(47) as
2
2 L0 W z
2 2
z n0 , n20 1 : (56)
2 2T
Consider the model (Shvartsburg & Maradudin, 2013)
h z
z
i1
W z cos + M sin : (57)
L L
This profile is characterized by two free parameters (scale L and dimen-
sionless quantity M), connected with the thickness of gradient layer d and the
depth of modulation of the content of polarized molecules determined by
the value W 2min:
d 1
2 arctgM , Wmin
2
: (58)
L 1 + M2
To solve the master equation (2) with the dielectric permittivity, given
by functions (9) and (57), it is worthwhile to introduce the new function f(z)
and variables and
z
f z
z p , W z1 dz1 , (59)
W z 0
2 z
3
6
1 + m tg p

7
1 +
p ln 4 2L
z 5, m 1 + M 2 M, (60)
L 1 + M2 1 m tg
2L
p
1 + M2
0 , 0 ln m + , jz0 0 , jzd 0 : (61)
L
The function W(z) can be expressed via the dimensionless coordinate
in an explicit form:
ch
W p , W 0 W 0 1: (62)
1 + M2
Bringing together the results (59)(62), we obtain from Equation (2)
equation, governing the function f(), in a form, familiar from quantum
mechanics (Schiff, 1968):

d2 f
q + 2 f 0,
2
(63)
d2 ch
" #
1 2 2L0 c
q 1+
2
, 0 , (64)
4 0
1 + M T
2 2 2 2n0 L
" 2 #
1 2
1 : (65)
4 0 2 2T
The general solution of Equation (63) is known to be presented via the

hypergeometric functions (Abramowitz & Stegun, 1968). However, some
physically meaningful results can be visualized even in a limiting case
0; the frequencies , obeying this condition, are:
s!
2 1 1 2T
1, 2 0
2
: (66)
2 4 20
Substitution of values 1,2 to the expression for q2 (64) gives the value
q2 > 0; thus, the generating function is expressed via the evanescent
and anti-evanescent waves

Aexpq + Q expq 2 1 1 L0
2
p , q 1+ : (67)
W z 4 1 + M2
Note, that when > T these evanescent modes arise in the spectral
range > T , located beyond of the band stop of the unperturbed material
T < < L0 .
To illustrate the spectral properties of a spatially varying stop band, one
can consider the light wave, which is impinging normally on the gradient
subsurface layer in dielectric ZnSe, possessing a polariton bandgap between
L0 and T. Taking into account, that this material in a natural state is
characterized by the refractive index n0 1.98 and a polariton gap in the
near IR range between the frequencies T 0:47
1015 rad=s,
L0 0:88
1015 rad=s (Deych et al., 1998). Using the model (57), one
can find the variation of n; thus, in a case of modulation depth Umin 2
0.1
(58) and thickness of gradient layer d 150 nm the values of parameter M
and characteristic scale L, calculated from (58), are: M 3, L 60 nm;
herein the characteristic frequency, occurred due to heterogeneity (64),
0 1:26
1015 rad=s The frequency of evanescent mode , determined
from Equation (66) is 1.151015 rad/s; parameter q for this mode (67)
is q 0.51; thus this mode arises in the perturbed layer in the transparency
range of unperturbed dielectric > L0 0:88
1015 rad=s. Note, that
the refractive index of the perturbed dielectric mode for this frequency is real:
substitution of this value to (56) yields the value n 2.1. Side-by-side with
the analysis, performed in the Section 2.3, this example shows the drastic
changes of optical properties of material, providing the formation of evanes-
cent modes, existing surprisingly in the spectral range with the positive
values of n2. The similar effect can arise in another dielectrics too, e.g., in
LiF, characterized by a polariton gap between T 0:365
1015 rad=s and
L0 1:08
1015 rad=s (Serebryannikov, Ozbay, & Nojima, 2014).
Progress in crystal growth techniques has made feasible the realization of
multilayer periodic dielectric transparent structures, composed from
ultrathin layers with alternating high- and low-refractive indices; the periods
of these artificial one-dimensional structures, so-called superlattices, signif-
icantly exceed the natural periods of crystal lattices, constituting these
structures (Yeh, 1985). The prescribed distribution of refractive index in
these structures formed the basis for the design of systems for spectral
(a)
1.0
2
0.9
|T |2 1
0.8
0.7
1.05 2.00 3.00
u
(b) 1
0.98
0.96
0.94
0.92
|Tm|2
0.9
0.88
0.86
m = 10
0.84
0.82
m = 20
0.8
1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
u
Figure 6 Transmittance spectra of periodical gradient structures (20) providing the
wave energy transport in the tunneling regimes. The values of transmission coefficient
jTj2 are plotted versus the normalized frequency u (24). Structures containing m
nanolayers with thickness d 100 nm are supposed to be deposited on the thick sub-
strate with refractive index n 2.3. (a) n0 2.2187, y 0.75, spectra 1 and 2 relate to
nanostructures, containing 1 and 20 nanolayers, respectively. (b) n0 1.8928, y 0.577.
filtration of radiation, e.g., transparency windows (Fratalocchi & Assanto,

2006; Li et al., 2007).
The use of gradient nanofilms opens the way to design the superlattices
for tunneling waves. Thus, the periodical system of dielectric nanolayers
with profile (z), defined in (20), can support energy transfer by evanescent
modes (Figure 6). Comparison of monotonic (1) and oscillating (2) spectral
curves on Figure 6a, characterizing, respectively, the single-layer and
multilayer structures, located on the same substrate, illustrates the decisive
role of interference of forward and backward evanescent modes in for-
mation of transmitted wave field; meanwhile the transmittance spectrum
of a single-layer gradient nanofilm is presented by a monotonically decaying
curve 1, the multilayer structure provides the oscillating spectrum jT j2 (2).
An important result of this interference is manifested by formation of the
discrete nonequidistant spectrum of frequencies, providing the complete
transmittance (jTj2 1) in the tunneling regime (curve 2). Comparison
of curve 2 (Figure 6a) and curve m 20 (Figure 6b) shows the dependence
of transmittance upon the refractive index n0 of nanofilms, the other param-

eters of superlattices being equal; inspection of the transmittance spectra for
superlattices with m 10 and m 20 on Figure 6b illustrates the influence of
the number of nanofilms on the location of windows of transparency,
where jT j2 1, in the tunneling regime.
Studies of optical superlattices have shown the important analogy
between the tunneling of electrons and photons. This analogy is connected
with resonant Zener tunneling of electrons in WannierStark ladders, con-
sisting of a set of equidistant energy levels in the electron band structure in
crystal, that is exposed to a stationary electric field (Wannier, 1955). The
optical counterpart of this effect can be observed in a one-dimensional opti-
cal superlattice, presented by the transparent multilayer dielectric structure.
The role of the constant electric field is played here by the linear refractive
index gradient of layers along the direction of wave propagation. The time-
resolved oscillations in light transmission were mimicking the electronic
Bloch oscillations in conducting crystals, where the WannierStark ladder
was obtained via an external electric field (Ghulinyan et al., 2005).
The flexibility of parameters of these dielectric structures can be
improved by the use of graded metamaterials, characterized by
frequency-dependent dielectric and magnetic parameters, providing both
positive and negative values of () and () over some spectral ranges.
Some examples of such improvement are discussed below.
3.3 Resonant Magnetic Modes in Dielectric Inclusions

Till now we were dealing with the tunneling phenomena, based on the
interplay of natural eigenfrequencies of a material (p, L0, T) and artificial
HID of dielectric permittivity of the metamaterial. However, over the past
decade substantial efforts have been put on the creation of metamaterials
with prescribed spatial distribution of magnetic permeability, operating in
the IR and visible optical ranges. These studies were based on the use of
materials, containing the dielectric cylindrical or spherical inclusions, which
were acting as resonators, supporting the resonant modes for formation of
magnetic moments along their specific axes. Side-by-side with the problems
of optically induced magnetization in homogeneous dielectric media (Rand,
Fisher, & Oliveira, 2008), attention was given to the task of magnetization of
dielectrics at the optical and IR frequencies, so-called magnetic light
(Miroshnichenko et al., 2012). The magnetic activity of media containing
the nanoinclusions can be controlled by means of Mie resonances, excited
due to scattering of incident waves on these inclusions (Zhao, Zhou,

Zhang, & Lippens, 2009).
Let us recall the Mie theory for a plane wave, scattered on an infinite
dielectric cylinder. Suppose that the wave is propagating in the z-direction
and its magnetic component Hy is oriented parallel to the cylinder (y-direc-
tion). The magnetic components of incident, scattered, and internal fields
H inc sc int
y , H y , N y are given by
X
Hyinc kr H0 im Jm kr expim,
m
X
Hysc kr H0 im am Hm1 kr expim, (68)
m
X
Hyint nkr H0 im cm Jm nkr expim,
m
where

1 dJm n dJm 2a
am Jm nJm n , , m 0, 1, 2 . ..
d d
(69)
dJm n dH 1 Jm am Hm1
Hm1 nJm n m , cm :
d d Jm n
Here, n and a are the refractive index and radius of the cylinder, is
the wavelength of the incident wave, Jm and H(1)
m are the Bessel and Hankel
functions of the first kind, respectively. These forms satisfy continuity of
the tangential components of the total electric and magnetic fields at the
cylindrical surface.
The origin of magnetic activity in this configuration is connected
!
with the displacement currents j d (OBrien & Pendry, 2002). These
currents, induced inside the cylinder in the planes perpendicular to the
!
cylinder axis due to temporal variations of electric displacement D
( Jackson, 1998)
!
! 1 @D
jd ; (70)
4 @t
can be greatly enhanced, when the size parameter (69) is resonant with an
eigenmode of the field oscillations inside the cylinder. A characteristic equa-
tion is obtained by setting the denominator of the coefficient am (69) equal
to zero. In this case (m-th order resonance), the relevant cm coefficient
dominates over the others and, the field inside the cylinder is described
mainly by the basis function Jm(nkr) (68). For the zeroth-order resonance,
only the azimuthally directed displacement current is enhanced and
results in the enhanced magnetic field along the cylinder axis. This effect
can provide the magnetic activity in the nonmagnetic dielectric inclusion;
herein the magnetic field is not confined inside the cylinder, but forms
some halo around it. In an array of identical cylinders, these localized mag-
netic modes could stimulate the macroscopic bulk magnetization of the
system, characterized by some frequency-dependent effective magnetic
permeability eff. This permeability can be presented in a dispersive form,
similar to the one, which was found at first for the arrays of split-ring metallic
resonators, embedded to the dielectric matrix (Pendry, Holden, Robbins, &
Stewart, 1999)
Y 2
eff 1 : (71)
2 20 + i
Here, the resonant frequency 0, determined from condition 0 (69),

depends upon the dielectric permittivity of cylinder and its radius a, is the
damping factor, Y is the geometrical factor connected with the fraction of
metamaterial, filled by dielectric cylinders. Note, that in some spectral range
the values Re eff can take on negative values; thus in a case > 0 this
metamaterial is characterized by a stop band, determined by an imaginary
p
part of the effective refractive index n eff ; the waves, belonging to this
band, can penetrate to the metamaterial as evanescent modes. This resonant
form of eff is mimicking the local dispersion, meanwhile the flexible non-
local dispersion can be provided by the prescribed variation of distances
between the neighboring cylinders or their arrays (Semouchkina,
Werner, Semouchkin, & Pantano, 2010).
The consideration above was focused on the magnetic activity of non-
conducting composites governed by Mie resonances in the buried dielectric
inclusions. If these inclusions are buried to the conducting medium then
another tunneling effects can come into play. Thus, the peculiar mechanism
of enhanced light transmission through a metallic film can be assisted by
tunneling between the resonating dielectric inclusions, which can sustain
localized resonances inside these films. An example of this effect was illus-
trated by light transmission through Ar films containing arrays of Si
nanospheres (Garcia de Abajo, Gomez-Santos, Blanco, Borisov, &
Shabanov, 2005). The resonant modes of the Si spheres can be found
through their Mie scattering coefficients by analogy with (68), herein the
characteristic equation, determining the eigenfrequencies of m-th mode for

these spheres, is (Smith & Fuller, 2002):
dHm1 dJm
2p Jm p1 Hm1 0: (72)
d d
Here, Jm and H(1) m are the spherical Bessel and Hankel functions, the
superscript p refers to either TM (p 1) or TE (p 2) wave polarization;
the size parameter and the relative refractive index are defined as
2anmed np
, ; (73)
nmed
np and nmed are the refractive indices of the particle and surrounding
medium, respectively, a is the particle radius. When these spherical Si inclu-
sions are arranged in square-lattice layers, the coupling between Mie modes
in the neighboring layers results in splitting and shifting of Mie resonances.
The refractive index nmed for Ar in the visible and near IR spectral ranges is
imaginary, so, the coupling among the nanospheres and between the
spheres and the film surfaces is mediated by evanescent modes. When
the imaginary part of dielectric function of Ar is small ImAr 1,
the transmittance of an Ar nanofilm with embedded Si spheres for some
wavelengths in the near IR range tends to 100% (Borisov, de Abajo, &
Shabanov, 2005).
3.4 Tunneling Phenomena in Negative and ENZ Metamaterials

The condition of formation of tunneling regime for EM waves in non-
magnetic dielectrics < 0 proves to be invalid for the metamaterials in
the spectral range, where both the dielectric permittivity and the magnetic
permeability have negative values. These composite, so-called negative,
media are known to possess a negative-refractive index n < 0 (Veselago,
1968). In the microwave spectral range such artificial materials have been
constructed by embedding the arrays of metallic split-ring resonators
(SRR) and wires in a host medium (Shelby, Smith, & Schultz, 2001). Unlike
the negative values of , provided by metallic components of these materials,
the design of their components, responsible for the negative values < 0,
seems to be a more sophisticated task; the creation of materials with spatially
distributed magnetic permeability is a hot problem too (Tretyakov, 2003).
Consider first the values and , varying in z-direction, and an EM wave,
propagating in z-direction too. To take into account the coordinate-
dependent values (z), distinguished from unity, as well as the case < 0,
it is worthwhile to introduce the generating function , generalizing the

representation of wave field components (1):
1 @ 1 @
Ex , Hy : (74)
c @t @z
Owing to the representation (74), the set of Maxwell equations for a
monochromatic wave is reduced to the equation
d 2 2 1 d d
+ 2 ; (75)
dz 2 c dz dz
Equation (75) is valid for arbitrary combinations of and , including, in
particular, so-called right-handed ( > 0, > 0), left-handed ( < 0,
< 0), and single-negative ( > 0, < 0 or < 0, > 0) materials.
The propagation of EM waves through sharp boundaries between
homogeneous right-handed (RHM) and left-handed (LHM) materials
was examined by Ziolkowski and Heyman (2001). To visualize some pecu-
liarities of the gradual transition layer between these media, it is expedient
to present the spectral and spatial dependences of and as
, z U z and , z U z; here U(z) is an arbitrary real
odd function, the dependence () in the metamaterials is usually pres-
ented in a Lorentz-like resonant form (Pendry et al., 1999). Introducing
the new variable (23), we can eliminate the function U(z) from the
Equation (75):
d2 2 n20
+ 2 0, n20 : (76)
d2 c
To describe the evolution of wave in the metamaterial, characterized

by gradual changes of and from positive to negative values, one can use
the simple function U z thz=L (Dalarsson, Jaksic, & Tassin, 2009);
substitution of this function to (19) brings the value of new variable
L ln chz=L . The wave propagating in this medium in z-direction
is written as
h z
iiqL n0
ch , q : (77)
L L
Some unusual properties of this propagating wave deserve to be
stressed:
i. Near the center of the transition layer jzj < L the distances between
zero-crossing points of Re are unequal, whereas deep in the media
jzj L these distances are equal, and expression (77) describes the
harmonic wave expiqz;
p
ii. The refractive index n0 and impedance Z = retain their
values for any given frequency in both half-spaces z 0 and z 0;
iii. The constancy of n0 and Z indicates that the wave travels from z 0
to z 0 without reflection from the plane z 0.
Note, that the generating function can be introduced, side-by-side with
(74), by an another representation
1 @ 1 @
Ex , Hy : (78)
@z c @t
The equation, governing this newly introduced function, coincides

with (75) after the replacements ! , ! . This feature of representa-
tions (75) and (78) can be viewed as a consequence of the invariance of
! ! ! !
Maxwells equations under the replacements B ! D , E !H (duality
principle) (Landau & Lifschitz, 1984).
The drastic changes of the aforesaid wave field structure (iiii) can
be stipulated by an inclined propagation of EM wave through the transi-
tion layer between RHM and LHM. These changes were shown by
Litchinitser, Maimistov, Gabitov, Sagdeev, and Shalaev (2010) for the
oblique propagation of TE wave in the structure, containing the homo-
geneous layers with positive index 0 , 0 and negative index
0 , 0 , separated by a transition layer of width 2L, where
both and are presented at the familiar forms , z U z
and , z U z; herein the function U(z) is the linear func-
tion of the coordinate z along the normal to the planes z 0 and
z 2L: U z 1 z=L. Unlike the normal impinging of EM wave,
the oblique propagation is characterized by resonant absorption of
wave and enhancement of its magnetic field in the vicinity of the point
z L, where the refractive index changes its sign. This field enhance-
ment, occurring near the zero refractive index, when ! 0, 1,
resembles a similar effect, well known in the physics of heterogeneous
plasmas (Ginzburg, 1970).
The unusual properties of EM waves in epsilon-near-zero (ENZ)
media (Silveirinha & Engheta, 2006) nowadays attract a growing
attention. Within the ENZ material, the wave possesses small spatial var-
iations, an extremely large phase velocity, and its wavelength tends to
infinity; in other words, this wave exhibits a static-like character;
the source of radiation, embedded to the ENZ medium, can generate
an extraordinarily directed wave beam. These materials have been
realized in the microwave (Liu et al., 2008), mid-infrared (Mocella,
Dardano, Rendina, & Cabrini, 2010), and visible (Vesseur, Coenen,
Caglayan, Engheta, & Polman, 2013) spectral ranges. However, the
ENZ materials having a permittivity near zero and a permeability
p
about unity, possess a near-zero refractive index n and high opti-
p
cal impedance Z = ; this combination results in a large reflec-

tion from the interfaces of such materials. In contrast, impedance
zero-matched all-dielectric metamaterials, in which both and tend
to zero, eliminate these strong reflections; moreover, these structures
can provide the isotropy of optical properties and small loss (Zhao
et al., 2009). These phenomena open the new avenues for effective
tunneling of EM waves through the ultranarrow channels and wave-
guide junctions filled with the ENZ material (Edwards, Alu, Young,
Silveirinha, & Engheta, 2008). Analogous tunneling effects have been
observed in 90 waveguide bends and U-shaped waveguides, filled with
-near-zero media (Alu & Engheta, 2007). Unlike the energy squeez-
ing achieved in a single-narrow region, the multi-passband tunneling
effect in multilayer ENZ channels with different tailored plasma fre-
quencies was considered by Lui, Hu, Zhao, and Luo (2009). Since
the frequency response in the wire-loaded waveguides can be modified
by variation of the lengths of wires, this tunneling effect proved to
be scalable over the wide frequency range (Siddique, Ramahi, &
Kashanianfard, 2010).
Note should be taken on the difference between the tunneling phenom-
ena in ENZ-based and wire-loaded metamaterials. The ENZ-based tunnel-
ing is characterized by tailored phase patterns (Alu, Silveirinha,
Salandrino, & Engheta, 2007), opening the possibility to transport an image
through the subwavelength hole (Silveirinha & Engheta, 2009). In contrast,
the wire-loaded waveguides are distinguished by large phase variations and
nonuniform spatial field distributions, resembling the classical FabryPerot
resonance (Zhang, Luo, & Mortensen, 2010).
3.5 Energy Transfer by Reactive Modes in the Near Zone

of the Source
Consideration of interference of evanescent and anti-evanescent modes in
the gradient layer has revealed the formation of a nonzero energy flux
through this layer (Section 2.2). This flux was shown to be formed by
the electric and magnetic components of a total field, presented by the super-
position of forward and backward waves inside the layer; it is essential that
the phase shift between these total components differs from /2. The similar
effect of energy transfer by evanescent modes can arise in the near zone
of radiation source, containing both propagating and nonpropagating
(reactive) components of EM field.
This effect can be visualized in the framework of the well-known
radiation pattern of a Hertzian dipole, oscillating with frequency along
the z-axis. The components of electric and magnetic fields, induced by these
oscillations, are written in the spherical coordinate system as

1 i 2 1 i
Er 2cos 2 2 k p, E sin 2 2 1 k2 p,
k r kr
k r kr
i p0 h !
i
!
H sin + 1 k2 p, E Hr H 0, p exp i k r t :
kr r
(79)
!
Here, p0 is the complex amplitude of dipole moment oscillations, r is the
radius vector, drawn from the dipole center to the observation point,
!
and are the angles between the z-axis and r and between the x-axis
!
and projection of r on the (x, y) plane, respectively. The total energy flow
I through the closed surface, surrounding the dipole, is determined by two
components of the dipole field, describing the radiation in the far zone
kr 1
4
E k2 p sin, H k2 p, I jp0 j2 : (80)
3c 3
The components of dipole field in a near zone, which can be viewed as
standing waves, surrounding the source of radiation,
2p cos p sin ikp sin
Er 2
, E 2 , H ; (81)
r r r
are characterized by the phase shift /2 between electric and magnetic compo-
nents; these reactive components do not contribute to the radiation flux I (80).
However, the radiation of two dipoles, oscillating with the same fre-
quency , can create the interference of reactive components of dipole
fields, providing the nonradiating energy transfer between dipoles. Consider
two dipoles, p1 p10 expit and p2 p20 expit, spaced by distance l
at the x-axis and oscillating along the z-axis and let us suppose the dipoles to
be located symmetrically with respect to the plane x 0 at the points with
coordinates l=2,0,0 and l=2, 0,0. The total energy flow, irradiated
by dipoles through the plane x 0, contains three terms:
Ix I1x + I2x + Iix ; the first and second terms
4 jp10 j2 4 jp20 j2
I1x , I2x ; (82)
6c 3 6c 3
define the energy fluxes, irradiated by each dipole independently, mean-
while the term Iix corresponds to the interference flux, determined by the
EM fields of both sources. If the distance l between dipoles is much less than
the radiation wavelength 2 c=, the main contribution to the interfer-
ence flux is provided by the reactive components of the near zone:
p10 + p20 3cos 2 1expit

Ez 0, y, z, t ,
R3 r
ikp10 p20 sincos expit l2
Hy 0, y, z, t , R + x2 + y2 :
R3 4
(83)
Substitution of field components (83) to the general formula for pointing

vector gives the expression for the flux of nonradiating energy transfer Iix
(Kolokolov & Skrotsky, 1992):
*

Im p10 p20
Iix : (84)
2l 3
Note, that the similar effect can occur, when a dipole is excited by the radi-
ation of an another source, and the induced dipole radiation proves to be
phase-shifted due to, e.g., dissipative phenomena in the surrounding medium.
The following salient features of these unusual regimes of energy extraction
from the reactive fields, surrounding the wave radiators, have to be stressed:
i. the flux Iix arises, when the dipoles oscillations are phase-shifted;
ii. the the flux Iix is localized mainly in the area between dipoles:
l=2 < x < l=2;

! !
iii. The spatial distributions of the E H components of electric dipole,

! !
oscillating along the z-axis, and H E components of magnetic
dipole, located in a plane, orthogonal to z-axis, are known to coincide
(Kong, 1986); thus the nonradiating energy flux, stipulated by the inter-
ference of reactive field components, can arise in a near zone between
the magnetic dipoles too.
It is remarkable that the concept of formation of a wave pattern, providing
the energy transfer by the nonpropagating near-field components associated
with the source, seems to be valid for the near zones of sources of different
physical nature. Thus, in the optical range, the lifetime of a molecule excited
state (molecular fluorescence) was shown to depend upon the nonradia-
ting decay due to energy dissipation into the surrounding environment
(Liaw, Chen, Chen, & Kuo, 2009). At the other range of electromagnetic
radiation spectrum, at the GHz range, the physically similar situation was
presented in the experiments of Ranfagni, Fabeni, Pazzi, and Mugnai
(1993): in these experiments with closely located transmitter and receiver
horns the mouth of the microwave horn launcher was considered as a source
of radiation, and the field between horns was viewed as a reactive field. The
analysis of the pulse delay on the way between two tilted horns, analogous to
the tunneling of microwaves in the sub-cutoff waveguide, was treated in
favor of the concept of energy transfer by reactive wave fields.
4. TUNNELING OF MICROWAVES THROUGH

GRADIENT TLs
A TL can be viewed as a unique combination of a widely used system
for transfer of electromagnetic energy with a flexible device for the demon-
stration and modeling of wave phenomena. The new trend in physics of TLs
is based on the use of metamaterials with negative values of and in the
design of these directional systems. The examples of this trend are the con-
trolled energy transfer through the TL, providing backward-wave radiation
(Grbic & Eleftheriades, 2002), an array of SRR inside the waveguide
(Baena, Jelinek, Marques, & Medina, 2005) or arbitrarily shaped plasmonic
channels and sharp bends (Alu & Engheta, 2007; Liu, Hu, Zhang, & Luo,
2009). TLs with continuously distributed dielectric and geometrical
parameters (for short, gradient TL) open one more new avenue for the con-
trol of microwave radiation flows. A theoretical analysis of these directional
systems illustrates the applicability of the concept of nonlocal HID to micro-
wave circuits. The propagation of EM radiation in these circuits can be
governed by the simultaneous action of geometrical dispersive effects, stip-
ulated by the shapes and sizes of perfectly conducting waveguides, and non-
local effects, arising due to heterogeneous distributions of capacity and/or
inductance in the circuit. Attention will be given below to the influence
of these phenomena on the tunneling of electromagnetic waves in
gradient TLs.
The series of electromagnetic tunneling phenomena have both optical
and microwave counterparts. These analogies promote the elaboration of
some general physical insight on these phenomena despite the drastic differ-
ence in their temporal and spatial scales. TLs can be used for the modeling of
effects of gradient nanophotonics, especially as the making of microwave
circuitry proves to be easier, than the fabrication of gradient nanofilms
(Nelin, 2007).
To emphasize the dispersive effects, let us consider a lossless TL. The
spatial-temporal variations of current I and voltage V are known to be
governed by the pair of equations, describing the propagation of wave in
z-direction (Kong, 1986):
@I @V @V @I
+ C z 0, + M z 0: (85)
@z @t @z @t
Here, C(z) and M(z) are the coordinate-dependent capacitance and
inductance per unit length, characterized by dimensionless functions
W 2(z) and F 2(z), respectively:
C z C0 W 2 z, M z M0 F 2 z: (86)
The heterogeneous part of a TL, occupying the region 0 z d, is
assumed to be located between its homogeneous regions z 0 and z d,
where the quantities C and M possess constant values.
The equations of a heterogeneous TL with continuously distributed
parameters (85) are used for analysis of HID of a coaxial waveguide, stipu-
lated by all-dielectric 3D gradient porous diaphragm, installed inside this
waveguide (Section 4.1); this structure which is characterized by some dis-
tribution of capacitance C(z), mimics the gradient nanofilm with the similar
profile of effective dielectric permittivity eff (20). Calculation of GHz
transmittance spectra for TEM mode for this structure, based on the exact
analytical solutions, describing the propagation of IR radiation through gra-
dient nanofilms, shows a good agreement with experimental measurement
of these spectra. The similar effect of HID is shown (Section 4.2) to provide
the formation of cutoff frequency and tunneling of waves in TL, character-
ized by different spatial distributions of capacitance and inductance. Tunnel-
ing of the fundamental TE10 mode through the gradient diaphragm with the
flexible distribution of eff, installed to the metallic waveguide, is considered
in Section 4.3; the reflectionless propagation of this evanescent mode,
analogous to the Brewster effect, is illustrated too.
The decrease of natural plasma frequencies of metals down to the GHz
range is considered in Section 4.4 as a perspective method for the creation
of a metamaterial with strong dispersion in the microwave spectral range.
The effect of nonradiating energy transfer by reactive waves in the near
zone of the source is illustrated in Section 4.4. The effective transmittance
of waves through a smoothly shaped subwavelength narrowing in the
waveguide is discussed in the Section 4.5. This analysis is continued by
the physically similar problem of tunneling of radiation through the
ultrathin slit, filled by a dielectric with constant permittivity, determined
by the geometrical parameters of narrowing. The last effect seems to be
rather intriguing, and the design of this configuration shows the richness
of possibilities offered by the transformation optics for control of the paths
of EM waves.
4.1 Microwave Analogy of a Gradient Photonic Barrier

Here, we take an approach, originating from the analogy between electro-
magnetic propagation on TL and plane wave propagation in an isotropic
medium with positive material parameters and . To use this analogy, it
makes sense to start from the simple case related to the TL with constant
inductance M M0 and coordinate-dependent distribution of capacity
C z C0 U 2 z. Introducing the generating function by equations
r
1 @ 1 @ 1 M0
V , I , v0 p , Z0 ; (87)
v0 @t Z0 @z C0 M0 C0
we will find that the second equation in (85) is reduced to identity, while the
function is governed by the first equation, which reads now as
@ 2 U 2 z @ 2
+ 0; (88)
@z2 v02 @t2
v0 and Z0 are the wave velocity and impedance in a homogeneous TL

(U 1) with the constant values of parameters M M0 and C C0 . The
replacement v0 ! c transforms Equation (88) to the one-dimensional version
of Equation (2); voltage and current waves (87) correspond, respectively, to
the electric and magnetic components of electromagnetic wave. Following
this analogy, one can expect a series of intriguing microwave effects in gra-
dient TL:
i. Formation of HID, both positive and negative, in the initially
dispersive-free TL;
ii. Non-Fresnel transmission/reflectance spectra for voltage and current
waves, described by the generalized formulae (28)(31) for the normal-
ized frequencies u =; to provide the characteristic frequency
inside the GHz range, the sizes of spatial distribution C(z) have to be
chosen in the cm scales.
iii. Nonreflection tunneling of voltage and current waves through the TL
in a low-frequency spectral range.
All these dispersive phenomena are caused by smooth spatial variations of its
impedance Z(z), determined by the distribution of capacity C(z). Attribut-
ing the values C0 and M0 to the empty segments of waveguides one can recall
the expressions for C0 and M0 (Collins, 1991)

20 0 r0
C0 , M0 ln : (89)
r0 2 r1
ln
r1
Here, r0 and r1 are the external and internal radii of the waveguide,
0 and 0 are the dielectric susceptibility and magnetic permittivity of
vacuum. Substitution of (89) to (87) yields the well-known result:
p1
v0 0 0 c.
To verify the formation of HID in the microwave range, let us consider
the interferometer-like system, built from two similar coaxial waveguides
(Shvartsburg et al., 2008). We will examine the propagation of fundamental
TEM modes, launched to this coaxial waveguide from the generator. This
regime is chosen due to the following salient features, approaching to the
properties of EM waves in free space:
Figure 7 Controlled formation of nonlocal dispersion in the coaxial GHz waveguide.

(a) Arrangement of plastic porous gradient diaphragm in the waveguide.
(b) Different porous gradient diaphragms, used for the creation of heterogeneity-
induced dispersion of coaxial transmittance.
i. The velocity of the TEM mode is known to coincide with the free-
space light velocity c.
ii. The TEM mode does not possess frequency dispersion.
iii. The polarization structure of the TEM mode contains only the trans-
versal electric and magnetic field components; unlike any other wave-
guide modes it is polarization does not contain any longitudinal
components.
One of these waveguides contains a plastic porous diaphragm, fabricated
from the dielectric with H with the air-filled pores (Figure 7a). The
ratio of plastic-filled and air-filled parts in each cross-section, varying along
the diaphragm, mimics the gradient profile of refractive index (20); unlike
the sputtered nanofilm the thickness of this diaphragm is about several cm, as
is illustrated by the ruler on Figure 7b. The effective value of the dielectric
permittivity for this porous structure eff can be found by means of the
Maxwell-Garnett formula (Maxwell-Garnett, 1904):
H PD + 1 P H + G1 P D H
eff , (90)
PD + 1 P H GP D H
here G is the volume part of medium, filled by the admixture with the value
of dielectric permittivity D ; in the problem under discussion H 2:3
and D 1. The polarization factor P, dependent upon the shape of admix-
ture particles, is known to possess the values P 0:5 and P 0:33 for the
lengthy cylinder and sphere, respectively.
To compare the microwave transmittance of a heterogeneous diaphragm
with the transmittance of a plane photonic barrier (20), the spatial distribu-
tion of eff (z) in the diaphragm has to follow the profile U2(z) (20):
eff n20 U 2 z, n20 H , (91)
herein the parameter G becomes coordinate-dependent G Gz. Using

Equation (90), one can write the spatial distribution G(z):
1 U 2 zP + 1 P H
Gz , (92)
H 1PU 2 z + 1 P
Our goal is to define the spectral range k < km , where the wave number
km is the lowest of the critical wave numbers kc, related to TE and TM
modes in the coaxial waveguide; herein the waveguide will support the
propagation of only fundamental TEM mode. The critical value kc for
TE modes is given by the first root of the dispersive equation
N0 kc r1 N0 kc r0
, (93)
J0 kc r1 J0 kc r0
J and N are the Bessel and Neuman functions, r0 and r1 are the internal and
external coaxial radii. The dispersive equation for the TM mode reads as:
N1 kc r1 N1 kc r0
: (94)
J1 kc r 1 J1 kc r0
Fixing the values r0 and r1, one can define the upper band of the spectral
range kc, related to the single-mode propagation of the TEM wave.
0.98
1
0.96
0.94
0.92
2
|T|2
0.9
0.88
0.86
0.84
0.82
0.8
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
f (GHz)
Figure 8 Transmittance spectra of coaxial waveguide, shown on Figure 7a, with porous
gradient diaphragm, mimicking the photonic barrier (20); the values of transmission
coefficient jTj2 are plotted versus the frequency f. Curves 1 and 2 relate to the exper-
imental and theoretical spectra.
Since the incident and transmitted TEM modes are propagating along
the coaxial with free-space light velocity v0 c , we can calculate the trans-
mittance spectra using equation (28), obtained for the photonic barrier, put-
ting there the refractive index of the substrate n 1. The transmittance
spectra of the coaxial waveguide r1 1 cm, r0 2:3 cm with this gradient
diaphragm, calculated for the layer with effective permittivity (90) by means
of standard continuity conditions for voltage and current and measured in
experiment, are shown on Figure 8; the smallest values of kc for TE and
TM modes, defined from Equations (93) and (94), are 2.63 and
4.88 cm1, respectively. Thus, choosing the value km 0:628 cm1 ,
related to the frequency m km =c 1:88
1010 rad=s ( fm 3 GHz),
one can consider the frequencies of spectral range f < fm , shown on
Figure 8; herein the frequencies in the observed range 13 GHz exceed
the critical frequency of barrier fc =2, where the value is calculated
by means of (22); for the parameters, related to Figure 7, fc 0.27 GHz.
Comparison of calculated and measured transmittance spectra on this figure
indicates a good agreement between the theoretical and experimental data,
visualizing the effect of HID of this microwave TL.
4.2 Evanescent Voltage and Current Waves in a Gradient TL

The generating function , introduced by the presentation (87), corre-
sponds to a TL with smoothly distributed capacity C(z) and constant induc-
tance M. The opposite caseconstant capacity C and distributed inductance
M(z)can be examined by means of an another method of introduction of
generating function
1 @ @
I , U Z0 : (95)
v0 @t @z
In this case, the first equation in (85) is reduced to an identity, while the
second one brings the equation, governing the function in a form, coin-
ciding with (88). However, to examine the combined influence of wave
propagation, produced by spatial distributions of both capacitance
C z C0 W 2 z and inductance M z M0 F 2 z of TL, special consid-
eration is needed.
To solve this problem, let us differentiate the first and second equations
in (85) with respect to the variables z and t, respectively, excluding the func-
tion V we will obtain one equation governing the unknown distribution of
current I:
@2I @2I 1 dC z @I
C zM z : (96)
@z2 @t 2 C z dz @z
Note, the similarity of Equations (96) and (75), describing the waves in a
metamaterial with coordinate-dependent parameters (z) and (z). The
replacement I ! V , C z ! M z in (96) brings the similar equation,
determining the distribution of voltage V. Due to using a new variable ,
distinguished from (23)
z
W 2 z1 dz1 , (97)
0
Equation (96) reads as
@ 2 I 2 F 2 z
+ I 0: (98)
@2 v2 W 2 z
By describing the distributions F2(z) and W2(z) inside the barrier of

width d with the help of characteristic lengths l1 and l2 (Shvartsburg &
Maradudin, 2013)

z z 1
W z 1 + , F z 1 + , (99)
l1 l2
we can study the effects caused by the increase or decrease of capacity and
inductance of TL in a general form, considering both positive and negative
values of the lengths l1 and l2 independently. To distinguish these lengths,
related to models of different physical quantities, from the lengths L1 and
L2, describing the distribution of one quantity, e.g., the dielectric permittiv-
ity in model (99), lowercase letters are used in models (99).
Substitution of function W2(z) (99) into (97) yields the explicit expres-
sion of variable via z:

z 1
z 1+ : (100)
l1
Owing to (100) the product of the functions W(z) and F(z) reads as a
function of :

1
F zW z U 1 + , (101)
l
l1 l2
l : (102)
l1 l2
By substituting (101) into (98), we can rewrite this equation in -space in
a form similar to (88):
d2 u 2 2
+ U u 0: (103)
d2 v02
This equation is easy to solve by means of the algorithm, used above in
the Section 2.1. Introducing the new variable

l1 z + l2
U 1 d1 l ln , (104)
0 l2 z + l1
we can present the solution of Equation (103) in the form of forward and
backward waves, traveling along the -axis
Ar expiq + Q expiq
u p : (105)
U
The wave number q in (105) corresponds to the plasma-like dispersion of
the gradient layer
s
2 v0
q 1 2, : (106)
v0 2jlj
The characteristic frequency in (106) depends via the parameter l on

the spatial scales of variations of the capacity and inductance l1 and l2. The
low frequencies < relate to the tunneling of voltage and current waves
in the gradient TL.
4.3 Brewster Effect for the TE10 Mode Tunneling Through

a Gradient Diaphragm in a Waveguide
Microwave guiding circuits are widely used for the research of EM wave
tunneling phenomena in the GHz range, providing the experimental mea-
surements in a very accessible temporal range of the order of nanoseconds.
The first experiments, visualizing the tunneling of the widely used TE10
mode in a rectangular perfectly conducting waveguide with step narrowing
(undersized waveguide) had simulated a quantum mechanical rectangular
potential barrier (Ranfagni et al., 1993). In contrast, this section illustrates an
another mechanism of tunneling of the same mode in the same waveguide,
containing, instead of a narrowing, the gradient all-dielectric diaphragm
characterized by some smooth distribution of the dielectric permittivity
(z) along the direction of wave propagation z; herein the waveguide
cross-section with sides a and b (b > a) remains unchanged. Choosing the
coordinate axes x and y parallel to these sides a and b, respectively, one
may characterize the polarization structure of TE10 mode by the field com-
ponents Ex, Hy, Hz (Collins, 1991). Unlike the normal incidence of TEM
mode on the diaphragm in the coaxial waveguide (Section 4.1), the config-
uration under discussion can be viewed as the inclined incidence of waves,
reflected from waveguide walls, on the diaphragm boundary, located at the
waveguide cross-section plane. While examining this configuration we have
to consider three problems:
i. To find the exact analytical solution for the wave components inside the
gradient diaphragm and to single out the spectral range, corresponding
to the tunneling regime;
ii. To calculate the complex reflection coefficient for the tunneling
regime;
iii. To visualize the conditions for reflectionless tunneling of the
TE10 mode.
Before the analysis of these problems, it is worthwhile to express the wave

field components Ex, Hy, Hz through the generating function . The equa-
tion governing the propagation of a wave with frequency in the TE10
mode inside the diaphragm coincides with (2). The diaphragm is supposed
to occupy the layer 0 z d; outside this layer we have 1 and the gen-
erating function for single-mode waveguide, satisfying to the boundary
conditions Ex jy0 Ex jyb 0, is:
y
Asin expiz t : (107)

b
The propagation constant and the cutoff frequency 0 for TE10
mode are
r

2 c
k2? , k? , 0 : (108)
c b b
Let us consider the distribution of the effective dielectric permittivity

(z) inside the diaphragm in the familiar form (9), containing, however,
an another function W(z) (Shvartsburg, Kuzmiak, & Petite, 2007):
1 W 2 z h z
z
i1
U 2 z 1 + , W z cos + M sin , (109)
g g L L
here the real positive quantities g, M, and L are the free parameters of dis-
tribution (109), dependent upon the diaphragm width d, minimum of
refractive index nmin, and half width of nmin. Considering the symmetrical
profile (W 0 W d 1), one can link these quantities with both physical
and geometrical parameters of the profile

d M2
2 arctgM, nmin n0 1
2 2
: (110)
L g1 + M 2
Compared to the exactly solvable model (z) (20), characterized by two

free parameters, the profile (109), having three free parameters, is more flex-
ible; in particular, the profile (109) is suitable for the description of different
slopes at the wings of the distribution (z), when its width d and the value
nmin are fixed (Figure 1, curve 2).
Now we can deal step-by-step with three abovementioned problems.
i. Let us seek the solution of Equation (2) in a form, similar to (107):
f z y
A p sin : (111)
W z b
The unknown function f(z) is governed due to substitution of pre-

sentation (111) and distribution (109) to equation (2). Introducing the
variable and normalized variable
z
W z1 dz1 , (112)
0
2 z
3
1 6
1 + m + tg
7 p
p ln 4 2L
z
5 , m 1 + M 2 M, (113)
L 1 + M2 1 m tg
2L
p
1 + M2
0 , 0 ln m + , jz0 0 , jzd 0 , (114)
L
one can write an explicit expression for the function W(z) via the
coordinate :
ch
W p , W 0 W 0 1: (115)
1 + M2
Finally, omitting the phase factor expit, we obtain the equa-
tion, governing the function f(), in a form, familiar from quantum
mechanics (Schiff, 1968):

d2 f
q + 2 f 0,
2
(116)
d2 ch

1 1 Ln0 2
q2 ,
4 g1 + M 2 c
(117)
1 2 Ln0 2 1
+ k? L 1 :
4 c g
Equation (116) is known to have a pair of linearly independent solu-
tions, presented by the hypergeometric functions (Abramowitz &
Stegun, 1968). These solutions have to be used for the calculations
of reflection spectra for all the waves propagating along the waveguide.
However, our goal is to reveal an example of tunneling in a wave-
guide through the gradient diaphragm; thus, let us turn aside from the
layout of general solution of Equation (116), focusing the attention on

the special case 0 (117); manipulations with equalities (117) show,
that in this case the value q2 is
" #
2
1 1 + 4 k ? L
q2 1 : (118)
4 g 11 + M 2
Moreover, we will consider here the case q2 > 0, when the solution
of Equation (116) is presented by evanescent and anti-evanescent waves
in -space. Under these conditions the generating function (111),
corresponding to the tunneling regime, can be written as
y
r
z
z
Asin cos + M sin expq + Q expq, (119)

b L L
ii. Substitution of the generating function (119) to (1) brings out the com-
ponents of the TE10 wave inside the diaphragm and hollow parts of
waveguide, respectively. Consider the wave incident on the boundary
of the diaphragm (plane z 0). Making use of continuity conditions on
this plane and on the back side of diaphragm, we will find the complex
reflection coefficient R:

R , t th2q0 ,
1 + i 2
p
t 42 L 2 + M 2 + 4q2 1 + M 2 4qM 1 + M 2 ,
p
1 t 42 L 2 M 2 4q2 1 + M 2 + 4qM 1 + M 2 , (120)
p
1 t 42 L 2 M 2 4q2 1 + M 2 + 4qM 1 + M 2 ,
p
2 4L 2q 1 + M 2 Mt :
Formulae (120) present the reflection coefficient R for the TE10

mode, which is propagating in the waveguide, traversing the diaphragm
in a tunneling regime.
iii. Putting in (120) the value 0, we will obtain the condition for
reflectionless tunneling through diaphragm for the mode, propagating
in the empty part of the waveguide with propagation constant (108)
p
4qM 1 + M 2
th2q0 2 : (121)
M + 4q2 1 + M 2 + 42 L 2
It is remarkable that the frequency of tunneling wave was, in fact, defined

from the condition 0 (116) by means of the profile parameters n0, g, L
and waveguide size b; this frequency does not depend upon the diaphragm
parameter M. For example, taking n0 1:55, g 2:5, L 0:510 cm, and
b 2:3 cm, we will find the frequency of tunneling wave
4:2
1010 rad=s; the cutoff frequency of this waveguide in TE10 mode
is 0 4:1
1010 rad=s. Formula (121) describes the wave reflection coef-
ficients from different dielectric diaphragms subject to the values M charac-
terizing the wave number q (118) and the depths of modulation of their
dielectric permittivity; thus, the value M 10 corresponds to the vanishing
of reflection; herein the thickness of diaphragm, calculated from (112), is
d 1.5 cm; the smallest value of refractive index at this profile is (110)
nmin 1:205.
The polarization structure of TE10 mode, containing Ex, Hy, Hz com-
ponents, may be compared with the s-polarized wave in free space, incident
on the interface under an angle , determined by the equation
c
sin : (122)
b
For the aforesaid values of and b, Equation (122) brings the value
77 .
The abovementioned reflectionless propagation of this mode through
a dielectric layer can be viewed as an analogy of the Brewster effect,
which is known to exist for TM waves at the interface between two
distinct homogeneous dielectric media. Nondestructive optical control,
based on Brewster angle monitoring, was developed by Lekner (2000).
This effect for TE waves was observed on the boundary of two magne-
tic media (Futterman, 1995) and on the interface of negative-index
materials (Fu et al., 2005). Unlike this, the Brewster effect is shown here
to arise on the interface of gradient dielectrics (Erokhin, Zueva, &
Shvartsburg, 2013) and, moreover, the microwave energy flow, incident
under the Brewster angle, can propagate in a reflectionless tunneling
regime.
Side-by-side with the porous dielectrics some conducting gradient mate-
rials have been regarded as promising candidates for tunable microwave
applications. Thus, the compositionally graded ferroelectric (Bax Sr1x)
TiO3 film (x 0, 0.2, 0.4, 0.6, 0.8, 1.0) was shown to be an effective
frequency-dependent phase shifter in the range 812 GHz (Lee et al.,
2003); these phase shifts are electrically tunable by a dc bias voltage. An
another trend in using of heterogeneous conducting media for creation of

dispersive structure in the microwave range is discussed below.
4.4 GHz Resonance in Periodic Arrays of Plasmonic Nanowires

The local dispersion of waves in some specimens proves to be dependent
upon the shape and sizes of these specimens. An example of such an
effectthe shift of the plasma frequency in the dielectric spheroid subject
to the ratio of its axeswas mentioned above (Section 3.1). Now, we will
consider the effect of depression of plasma frequencies in metals from visible
and near UV ranges into the GHz range. This effect can be achieved by
means of a periodic structure, consisting of thin parallel metallic wires with
radius r and electron density N; the spacing between wires is a (Pendry,
Holden, Stewart, & Young, 1996). The current I, produced by electron
motion with velocity v along the wire (I r 2 nev) creates a magnetic field
H, circling the wire
2r 2 nev
H R , (123)
cR
where R is the distance from the wire axis in the plane, orthogonal to this
!
axis. Expressing the magnetic field in terms of the vector potential A
! !
H R rot A , (124)
and substituting from (124) into (123) one can find the vector potential
(Pendry et al., 1996):
2r 2 nev a
A R ln : (125)
c r
Electrons in a magnetic field are known to have an additional contribu-
tion to their momentum of eAc 1 , and therefore their momentum per unit
length of the wire is
2 2 e2 n2 r 4 a
r 2 neAc 1 ln : (126)
c2 r
Introducing the effective mass of the electron in the current-carrying
wire meff and equating (77) to the another expression for the momentum
per unit length of the wire meff r2nv, we find meff:
2e2 nr 2 a
meff ln : (127)
c2 r
This effective mass, stipulated by the wire self-inductance, can exceed
drastically the value of the electron rest mass me: thus, in case of thin
aluminum wires (r 106 m, n 1.8
1029 m3, a 5
103 m) the effec-
tive mass is meff 2:72
104 me . The effective density of electrons in the
periodic structure as a whole is
r 2
neff n : (128)
a2
Substituting both the effective mass meff and effective density neff into the
definition of the plasma frequency 2p 4e2 neff =meff , we can calculate the
frequency, characterizing this plasma-like medium:
2c 2
2p a
: (129)
a2 ln
r
Thus, the large self-inductance of a thin wire is mimicking the increase of
electron mass. The periodic structure composed of infinite wires arranged in
a cubic lattice can be considered as an effective plasma-like medium. This
man-made medium, characterized by effective dielectric permittivity eff
and plasma frequency p
2p
eff 1 , (130)
2
possesses the following properties:
i. This value p for the effective medium, determined by the geometrical
factors and independent of the electron density of metallic wires, can be
several orders of magnitude less than the plasma frequency in metals;
thus, for the aforesaid values of geometric factors a and r we have from
(129): p 5:15
1010 rad= s ( f 8.2 GHz).
ii. Until the wavelength of external radiation, incident on this microstruc-
ture, exceeds essentially the period a, this regular array of thin wires
appears as a homogeneous dielectric medium, characterized by effective
dielectric permittivity (130); thus in the aforesaid case the wavelength ,
corresponding to p, is 35 mm, which is much greater than the
lattice spacing a 5 mm.
iii. The microstructure under discussion can be viewed as a high-pass filter

for microwaves; its opacity for low frequencies < p resembles the
cutoff effect, inherent to metallic films in the UV range.
4.5 Propagation of Microwaves Through a Subwavelength

Narrowing in a Waveguide
The consideration of tunneling phenomena in Section 2 was based on the
concept of plane waves, where the optical properties of a medium in the
direction, orthogonal to the direction of propagation, were supposed to
be uniform. The radiation losses, governed by the energy transfer by evanes-
cent modes along the plane wave front, do not occur in the framework of
this concept. An opposite situation arises, when the radiation flow is guided
by a narrow channel, formed inside the dielectric by its optically denser part.
If this channel contains a smoothly shaped subwavelength narrowing, some
nondissipative energy losses, stipulated by the energy transfer by evanescent
modes outside the channel, will appear. To evaluate this effect, we will
examine the propagation of radiation along a cylindrical fiber, fabricated
from a uniform dielectric with smoothly variable cross-section, having
the diameter of narrowing noticeably less than the radiation wavelength
(Vahala, 2003). The radiation losses caused by the deformation of an optical
fiber are known to grow drastically in a case of subwavelength narrowings,
so-called nanotapers (NT). Herein, the dynamics of percolation of evanes-
cent modes through the transparent NT is very sensitive to its shape and spa-
tial sizes. The radiation losses are evaluated below for the adiabatic tapering
of the waveguide, assuming slow dependence of its radius r(z) upon the
coordinate z along the fiber axis, the cylindrical symmetry being conserved.
Let us consider the fundamental mode of a fiber with NT, propagating
primarily outside the narrowed area. The propagating constant of this mode
(z), which is close to the propagation constant of an unperturbed fiber 0, is
a slowly varying function of coordinate z:
q
2 z
z 20 + 2 z 0 + : (131)
20
Here, (z) is the transversal component of propagation constant, which is
supposed to be relatively small, z 0 . The problem of losses under the
condition of adiabatic variation of (z) (131) was treated analytically for the
Laguerre-Gaussian mode in an NT of biconical shape (Sumetsky, 2007),
where
0
z q : (132)
1 + z=L 2
Here, L and 0 are, respectively, the characteristic scale of nonuniformity

of the NT and the value of the transverse component of propagation con-
stant in its middle. The influence of the distribution 2(z) (132) resembles the
effect of a parabolic potential barrier. This biconical NT was shown to
exhibit exponentially small radiation losses, which can be determined as a
tunneling rate through some effective potential barrier. The attenuation
parameter (z) describes radiation power at the cross-section of the NT with
coordinate z:
z
I z I z0 exp zdz : (133)
z0
Calculation of the radiation flow density along the radial direction for the
NT with a biconical shape (132) yields the following expression for the
parameter (z):

2 2 z 20 L
z exp : (134)
0 20
Thus, the radiation losses are distributed along the NT. This analysis illus-
trates the weakening of transparency of smoothly shaped subwavelength
narrowings in a dielectric waveguide, stipulated by the evanescent tails of
a guided mode, dependent upon the geometric parameters of narrowing,
L and 0. This weakening is an important factor, restricting the efficiency
of wave flow transfer through the subwavelength slit in a transparent material.
Another family of heterogeneous wave barriers, supporting the resonant
tunneling of light, may be found in directional systems, containing a
narrowing, confined by curvilinear interfaces. This peculiar type of
heterogeneity-induced tunneling proves to be controlled by the geometrical
parameters of the confining interfaces. To examine this effect, let us consider
again the propagation of the TE10 mode with EM field components
Ey, Hx, Hz along a single-mode waveguide with axis z and thickness
d d=2 x d=2, filled by a dielectric with dielectric permittivity ;
to avoid any confusion with the tunneling problem, considered above in
the Section 4.3, a slightly changed notation will be used below. The
waveguide contains a symmetric smoothly shaped narrowing in the area
b=2 z b=2, and the planes x d=2 and x d=2 are assumed to
(a) Hx
Hz
Ey
X
b0 x = d/2
a<0 a>0
z
O
b 0
x = d/2
b
(b)
1 2 3
1
0.8 4
|T|2
0.6
0.4
0.2
0
700 750 800 850 900 950 1000 1050 1100 1150 1200
l
Figure 9 Tunneling of fundamental mode through the smooth narrowing in the
waveguide. (a) Geometry of the narrowed waveguide. The following geometrical
parameters are indicated: the narrowing length b, the distance between the wave-
guide walls d, and the distance between the tops of the coordinate lines 0 and
0 determines the minimal width s. The polarization structure of the guided wave
is shown on the inset. (b) Dependence of the transmittance upon the slit width s in
the range 225 nm < s < 300 nm, while b 1400 nm, d 500 nm are kept fixed; curves
1, 2, 3, and 4 relate to the values of width s 225, 250, 275, and 300 nm, respectively.
be perfectly conducting; the least width of narrowing is less than the wave-
length in the waveguide (Figure 9a). To avoid massive algebra, the analysis
below is restricted to the consideration of a single mode (Shvartsburg &
Kuzmiak, 2010).
Let us find the reflection coefficient of the narrowing, formed by ellip-

tical surfaces. To do this, it is convenient to introduce in the (x, z) plane the
curvilinear coordinate system (, ), formed by mutually orthogonal ellipses
and hyperbolas (Abramowitz & Stegun,1968):
x ach sin , z ash cos: (135)
The values < 0 ( > 0) relate to the half-plane z < 0 (z > 0), the line
0 corresponds to the x-axis. Analogously, the values > 0 ( < 0) relate
to the upper (lower) half-planes, the line 0 coincides with the z-axis. Let
us consider, e.g., the slit, formed by symmetrical hyperbolas, located
between the points x d=2, z b=2 and x d=2, z b=2. In the
coordinate system (, ) these hyperbolas coincide with the coordinate lines
0 and 0. Designating the smallest width of the slit (the distance between
the tops of hyperbolas) as s, one can define the parameters 0 and a:
p!
d 2 s2 s
0 arctg , a : (136)
b 2jsin 0 j
The wave equation governing the generating function inside the slit is
derived from the Maxwell equations. Using the (, ) coordinates and sep-
arating the variables F f , one can find the equations, determining
the unknown functions F() and f():
d2 F 2 2 a2
+ p 2
ch A F 0, p 2
, (137)
d2 c2
d2 f
2 + p A sin f 0:
2 2
(138)
d
Here, A is some dimensionless constant, which is determined from
the boundary conditions. Rigorously speaking, the regular solutions of
Eqs. (137) and (138) can be written in terms of Mathieu functions
(Abramowitz & Stegun, 1968). However, for our goals, it is convenient to
present the solutions of these equations directly by power series. Calculation
of the reflection/transmission coefficients R and T is based on the match-
ing of field components beyond of the narrowing z b=2, z b=2
with the field components inside the narrowing (b=2 z b=2). The
obtained expression for jT j2 illustrates the influence of each of the geo-
metric parameters of the slit (distance d, narrowing length b, and its width s)
on the tunneling of waves through the slit. Thus, the transmittance
spectra jT j2 as functions of the variations of slit width, while the other

parameters are kept fixed, are shown in Figure 9b. The narrow peaks in
transmittance with jT j2 ! 1 belonging to the fundamental TE10 mode
start to appear at the lower end of the wavelength range close to
700 nm, when s 225 nm (curve 1); the other peaks arise, when the
width s is increased, at the longer wavelengths (curves 2 and 3). The subse-
quent increase of the slit width results in the decrease of the transmittance
(curve 4).
The effect of the resonant tunneling of the fundamental mode of radia-
tion through a subwavelength slit, formed by a smoothly shaped narrowing
in the waveguide, is caused by the interference of waves, reflected from
the different parts of the curvilinear slit with the different phases. Unlike
the exponential weakening of the fundamental mode, tunneling through
the rectangular undersized segment in the waveguide (Ranfagni,
Mugnai, Fabeni, & Pazzi, 1991), an effective energy transfer by evanescent
waves proves to become possible due to the geometry of the curvilinear slit.
This interference of evanescent and anti-evanescent waves is shown to pro-
vide the possibility of nonreflection transmittance for waves with wave-
lengths 2.53
longer than the width of the slit.
An another method of tunneling of electromagnetic energy through
subwavelength channels is based on transformation optics (Leonhardt,
2006). This transformation, which compresses the original space inside
the waveguide into a much narrower region, provides the confinement
of EM waves in a channel, whose cross-section is much less than the
free-space wavelengths. Owing to this transformation one can determine
the material parameters, which can be achieved using natural materials.
Let us consider this approach in order to obtain the values of material param-
eters, required to squeeze and guide the electromagnetic radiation in a sub-
wavelength channel. For simplification, we will restrict our analysis to
the case of a waveguide, formed by two parallel plates, separated by a dis-
tance h; the waves are traveling between these plates in the z-direction.
A trapezoidal scatterer, formed by the broken straight lines, with height
h1, short side 2l and long side 2l + d is embedded in this waveguide across
the z-direction and along the x-direction; herein the thickness of the chan-
nel between the short side of scatterer 2l and the waveguide wall h h1 is
much smaller than the free-space wavelength. The boundaries of the wave-
guide and the trapezoidal sides are supposed to be perfectly conducting.
Introducing the coordinate y in the direction, normal to the conducting
plates, we will consider the transformation, described by the following

equations (Zhang et al., 2010):

0 h1 l+z
y y 1 + h1 1 + , f or l + d < z < l,
h d

0 h1 lz
y y 1 + h1 1 + , f or l < z < l + d, (139)
h d

0 h1
y y 1 + h1 , f or l < z < l,
h
x0 x, z0 z:
Using the standard algorithm for the calculation of the transformed i,g
and i,g tensors (Pendry, Schurig, & Smith, 2006), we find all the compo-
nents of these tensors, which prove to be positive constants, depending upon
the geometrical parameters of waveguide and scatterer h and h1:
2
h
xx xx , zz zz yy yy 1: (140)
h h1
For a certain TE (TM) mode, only xx, zz, yy (xx, zz, yy) enter into
the Maxwell equations; thus the material in the channel can be replaced
by the isotropic material with parameters
h2 h h1 2 , 1 or h2 h h1 2 , 1, (141)
To examine the possibility of tunneling through this narrow channel,
one has to recall the dispersive equation for the waveguide
2 m
k2z + k2y 2
, ky , m 1,2,3 . .. (142)
c h
the values and are given in (141). When becomes large enough, the
value ky will change from imaginary to real, providing the tunneling of
the wave through the channel.
Moreover, the EM transmission in a hollow waveguide, periodically
loaded with SRR, may become feasible within a certain frequency band,
even if the transverse dimensions of the waveguide are much smaller than
the associated free-space wavelength (Marques, Martel, Mesa, & Medina,
2002). The systematic analysis of similar problems in the framework of trans-
formation electromagnetics was presented by Ochiai, Leonhardt & Nacher
(2008).
5. CONCLUSIONS: THE MYSTERY OF PHOTON

TUNNELINGPARADOXES AND PERSPECTIVES
Although the simplest manifestation of light tunnelingTIR
was examined by means of the Maxwell equations more than a century
ago (Eikhenwald, 2009), the mystery of tunneling still gives to hot debates
and controversy. These debates are sharpened due to the link between the
tunneling processes and their superluminal aspects. To compare and contrast
the different concepts in the underlying physics of tunneling, some view-
points concerning the evanescent waves in all-dielectric structures are briefly
marked. The new trends appearing due to controllable heterogeneity of
these structures are stressed below.
5.1 Tunneling and Nonlocality

Consideration of HID (Section 2.3) has shown the significance of non-
locality effects for the tunneling of waves in gradient dielectrics: the charac-
teristic frequency (22), separating the spectral ranges of propagating and
evanescent waves, was shown to depend upon the first and second deriva-
tives of the spatial distribution of the dielectric permittivity (20). Together
with the formation of (22), the series of electromagnetic tunneling phe-
nomena prove to be influenced by nonlocality. Thus, an important example
of this influence was illustrated by the determination of velocity of energy
transfer through a photonic barrier by evanescent modes. This velocity vE
can be found for normal incidence from the z-component of the Poynting
vector (27) and the energy density W for a dispersive nonmagnetic medium
(Landau & Lifschitz, 1984)

Pz 1 d !2 ! 2
vE , W E + H : (143)
W 16 d
Thus, considering the model of a homogeneous plasma layer (plasma
frequency p), deposited on a thick substrate with constant refractive index
n, one can find for the velocity vE for the wave with frequency , tunneling
through this layer ( < p )
!
2 1
vE p
2n 1 + n2 + 2 : (144)
c
0.6
0.5
1
0.4
2
V (x)
0.3
3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
Figure 10 Spatial distributions of normalized velocity of energy transfer V x vE =c by
waves tunneling through the gradient barrier (y 0.75, n0 2.3) located on a homoge-
neous substrate with refractive index n 1.47, x z=dthe normalized coordinate
across the barrier. Curves 1, 2, and 3 relate to the normalized frequencies u 1, 1.5,
and 2.
Note, that the velocity of tunneling through the homogeneous barrier vE

is subluminal and coordinate-independent (Frias, Smolyakov, &
Hirose, 2011).
In contrast, the velocity of energy transfer in a tunneling regime through
the gradient barrier, deposited on the substrate with constant refractive index
n, proves to be coordinate-dependent. Thus, substitution of expressions for
modes (26), tunneling through the dielectric barrier (20), to formulae (27)
and (143) yields the value for the normalized velocity of nonlocal energy
transfer by an evanescent wave V (Shvartsburg & Maradudin, 2013):
vE 4 z
V x ; x ; 0 x 1; (145)
c U xx d
where is the profile of refractive index inside the barrier (20), the dimension-
less function is determined by the spatial distribution of field components.
A graph of V V x, presented on Figure 10 for different frequencies,
shows, that the velocity vE is spatially variable and highly dispersive. Note,
that vE depends on the global properties of the photonic barrier (20) via the
parameters y L2 =2L1 , with n being a local property. The difference
between the nonlocal velocity of energy transfer and the group velocity
in finite photonic structures was stressed by de Dios-Leyva and Drake-

Perezrez (2012).
The velocity of energy transfer by evanescent modes can be viewed as a
tunneling velocity of light in the forbidden zone (Li & Wang, 2001). It is
essential that the time tE of traversing the photonic barrier with thickness
d in the tunneling regime
d
dz d
tE < t0 , (146)
0 vE z c
is subluminal in this approach. This definition eliminates the problem of
superluminality, inherent to some other theories of tunneling of light
(see, e.g., Recami, 2004, and the references therein).
While considering the tunneling phenomena in the gradient dielectrics,
we have focused attention on the nonlocal effects in the far zone of the
source, e.g., on the formation of a cutoff frequency in the transparent layer
(20) or on the deformation of the polariton gap (56). However, there is an
another group of tunneling phenomena, inherent to the near-field zone of
the source kjzj 1. These phenomena, independent of the macroscopic
! !
parameters r and/or r , originate from an inability to localize photons in
space. This inability results in nonlocal interaction between the source and
receiver of photons by means of evanescent waves. The source region of the
photons encompasses the near-field zone; for an electric point dipole in a
vacuum this zone has a R3 profile (see Section 3.5). Optical tunneling phe-
nomena between the source and detector may appear, if these systems
are brought into near-field contact. Herein, the photons are known to be
!
associated with the transverse part E T of the electromagnetic vector field
!
E (Cohen-Tannoudji, Dupont-Roc, & Grynberg, 1989); however, the
! !
total E field in the near zone contains the longitudinal component E L , so
! ! !
E E T + E L . Recalling the formulae (81) determining the transverse E
and longitudinal Er components of the electric point dipole near-field zone
!
one can see, that in this zone the total field E satisfies simultaneously the two
equations:
! !
div E 0, rot E 0: (147)
Note, that the conditions (147) are fulfilled only for the evanescent
waves, damping at the vicinity of dipole in vacuum as R3 . Although the
!
near-field zone is matter-free and the divergence of the total field E is zero,
!
the total field E is rotational-free, since it is curl in this zone is vanishing too;
! ! !
hence, E is not a transverse vector field. The components E T and E L are
linked by conditions (147), and the disappearance of longitudinal field indi-
cates that the transverse field is vanishing too; herein the system proves to be
in a stationary state, incapable to irradiate or absorb photons. The source for
the transverse field is shown to extend not only over the current-carrying
region, but also over the near-field zone of the source. Note, that a similar
analysis can be performed for the evanescent magnetic components of a near
field of magnetic dipole, presented by an aerial frame; the spatial depen-
dences of these components Hr and H are known to coincide with the spa-
tial dependences of the Er and E components of the electric dipole field
(Kong, 1986). Hence, the evanescent magnetic field in the near zone of a
!
magnetic dipole satisfies simultaneously the Maxwell equation div H 0
!
as well as the condition rot H 0. These results link the photon-tunneling
process to impossibility of localizing photons in space (Keller, 2001),
highlighting the nonlocality as one of the most intriguing aspects of the elec-
tromagnetic wave theory.
5.2 Tunneling of Light as a Stochastic Process

Tunneling processes are treated in the framework of this concept as the sto-
chastic motions of photons in the forbidden zone connected with multiple
reflections of photons from the boundaries of the zone. Consider a photonic
barrier, consisting of transparent layers interleaved with identical thin
opaque layers at regular separations. The photons, impinging on each layer,
are characterized by their probabilities to be transmitted or to be reflected.
The photons which have not been reflected even once possess the smallest
attenuation; photons with a single reflection are attenuated more and so on.
Hence only the leading edge of the incident pulse survives without being
strongly attenuated. Owing to this reshaping, the peak of the transmitted
pulse seems to outstrip the peak of reference pulse traveling in vacuum
(Chiao & Steinberg, 1997). This effect, which can be treated in favor of
superluminal tunneling, is related to the irregular paths in quantum mechan-
ics predicted as long ago as 1965 by Feynman and Hibbs.
Unlike the approach based on Maxwell equations, the dynamics of
photons in processes of multiple reflections is described by telegraph
equation with imaginary characteristic time iT0 (De-Witt Morette, &

Foong, 1989)
@2 1 @2 2 @
2 2 : (148)
@z 2 v @t iT0 v2 @t
In contrast to the traditional form of the telegraph equation with a real

characteristic time, connected with dissipation processes, Equation (148)
relates to the lossless medium; herein the imaginary dissipation corre-
sponds to the reactive fields, mentioned in Section 3.5. The salient features
of this stochastic model can be treated by means of the concept of a complex
tunneling time p + iT (Balqou & Dutriaux, 1997), determined via the
complex transmission coefficient T jT jexpit :
@t @ ln jT j
p , T : (149)
@ @
The first equality in (149) defines so-called phase time p; note, that
under the barrier the phase is not accumulated, and the phase shift t comes
mainly from the boundaries. However, using this concept in the interpre-
tation of measurements of penetration times for GHz microwaves in a sub-
cutoff waveguide (Ranfagni, Ruggeri, Susini, Agresti, & Sandri, 2001) has
revealed a paradoxical result; the dependencies of both the real p and imag-
inary T tunneling times upon the length of barrier contained two ascending
branches, separated by one descending branch. To understand the origin of
the maximum and minimum on the graphs of p and T, restricting this
descending branch, further investigations are needed.
The interpretation of evanescent modes as the stochastic motions of pho-
tons admits the generalization to virtual photons. Some peculiar properties
of these modes, resembling the virtual particles in quantum electrodynamics,
can be stressed:
i. The attempts to quantize the evanescent modes in a medium with con-
stant frequency-independent refractive index has revealed an important
feature of these modes; unlike the propagating waves the commutator
of field operators between two space-like separated points in the near-
field zone does not vanish (Stahlhofen & Nimtz, 2006).
ii. To provide a valid recording of the tunneling signal, it is necessary that
its power be above the thermal noise level of the receiver. The power of
this noise, generated under the condition of thermal equilibrium, is
given by the Nyquist formula (Landau & Lifschitz, 1984)
PN kT , (150)
here k, T, and are the Boltzmann constant, temperature, and the

frequency range of the signal, respectively. This equation allows one to
evaluate the smallest power of tunneling signals, which can be recorded
by the receiver.
iii. Propagation of light pulse through a periodic system of -like spikes
results in the formation of a delay, which can be larger than the uncer-
tainty stipulated by the length of the pulse itself (Aharonov, Erez, &
Reznik, 2002).
These results, based on presentation of fields via Fourier harmonics, were
obtained for vacuum or for homogeneous transparent media; however,
the waves in gradually heterogeneous dielectrics are spatially nonharmonic.
This circumstance impedes the use of the standard approach for the quan-
tization of fields in gradient nanophotonics. Quantization of evanescent
waves in gradient media still remains an open problem.
5.3 Tunneling of Wave Pulses as a Quasistatic Process

This concept was developed by Winful (2003) for light pulses traversing a
periodic dielectric lossless nondispersive structure, characterized by some
cutoff frequency c. The pulses are supposed to be long enough, such that
the temporal variations of the pulse envelope can be assumed to be slow as
compared with the transit time of light across the barrier. Within this,
so-called quasistatic, approximation the Maxwell equations are reduced to
a pair of coupled-mode first-order equations for the forward and backward
waves. A theoretical consideration of harmonic modulation of the input
pulse envelope E 0, t E0 cos t, where < c , shows that the input field
merely modulates the amplitude of a standing wave formed due to the inter-
ference of forward and backward waves inside the barrier; herein the output
envelope follows the input adiabatically. Calculation of the energy density of
the total field, created by this interference, shows that under the quasistatic
conditions the intrabarrier energy density distribution decreases monoton-
ically from the input to the output even if the incident pulse goes through a
maximum. The absence of any maximum in this distribution may be con-
sidered as an indication that the incident peak did not propagate through the
barrier, and, therefore, the notion of a transit time is meaningless
(Winful, 2003).
In this conjunction, one can recall the FTIR experiment (Haibel,

Nimtz, & Stahlhofen, 2001); two similar receivers were placed symmetri-
cally at equal distances from the opposite sides of the opaque barrier.
The light pulse impinging on this barrier was reflected and partially trans-
mitted. The times of arrival of reflected (tR) and transmitted (tT) pulses to the
corresponding receivers were recorded; these measurements have shown the
simultaneous arrival of both pulses. This equality (tR tT ) inspired the dis-
crepant conclusion about the zero transit time for the tunneling wave.
On the other hand, the definition of speed of tunneling marked by the
instant of arrival of its peak seems to be not valid due to pulse peak advance-
ment as a result of reshaping phenomena, in which the trailing edge of the
pulse is more attenuated than the leading one; herein the registration of this
instant is impeded due to the slowness of the pulse envelope variation
namely nearby the peak. Moreover, to identify the pulse signal by a detector,
some additional time, dependent upon the detecting scheme, has to be
spent. Note that the signal speed can be considered as the speed of propa-
gation of an abrupt disturbance of envelope; according to Buttiker and
Washburn (2003) one has to conclude that the signal speed is always sub-
luminal, whereas the other characteristic times, e.g., the dwell time or
group-delay time may become superluminal.
This coexistence of different and even opposite viewpoints on the
nature of tunneling of EM waves in transparent media indicates the
novelty, actuality, and fast development of this branch of physics, important
for both academic researches and applications. Application of evanescent
waves for energy and information transfer is expected to create new insights
in the different wave processes. It is expedient to emphasize some problems,
which, though being hot and offer rich perspectives, are worked out
insufficiently:
i. Electromagnetics of three-dimensional evanescent fields in gradient
dielectric miniaturized devices, which provides controllable formation
of spatial and polarization structures of trapped fields, such as, e.g., the
split-ball cavity, which can be viewed as a 3D analogy of a planar split-
ring resonator (Miroshnichenko et al., 2012).
ii. Nonlinear optics of evanescent waves, characterized by giant enhance-
ment of near-zone fields (Lukyanchuk & Qiu, 2008), power-
dependent optical properties of composite films (Huang & Yu,
2005), and the possibilities of interplay between near-field and
far-field interference phenomena (Lukyanchuk, Miroshnichenko, &
Kivshar, 2013).
iii. Tunneling phenomena in the electrodynamics of displacement currents

in gradient all-dielectric structures reveal the unwanted features of these
structures as compared to plasmonicsbased devices (Alu et al., 2007).
Moreover, it has to be emphasized that the parallel researches of tunneling
effects for EM waves of different spectral ranges (e.g., optical and microwave
ranges) and for wave fields of different physical nature seems to be a fruitful
and promising approach. Thus, the universality of wave patterns nearby the
maxima/minima of the gradient of a heterogeneous continuum, common
for optics and quantum mechanics, was revealed (Kaplan, 2012). The same
approach allows to transfer the concept of HID from gradient optics, to, e.g.,
acoustics of heterogeneous solids with spatially distributed densities and/or
Youngs modulus. This far reaching analogy between electromagnetic and
acoustic waves in heterogeneous media has stimulated the search of an
unusual effectthe tunneling of sound through gradient acoustical barriers
in compositionally graded alloys, composite materials, and porous structures
(Shvartsburg & Maradudin, 2013).
ACKNOWLEDGMENTS
We are much obliged to Professor E. Wolf for his kind suggestion to present this paper to
Progress in Optics. The authors appreciate Prof. A. Alu, N. Engheta, N. Erokhin,
V. Fortov, J. Pendry, N. Silin, L. Vazquez and L. Zelenii for their immutable interest to
these topics and valuable discussions. The work of A.S. was supported by the Far Eastern
Federal University, Project No. 14-08-2/3-20.
REFERENCES
Abramowitz, M., & Stegun, I. (1968). Handbook of mathematical functions. NY: Dover
Publications.
Adachi, S. (1988). Model dielectric constants of Si and Ge. Physical Review B, 38, 12966.
Aharonov, Y., Erez, N., & Reznik, B. (2002). Superoscillations and tunneling times. Physical
Review A, 65, 052124.
Alu, A., & Engheta, N. (2007). Light squeezing through arbitrarily shaped plasmonic chan-
nels and sharp bends. Physical Review B, 78, 035440.
Alu, A., Silveirinha, M. G., Salandrino, A., & Engheta, N. (2007). Epsilon-near-zero
metamaterials and electromagnetic sources: Tailoring the radiation phase pattern. Physical
Review B, 75, 155410.
Azzam, R. M. A. (2006). Phase shifts in frustrated total internal reflection and optical
tunneling by an embedded low-index thin film. Journal of the Optical Society of America A,
23(4), 960.
Baena, F. D., Jelinek, L., Marques, R., & Medina, F. (2005). Near-perfect tunneling and
amplification of evanescent electromagnetic waves in a waveguide filled by a
metamaterial: Theory and experiments. Physical Review B, 72, 075116.
Balqou, Ph., & Dutriaux, L. (1997). Dual optical tunneling times in frustrated total internal
reflection. Physical Review Letters, 78(5), 851.
Born, M., & Wolf, E. (1999). Principles of optics (7th ed.). Cambridge University Press.
Borisov, A. G., de Abajo, Garsia, & Shabanov, S. V. (2005). Role of electron trapped
modes in extraordinary transmission in nanostructured materials. Physical Review B,
71, 075408.
Bulgakov, A. A., Bulgakov, S. A., & Nieto Vesperinas, M. (1998). Inhomogeneous waves
and energy localization in dielectric superlattices. Physical Review B, 58(8), 4438.
Buttiker, M., & Washburn, S. (2003). Ado about nothing much? Nature, 422, 271272.
Chiao, R., & Steinberg, A. M. (1997). Tunneling times and superluminality. In E. Wolf
(Ed.), Progress in optics: Vol. 37 (p. 345).
Cohen Tannoudji, C., Dupont Roc, J., & Grynberg, G. (1989). Photons and atoms: Intro-
duction to quantum electrodynamics. John Viley & Sons, Canada, Ltd.
Collins, R. (1991). Field theory for guided waves. New York, NY: John Wiley.
Dalarsson, M., Jaksic, Z., & Tassin, P. (2009). Exact analytical solution for oblique
incidenceon a graded index interface between a right-handed and left-handed material.
Journal of Optoelectronics and Biomedical Materials, 1(4), 345.
Darmanyan, S. A., Neviere, M., & Zayats, A. V. (2004). Light tunneling via resonant surface
plasmon polariton state. Physical Review B, 70, 075103.
de Dios-Leyva, M., & Drake-Perez, J. C. (2012). Group velocity and non-local energy trans-
port velocity in finite photonic structures. Journal of the Optical Society of America B, 9,
2275.
de Fornel, F. (2001). Evanescent waves: From Newtonian optics to atomic optics.
Springer ser. in optical sciences 73. Berlin: Springer.
De Witt Morette, & Foong, S. K. (1989). Path integral solutions of wave equations with
dissipation. Physical Review Letters, 62, 2201.
Deych, I., Livdan, L., & Lisyansaky, A. A. (1998). Resonant tunneling of electromagnetic
waves through polariton gaps. Physical Review E, 57, 7254.
Dresselhaus, G. (1955). Spin-orbit coupling in III-IV semiconductor heterostructures. Phys-
ical Review, 100(2), 580.
Dunster, T. M. (1990). Bessel functions of purely imaginary order. SIAM Journal on Mathe-
matical Analysis, 21, 995.
Edwards, B., Alu, A., Young, M. E., Silveirinha, M., & Engheta, N. (2008). Experimental
verification of epsilon-near-zero metamaterial coupling and energy squeezing using a
microwave waveguide. Physical Review Letters, 100, 033903.
Engheta, N. (2007). Circuits with light at nanoscales: Optical nanocircuits inspired by
metamaterials. Science, 317, 1698.
Efimov, V. V., Ivanov, O. V., & Sementsov, D. I. (2001). Phase relation for electromagnetic
waves reflected and transmitted through a dielectric slab. Journal of Optics A: Pure and
Applied Optics, 3, 514.
Eikhenwald, A. A. (2009). Energy flow with total internal reflection. Journal of Russian
Physical Chemical Society, 41, 131.
Erokhin, N. S., & Zakharov, V. E. (2011). Non-reflective propagationof electromagnetic
wave through heterogeneous plasma layer. Plasma Physics Reports, 37(9), 818.
Erokhin, N. S., Zueva, Yu. M., & Shvartsburg, A. B. (2013). Polarization phenomena in
gradient nanooptics. Quantum Electronics, 43(9), 785.
Fratalocchi, A., & Assanto, G. (2006). Dispersion spectroscopy of photonic barrier. Optics
Letters, 31(22), 3351.
Frias, W., Smolyakov, A., Hirose, A., Frias, W., Smolyakov, A., Hirose, A., et al. (2011).
Non-local energy transport in tunneling and plasmonic structures. Optics Express,
19(16), 15281.
Fu, C., Zhang, Z. M., & Tanner, D. B. (2005). Planar heterogeneous structures for coherent
emission of radiation. Optics Letters, 30(14), 1873.
Futterman, J. (1995). Magnetic Brewster angle. American Journal of Physics, 63, 471.
Gamow, G. (1928). Zur quantentheorie des atomkernes. (On quantum theory of atomic
nuclei). Zeitschrift f
ur Physik, 51, 204.
Garcia de Abajo, F. J., Gomez-Santos, G., Blanco, L. A., Borisov, A. G., & Shabanov, S. V.
(2005). Tunneling mechanism of light transmission through metallic films. Physical
Ghulinyan, M., Oton, C. J., Gaburro, Z., Pavesi, L., Toninelli, C., & Wiersma, D. S. (2005).
Zener tunneling of light waves in the optical superlattice. Physical Review Letters, 94,
127401.
Ginzburg, V. L. (1970). Propagation of electromagnetic waves in plasma. Oxford: Pergamon Press.
Grbic, A., & Eleftheriades, G. V. (2002). Experimental verification of backward-wave radiation
from a negative refractive index metaqmaterial. Journal of Applied Physics, 92(10), 5930.
Haibel, A., Nimtz, G., & Stahlhofen, A. A. (2001). Frustrated total reflection: The double -
prisms revisited. Physical Review E, 63, 047601.
Hartman, T. E. (1962). Tunneling of a wave packet. Applied Physics, 33, 3427.
Hayata, K., & Koshiba, M. (1995). Algebraically decaying modes of dielectric planar wave-
guides. Optics Letters, 20(10), 1131.
Huang, J. P., & Yu, K. W. (2005). Effective nonlinear optical properties of graded metal-
dielectric composite films of anisotropic particles. Journal of the Optical Society of America
B, 22(8), 1640.
Jackson, J. D. (1998). Classical electrodynamics. New York, NY: Wiley.
Kaplan, A. E. (2012). Hysteresis in cyclotron resonance based on weak-relativistic mass-effect
of the electron. Physical Review Letters, 109, 153901.
Keller, O. (2001). Optical tunneling: A fingerprint of the lack of photon localizability. Journal
of the Optical Society of America B, 18(2), 206214.
Kim, K. (2008). Excitation of S-polarized surface electromagnetic waves in homogeneous
dielectric media. Optics Express, 16(17), 13354.
Kolokolov, A. A., & Skrotsky, G. V. (1992). Interference of reactive components of an elec-
tromagnetic field. Soviet PhysicsUspekhi, 162, 165.
Kong, J. A. (1986). Electromagnetic wave theory. New York, NY: Wiley-Interscience.
Kuzelev, M. V., & Rukhadze, A. A. (2009). Methods of waves theory in dispersive media.
Singapore: World Scientific Publishing Co.
Lakhtakia, A., & Messier, R. (2005). Sculptured thin films: Nanoengineered morphology and optics.
Bellinghem, Washington: SPIE Press.
Landau, L. D., & Lifschitz, E. M. (1984). Electrodynamics of continuous media. Amsterdam:
Elsevier.
Lee, S. J., Moon, S. E., Ryu, Y. C., Kwak, M. H., Kim, E. T., & Han, S. K. (2003). Micro-
wave properties of compositionally graded (Ba, SR)TiO3 thin films according to the
direction of the composition gradient for tunable microwave applications. Applied Physics
Letters, 82(13), 2133.
Lekner, J. (2000). Omnidirectional reflection by multilayer dielectric mirrors. Journal of Optics
A: Pure and Applied Optics, 2, 349.
Leonhardt, U. (2006). Optical conforming mapping. Science, 312, 1777.
Li, M., Liu, X., Zhang, J., Ma, X., Sun, X., Li, Y., et al. (2007). Rigorous analysis of the
spatial dispersion effect in multilayer thin-film coatings. Journal of the Optical Society of
America A, 24(8), 2328.
Li, C.-F., & Wang, Q. (2001). Duration of tunneling photons in a frustrated-total-internal-
reflection structure. Journal of Optical Society of America B, 18(8), 1174.
Liaw, J. W., Chen, J.-H., Chen, C.-S., & Kuo, M.-K. (2009). Purcell effect on nanoshell
dimmer on single molecules fluorescence. Optics Express, 17, 13532.
Litchinitser, N. M., Maimistov, A. I., Gabitov, I. R., Sagdeev, R. Z., & Shalaev, V. M.
(2010). Metamaterials: Electromagnetic enhancementat zero-index transition. Optics
Letters, 33(20), 2350.
Liu, R., Cheng, Q., Hand, T., Mock, J. J., Cui, T. J., Cummer, S. A., et al. (2008). Exper-
imental demonstration of electromagnetic tunneling through an epsilon-near-zero
metamaterial at microwave frequencies. Physical Review Letters, 100, 023903.
Liu, L., Hu, C., Zhang, Z., & Luo, X. (2009). Multi-passband tunneling effect in multilay-
ered epsilon-near-zero metamaterials. Optics Express, 17(14), 12183.
Longhi, S. (2005). Resonant tunneling in frustrated total internal reflection. Optics Letters, 30,
2781.
Lui, L., Hu, C., Zhao, Z., & Luo, X. (2009). Multipassband effect in multilayered epsilon
near zero metamaterial. Optics Express, 17(14), 12183.
Lukyanchuk, B. S., & Qiu, C.-W. (2008). Peculiarities of light scattering by spherical par-
ticles with radial anisotropy. Journal of the Optical Society of America A, 25, 1623.
Lukyanchuk, B. S., Miroshnichenko, A. E., & Kivshar, Yu. S. (2013). Fano resonances and
topological optics: An interplay of far and near field interference phenomena. Journal
of Optics, 15, 073001.
Maier, S. A. (2007). Plasmonics: Fundamentals and applications. New York, NY: Springer.
Marques, R., Martel, J., Mesa, F., & Medina, F. (2002). Left-handed media simulation and
transmission of EM waves in subwavelength split-ring-resonator-loaded metallic wave-
guides. Physical Review Letters, 89(18), 183901.
Maxwell-Garnett, J. C. (1904). Colours in metal glasses and in metallic films. Philosophical
Transactions of the Royal Society A, 203, 385.
Miroshnichenko, A. E., Kuznetsov, A. I., Wei, L., Fu, Y. H., Neshev, D., Lukyanchuk, B.,
et al. (2012). Magnetic light: Optical magnetism of dielectric nanoparticles. Optics &
Photonics News, 23(12), 35.
Mocella, V., Dardano, P., Rendina, I., & Cabrini, S. (2010). An extraordinary directive radi-
ation based on optical antimatter at near infrared. Optics Express, 18(24), 25068.
Nelin, E. A. (2007). Impedance model for quantum mechanical barrier problems. Physics
Uspekhi, 50, 293.
OBrien, S., & Pendry, J. P. (2002). Photonic band-gap effectsand magnetic activityin dielec-
tric composites. Journal of Physics: Condensed Matter, 14, 4035.
Ochiai, T., Leonhardt, U., & Nacher, J. C. (2008). A novel design of dielectric perfect invis-
ibility devices. Journal of Mathematical Physics, 49, 032903.
Olkhovsky, V. S., Recami, E., & Jakiel, J. (2004). Unified time analysis of photon and particle
tunneling. Physics Reports, 398, 133.
Pendry, J. B., Holden, A. J., Robbins, D. J., & Stewart, W. J. (1999). Magnetism from
conductors and enhanced nonlinear phenomena. Microwave Theory and Techniques,
47, 2075.
Pendry, J. B., Holden, A. J., Stewart, W. J., & Youngs, I. (1996). Extremely low frequency
plasmons in metallic mesostructures. Physical Review Letters, 76, 4773.
Pendry, J. B., Schurig, D., & Smith, D. (2006). Controlling electromagnetic fields. Science,
312, 1780.
Rand, S. C., Fisher, W. M., & Oliveira, S. L. (2008). Optically induced magnetization in
homogeneous, undoped dielectric media. Journal of the Optical Society of America B,
25(7), 1106.
Ranfagni, A., Mugnai, D., Fabeni, P., & Pazzi, G. P. (1991). Time-delay measurements in
narrowed waveguides as a test of tunneling. Applied Physics Letters, 58(7), 774.
Ranfagni, A., Fabeni, F., Pazzi, G. P., & Mugnai, D. (1993). Anomalous pulse delay in
microwave propagation: A plausible connection to the tunneling time. Physical Review
E, 48, 1473.
Ranfagni, A., Ruggeri, R., Susini, C., Agresti, A., & Sandri, P. (2001). Experimental evi-
dence of tunneling as a stochastic process. Physical Review E, 63, 025102.
Rayleigh, J. W. S. (1937). The theory of sound. London: McMillan & Co.
Recami, E. (2004). Superluminal tunneling through successive barriers: Does quantum
mechanics predict infinite group velocities? Journal of Modern Optics, 51, 913.
Sarid, D., & Challener, W. (2010). Modern introduction to surface plasmons: Theory metamaterial
modelling and applications. NY: Cambridge University Press.
Schiff, L. I. (1968). Quantum mechanics. Tokyo: Mc Grow Hill.
Semouchkina, E., Werner, D. H., Semouchkin, G. B., & Pantano, C. (2010). An infrared
invisibility cloak composed of glass. Applied Physics Letters, 96, 233503.
Serebryannikov, A. E., Ozbay, E., & Nojima, S. (2014). Asymmetric transition of terahertz
waves using polar dielectrics. Optics Express, 22(3), 3075.
Shelby, R. A., Smith, D. R., & Schultz, S. (2001). Experimental verification of a negative
index of refraction. Science, 292, 77.
Shvartsburg, A. B., & Kuzmiak, V. (2010). Reflectionless tunneling of microwave radiation
through subwavelength slit in a waveguide. Journal of Optical Society of America B, 27(12),
2766.
Shvartsburg, A. B., Kuzmiak, V., & Petite, G. (2007). Optics of subwavelength gradient
nanofilms. Physics Reports, 452, 33.
Shvartsburg, A. B., & Maradudin, A. A. (2013). Waves in gradient metamaterials. Singapore:
WSPC.
Shvartsburg, A. B., Marklund, M., Brodin, G., & Stenflo, L. (2008). Superluminal tunneling
of microwaves in smoothly varying transmission lines. Physical Review E, 78, 016601.
Shvartsburg, A. B., & Petite, G. (2005). Concave and convex photonic barriers in gradient
optics. European Physical Journal D, 36, 111.
Siddique, O. F., Ramahi, O. M., & Kashanianfard, M. (2010). Frequency-selective energy
tunneling in wire-roaded narrow waveguide channels. Progress in Electromagnetic Research
Letters, 15, 153.
Silveirinha, M. G., & Engheta, N. (2006). Tunneling of electromagnetic energy through sub-
wavelength channels and bends using 0-near-zero materials. Physical Review Letters, 97,
157403.
Silveirinha, M. G., & Engheta, N. (2009). Transporting of image through the subwavelength
hole. Physical Review Letters, 102, 103902.
Smith, D. D., & Fuller, K. A. (2002). Photonic bandgaps in Mie scattering by con-
centrationally stratified spheres. The Journal of the Optical Society of America B, 19(10),
2449.
Smith, D. R., Vier, D. C., Koschny, T., & Soukolis, M. C. (2005). Electromagnetic param-
eter retrieval from inhomogeneous metamaterials. Physical Review E, 71, 036617.
Stahlhofen, A. A., & Nimtz, G. (2006). Evanescent modes and virtual photons. Europhysics
Letters, 76(2), 189.
Sumetsky, M. (2007). Radiation loss of a nanotaper: Singular Gaussian beam model. Optics
Express, 15(4), 1480.
Tretyakov, S. A. (2003). Analytical modelling in applied electromagnetics. Norwood, MA, USA:
Artech House, INC.
Vahala, K. J. (2003). Optical microcavities. Nature, 424, 839.
Veselago, V. G. (1968). The electrodynamics of substanceswith simultaneously negative
and . Soviet PhysicsUspekhi, 10, 509.
Vesseur, E., Coenen, T., Caglayan, H., Engheta, N., & Polman, A. (2013). Experimental
verifications of n 0 structures for visible light. Physical Review Letters, 110, 013902.
Wang, J., Li, L., & Jia, S. (2008). Nonlinear tunneling of optical similaritons in nonlinear
waveguides. Journal of the Optical Society of America B, 25(8), 1254.
Wannier, G. H. (1955). Elements of solid state theory. London: Cambridge University Press.
Wilson, D. W., Glytsis, E. N., & Gaylord, T. X. (1993). Electron waveguiding characteristics
and ballistic current capacity of semiconductor quantum slabs. IEEE Journal of Quantum
Electron, 29, 1364.
Winful, H. G. (2003). Nature of superluminal barrier tunneling. Physical Review Letters,
90(2), 023901.
Yang, F., & Sambles, J. R. (2002). Resonant transition of microwaves through a narrow
metallic slit. Physical Review Letters, 89, 063901.
Yeh, P. (1985). Resonant tunneling of electromagnetic radiation in superlattice structures.
Journal of the Optical Society of America A, 2(4), 568.
Zayatz, A. V., Smolyaninov, I. I., & Maradudin, A. A. (2005). Nano-optics of surface
plasmon polaritons. Physics Reports, 408, 131.
Zhang, J., Luo, Y., & Mortensen, N. A. (2010). Transmission of electromagnetic waves
through sub-wavelength channels. Optics Express, 18(4), 3864.
Zhao, Q., Zhou, J., Zhang, J., & Lippens, F. (2009). Mie resonance- based dielectric
metamaterial. Materials Today, 12, 60.
Ziolkowski, R. W., & Heyman, E. (2001). Negative effective permeability and left-handed
materials at optical frequencies. Physical Review E, 64, 056625.
AUTHOR INDEX
Note: Page numbers followed by f indicate figures.
A Ansmann, M., 358

Abdallah, N., 141, 244245 Arakawa, Y., 60
Abraham, M., 440 Aranda, F.J., 171172, 179
Abram, I., 2, 72, 79 Armitage, D., 179
Abram, L., 46 Arnold-Bos, A., 125126, 190191,
Abramowitz, M., 513514, 517518, 194195
548549 Arrad, G., 426427
Abrams, D.S., 426427 Ashkin, A., 438
Acin, A., 426427 Aspachs, M., 422
Adachi, S., 511 Aspect, A., 7273
Adamo, Q., 153 Assanto, G., 518519
Adesso, G., 353, 356357 Atat
ure, M., 24, 89
Aers, G.C., 58 Atchia, Y., 144, 146147
Afek, I., 390 Atkinson, P., 32, 53, 79
Aggoun, A., 127128, 141 Aucsmith, D., 139
Agresti, A., 555 Awatsuji, Y., 173174, 239
Aharonov, D., 426427 Azzam, R.M.A., 497
Aharonovich, I., 61, 68
Ahmed, N., 137138 B
Aichele, T., 59 Babinec, T.M., 65, 82
Akhmedzhanov, R., 96 Babkina, T.M., 171
Akimov, A.V., 8283 Bacher, G., 5960
Akimov, I., 59 Bachor, H.-A., 355357
Akkara, J.A., 171172 Badolato, A., 89
Alam, M.S., 119120, 125126, 141, 153, Baehr-Jones, T., 96
155, 169170, 190191, 194195, Baena, F.D., 529530
244245 Bagan, E., 422
Alfalou, A., 119246 Bajcsy, M., 89
Allais, A.G., 210211, 219221 Baker, A., 131132
Alleaume, R., 94 Balabas, M.V., 373
Allen, L., 467468 Ballester, D., 82
Altewischer, E., 8283 Bamba, M., 89
Alu, A., 525526, 529530, 558 Banaszek, K., 350353, 356358, 400,
Amann, M.C., 56 404408
Ambar, O., 390 Banerjee, P.K., 153
Anan, T., 13 Banerjee, P.P., 119120
Anderson, B.P., 373 Bara, S., 128
Andonian, A.T., 127128 Barakat, R., 212213
Andre, A., 94, 427 Baraniuk, R.G., 133f
Andre, R., 59 Barat, C., 209
Anger, P., 59 Barbieri, M., 427
Anisimov, P.M., 349350, 396397 Barclay, J., 170
565
566 Author Index
Barlaud, M., 133135 Bommer, M., 58

Barnard, K.J., 131132 Bondo, T., 3637
Barnett, S.M., 348350 Bondurant, R.S., 347
Barrett, M.D., 369, 427 Bonneau, D., 96
Barrett, P., 140 Boorson, S.D., 13
Barros, H.G., 38 Boozer, A.D., 40, 89
Barter, O., 96 Borgstrom, M.T., 59
Bartlett, S.D., 387, 391393, 400401 Boring, C.C., 170
Bartolini, L., 209211, 219221 Borisov, A.G., 522523
Becher, C., 4344 Born, M., 208, 211212, 219220
Behbood, N., 373 Boto, A.N., 426427
Ben Eliezer, E., 164165 Bougerol, C., 59
Benatti, F., 359360, 399 Boulvert, F., 210211, 219221
Beneteau, L., 241 Bouzidi, F., 156
Bengtsson, I., 382, 410 Bovino, F., 196
Benjamin, S.C., 426427 Boyd, R.W., 29
Benkelfat, B.-E., 125126, 156 Bradford, M., 60
Bennett, A.J., 32, 51, 53 Brady, P., 210211, 219221
Benoudjit, N., 242 Branning, D., 78
Benson, O., 31, 41, 59 Brask, J.B., 426427
Benzid, R., 242 Brattke, S., 3336
Berglund, A.J., 94 Brauchle, C., 176177
Bergquist, J.C., 369, 371373, 426427 Braun, D., 353, 359360
Berriel, R., 131132 Braunstein, S.L., 17, 347, 383,
Berry, D.W., 349350, 390392, 396, 426427
400401, 424425 Breki-Iovskhikh, G.L., 176177
Bertolotti, M., 196 Brendel, J., 94
Bescos, J., 131132 Brennan, M.J., 210211, 219221
Beveratos, A., 2, 64, 80, 94 Brevik, I., 451
Bhattacharya, P., 96 Brillouin, L., 297, 312314, 471472
Bialczak, R.C., 358 Britton, J., 369, 427
Bichler, M., 56 Brodin, G., 493, 532533
Bicout, D., 209210 Brokmann, X., 59
Biham, E., 238 Brooker, G., 119120
Billingsley, F.C., 124125 Brosseau, C., 119246
Birch, P., 131132 Brouri, R., 2, 64
Birnbaum, K.M., 89 Bruckstein, A.M., 128
Bissell, L.J., 29 Brune, M., 17, 3940, 9294
Bjorkholm, J.E., 438 Buck, J.R., 40, 136137
Blanco, L.A., 522523 Budgett, D., 131132
Blatt, R., 38 Budker, D., 369, 373, 426427
Bleistein, N., 288289, 301302 Bulu, I., 13, 82
Blinn, J.F., 137138 Buratto, S.K., 59
Bloom, J., 139 Burnett, K., 347, 367369
Boca, A., 40, 89 Burns, M.M., 438439
Boer, G., 477 Burrus, C., 136137
Boffety, M., 210211, 219221 Buytaert, J.A.N., 127128
Bollinger, J.J., 347349, 371373, 390 Buzek, V., 70, 349351
Author Index 567
C Chin, A.W., 426427

Cable, A., 438 Chiribella, G., 387, 390391
Cabrini, S., 525526 Chiruvelli, A., 349350, 396
Caglayan, H., 13, 525526 Choi, M.-D., 412
Calsamiglia, J., 422 Choy, J.T., 65, 82
Canet-Ferrer, J., 58 Chu, S., 438
Cao, F., 166167 Chuang, I.L., 382, 403, 406407
Cao, J.P., 396 Chung, M.S., 128
Caprari, R.S., 210211, 219221 Chwede nczuk, J., 373
Cariou, J., 210211, 219221 Chyi, J.-I., 82
Carlisle, C.B., 131132 Ci, X., 61, 70
Carnal, O., 18 Cirac, J.I., 350, 407408
Carson, P.J., 59 Ciuti, C., 89
Cartwright, N.A., 264343 Claret-Tournier, F., 131132
Carusotto, I., 89 Clarke, E., 96
Casasent, D., 142 Claudon, J., 33, 59
Castelletto, S., 61, 6869, 75, 78 Clauser, J.F., 3
Castillo, I.M.D., 179 Clemenceau, P., 210211, 219221
Catrysse, P.B., 144 Cochenour, B., 210211, 219221
Caves, C.M., 347, 350351, 366367, 383, Coenen, T., 525526
425f Cogan, D., 96
Centurion, M., 239 Colby, W., 339
Cerf, N.J., 2 Coles, R.J., 96
Chaneliere, T., 94 Collins, R., 532, 538539
Chang, D.E., 8283, 89 Collins, S.A., 162163
Chang, H.-S., 82 Constanza, M., 170
Chang, P.C.Y., 210211, 219221 Cooley, J.W., 136137
Chang, T.Y., 175 Cooper, D.E., 131132
Chang, W.H., 82, 136137 Cooper, K., 31, 79
Chavel, P., 182184 Corbie`re, E., 144145
Chaves, R., 426427 Cox, I.J., 139, 166167
Chekhova, M., 351 Cristobal, G., 141, 205, 235236
Chen, C.-S., 529 Cronin, A.D., 369, 373, 426427
Chen, G., 61, 70 Crowley, J.L., 157
Chen, H., 239, 241242 Crowley, P.J.D., 427
Chen, H.C., 137138 Cui, B., 180
Chen, J.-H., 529 Cui, T.J., 467469, 525526
Chen, M.-C., 58 Cummer, S.A., 525526
Chen, P.H., 171172 Cummings, F.W., 1415
Chen, W., 96, 137138 Cummings, M.E., 210211, 219221
Chen, W.Y., 82
Chen, Z.-B., 358 D
Chenault, D.B., 210211, 219221, 246 DAriano, G.M., 353, 387, 390391,
Cheng, D., 144 396397
Cheng, Q., 525526 Dada, A.C., 96
Chester, C., 288289 Daemen, J., 238
Chiao, R.Y., 554 Dagenais, M., 8, 23
Chiaverini, J., 369, 427 Dahan, M., 59
568 Author Index
Dalacu, D., 58, 96 Dodd, G.D., 170

Dalarsson, M., 524 Dogariu, A., 190191, 210212, 219221,
Daly, P., 446 246
Darakis, E., 173174, 239 Dolan, G.J., 89
Dardano, P., 525526 Dorner, U., 350351, 404405, 408,
Darmanyan, S.A., 512 426427
Das, P., 141 Dotsenko, I., 17, 358
Datta, A., 350, 427 Douak, F., 242
Datta, A.K., 153 Dousse, A., 80
Davanco, M., 90 Dowling, J.P., 347350, 358, 373, 390,
Davidovich, L., 350351, 384, 408409, 396397, 426427
417 Drabenstedt, A., 61
Davies, G., 61 Dresselhaus, G., 515
Davis, C.C., 41 Drummond, P.D., 357
Dawes, M.C., 321323 Duan, L.M., 18
de Abajo, Garsia, 523 Dubost, B., 373
De Bougrenet De la Tocnaye, J.L., 153 Dubreuil, M., 190191, 212,
De Dominicis, L., 209211, 219221 219221
De Martini, F., 25, 94, 404405 Dudin, Y.O., 42
de Matos Filho, R.L., 350351, 384, Dufour, S., 144, 146147
408409, 417 Dumeige, Y., 94
DeCusatis, C., 141 Duncan, R.C., 209
Degiovanni, I.P., 75 Duntley, S.Q., 188
Dehmelt, H., 13 D
ur, W., 426427
Dekker, M., 153 Duraffourg, L., 241
Delacetaz, G., 477 Durkin, G.A., 350351
Delrot, P., 190191, 212, 219221 Dusek, M., 2
Demkowicz-Dobrza nski, R., 346427 Duvaut, P., 130, 136137, 232
Demler, E.A., 89 Dziedzic, J.M., 438
Deppe, G.D., 24
Dereniak, E.L., 210211, 219221 E
Derka, R., 349350 Eades, W.D., 179
Desbiolles, P., 59 Eberly, J.H., 13
Deshpande, S., 96 Edwards, B., 525526
Deutschet, M., 41, 89 Efimov, V.V., 497
Dewhurst, S., 32 Egami, C., 171172
Dholakia, K., 438439 Egan, J.P., 157
Di Giuseppe, G., 25 Ehrenfreund, E., 79
Diamanti, E., 20 Eichfelder, M., 32
Dickey, T.D., 187188 Einstein, A., 2122, 75, 451452
Diddams, S.A., 369, 371373, 426427 Eisaman, M.D., 94
Diecks, D., 2 Ekert, A.K., 350, 407408
Diedrich, F., 23 Elbouz, M., 125126, 141, 156, 205,
Dilley, J., 96 235236
Dillrey, J., 96 Eleftheriades, G.V., 529530
Ding, J.J., 162163 Ellis, D.J.P., 32
Dirckx, J.J.J., 127128 Engheta, N., 525526, 529530, 558
Dixon, J.E., 96 Engin, E., 96
Author Index 569
Englund, D., 89 Fournier, G., 188190

Epifano, C.M., 125126, 153, 193194, Fournier, J.M., 438439, 477
244245 Fournier-Prunaret, D., 241
Erlick, C., 209210 Fox, M., 346347
Erni, D., 8283 Fralick, S., 137138
Erokhin, N.S., 507, 542 Francucci, M., 209211, 219221
Escher, B.M., 350351, 384, 408409, 417 Franson, J.D., 78, 88
Espinosa, J., 144 Fratalocchi, A., 518519
Esteve, J., 373 Frauel, Y., 173174, 239
Freivald, P., 29
F Frias, W., 552
Fabeni, F., 538539 Fridrich, J., 139
Fabre, C., 353 Friedler, I., 59
Faddeev, L.D., 3 Friedman, B., 288289
Fade, J., 353 Frins, E.M., 171172
Fales, C.L., 166167 Frost, T., 96
Fan, H., 396 Fr
owis, F., 426427
Faraon, A., 89 Fry, E.S., 188, 210211, 219221
Farrer, I., 47, 51 Fu, C., 492, 542
Fasel, S., 8283 Fu, Y.H., 520521, 557
Fattal, D., 47, 51, 54 Fujii, M., 239
Fazio, P., 127128 Fujiwara, A., 350351
Fazio, R., 89 Fuller, K.A., 522523
Fejer, M.M., 96 Fulton, T.A., 89
Fellgett, P.B., 166167 Fushman, I., 89
Felsen, L.B., 288289, 302303 Futterman, J., 542
Ferrari, J.A., 171172, 195196
Ferraro, P., 173174, 239 G
Ferri de Collibus, M., 209210 Gabrielse, G., 13
Ferrier, J.L., 176177 Gaburro, Z., 520
Figue, J., 192193 Gacoin, T., 2, 64
Filip, R., 96 Gali, A., 69
Finizio, A., 173174, 239 Galland, C., 30
Fisher, W.M., 520521 Galland, F., 210211, 219221
Fitzsimons, J., 426427 Galli, M., 96
Flagg, E.B., 24, 51 Gan, C.H., 82
Fleischhauer, M., 13, 94 Gantz, L., 96
Fleury, L., 61 Garbusi, E., 171172
Floreanini, R., 399 Garcia de Abajo, F.J., 522523
Flores, J.L., 195196 Garca, R.R., 179
Flurasek, J., 2 Garca-Calzada, R., 58
Foden, C.L., 33 Garfinkel, L., 170
Forand, J., 188190 Gauthier, D.J., 306307, 321323
Forchel, A., 5960 Gaylord, T.X., 515516
Fornetti, G., 209211, 219221 Gazengel, J., 176177
Foty, D.P., 297 Genoni, M.G., 404
Fouge`res, A., 348 Gerace, D., 89
Founta, S., 24 Gerard, J.M., 33, 46, 59, 79
570 Author Index
Gerry, C., 70 Gross, M., 13

Gershoni, D., 79, 96 Gruber, A., 61
Ghael, S.P., 133f Gruenberg, H., 446
Ghasemi, H., 128 Grzegorczyk, T.M., 446, 450, 454, 462,
Ghirardi, G.C., 2 464468
Ghosh, S., 94 Guarneri, M., 209211, 219221
Ghulinyan, M., 520 Guern, Y., 210211, 219221
Giacobino, E., 59 Guerreiro, T., 96
Gianino, P.D., 131132, 205 Guglielmi, V., 241
Gilbert, G., 350 Guichard, F., 166167
Gilbert, G.D., 209210 Gulinatti, A., 96
Gilchrist, A., 424425 Gungor, A., 41
Gilerson, A.A., 210211, 219221 Guo, Q., 144, 239
Ginzburg, V.L., 490491, 525 Gurov, I.P., 192193
Giovannetti, V., 89, 348350, 359360, Guta, M., 350351, 384385, 408412, 415
389390, 394396, 404405 Guthohrlein, G.R., 35
Gisin, N., 2, 23, 8283, 94
Givens, C.R., 212213 H
Glauber, R.J., 354355 Hagley, E., 39, 9293
Glenn, D.B., 136137 Hall, D., 157
Glennon, B., 144, 148150 Hall, M.J.W., 396, 426
Gleyzes, S., 17, 358 Hamer, M.F., 61
Gl
uckstad, J., 179 Hampp, N., 176177
Glytsis, E.N., 515516 Hamrick, M., 350
Goedgebuer, J.P., 241 Han, B., 29
Gogyan, A., 96 Han, S.K., 542543
Goiffon, V., 144145 Han, Z.-F., 96
Goldstein, D.L., 210211, 219221, 246 Hand, T., 525526
Golovchenko, J., 438439 Handelsman, R.A., 288289, 301302
Gomez-Santos, G., 522523 Hanzawa, H., 61
Gonzalez, L.A., 179 Hara, T., 179
Goodman, J.W., 128, 130, 141142, Hardie, R.C., 131132
147149, 153, 156, 169171, 175 Hariharan, P., 346
Gordon, J.P., 438 Haroche, S., 13, 17, 40, 94
Gorman, G.A., 210211, 219221 Harris, N.C., 96
G
otzinger, S., 60 Hartman, T.E., 491
Gourier, N., 157 Hartmann, G.K., 209
Goy, P., 13, 39, 9293 Hartschuh, A., 29, 59
Grangier, P., 2, 64, 68, 7273 Hasegawa, Y., 390391
Granier, B., 192193 Haus, A.G., 170
Grassani, D., 96 Haus, H.A., 441, 459460
Gray, D., 188190 Hausmann, B., 65, 82
Grbic, A., 529530 Hayasaka, K., 38
Greentree, A.D., 61 Hayata, K., 509
Gregersen, N., 33 Hazari, A., 96
Grier, D.G., 438439 He, Y., 58, 89
Griffiths, D.J., 459460 He, Y.-M., 58
Gross, C., 373 Healy, J.J., 144
Author Index 571
Heath, C.W., 170 Hough, J., 346347, 366367, 388

Hecht, E., 141 Hradil, Z., 910, 390391
Heckenberg, N.R., 451 Hsieh, T.P., 82
Heerlein, R., 3335 Hsu, T.-M., 82
Heideman, M., 136137 Hu, C., 529530
Heindel, T., 59 Hu, E.L., 89
Heinrich, J., 59 Hua, H., 144
Heinzen, D.J., 347349, 371373, 390 Huang, G., 180181
Helstrom, C.W., 348, 383, 395 Huang, K.S., 180, 182184
Hemmer, P.R., 65, 8283 Huang, L., 59
Hendrych, M., 2 Huck, F.O., 166167
Henneberger, F., 59 Huelga, S.F., 350, 407408, 426427
Hennelly, B.M., 162163 Huggenberger, A., 59
Hennessy, K.J., 89 Hugonin, J.P., 59, 82
Hennrich, M., 3637 Humphreys, P.C., 78, 427
Hermier, J.P., 59 Huver, S.D., 349350, 396
Hesselink, L., 175 Hwang, S.Y., 197200
Higgins, B.L., 391392, 400401, 424425 Hyllus, P., 373, 399400, 424425
Hirose, A., 552
Hnizd, V., 459460 I
Hofheinz, M., 358 Ichioka, Y., 131132
H
ofling, S., 59, 89 Igeta, K., 13
Hofmann, H.F., 349350, 396397, Ikeda, O., 131132
424425 Imai, F.H., 144
Hogele, A., 30 Imai, H., 350351
Holberg, L., 438 Imamoglu, A., 3031, 41, 59, 69, 89
Holden, A.J., 543 Inbar, H., 131132, 138
Holevo, A.S., 348350, 383, 386387, Ippen, P., 13
390391 Itano, W.M., 347349, 369, 371373, 390,
Holland, M.J., 347, 367369 427
Holleczek, A., 96 Ito, K., 239
Hommel, D., 5960 Ivady, V., 69
Hong, C.K., 50, 7273, 76 Ivanov, O.V., 497
Hong, J., 239
Hong, J.H., 175 J
Hopcraft, K.I., 210211, 219221 Jackson, J.D., 471472, 521522
Horikoshi, Y., 13 Jacobs, B.C., 78, 88
Horner, J.L., 157, 205 Jacquot, P., 477
Hornung, H., 166167 Jaffe, J., 188, 210211, 219221
Horodecki, K., 359360 Jakeman, E., 7273
Horodecki, M., 359360 Jakiel, J., 491
Horodecki, P., 359360, 406407 Jaksic, Z., 524
Horodecki, R., 359360 Jamiokowski, A., 412
Horrom, T., 373 Jaroszewicz, Z., 128
Hoshino, K., 60 Jarzyna, M., 346427
Hou, H., 137138 Javidi, B., 173174, 239242
Hou, S., 154 Jaynes, E.T., 1415
Hou, W., 188190 Jefferts, S.R., 369, 371373, 426427
572 Author Index
Jeffrey, E., 77 Katz, B., 119120

Jelezko, F., 78 Katz, P., 125126, 244245
Jelinek, L., 529530 Kaushik, S., 350
Jenkins, B.K., 180, 182184 Keller, M., 35, 38
Jenkins, S.D., 94 Keller, O., 553554
Jennewein, T., 94 Kemp, B.A., 437485
Jensen, K., 373 Kennedy, T.A.B., 94
Jeong, H., 306307, 321323 Kerlikowske, K., 170
Jetter, M., 32, 58 Keryer, G., 125126, 153, 205
Ji, Z., 29 Kessler, C., 32
Jian, P., 353 Ketterle, W., 373
Jiang, P., 96 Khalesifard, H.R., 128
Jiang, S.S., 131132 Khalfin, L.A., 3
Jiang, X., 58 Khalil, A.M., 127128
Jin, G., 144, 180181 Khan, M., 65
Jin, G.R., 396 Khoo, I.C., 171172
Jin, Y., 144 Khoury, J., 131132
Joens, K.D., 96 Kilin, S.Ya., 2
John, S., 13 Kim, A.D., 210211, 219221
Johnson, B.C., 69 Kim, E.T., 542543
Joo, J., 394395, 422 Kim, J., 31, 41, 128
Jordan, D.L., 209210 Kim, M.S., 70
Joseph, J., 171172 Kim, Y., 128
Jridi, M., 135137, 141, 154, 244245 Kimball, B., 176177, 179
Jung, J.H., 128 Kimble, H.J., 8, 18, 23, 89, 350
Jur, B., 910 Kiraz, A., 4344
Jutamulia, S., 153 Kitagawa, M., 371373
Klauder, J.R., 347, 363364, 367369
K Klude, M., 5960
Kacprowicz, M., 404405 Knight, P.L., 70, 351
Kajzar, F., 179 Knill, E., 12
Kako, S., 60 Knox, R.P., 29
Kakue, T., 173174, 239 Knysh, S., 350351
Kalker, T., 139 Kobayashi, M., 61
Kan, H., 31, 41 Kohlgraf-Owens, T., 210211, 219221, 246
Kanda, H., 61 Kojima, H., 131132
Kaniber, M., 56 Kok, P., 78, 348349, 358, 390, 396,
Kaplan, A.E., 558 426427
Kardynal, B.E., 31 Kolba, D., 136137
Karpinski, M., 406407 Koody nski, J., 346427
Kasevich, M.A., 373 Kolodziejczyk, A., 128
Kashanianfard, M., 525526 Kolokolov, A.A., 528
Kastner, M.A., 89 Komarov, I.V., 3
Kato, H., 171 Komorowska, K., 179
Kato, J.I., 127128 Kondo, K., 131132
Kato, U.I., 171172 Konforti, N., 141
Kattawar, G.W., 187188, 210211, Kong, J.A., 446448, 454, 456457, 459462,
219221 467472, 475, 529530, 553554
Author Index 573
Korchemskaya, E.Y., 171172 Lapointe, J., 58

Korlacki, R., 29 Laskowski, W., 399400
Koschorreck, M., 373 Laub, J., 451452
Koshiba, M., 509 Laucht, A., 56
Kostinski, A.B., 212213 Le Brun, G., 210211, 219221
Kothapalli, S.-R., 169170, 175177, 179 Le Jeune, B., 210211, 219221
Kothiyal, M.P., 179 Leadbeater, M.L., 33
Koukoulas, T., 131132 Leanhardt, A.E., 373
Kouzoubov, A., 210211, 219221 Lebanon, G., 128
Kowarschik, R., 179 Lee, B., 128
Kramers, H.A., 16 Lee, C., 82
Krasinski, J.S., 352353 Lee, H., 348350, 390, 396397
Kraus, B., 426427 Lee, J., 82
Krause, J., 16 Lee, S.H., 197200
Krauter, H., 373 Lee, S.J., 542543
Krebs, O., 80 Lefevre, V., 17
Kremer, P.E., 96 Legero, T., 3637
Kress, A., 56 Lei, L., 144, 239
Krischek, R., 399400, 424425 Lei, M., 171172
Kudriavtseva, A.D., 176177 Leibfried, D., 369, 427
Kuhn, A., 3639, 96 Leitch, A.M., 170
K
uhn, S., 64 Lemaitre, A., 80
Kumar, P., 96 Leonard, I., 125126, 156, 188, 190191,
Kumar, R., 96 193195, 209210, 212, 219221,
Kumar, S., 96 230231
Kuo, M.-K., 529 Leonhardt, U., 440, 549550
Kurtsiefer, C., 77 Lerner, A., 209210
Kutay, M.A., 131132 Levi, O., 144, 146147
Kuwashima, S., 127128 Levie, H.S., 209
Kuzmiak, V., 539, 546547 Levy, H., 144, 146147
Kuzmich, A., 18, 40, 42, 94 Lewis, G.D., 209210
Kuznetsov, A.I., 520521, 557 Li, C.-F., 553
Kwak, M.H., 542543 Li, M., 96, 518519
Kwiat, P.G., 7778, 94 Li, P., 239, 241242
Kyriienko, O., 96 Li, S., 239
Li, W., 137138
L Li, Y., 518519
Labit, C., 133135 Liaw, J.W., 529
Ladd, T.D., 78 Liew, T.C.H., 89, 96
Lading, L., 179 Lifschitz, E.M., 525, 551, 555556
Laflamme, R., 12, 78 Lifshitz, E.M., 467468, 471472
Lalanne, P., 59, 82 LIGO Collaboration, 346347, 366367,
Lan, S.-Y., 94 388, 400, 425426
Landau, L.D., 467468, 471472, 525, 551, Likharev, K.K., 89
555556 Lim, J.S., 128
Lange, B., 38 Lin, C., 239, 241242
Lange, W., 35, 38 Linfoot, E.H., 166167
Langrock, C., 96 Ling, A., 51
574 Author Index
Linne von Berg, D.C., 144 Mahon, R., 210211, 219221

Linver, M., 170 Maier, S.A., 511513
Lippens, F., 520521, 525526 Matre, X., 39, 9293
Liu, L., 144, 180, 529530 Majumdar, A., 89
Liu, R., 525526 Makhonin, M.N., 96
Liu, T., 239, 241242 Malakyan, Y., 96
Liu, W., 239 Mandel, L., 6, 89, 23, 50, 7273, 76, 348,
Liu, W., 239 351352
Liu, W.M., 396 Manin, L., 46, 79
Liu, X., 518519 Mansour, A., 125126, 141, 205, 235236
Liu, Y., 61, 70, 144 Mansuripur, M., 439, 450452, 454,
Liu, Z., 239, 241242 459462
Ljunggren, D., 96 Maradudin, A.A., 503504, 509511, 516,
Lloyd, S., 348349, 359360, 389390 536537, 552553, 558
Lobo, C.J., 31, 47 Marcuvitz, N., 288289, 302303
Lodahl, P., 58 Markiewicz, M., 426427
Lohmann, A.W., 119120 Marklund, M., 493, 532533
Loncar, M., 65, 82 Marom, E., 131132, 138, 164165
Longhi, S., 501 Marozas, J.A., 330331
Lotrian, J., 210211, 219221 Marques, R., 529530, 550
Loudon, R., 12, 266, 406, 467468 Marrocco, M., 25
Lu, C.-Y., 24, 358 Marte, P., 18
Lucero, E., 358 Martel, J., 550
Lucivero, V.G., 394395, 404405 Martin, A., 96
Luis, A., 349350, 392, 396 Martinez, A.S., 209210
Lukyanchuk, B., 520521, 557 Marzolino, U., 399
Lukin, M.D., 41, 89, 94, 427 Mas, D., 144
Lukishova, S.G., 29 Masmoudi, N., 135
Lundeen, J.S., 350351, 404405, 408, Mason, M.D., 59
424425 Massar, S., 349350
Luo, X., 529530 Massou, F., 94
Luo, Y., 526, 549550 Matsukevich, D.N., 94
Lutkenhas, N., 2 Matsuzaki, Y., 426427
Luxmoore, I.J., 96 Maxwell-Garnett, J.C., 533534
Lvovsky, A.I., 8889 Mazurenko, Y., 241
Lynch, R., 348 McCall, S.L., 347, 363364, 367369
McCusker, K.T., 78
M McDonnell, S., 144, 148150
Ma, J., 371373, 423 McGinis, L., 170
Ma, L., 90 McGlamery, B.L., 209210
Ma, X., 518519 McKeever, J., 40
Ma, Y., 96 McNamara, A.J., 29
Macchiavello, C., 350, 353, 396397, Medina, F., 529530, 550
407408 Meher, P.K., 137138
Maccone, L., 348351, 359360, 389390, Meixner, A.J., 29
394396, 404405, 408409 Memmolo, P., 173174, 239
Machida, S., 13 Mendlovic, D., 131132, 141, 150152,
Magnan, P., 144145 241242
Author Index 575
Menke, N., 171172 Mugnai, D., 538539

Merolla, J.M., 241 Mukherjee, A.Yu., 8283
Mesa, F., 550 Mullen, L., 210211, 219221
Meschede, D., 33 Muller, A., 24, 51
Messin, G., 59, 94 Muller, E., 59
Metcalf, B.J., 78 M
uller, G., 33
Meyer, R., 56 Munoz Tapia, R., 422
Meystre, P., 14, 17, 3334 Munoz-Matutano, G., 58
Miccio, L., 173174, 239 Munro, W.J., 358, 394395, 422
Michler, P., 4344, 5860 Muth, J., 210211, 219221
Mie, G., 209210 Myska, R., 390391
Migdall, A., 51, 75
Migdall, A.L., 78
Mikhailov, E.E., 373 N
Milburn, G.J., 12, 33, 76, 347, 350352, Nagata, T., 390, 424425
354358 Naik, D.S., 94
Miller, D.A., 210211, 219221 Nakajima, N., 144
Miller, D.A.B., 166167 Nakamura, T., 127128
Miller, M., 139 Nakamura, Y., 78
Miller, P.C., 210211, 219221 Nakashima, M., 171172
Miller, R., 40, 89 Napolitano, M., 373
Miniewicz, A., 179 Natarajan, T., 137138
Minkowski, H., 440 Naughton, T.J., 173174, 239
Mirin, R.P., 53 Neeley, M., 358
Miroshnichenko, A.E., 520521, 557 Nehmetallah, G., 119120
Mitchell, J.L., 134135, 137138 Neji, N., 135
Mitchell, M.W., 373, 400, 424425 Nelin, E.A., 515516, 530
Mitchell, O., 180 Nelson, D.F., 446, 467468
Mizrahi, U., 79 Nemoto, K., 358
Mlynek, J., 8889 Neshev, D., 520521, 557
Mnaymneh, K., 58 Neviere, M., 512
Mocella, V., 525526 Nguyen, T.Q., 136137
Mock, J.J., 525526 Nicklas, E., 373
Moerner, W.E., 26 Nicoll, C.A., 32, 51
Molina, R., 144145 Nielsen, M.A., 382, 403, 406407
Mollow, B.R., 24 Nielsen, T.R., 33
Mlmer, K., 393 Nieminen, T.A., 451
Monroe, C., 78, 94 Nimtz, G., 555
Mookherjea, S., 96 Nisbet-Jones, P.B.R., 96
Moon, S.E., 542543 Nishchal, N.K., 239
Moore, F.L., 347 Nishi, K., 13
Moore, M., 78 Nishio, K., 173174, 239
Moreau, E., 46, 79 Nisida, Y., 61
Moreno, E., 8283 Nogues, G., 39, 9293
Moreno, V., 128 Noh, J.W., 348
Mrk, J., 33 Nojima, S., 518
Mortensen, N.A., 526, 549550 Nori, F., 371373, 423
Moscoso, M., 210211, 219221 Northup, T.E., 38, 89
576 Author Index
O Passow, T., 59
OBrien, J.L., 78, 390, 424425 Patel, R.B., 51
OBrien, S., 521522 Paternostro, M., 82
Oates, C.W., 369, 371373, 426427 Paturzo, M., 173174, 239
Oberthaler, M.K., 373 Pavesi, L., 520
Obod, Y.A., 490558 Paz, S., 144
Ochoa, E., 175 Pazzi, G.P., 538539
Oesterhelt, D., 176177 Peaudecerf, B., 358
Ohnishi, K., 127128 Pedrosa, H.N., 59
Okada, M., 61 Pegg, D.T., 348350, 392
Okamoto, N., 171172 Pelc, J.S., 96
Okamoto, R., 390, 424425 Pellizzari, T., 350, 407408
Okamoto, T., 175 Pelton, M., 45, 47, 49
Olivares, S., 353, 367, 404 Pendry, J.B., 543, 550
Oliveira, S.L., 520521 Pendry, J.P., 521522
Olivos-Perez, L.I., 179 Penfield, P., 441, 459460
Olkhovsky, V.S., 491 Pennebaker, W.B., 134135, 137138
Olver, F.W.J., 290292 Pepper, M., 33
Ong, J.R., 96 Perez, J., 144
Ono, T., 396397 Perez-Delgado, C.A., 396
Oppenheim, A.V., 136137 Perina, J., 911, 1718, 349350, 390392
Oravec, M., 119120, 125126, 141, 153, Pernicka, J.C., 209210
169170, 239, 244245 Peters, N.A., 77
Orwa, J., 68 Peterson, C.G., 94
Oton, C.J., 520 Peterson, J., 59
Ou, Z.Y., 50 Petite, G., 492, 501502, 539
Ouerhani, Y., 136138, 154 Petroff, P.M., 4344, 79, 89
Oughstun, K.E., 264343 Petz, D., 385386
Ozaktas, H.M., 131132, 241242 Pezze, L., 359360, 367, 389390, 396397,
Ozbay, E., 13, 518 399, 424425
Pfeifer, R.N.C., 451
Pfitzmann, A., 139
P Phlypo, R., 209
Palombini, C.L., 341343 Phong, B., 193194
Pan, J.-W., 358 Piazza, F., 373
Pandalai, S.G., 209210 Piederrie`re, Y., 210211, 219221
Panotopoulos, G., 239 Piestun, R., 166167
Pant, M., 96 Pinel, O., 353
Pantano, C., 522 Pinel, P., 241
Papoulis, A., 205 Priz, G., 171172
Paris, M.G.A., 353, 367, 396397, 404 Pitaevskii, L.P., 467468, 471472
Park, G., 128 Pitkin, M., 346347, 366367, 388
Park, H., 8283 Pittman, T.B., 78, 88
Park, H.M., 197200 Plant, J., 47
Parka, J., 179 Plenio, M.B., 350, 407408, 426427
Parkins, A.S., 18 Plick, W.N., 349350, 396
Parks, T., 136137 Podolky, B., 75
Pasquini, T.A., 373 Poggi, C., 210211, 219221
Author Index 577
Poizat, J.-P., 2, 59, 64 Reitzenstein, S., 59

Polman, A., 525526 Rempe, G., 3334, 3639
Polyakov, S.V., 51 Ren, L., 171172
Polzik, E.S., 373 Rendina, I., 525526
Poole, P.J., 58, 96 Renema, J.J., 373
Prawer, S., 61 Resch, K.J., 424425
Pregnell, K.L., 424425 Retzker, A., 426427
Pressman, P., 170 Rhodes, W.T., 241
Prevedel, R., 424425 Ribordy, G., 2, 23
Pritchard, D.E., 369, 373, 426427 Richardson, E.G., 133135
Pryde, G.J., 391392, 400401, 424425 Rijmen, V., 238
Psaltis, D., 239 Rio, K.R., 137138
Riris, H., 131132
Q Ritchie, D.A., 32, 47, 51, 53, 79
Qasmi, S., 125126, 156 Rivas, A., 396
Qian, H., 59 Rivas, D., 58
Quaglia, A., 125126, 193194, 244245 Rivoire, G., 176177
Robert, J.M., 46, 79
R Roberts, P.J., 209210
Rabeau, J.R., 68 Robertson, J.W., 24
Rabinovich, W., 210211, 219221 Robin, F., 8283
Radzewicz, C., 352353, 406407 Roch, J.F., 94
Raimond, J.M., 13, 17, 40, 94 Rodrguez-Ortiz, M.M., 179
Raindel, S., 96 Roger, G., 68, 7273
Rajput, S.K., 239 Rohner, J., 477
Rakher, M.T., 90 Roig, A.B., 144
Rakovic, M.J., 210211, 219221 Rolando, S., 144145
Ralph, T.C., 355358 Romalis, M., 369, 373, 426427
Ramahi, O.M., 525526 Rosen, J., 119120
Ramakrishna, S.A., 467468 Rosen, N., 75
Rand, S.C., 520521 Rossbach, R., 32, 58
Ranfagni, A., 538539, 555 Rouabhi, S., 241
Rao, D.V.G.L.N., 169172, 175177, 179 Rowan, S., 346347, 366367, 388
Rarity, J.G., 7273 Royall, B., 96
Rasouli, S., 128 Rubin, M.A., 350
Raterman, G.M., 349350, 396 Rubinsztein-Dunlop, H., 451
Rauch, H., 390391 Rudolph, T., 387, 393
Rayleigh, J.W.S., 501 Ruggeri, R., 555
Raymer, M.G., 86 Rundquist, A., 89
Recami, E., 491 Russ, J.C., 170
Refregier, P., 192193, 239242 Russo, C., 38
Regelman, D.V., 79 Rybarczyk, T., 358
Rehn, H., 179 Ryle, J.P., 144, 148150
Reid, M.D., 357 Ryu, Y.C., 542543
Reid, S., 346347, 366367, 388
Reimer, M.E., 96 S
Reinhard, A., 89 Saba, M., 373
Reischle, M., 32 Sabbah, S., 209210
578 Author Index
Sacchi, M.F., 387, 390391 Schurig, D., 550

Sagnes, I., 80 Schwartz, I., 96
Sahin, A., 241242 Schwemmer, C., 399400, 424425
Sai, S.S., 176177 Schwinger, J., 363364
Salandrino, A., 526, 558 Sciarrino, F., 394395, 404405
Salathe, R.P., 477 Scully, M.O., 16
Saldanha, P.L., 459460 Sebald, K., 59
Sallen, G., 59 See, P., 32
Salomon, D., 133135 Sementsov, D.I., 497
Salter, P.S., 78 Semouchkin, G.B., 522
Samms, R.W., 166167 Semouchkina, E., 522
Sanchez-de-la-Llave, D.D., 179 Sendhil, K., 179
Sanders, B.C., 347 Seravalli, L., 58
Sandri, P., 555 Serebryannikov, A.E., 518
Sanguinetti, B., 96 Seshadreesan, K.P., 396397
Santamato, A., 96 Sewell, R.J., 373
Santisteban, A., 131132 Shabanov, S.V., 522523
Santori, C., 47, 49, 51, 54, 60 Shabtay, G., 141
Sarovar, M., 350 Shaji, A., 350
Sasaki, K., 390, 424425 Shamir, A., 238
Sato, T., 131132 Shankar, R., 359
Sauvan, C., 59 Shapiro, J.H., 78, 347
Savage, C.M., 17 Shashar, N., 209210
Savanier, M., 96 Shaw, J.A., 210211, 219221, 246
Savona, V., 89 Shelby, R.A., 523524
Sawchuk, A.A., 131132, 180, 182184 Shelykh, I.A., 96
Sayeed, A.M., 133f Shemer, A., 164165
Sayood, S., 133135 Shen, J.-T., 60
Sayrin, C., 17, 358 Shen, L.C., 475
Sazonova, V., 58 Shen, Y., 60
Schaetz, T., 369, 427 Sheppard, C.J., 446
Schafer, R.W., 136137 Sheppard, C.J.R., 166167
Scharf, L.L., 126 Sheridan, J.T., 144, 148150, 162163
Schechner, Y.Y., 210211, 219221 Sherman, G.C., 274275, 280282, 285,
Schelkens, P., 141, 205, 235236 287289, 292, 295297, 299, 301302,
Schiff, L.I., 517, 540 304305, 314316, 338340, 342343
Schleich, W.P., 352353 Sheynihovich, D.V., 192193
Schmid, A.W., 29 Shields, A.J., 3132, 4243, 47, 51, 53, 79
Schmidgall, E.R., 96 Shih, F., 180
Schmidt, H., 41, 89 Shih, M.Y., 171172
Schmidt, P.O., 38 Shimozato, Y., 173174, 239
Schmiedmayer, J., 369, 373, 426427 Shin, Y., 373
Schmitt, J.M., 209210 Shishido, A., 171172
Schnabel, R., 350351, 400, 404405 Shvartsburg, A.B., 490558
Schneider, C., 89 Sibilia, C., 196
Schoenfeld, W.V., 4344, 79 Siddique, M.A., 154
Schultz, S., 523524 Siddique, O.F., 525526
Schulz, W.-M., 32, 58 Silberberg, Y., 390
Author Index 579
Silveirinha, M.G., 525526, 558 Stevenson, R.M., 3132, 47, 79

Silverstone, J.W., 96 Stewart, W.J., 543
Simbula, A., 96 Stobbe, S., 58
Simon, C., 94 Stoev, K.N., 128
Simon, J., 94 Stoltz, N., 89
Simova, E.S., 128 Strang, G., 137138
Simpson, D.A., 61, 68 Stratton, J.A., 264266
Singh, R., 373 Strauf, S., 5960
Singh, S., 11 Stroud, C.R., 29
Sisakyan, N., 96 Strouse, G.F., 59
Skauli, T., 144, 167f, 168169 Stryewski, J., 210211, 219221
Skotiniotis, M., 426427 Stute, A., 38
Skrotsky, G.V., 528 Su, C.H., 61
Slattery, O., 90 Suffczynski, J., 80
Slawecki, D., 137138 Sugihara, O., 171172
Slosser, J.J., 17 Sullivan, J.M., 210211, 219221
Smelyanskiy, V.N., 350351 Sumetsky, M., 545546
Smerzi, A., 359360, 367, 373, 389390, Summy, G.S., 348, 392
396397, 399 Sun, C., 371373, 423
Smith, B.J., 350351, 404405, 408 Sun, X., 518519
Smith, C., 137138 Sungwook, Y., 137138
Smith, D.D., 522523 Susini, C., 555
Smith, D.R., 523524, 550 Suzuki, T., 131132
Smith, R.A., 170 Suzuki, Y., 171172
Smolyakov, A., 552 Swartziander, E., 137138
Smolyaninov, I.I., 41, 510511
Sogandares, F.M., 188 T
Solomon, G.S., 45, 47, 49, 51, 54 Taha, A.K., 241
Son, W., 70 Tahara, T., 173174, 239
Song, M., 61, 70 Tajahuerce, E., 173174, 239
Soraghan, J.J., 173174, 239 Takeuchi, S., 390, 424425
Srensen, A.S., 89, 427 Talyanskii, V.I., 33
Sorensen, H., 136137 Tame, M.S., 82
Spagnolo, N., 394395, 404405 Tan, S., 131132
Specht, H.P., 3839 Tanaka, H., 171172
Spekkens, R.W., 387, 393 Tang, J., 144
Spiller, T.P., 394395, 422 Tang, X., 90
Spring, J.B., 78 Tanji, H., 94
Srinivasan, K., 90 Tanner, D.B., 492, 542
Stacey, A., 68 Tapster, P.R., 7273
Stahlhofen, A.A., 555 Tassin, P., 524
Stavrias, N., 69 Tavassoly, M.T., 128
Stegun, I., 513514, 517518, 548549 Testorf, M., 119120
Steinberg, A.M., 424425, 554 Thackara, J.l., 179
Steiner, M., 29 Thew, R., 2
Stenflo, L., 493, 532533 Thienpont, H., 141, 205, 235236
Stepanchikov, D.A., 171172 Thierry-Mieg, V., 46, 79
Stern, A., 162163 Thoma, R., 176177
580 Author Index
Thomas, J.C., 210211, 219221 Vesseur, E., 525526

Thomas-Peter, N., 78, 350 Vest, C.M., 164165
Thomay, T., 58 Vijayan, C., 179
Thompson, R.M., 47 Villing, A., 2, 64
Tietz, C., 61 Vinkler, Y., 426427
Tisse, C.-L., 166167 Vitelli, C., 394395, 404405
Tittel, W., 2, 23, 94 Voelz, D.G., 156
Toffoli, L., 404405 Vogel, W., 1718
Toninelli, C., 520 Voloshinov, V.B., 171
Tonizzo, A., 210211, 219221 Volpian, O.D., 490558
Torres, J.P., 351, 358 Volz, J., 77
Toth, G., 385386, 399400 Volz, T.B., 89
Toyoda, H., 179 von Borczyskowski, C., 61
Treibitz, T., 210211, 219221 Voss, K.J., 187188, 210211, 219221
Treps, N., 353 Vuckovic, J., 47, 51, 54, 89
Tretyakov, S., 523524 Vucokovic, J., 89
Treussart, F., 68, 94 Vuletic, V., 94
Trevisi, G., 58
Tribu, A., 59
Truver, S., 209 W
Tsang, M., 396 Waks, E., 20
Tu, B., 144 Walker, J.G., 210211, 219221
Tukey, J.W., 136137 Walls, D.F., 8, 17, 76, 351352, 354357
Tulet, M., 144145 Walmsley, I.A., 86, 350351, 358, 427
Tulino, A., 173174, 239 Walther, H., 16, 23, 3336, 38
T
ureci, H.E., 89 Walther, T., 16
Tyo, J.S., 210211, 219221, 246 Wang, D., 96
Wang, J., 96
U Wang, J.M., 182184
Ueda, M., 371373 Wang, Q., 144, 239, 553
Uemori, T., 171172 Wang, X., 371373, 423
Umeda, T., 69 Wang, Y., 144, 171172, 239
Umemura, N., 61 Wannier, G.H., 520
Unitt, D.C., 53 Ward, M.B., 32
Ursell, F., 288289 Ware, M., 75
Warren, R.E., 131132
V Wasilewski, W., 350351, 373, 404405,
Vahala, K.J., 545 408
Vaillancourt, Y., 179 Weaver, C.S., 153
Vamivakas, A.N., 24 Weber, A.G., 182184
van der Vaart, A.W., 381 Weber, T., 2
Van Exter, M.P., 8283 Webster, S.C., 3839
Vander Lugt, A., 153 Wei, L., 520521, 557
Varcoe, B.T.H., 3336 Wei, Y.-J., 58, 89
Vasilev, G.S., 96 Weidemann, A., 188190
Verger, A., 89 Weidinger, M., 3336
Versteegh, M.A.M., 96 Weig, E.M., 358
Veselago, V.G., 472, 523524 Weihs, G., 94
Author Index 581
Weinfurter, H., 77, 94, 358, 399400, Xiang, G.-Y., 424425

424425 Xiao, M., 24, 350
Weinstein, Y.S., 350 Xiong, F.-L., 58
Wellentowitz, S., 1718 Xu, L., 239, 241242
Wells, W.H., 188 Xuan, N.P., 176177
Werner, D.H., 522
Werner, M.J., 31, 41 Y
Westfeld, A., 139 Yablonovitch, E., 13
White, A.G., 94 Yaffe, M.J., 170
White, S., 137138 Yamada, H., 13
Wieczorek, W., 399400, 424425 Yamagata, K., 175
Wiener, N., 232 Yamaguchi, I., 127128, 171175
Wiersma, D.S., 520 Yamamoto, Y., 13, 20, 3031, 41, 45, 47,
Wigner, E.P., 352353 49, 51, 54, 60, 69
Wildfeuer, C.F., 350 Yan, Y., 180181
Wilk, T., 3839 Yang, Y., 13
Williams, C.P., 426427 Yao, B., 171172
Wilson, D.W., 515516 Yao, Z., 180
Wineland, D.J., 347349, 371373, 390 Yaroslavsky, L.P., 125, 130, 132
Winful, H.G., 556 Yeh, P., 175, 518519
Winger, M., 30, 89 Yelleswarapu, C.S., 169170, 175177, 179
Wiseman, H.M., 349350, 390392, 396, Yin, Z.-Q., 96
400401, 424426 Yokoyama, H., 13
Wodkiewicz, K., 352353, 356357 Yokozeki, S., 127128
Woerdman, J.P., 8283 You, Y., 210211, 219221
Wolf, E., 6, 9, 141, 208, 211212, 219220, Young, M.E., 525526
351352, 358, 554 Young, R., 131132
Wolf, G., 139 Young, R.J., 79
Wong, F.N., 78 Youngs, I., 543
Wood, M.V., 171172 Yu, F.T.S., 153
Woods, C.L., 131132 Yuan, Z.L., 3132
Woods, G., 41, 89 Yuen, H.P., 5
Wootters, W.K., 2 Yurke, B., 347, 363364, 367369
Worschech, L., 59
Wrachtrup, J., 61 Z
Wright, E.M., 17, 439, 450452 Zagury, N., 17, 384, 417
Wu, B., 61, 70 Zaitsev, A.M., 61
Wu, D., 89 Zakharian, A.R., 439, 450452
Wu, E., 61, 68, 70 Zakharov, V.E., 507
Wu, L.-A., 350 Zalevsky, Z., 131132, 141, 144, 164165
Wu, M., 180181 Zapletal, P., 96
Wu, P., 175177 Zawisky, M., 390391
Wunderlich, C., 39, 9293 Zayats, A.V., 41, 82, 512
Wyns, P., 297 Zayatz, A.V., 510511
Zbinden, H., 2, 23, 8283, 94
X Zeilinger, A., 94, 358
Xia, P., 173174, 239 Zemnek, P., 438439
Xia, S.F., 142 Zeng, H., 68
582 Author Index
Zhang, B., 47 Zhou, X., 358

Zhang, J., 518521, 525526, 549550 Zhouand, S., 154
Zhang, L., 4344, 350 Zhu, F., 96
Zhang, T., 173175 Zhu, S.-Y., 13
Zhang, X., 180 Zibold, T., 373
Zhang, Y.R., 239, 396 Zibrov, A.S., 8283, 94
Zhang, Z., 180, 529530 Zlotnik, A., 144, 164165
Zhang, Z.M., 180, 492, 529530, 542 Zoller, P., 18
Zhao, L., 144, 180 Zueva, Yu.M., 542
Zhao, Q., 520521, 525526 zukowski, M., 358
Zhao, Y., 24 Zurek, W.H., 2
Zheng, Y., 29, 171172 Zwierz, M., 396
Zhou, C.Y., 68 Zwiller, V., 59
Zhou, J., 520521, 525526 Zyczkowski, K., 382, 410
SUBJECT INDEX
Note: Page numbers followed by f indicate figures and t indicate tables.
A B
Abraham formulation, 450451 Barbaras spectrum, 155, 155f
AbrahamMinkowski debate, 440 Bayesian approach
Abraham momentum density decoherence-free interferometry,
field momentum, 458459 390392
transmitted, 457 definition, 349350
Adapted all-numerical correlator estimation theory
AUC, 157159, 158f classical parameter, 377379
comparison, FFT and NO_FT algorithm, quantum parameter, 386387
161162, 163f realistic interferometry
correlation results, 159161, 160f, 162f phase diffusion, 420421
drawbacks and stringent requirements, photonic losses, 419420
156 Bayesian inference, 348
face database, 157, 158f BEC interferometry, 373
numerical algorithm, 156157, 157f Brewster effect
reconstructed image, 159, 159f generating function, 541
reconstruction, input plane, 160161, polarization structure, 542
161f problems, 538
ROC curve, 157159, 158f Brillouin precursor
Adaptive block scanning, 242 application, 301302
Antibunching asymptotic analysis, 294
example, 9f characteristic, 302303
Hanbury Brown and Twiss dynamic behavior, 302303, 303f
experiment, 8f first-order saddle points, 304305,
optical microcavity, 25 307308
resonant fluorescence, 24 limited validity, 300301
second-order correlation function, 78 second-order saddle point, 306307
Area under curve (AUC), 157159, 158f
Asymmetric cryptography, 235 C
Asymptotic analysis Cauchys residue theorem, 289
Brillouin precursor field, 294 Chu formulation
Cauchys residue theorem, 289 electrodynamics, moving media, 443444
dynamical behavior, 294 field and material contributions, 468470
FourierLaplace integral representation, Cipher, digital encryption, 140
289 Circular polarimeter, 220221, 220221f
Olver-type path, 290, 291f Classical electrodynamics
single-resonance Lorentz model Abraham formulation, 450451
dielectric, 292 EinsteinLaub formulation, 450451
Sommerfeld precursor, 292 energy and momentum continuity, 441
space-time intervals, 295 momentum density and stress tensors,
Autonomous underwater vehicles 451, 451t
(AUVs), 209 moving media
583
584 Subject Index
Classical electrodynamics (Continued ) imperfect visibility, 402403f, 406408

Amperian formulation, 444445 Kraus operators, 403
Chu formulation, 443444 phase diffusion, 402403f, 404
Minkowski formulation, 442443 photonic losses, 402403f, 404406
quasi-stationary approximation Decoherence process, 350351
Doppler shifted photon, 448449, 449f Deconvolution, 130
plane wave, 446448 Decryption algorithm, 139
Classical interferometry, 346 Definite photon-number state
Classical optical tweezers, 476477, 476f interferometry, 367369
Classical simulation (CS), 409412, 411f Degree of polarization (DOP)
Coherent + squeezed vacuum strategy calculation
conjugated map, 423 correlation peak, 218219
gravitational-wave interferometry nonuniform lighting, 218, 218f
experiments, 425426 Stokes images, 217
Heisenberg picture transformation, 423 Denoising method, 154
interferometry, 366367 Depth-of-field (DOF)
NOON states, 424425 digital in-line holographic microscopy
phase estimation precision, 424425, 425f system, 148149
Coherent state interferometry, 364365 glass lattice three-dimensional mapping,
Collective dephasing. See Phase diffusion 164165
Color image encryption algorithm, 241242 Diamond-plasmonic system, 86, 87f
Compensator, 212213, 214t Dielectric fluids, 462464
Compression Digital encryption
block, 135, 136f cipher, 140
image transmission, 236 cryptovariable, 139140
Conjugated map, 423 decryption algorithm, 139
Coregistration error and pixel count, encoding and decoding keys, 139
166168, 167f experimental setup, 139, 139f
Coulomb blockade, 89 key, 139140
Coupled QD, 84, 85f RSA algorithm, 140
Cryptovariable, digital encryption, 139140 steganography, 139
Digital holography (DH), 148149
D Digital image processing (DIP)
Data compression, 134135 compression block, 135, 136f
Deblurring method, 132, 132f data compression, 134135
de Broglie waves, 490 DCT, 137138, 138f
Decoherence-free interferometry digital encryption
Bayesian approach, 390392 cipher, 140
entanglement, 399400 cryptovariable, 139140
indefinite photon-number states decryption algorithm, 139
Gaussian state, 396398, 399f encoding and decoding keys, 139
local oscillator, 393 experimental setup, 139, 139f
optimal states, 395396 key, 139140
reference beam, 393395 RSA algorithm, 140
multipass protocols, 400401, 401f steganography, 139
QFI approach, 388390 DTF, 136137
Decoherence models, realistic image compression, 134135
interferometry JPEG, 135
Subject Index 585
LempelZivWelch technique, 135 DOF, 164165

lossless compression, 135 EDOF, 165, 166f
lossy compression, 135136 RT, 164165
merits and drawbacks hybrid methods, 143144
advantages, 124 image detection, identification, and
dynamic range, 124 localization
filtering principle, 126, 126f Barbaras spectrum, 155, 155f
flexibility, 124 denoising method, 154
low light sensitivity, 124 drawbacks and stringent requirements,
nonuniform camera-response surveyor, 154
125 Lenas spectrum, 155, 155f
object localization, 125, 126f OC, 153
OC, 125126 reprogrammable targets, 154
preprocessing and postprocessing steps, target image, 155, 155f
125 imaging systems and applications
SNR, 124 digital in-line holographic microscopy
Moire correction system, 148150
FT, 128 image quality, 144
moire effect, 128 image sensor, 144146
moire noise, 128, 129f in vivo brain optical imaging system,
spectral filtering, 127 146148
superimposed pattern, 127f, 128 light to image formation
receiver, 134, 134f image degradation, 122123
redundance, 134135 image restoration, 123124
transmission channel, 134, 134f limitations, 122123
transmitter, 133, 134f limitations, 142
Wiener filtering need for, 141
deblurring method, 132, 132f numerical approach, 144
deconvolution, 130 numerical implementation
features, 131 adapted all-numerical correlator,
Gaussian white noise, 132, 133f 156162
image denoising method, 132 unitary discrete LCT, 162164
noise suppression, 131132, 131f optical spatial filtering, 142
SNR, 131132 SLM, 142
Digital in-line holographic microscopy spectral imager
system coregistration error and pixel count,
advantage, 148149 166168, 167f
DH, 148149 information capacity, 168f, 169
DOF, 148149 spatial coregistration, 166167
Digital processing (DP), 120 super imaging system
Digital processing vs. optical processing light field photography, 151152, 152f
advantages, optics, 141142 multiexposure camera, 152, 152f
correlation techniques, 122 multiple aperture cameras, 150151,
DIP, 124141 151f
drawbacks and stringent requirements, optical setup, 150, 150f
121 subchannel, 150151
Fourier methods, 141 Discrete cosine transform (DCT)
glass lattice three-dimensional mapping DIP, 137138, 138f
586 Subject Index
Discrete cosine transform (DCT) beam splitter, 7071, 71f

(Continued ) correlation, 72
optical compression, 239240 FranckHertz scattering, 72
Discrete FT (DTF), 136137 heralding photon, 76
Dispersive media Michelson interferometer, 8081, 82f
field and material contributions, 468470 micropillar cavity, QD, 80, 81f
time-averaged energy and momentum parametric down-conversion process, 78
propagation, 470472 photoelectric effect, 7576
Dispersive signal propagation photoluminescence spectrum, QD, 79,
singular dispersion limit 80f
algebraic power, 323 quantum dot biexciton, 79
below-resonance angular carrier signal and idler photons, 73, 73f
frequency, 322f, 323 single-mode optical fiber, 77
numerically determined peak two-photon system, 7273
amplitude decay, 320321, 321f vertical/horizontal polarization, 7475,
weak dispersion limit 74f
advantage, 327 Entanglement, 33
propagated wave field, 325326, 325f Estimation theory
propagation problems, 324325 classical parameter estimation
Distant saddle points, 272273 Bayesian approach, 377379
Dominant saddle points FI approach, 374377
angular frequency, 288 transmission coefficient estimation,
complex phase behavior, 285287 379382
critical space-time value, 287288 quantum parameter estimation
definition, 285 Bayesian approach, 386387
Doppler shifted photon, 448449, 449f QFI approach, 383386
Dynamics of electronic systems (DES), 238 Evanescent waves
definition, 496
E interface reflection phase shifts, 497
EinsteinLaub formulation, 450451 tunneling, features of, 497498
Electromagnetic fields, 439 Extended depth of field (EDOF), 165, 166f
Electromagnetic force
equivalence of total force, 453454, 453f F
equivalence of total time-average force, FabryPerrot interferometer, 369370, 370f
453f, 454455 First-order saddle points, 272
linear media, 439, 439t Fisher Information (FI) approach,
Electromagnetic momentum 374377
field momentum, 458460 Fock states
momentum densities, comparison of, annihilation and creation operators, 34
455458 coherent states, 56
Electromagnetic stress definition, 3
dielectric fluids, 462464 interferometry, 365366
momentum flow into media, 460462 microcavity and atoms, 35
radiation pressure, 464467, 465f, 467f photon number operator, 4
Electromagnetic wave theory, 437438 squeezed state, 6
Encryption. See Digital encryption; Optical FourierLaplace integral representation, 289
encryption Fraunhofer diffraction
Entangled states MATLAB function, 157
Subject Index 587
numerical implementation, VLC and polarization structure, 542

optical FT, 155164 problems, 538
Frustrated total internal reflection (FTIR) evanescent voltage and current waves,
evanescent waves 536538
definition, 496 GHz resonance
interface reflection phase shifts, 497 effective mass, 544
tunneling, features of, 497498 plasma frequency, 543
HID, 495496 properties, 544545
homogeneous lossless opaque layer heterogeneous part, 530531
reflection coefficient, 508 HID, 530531
refractive index, 507508 microwave analogy, gradient photonic
transition point, 509511 barrier
transparent gradient layer dispersive equation, 534
dispersive barrier, 501 microwave effects, 532
experimental and theoretical nonlocal dispersion, coaxial GHz
transmittance spectra, 505f, 506 waveguide, 533534, 533f
frequency, 502 properties, 532533
reflectance spectra, 505f, 506 transmittance spectra, 535, 535f
Fusion, 233 subwavelength narrowings
heterogeneity-induced tunneling,
G 546547
Gamows approach, 490 LaguerreGaussian mode, 545546
Gaussian beam, 475476, 476f NT, 545
Gaussian states radiation losses, 545
decoherence-free interferometry, transformation optics, 549550
396398, 399f theoretical analysis, 529530
general two-mode, 357 Gravitational-wave interferometry
indefinite photon-number states, experiments, 425426
396398, 399f
Gaussian white noise, 132, 133f
GHz resonance H
effective mass, 544 Hanbury Brown and Twiss experiment
plasma frequency, 543 antibunching, 8f
properties, 544545 microcavity and atoms, 39
Glass lattice three-dimensional mapping SPPs, 8486, 85f
DOF, 164165 Hartman paradox, 491
EDOF, 165, 166f Heisenberg limit
RT, 164165 NOON states, 390
Gradient photonic barrier QFI, 348349
dispersive equation, 534 Heisenberg picture transformation, 423
microwave effects, 532 Heisenberg scaling, 367369
nonlocal dispersion, coaxial GHz Heterogeneity-induced dispersion (HID)
waveguide, 533534, 533f FTIR, 495496
properties, 532533 gradient TL, 530531
transmittance spectra, 535, 535f Hidden momentum, 457
Gradient transmission line (TL) Hilbert space
Brewster effect, TE10 mode light, 351352
generating function, 541 photonic losses, 405406
588 Subject Index
Hybrid GaP-diamond whispering gallery multi-image, 244

mode cavity, 91, 92f multiple-image parallel, 243
multiple-parameter discrete fractional
I FT, 243
Ideal linear polarizer, 212213, 214t one-step multiplexing, 243
Image public-key encryption, 242
formation simultaneous compression and
degradation, 122123 encryption, 244245
limitations, 122123 symmetric key algorithms, 235
restoration, 123124 Imperfect visibility
identification, and localization decoherence models, 402403f, 406408
Barbaras spectrum, 155, 155f QFI approach, 409412
Lenas spectrum, 155, 155f Incident optical energy, 456
sensor Indefinite photon-number states
CMOS, 145146 Gaussian state, 396398, 399f
Meteosat Third-generation Imager local oscillator, 393
satellite, 144145 optimal states, 395396
triggered event (TE) detectors, reference beam, 393395
144145 Individual NV color centers
systems and applications autocorrelation function, 64, 64f
digital in-line holographic microscopy color centers, 61
system, 148150 confocal microscope scanning, 62, 63f
image quality, 144 coupling, 67
image sensor, 144146 Cr-related center (UM2), 69
in vivo brain optical imaging system, diamond nanocrystals, 64
146148 electroluminescence, 66
Image compression, 134135 limitation, 62
Image denoising method, 132 NE8 (nickelnitrogen complex), 67, 67f
Image transmission, OIP nickel centers, 68
asymmetric cryptography/public-key NV defect center, 61
cryptography, 235 photon emission, 62
compression, 236 single NV centers, 64
encryption, 237 type Ib synthetic diamond, 61
lossless compression, 234 Information capacity, 168f, 169
lossy compression, 234 In vivo brain optical imaging system
optical compression advantages, 146
DCT, 239240 LSCI technique, 147
digital Fresnel fields, 241 VCSEL, 147148
principle, 239
PSI-DH hologram, 240241 J
RT, 240 Joint transform correlator ( JTC), 153
video applications, 240 Jones and MuellerStokes formalisms,
optical encryption 212213, 214t
adaptive block scanning, 242 JordanSchwinger map, 363364
color image encryption algorithm, JPEG, 135
241242
DRP encoding method, 242243 K
Fourier spectrum, 244, 244f Kerckhoffs principle, 237238
Subject Index 589
L classical electrodynamics
LempelZivWelch technique, 135 Abraham formulation, 450451
Lenas spectrum, 155, 155f Amperian formulation, 444445
Light interference, 346 Chu formulation, 443444
Linear canonical transforms (LCTs), Doppler shifted photon, 448449, 449f
162164 EinsteinLaub formulation, 450451
Linearly crossed polarized (LCP) images, energy and momentum continuity, 441
230231, 230f Minkowski formulation, 442443
Linearly parallel polarized (LPP) images, plane wave, 446448
230231, 230f electromagnetic fields, 439
Linear polarimeter, 219220, 219220f electromagnetic force
Linear polarizer, 212213, 214t equivalence of total force, 453454,
Livens momentum density, 457 453f
Lorentz model dielectric, 264266 equivalence of total time-average force,
Lossless compression 453f, 454455
image transmission, OIP, 234 linear media, 439, 439t
LempelZivWelch technique, 135 electromagnetic momentum
Lossy compression field momentum, 458460
image transmission, OIP, 234 momentum densities, comparison of,
JPEG, 135 455458
lossy encoding, 136 electromagnetic stress
Lossy interferometer. See Photonic losses dielectric fluids, 462464
momentum flow into media, 460462
radiation pressure, 464467, 465f, 467f
M electromagnetic wave theory, 437438
MachZehnder (MZ) interferometer fundamental aim, 440
BEC, 373 Minkwoski stress tensor, 474475
classical theory, 346 momentum and stress, dispersive media
coherent + squeezed-vacuum field and material contributions,
interferometry, 366367 468470
coherent state interferometry, 364365 time-averaged energy and momentum
definite photon-number state propagation, 470472
interferometry optical manipulation
Heisenberg scaling, 367369 classical optical tweezers, 476477,
simple error propagation formula, 369 476f
twin-Fock state, 367369 description, 438439
FabryPerrot interferometer, 369370, Gaussian beam, 475476, 476f
370f interference, 477, 478f
Fock state interferometry, 365366 modeling process, 475
Michelson interferometer, 369, 370f radiation pressure, 477478
phase-sensing uncertainty optical tweezers, 438
JordanSchwinger map, 363364 plane wave scattering
standard configuration, 362363, 362f cylinder, 478481
Ramsey interferometry, 371373, 372f sphere, 481485
shot noise, 347 radiation pressure, 438
Macroscopic theory, optical momentum stationary media, 473
AbrahamMinkowski debate, 440 total average electromagnetic force, 474
causality requirements, 474 Magnetometry interferometer, 369
590 Subject Index
Medical image processing Loudons energy velocity

application, 175176 Brillouins analysis, 264
nonlinear transmission, 176179, 178f dispersive medium, 267
photoinduced anisotropy, 171173, Lorentz model dielectric, 264266
172173f Poyntings theorem, 264266
spatial filtering, 170171, 171f phase function and saddle point
Mesoscopic p-n junctions, single-photon approximate local behavior, 277278
states critical space-time point, 283284
acoustoelectric effect, 33 distant saddle points, 272273
GaAs-based LEDs, 31 dominant saddle points, 285288
InAs QD, 31 dynamical behavior, 280281
outcoupling efficiency, 32 final space-time domain, 284285
photon blockade effect, 31 first-order saddle points, 272
Metatronics, 493 initial space-time domain, 281282
Michelson interferometer, 369, 370f isotimic contours, 269271f
Microcavity and atoms limitations, 267268
atomic inversion, 35 limiting behavior, 275276
entanglement, 33 numerical results, 273274
Fock state, 35 saddle point method, 267268
Gaussian transverse profile, 3637 second-order saddle point, 273
Hanbury Brown and Twiss single-resonance Lorentz model
configuration, 39 dielectric, 274275
laser pulses, 40 space-time parameter, 272
photon blockade, 41 pulsed plane-wave transmission
Raman transitions, 3839 asymptotic description, 332337
Rubidium atoms, 3, 34f, 35 formulation, 329332
Rydberg atoms, 42 transmitted signal evolution,
STIRAP, 3637 338341
trapping states, 3334 superluminal pulse velocity, 341342
Microwave analogy, gradient photonic uniform asymptotic expansions
barrier asymptotic analysis, 295300
dispersive equation, 534 Brillouin precursor, 300308
microwave effects, 532 signal arrival and signal velocity,
nonlocal dispersion, coaxial GHz 312319
waveguide, 533534, 533f signal contribution, 308312
properties, 532533 Sommerfeld precursor field, 295300
transmittance spectra, 535, 535f total field, 312
Mie theory, 481482 Moire correction
Minimization over channel purifications FT, 128
method, 408409, 417419 moire effect, 128
Minkowski formulation, 442443 moire noise, 128, 129f
Minkowski momentum density, 457 spectral filtering, 127
Minkwoski stress tensor, 474475 superimposed pattern, 127f, 128
Modern asymptotic theory Mollow triplet, 24
dispersive medium, 343 Moving media, classical electrodynamics
dispersive signal propagation Amperian formulation, 444445
singular dispersion limit, 320323 Chu formulation, 443444
weak dispersion limit, 323327 Minkowski formulation, 442443
Subject Index 591
Multi-image optical encryption, 244 video applications, 240

Multipass interferometric protocols, Optical correlation (OC)
400401, 401f digital processing vs. optical processing,
Multiple-parameter discrete fractional FT, 153
243 DIP, 125126
Optical encryption
N adaptive block scanning, 242
Nanotapers (NT), 545 color image encryption algorithm,
Nitrogen-vacancy (NV) 241242
applications, 90 DRP encoding method, 242243
SPPs, 8384 Fourier spectrum, 244, 244f
Noise suppression, 131132, 131f multi-image, 244
Nonlocal dispersion multiple-image parallel, 243
gradient TL, 533534, 533f multiple-parameter discrete fractional FT,
transparent gradient dielectrics 243
applications, 490491 one-step multiplexing, 243
barrier fabrication, 493 public-key encryption, 242
de Broglie waves, 490 Optical FT (OFT)
energy transfer, 493 adapted all-numerical correlator
Gamows approach, 490 AUC, 157159, 158f
Hartman paradox, 491 comparison, FFT and NO_FT
heterogeneous dielectrics, 492 algorithm, 161162, 163f
metatronics, 493 correlation results, 159161, 160f, 162f
optoelectronic devices, 492 drawbacks and stringent requirements,
scalability of results, 493 156
NOON states face database, 157, 158f
Bayesian approach, 392 numerical algorithm, 156157, 157f
coherent + squeezed vacuum strategy, reconstructed image, 159, 159f
424425 reconstruction, input plane, 160161,
Heisenberg limit, 390 161f
multipass protocols, 400401 ROC, 157159, 158f
realistic interferometry, 349350, 408 unitary discrete LCT, 162164
Optical image processing (OIP)
O applications, 119120, 245246
Olver-type path vs. digital method
asymptotic analysis, 290, 291f advantages, optics, 141142
signal arrival and signal velocity, correlation techniques, 122
312314 DIP, 124141
One-axis/two-axis spin-squeezed states, drawbacks and stringent requirements,
371373 121
One-step multiplexing optical encryption, Fourier methods, 141
243 glass lattice three-dimensional
Optical compression mapping, 164165
DCT, 239240 hybrid methods, 143144
digital Fresnel fields, 241 image detection, identification, and
principle, 239 localization, 152155
PSI-DH hologram, 240241 imaging systems and applications,
RT, 240 144150
592 Subject Index
Optical image processing (OIP) (Continued ) optical implementation, mathematical

light to image formation, 122124 operations
limitations, 142 gray image processing, 180182,
need for, 141 182183f
numerical approach, 144 optical neighborhood operation binary
numerical implementation, 155164 image processor (ONOBIP),
optical spatial filtering, 142 182187, 185186f
SLM, 142 PO
spectral imager, 166169 air, optical setup polarization
super imaging system, 150152 measurement, 215217, 215216f
DP, 120 attenuation of light, 209210
historical and technical perspectives, 120 AUVs, 209
image denoising, fusion, and backscattering, 208210, 210f
reconstruction circular polarimeter, 220221,
binary morphological operation IP, 220221f
234 ideal linear polarizer, 212213, 214t
fusion, 233 Jones and MuellerStokes formalisms,
gray image, 232, 232f 212213, 214t
PRAs, 233, 233f linear polarimeter, 219220, 219220f
Wiener filtering, 232 linear polarizer, 212213, 214t
image transmission metal object detection, 221222,
asymmetric cryptography/public-key 222223f
cryptography, 235 minefields, 209
compression, 236 optical setup, polarization
encryption, 237 measurement, 212, 212213f
lossless compression, 234 principles, 208
lossy compression, 234 quarter-wave plate, 212213, 214t
optical compression, 239241 Rayleigh scattering, 211212
optical encryption, 241244, 244f Stokes polarimeter, 213214
simultaneous compression and underwater imaging, 222231, 223f,
encryption, 244245 231f
symmetric key algorithms, 235 wave plate, 212213, 214t
increasing image quality Optical manipulation
edge enhancement, 195196, classical optical tweezers, 476477, 476f
197198f description, 438439
enlightening correction, 187195, Gaussian beam, 475476, 476f
189f, 191193f, 195f interference, 477, 478f
image blur correction, 197200, 199f modeling process, 475
image restoration and edge detection, radiation pressure, 477478
202208, 205207f Optical tweezers, 438
moire correction, 200202, 202203f
OFT P
Fourier holography, 173176, 174f Particle entanglement, 399400
nonlinear transmission, 176179, 178f Peak-to-correlation energy (PCE),
phase contrast imaging, 179, 180f 225229, 226f, 228f
photoinduced anisotropy, 171173, Pegg-Barnett phase operator, 349350
172173f Perfect electrical conductor (PEC), 464
spatial filtering, 170171, 171172f Perfect magnetic conductor (PMC), 464
Subject Index 593
Phase diffusion Poissonian statistics, 346347

Bayesian approach, 420421 Polarization optics (PO)
decoherence models, 402403f, 404 air, optical setup polarization
QFI approach, 416419 measurement
Phase function and saddle point DOP, 217219
approximate local behavior, 277278 Stokes image, 216, 216f
critical space-time point, 283284 attenuation of light, 209210
distant saddle points, 272273 AUVs, 209
dominant saddle points backscattering, 208210, 210f
angular frequency, 288 circular polarimeter, 220221, 220221f
complex phase behavior, 285287 ideal linear polarizer, 212213, 214t
critical space-time value, 287288 Jones and MuellerStokess formalisms,
definition, 285 212213, 214t
dynamical behavior, 280281 linear polarimeter, 219220, 219220f
final space-time domain, 284285 linear polarizer, 212213, 214t
first-order saddle points, 272 metal object detection
initial space-time domain, 281282 disk-shaped materials, 222, 223f
isotimic contours, 269271f Scotch pieces, 221, 222f
limitations, 267268 minefields, 209
limiting behavior, 275276 optical setup, polarization measurement,
numerical results, 273274 212, 212213f
saddle point method, 267268 principles, 208
second-order saddle point, 273 quarter-wave plate, 212213, 214t
single-resonance Lorentz model Rayleigh scattering, 211212
dielectric, 274275 Stokes polarimeter, 213214
space-time parameter, 272 underwater imaging
Phase noise. See Phase diffusion glass tank, 223229, 224228f
Phase-retrieval algorithms (PRAs), 233, 233f water tank, 229f, 230f, 231f, 229231
Photon wave plate, 212213, 214t
definition, 2 Polymethyl methacrylate (PMMA), 2829
number operator, 6 Poyntings theorem, 264266, 471472
statistics and spectral purity, 912 Projection noise, 371373
application, 12 PSIDH hologram, 240241
BoseEinstein distribution, 10 Public-key cryptography, 235, 237238,
density matrix, 10 242
Fano factor, 11 Pulsed plane-wave transmission
wave function, 67 asymptotic description
Photonic losses Brillouin precursor, 334336
Bayesian approach, 419420 signal contribution, 336337
decoherence models, 402403f, 404406 Sommerfeld precursor, 334
QFI approach, 412416 total transmitted field, 337
Plane wave formulation
classical electrodynamics, 446448 frequency-dependent expression,
scattering 329330
cylinder, 478481 single-resonance Lorentz material,
sphere, 481485 331332
Plasmonic sources. See Surface plasmon SommerfeldBrillouin problem,
polaritons (SPPs) 331332
594 Subject Index
Pulsed plane-wave transmission (Continued ) ZnSe quantum, 5960

transverse electric (TE)-polarized SPPs, 83, 84f
plane, 330331, 330f structure, 92, 93f
Lorentz half-space, 327329 Quantum-enhanced interferometry, 348
practical applications and experiments, Quantum estimation theory, 348
327328 Quantum Fisher Information (QFI)
transmitted signal evolution decoherence-free interferometry,
Brillouin precursor, 339340 388390
relative velocity, 341, 341f Heisenberg limit, 348349
Sommerfeld precursor, 338339 quantum estimation, 348
Purcell effect realistic interferometry
advantage, 96 classical and quantum simulation
Fermi golden rule, 12 methods, 408409
semiconductor systems, 13, 44 CS, 409412, 411f
single-photon states, 3 imperfect visibility, 409412
minimization over channel
Q purifications method, 408409,
Quantum cryptography, 94 417419
Quantum dots (QDs) NOON states, 408
applications, 8990 phase diffusion, 416419
CdSe, 84 photonic losses, 412416
coupled, 84, 85f QS, 409, 413416, 414f
InAs, 31 Quantum Key distribution process, 8688
micropillar cavity, 80, 81f Quantum limits, optical interferometry
photoluminescence spectrum, 79, 80f decoherence-free interferometry
self-assembled strain-tunable, 96 Bayesian approach, 390392
semiconductor entanglement, 399400
Bragg reflectors, 46 indefinite photon-number states,
CdTe/ZnTe, 60 393398
correlation function, 4445, 45f multipass protocols, 400401
GaAs matrix, 4748 QFI approach, 388390
green photons, 59 estimation theory
IIIV, 43 classical parameter estimation, 374382
InAs, 47 quantum parameter estimation,
light extraction efficiency, 5758 382387
microdisk structure, 4344, 44f light
nanowires, 59 coherent states, 353355
nitride-based, 60 definite photon-number states, 358
photon correlation data, 49f, 50 general two-mode Gaussian states, 357
pillar microcavities, 46, 5253 Hilbert space, 351352
polarization switching, 5657 mode vs. particle entanglement,
Purcell effect, 44 359362
single-photon horn, 56 output modes, 352
single-photon states, 4260, 70 particle description, 358359
spontaneous emission, 43 single-mode squeezed states, 355356
strain-driven StranskiKrastanov two-mode squeezed states, 356357
growth, 52 Wigner function, 352353
two-photon interference, 51 MZ interferometer
Subject Index 595
BEC, 373 Rubidium atoms, 94

classical theory, 346 structure, QD, 92, 93f
coherent + squeezed-vacuum entangled states
interferometry, 366367 beam splitter, 7071, 71f
coherent state interferometry, 364365 correlation, 72
definite photon-number state FranckHertz scattering, 72
interferometry, 367369 heralding photon, 76
FabryPerrot interferometer, 369370, Michelson interferometer, 8081, 82f
370f micropillar cavity, QD, 80, 81f
Fock state interferometry, 365366 parametric down-conversion
Michelson interferometer, 369, 370f process, 78
phase-sensing uncertainty, 362364 photoelectric effect, 7576
Ramsey interferometry, 371373, 372f photoluminescence spectrum, QD, 79,
shot noise, 347 80f
realistic interferometry quantum dot biexciton, 79
Bayesian approach, 419421 signal and idler photons, 73, 73f
bounds, QFI approach, 408419 single-mode optical fiber, 77
decoherence models, 402403f, two-photon system, 7273
403408 vertical/horizontal polarization, 7475,
phase encoding transformation, 74f
402403 Fock states
practical schemes, 421426 annihilation and creation operators,
resource limitations, 426427 34
single phase parameter estimation, 427 coherent states, 56
Quantum phase operator, 348 definition, 3
Quantum simulation (QS), 409, 413416, photon number operator, 4
414f squeezed state, 6
Quantum state engineering photon
applications definition, 2
beam splitters entangling property, number operator, 6
8889 statistics and spectral purity, 912
catalyst, 8889 wave function, 67
cavity quantum electrodynamics, 88 Purcell effect
coherent quantum, 94 advantage, 96
controllable transfer, 9293 Fermi golden rule, 12
Coulomb blockade, 89 semiconductor systems, 13
entanglement, 86 self-assembled strain-tunable QD, 96
hybrid GaP-diamond whispering single-photon states
gallery mode cavity, 91, 92f antibunching, 79, 2425
hybrid quantum systems, 90 applications, 2
inputoutput coupling, 90 atomic resonance fluorescence, 24
NV, 90 attenuated Poisson source, 23
optical waveguides, 91 deterministic source, 70
QD, 8990 first generation, 23
quantum cryptography, 94 ideal source, 23
quantum interference, 88 individual nitrogen-vacancy color
Quantum Key distribution process, centers, 6069
8688 intensity correlation, 25, 26f
596 Subject Index
Quantum state engineering (Continued ) Radon transform (RT), 164165

mesoscopic p-n junctions, 3033 Ramsey interferometry, 371373, 372f
microcavity and atoms, 3342 Rayleigh scattering, 211212
Mollow triplet, 24 Realistic interferometry, quantum limits
Nd:YAG laser, 28 Bayesian approach
oxygen-depleted liquid crystal, 29 phase diffusion, 420421
photon gun, 23, 69 photonic losses, 419420
photonic crystal, 2930 bounds, QFI approach
photon source on demand, imperfect visibility, 409412
definition, 69 phase diffusion, 416419
PMMA, 2829 photonic losses, 412416
preparation of, 1320 decoherence models
principal problems, 69 imperfect visibility, 402403f, 406408
probabilistic source, 70 Kraus operators, 403
Purcell effect, 3 phase diffusion, 402403f, 404
Rabi frequency, 25 photonic losses, 402403f, 404406
realization of, 20 phase encoding transformation, 402403
review papers, 3 practical schemes
semiconductor QDs, 4260, 70 coherent + squeezed vacuum strategy,
single atomic Mg+ ions, 24 423426
singlet and a triplet state, 25 indefinite photon-number states, 422
Stark effect, 2627 Receiver, 134, 134f
strongly attenuated sources, 2123 Receiver operating characteristic (ROC)
terrylene, 2728 curve, 157159, 158f
SPPs Redundance, 134135
advantages, 83 Refractive index
coupled QD, 84, 85f homogeneous lossless opaque layer,
definition, 8283 507508
diamondplasmonic system, 86, 87f Retarder, 212213, 214t
fluorescence properties, 83 Rivest, Shamir, and Adleman (RSA)
Hanbury BrownTwiss experiment, algorithm, 140
8486, 85f encryption, 238239
NV center, 8384 Root-mean-square (RMS), 245
QD, 83, 84f Rubidium atoms
single-photon excitation, 83 applications, 94
Quarter-wave plate, 212213, 214t microcavity and atoms, 3, 34f, 35
Quasi-stationary approximation, classical Rydberg atoms
electrodynamics microcavity and atoms, 42
Doppler shifted photon, 448449, 449f preparation of, 16
plane wave, 446448
S
R Saddle point method, 267268
Rabi frequency, 25 Second-order saddle point, 273
Radiation pressure Self-assembled strain-tunable QD, 96
beam propagation direction, 438 Semiconductor QDs
electromagnetic stress, 464467, 465f, Bragg reflectors, 46
467f CdTe/ZnTe, 60
optical manipulation, 477478 correlation function, 4445, 45f
Subject Index 597
GaAs matrix, 4748 NE8 (nickelnitrogen complex), 67, 67f

green photons, 59 nickel centers, 68
IIIV, 43 NV defect center, 61
InAs, 47 photon emission, 62
light extraction efficiency, 5758 single NV centers, 64
microdisk structure, 4344, 44f type Ib synthetic diamond, 61
nanowires, 59 intensity correlation, 25, 26f
nitride-based, 60 mesoscopic p-n junctions
photon correlation data, 49f, 50 acoustoelectric effect, 33
pillar microcavities, 46, 5253 GaAs-based LEDs, 31
polarization switching, 5657 InAs QD, 31
Purcell effect, 44 outcoupling efficiency, 32
single-photon horn, 56 photon blockade effect, 31
single-photon states, 4260, 70 microcavity and atoms
spontaneous emission, 43 atomic inversion, 35
strain-driven StranskiKrastanov entanglement, 33
growth, 52 Fock state, 35
two-photon interference, 51 Gaussian transverse profile, 3637
ZnSe quantum, 5960 Hanbury Brown and Twiss
Shot noise, 346347 configuration, 39
Signal-to-noise ratio (SNR) laser pulses, 40
DIP, 124 photon blockade, 41
Wiener filtering, 131132 Raman transitions, 3839
Simultaneous compression and encryption Rubidium atoms, 3, 34f, 35
image transmission, OIP, 244245 Rydberg atoms, 42
Single-photon states STIRAP, 3637
antibunching trapping states, 3334
example, 9f Mollow triplet, 24
Hanbury Brown and Twiss Nd:YAG laser, 28
experiment, 8f oxygen-depleted liquid crystal, 29
optical microcavity, 25 photon gun, 23, 69
resonant fluorescence, 24 photonic crystal, 2930
second-order correlation function, 78 photon source on demand, definition, 69
applications, 2 PMMA, 2829
atomic resonance fluorescence, 24 preparation of
attenuated Poisson source, 23 adiabatic theorem, 16
deterministic source, 70 atomcavity system, 14
first generation, 23 definition, 1314
ideal source, 23 Fock states, 14, 16
individual NV color centers JaynesCummings model, 15
autocorrelation function, 64, 64f Kerr medium, 1920
color centers, 61 Krause method, limitation, 16
confocal microscope scanning, 62, 63f quality factor, 15
coupling, 67 Rydberg atoms, 16
Cr-related center (UM2), 69 Schrodinger cat state, 17
diamond nanocrystals, 64 STIRAP, 19
electroluminescence, 66 Zeeman electronic levels, 18
limitation, 62 principal problems, 69
598 Subject Index
Single-photon states (Continued ) Spectral imager

probabilistic source, 70 coregistration error and pixel count,
Purcell effect, 3 166168, 167f
Rabi frequency, 25 information capacity, 168f, 169
realization of, 20 spatial coregistration, 166167
review papers, 3 Stark effect, 2627
semiconductor QDs Stationary media, 473
Bragg reflectors, 46 Steganography, 139
CdTe/ZnTe, 60 STImulated RAman scattering involving
correlation function, 4445, 45f Adiabatic Passage (STIRAP)
GaAs matrix, 4748 microcavity and atoms, 3637
green photons, 59 single-photon states, 19
IIIV, 43 Stokes polarimeter, 213214
InAs, 47 Sub-Poissonian statistics, 347
light extraction efficiency, 5758 Subwavelength narrowings, gradient TL
microdisk structure, 4344, 44f heterogeneity-induced tunneling,
nanowires, 59 546547
nitride-based, 60 LaguerreGaussian mode, 545546
photon correlation data, 49f, 50 NT, 545
pillar microcavities, 46, 5253 radiation losses, 545
polarization switching, 5657 transformation optics, 549550
Purcell effect, 44 Super imaging system
single-photon horn, 56 light field photography, 151152, 152f
spontaneous emission, 43 multiexposure camera, 152, 152f
strain-driven StranskiKrastanov multiple aperture cameras, 150151, 151f
growth, 52 optical setup, 150, 150f
two-photon interference, 51 subchannel, 150151
ZnSe quantum, 5960 Surface plasmon polaritons (SPPs)
single atomic Mg+ ions, 24 advantages, 83
singlet and a triplet state, 25 coupled QD, 84, 85f
Stark effect, 2627 definition, 8283
strongly attenuated sources, 2123 diamondplasmonic system, 86, 87f
terrylene, 2728 fluorescence properties, 83
Single-resonance Lorentz model dielectric Hanbury BrownTwiss experiment,
asymptotic analysis, 292 8486, 85f
phase function and saddle point, NV center, 8384
274275 QD, 83, 84f
signal arrival and signal velocity, 312316 single-photon excitation, 83
Sommerfeld precursor field Surface plasmons. See Surface plasmon
asymptotic analysis, 292 polaritons (SPPs)
characteristics, 297 Symmetric key algorithms
limited validity, 295
signal arrival and signal velocity, 317 T
transformation, 295297 TE10 mode
uniform asymptotic expansions, 295300 Brewster effect, 542
Spatial coregistration, 166167 generating function, 541
Spatial filtering, 170171, 171172f polarization structure, 542
Spectral filtering, 127 problems, 538
Subject Index 599
Total internal reflection (TIR) theoretical analysis, 529530

FTIR nonlocal dispersion, transparent gradient
evanescent waves, 496498 dielectrics
gradient half-space, 498500 applications, 490491
HID, 495496 barrier fabrication, 493
homogeneous lossless opaque layer, de Broglie waves, 490
507509, 508f energy transfer, 493
transition point, 509511 Gamows approach, 490
transparent gradient layer, 501506 Hartman paradox, 491
Maxwells equations, 490 heterogeneous dielectrics, 492
Transmission channel, 134, 134f metatronics, 493
Transmission coefficient estimation, optoelectronic devices, 492
379382 scalability of results, 493
Transmitted Abraham momentum density, nonlocality and, 551
457 stochastic process, 554556
Transmitted Minkowski momentum wave pulses, quasistatic process, 556558
density, 457 Twin-Fock state, 367369
Transmitter, 133, 134f
Tunneling, electromagnetic waves U
artificial and natural dispersion Underwater imaging
metamaterials, 512 glass tank
negative and ENZ metamaterials, active polarization imaging, 223224
523526 adapted polarization imaging, 229
plasmonpolariton resonances, 512 CRL image, 223224, 224f
polariton gap, transparent dielectric, PCE, 225229, 226f, 228f
516520 water tank
reactive modes, energy transfer, color image intensity, 230231, 231f
527529 digital underwater camera, 229230,
resonant magnetic modes, 520523 229f
transition plasma layer, TIR, 512516 extinction coefficient, 229230
FTIR LCP images, 230231, 230f
evanescent waves, 496498 LPP images, 230231, 230f
gradient half-space, 498500 Uniform asymptotic expansions
HID, 495496 asymptotic analysis
homogeneous lossless opaque layer, Brillouin precursor field, 294
507509, 508f Cauchys residue theorem, 289
transition point, 509511 dynamical behavior, 294
transparent gradient layer, 501506 FourierLaplace integral
gradient TL representation, 289
analogy, gradient photonic barrier, Olver-type path, 290, 291f
531535 single-resonance Lorentz model
Brewster effect, TE10 mode, 538543 dielectric, 292
evanescent voltage and current waves, Sommerfeld precursor, 292
536538 spacetime intervals, 295
GHz resonance, 543545 Brillouin precursor
heterogeneous part, 530531 application, 301302
HID, 530531 characteristic, 302303
subwavelength narrowings, 545550 dynamic behavior, 302303, 303f
600 Subject Index
Uniform asymptotic expansions (Continued ) adapted all-numerical correlator

first-order saddle points, 304305, AUC, 157159, 158f
307308 comparison, FFT and NO_FT
limited validity, 300301 algorithm, 161162, 163f
second-order saddle point, 306307 correlation results, 159161, 160f, 162f
signal arrival and signal velocity drawbacks and stringent requirements,
angular frequency dependence, 156
318319, 319f face database, 157, 158f
anterior presignal velocity, 317318 numerical algorithm, 156157, 157f
deformed contour of integration, reconstructed image, 159, 159f
312314, 314f reconstruction, input plane, 160161,
main signal velocity, 317318 161f
Olver-type path, 312314 ROC, 157159, 158f
posterior presignal velocity, 317318 advantages, 153
single-resonance Lorentz model principle, 153f
dielectric, 312316 protocol, 153
Sommerfeld precursor field, 317 unitary discrete LCT, 162164
spacetime value, 314315 Vertical-cavity surface-emitting laser
signal contribution (VCSEL), 147148
characteristic, 311
complementary error function,
309310 W
simple pole contribution, 312 Wave plate, 212213, 214t
Sommerfeld precursor field Wiener filtering
characteristics, 297 deblurring method, 132, 132f
limited validity, 295 deconvolution, 130
transformation, 295297 features, 131
total field, 312 Gaussian white noise, 132, 133f
Unitary discrete LCT, 162164 image denoising method, 132
noise suppression, 131132, 131f
V OIP, 232
Vander Lugt correlator (VLC) SNR, 131132
CONTENTS OF PREVIOUS VOLUMES
VOLUME 1 (1961)
1 The modern development of Hamiltonian optics, R.J. Pegis 1 29
2 Wave optics and geometrical optics in optical design, K. Miyamoto 31 66
3 The intensity distribution and total illumination of aberration-free diffraction
images, R. Barakat 67 108
4 Light and information, D. Gabor 109 153
5 On basic analogies and principal differences between optical and electronic
information, H. Wolter 155 210
6 Interference color, H. Kubota 211 251
7 Dynamic characteristics of visual processes, A. Fiorentini 253 288
8 Modern alignment devices, A.C.S. Van Heel 289 329
VOLUME 2 (1963)
1 Ruling, testing and use of optical gratings for high-resolution spectroscopy,
G.W. Stroke 1 72
2 The metrological applications of diffraction gratings, J.M. Burch 73 108
3 Diffusion through non-uniform media, R.G. Giovanelli 109 129
4 Correction of optical images by compensation of aberrations and by spatial
frequency filtering, J. Tsujiuchi 131 180
5 Fluctuations of light beams, L. Mandel 181 248
6 Methods for determining optical parameters of thin films, F. Abele`s 249 288
VOLUME 3 (1964)
1 The elements of radiative transfer, F. Kottler 1 28
2 Apodisation, P. Jacquinot, B. Roizen-Dossier 29 186
3 Matrix treatment of partial coherence, H. Gamo 187 332
VOLUME 4 (1965)
1 Higher order aberration theory, J. Focke 1 36
2 Applications of shearing interferometry, O. Bryngdahl 37 83
3 Surface deterioration of optical glasses, K. Kinosita 85 143
4 Optical constants of thin films, P. Rouard, P. Bousquet 145 197
5 The Miyamoto-Wolf diffraction wave, A. Rubinowicz 199 240
6 Aberration theory of gratings and grating mountings, W.T. Welford 241 280
7 Diffraction at a black screen, Part I: Kirchhoffs theory, F. Kottler 281 314
VOLUME 5 (1966)
1 Optical pumping, C. Cohen-Tannoudji, A. Kastler 1 81
2 Non-linear optics, P.S. Pershan 83 144
Volumes IXL were previously distinguished by roman rather than by arabic numerals.
601
602 Contents of Previous Volumes
3 Two-beam interferometry, W.H. Steel 145 197

4 Instruments for the measuring of optical transfer functions, K. Murata 199 245
5 Light reflection from films of continuously varying refractive index,
R. Jacobsson 247 286
6 X-ray crystal-structure determination as a branch of physical optics,
H. Lipson, C.A. Taylor 287 350
7 The wave of a moving classical electron, J. Picht 351 370
VOLUME 6 (1967)
1 Recent advances in holography, E.N. Leith, J. Upatnieks 1 52
2 Scattering of light by rough surfaces, P. Beckmann 53 69
3 Measurement of the second order degree of coherence,
M. Frantcon, S. Mallick 71 104
4 Design of zoom lenses, K.Yamaji 105 170
5 Some applications of lasers to interferometry, D.R. Herriot 171 209
6 Experimental studies of intensity fluctuations in lasers,
J.A. Armstrong, A.W. Smith 211 257
7 Fourier spectroscopy, G.A. Vanasse, H. Sakai 259 330
8 Diffraction at a black screen, Part II: electromagnetic theory, F. Kottler 331 377
VOLUME 7 (1969)
1 Multiple-beam interference and natural modes in open resonators,
G. Koppelman 1 66
2 Methods of synthesis for dielectric multilayer filters, E. Delano, R.J. Pegis 67 137
3 Echoes at optical frequencies, I.D. Abella 139 168
4 Image formation with partially coherent light, B.J. Thompson 169 230
5 Quasi-classical theory of laser radiation, A.L. Mikaelian, M.L. Ter-Mikaelian 231 297
6 The photographic image, S. Ooue 299 358
7 Interaction of very intense light with free electrons, J.H. Eberly 359 415
VOLUME 8 (1970)
1 Synthetic-aperture optics, J.W. Goodman 1 50
2 The optical performance of the human eye, G.A. Fry 51 131
3 Light beating spectroscopy, H.Z. Cummins, H.L. Swinney 133 200
4 Multilayer antireflection coatings, A. Musset, A. Thelen 201 237
5 Statistical properties of laser light, H. Risken 239 294
6 Coherence theory of source-size compensation in
interference microscopy, T. Yamamoto 295 341
7 Vision in communication, L. Levi 343 372
8 Theory of photoelectron counting, C.L. Mehta 373 440
VOLUME 9 (1971)
1 Gas lasers and their application to precise length measurements, A.L. Bloom 1 30
2 Picosecond laser pulses, A.J. Demaria 31 71
3 Optical propagation through the turbulent atmosphere, J.W. Strohbehn 73 122
Contents of Previous Volumes 603
4 Synthesis of optical birefringent networks, E.O. Ammann 123 177

5 Mode locking in gas lasers, L.Allen, D.G.C. Jones 179 234
6 Crystal optics with spatial dispersion, V.M. Agranovich, V.L. Ginzburg 235 280
7 Applications of optical methods in the diffraction theory of elastic waves,
K. Gniadek, J. Petykiewicz 281 310
8 Evaluation, design and extrapolation methods for optical signals, based on use
of the prolate functions, B.R. Frieden 311 407
VOLUME 10 (1972)
1 Bandwidth compression of optical images, T.S. Huang 1 44
2 The use of image tubes as shutters, R.W. Smith 45 87
3 Tools of theoretical quantum optics, M.O. Scully, K.G. Whitney 89 135
4 Field correctors for astronomical telescopes, C.G. Wynne 137 164
5 Optical absorption strength of defects in insulators, D.Y. Smith, D.L. Dexter 165 228
6 Elastooptic light modulation and deflection, E.K. Sittig 229 288
7 Quantum detection theory, C.W. Helstrom 289 369
VOLUME 11 (1973)
1 Master equation methods in quantum optics, G.S. Agarwal 1 76
2 Recent developments in far infrared spectroscopic techniques, H. Yoshinaga 77 122
3 Interaction of light and acoustic surface waves, E.G. Lean 123 166
4 Evanescent waves in optical imaging, O. Bryngdahl 167 221
5 Production of electron probes using a field emission source, A.V. Crewe 223 246
6 Hamiltonian theory of beam mode propagation, J.A. Arnaud 247 304
7 Gradient index lenses, E.W. Marchand 305 337
VOLUME 12 (1974)
1 Self-focusing, self-trapping, and self-phase modulation of laser beams,
O. Svelto 1 51
2 Self-induced transparency, R.E. Slusher 53 100
3 Modulation techniques in spectrometry, M. Harwit, J.A. Decker Jr 101 162
4 Interaction of light with monomolecular dye layers, K.H. Drexhage 163 232
5 The phase transition concept and coherence in atomic emission, R. Graham 233 286
6 Beam-foil spectroscopy, S. Bashkin 287 344
VOLUME 13 (1976)
1 On the validity of Kirchhoffs law of heat radiation for a body in a
nonequilibrium environment, H.P. Baltes 1 25
2 The case for and against semiclassical radiation theory, L. Mandel 27 68
3 Objective and subjective spherical aberration measurements of the human
eye, W.M. Rosenblum, J.L. Christensen 69 91
4 Interferometric testing of smooth surfaces, G. Schulz, J. Schwider 93 167
5 Self focusing of laser beams in plasmas and semiconductors,
M.S. Sodha, A.K. Ghatak, V.K.Tripathi 169 265
6 Aplanatism and isoplanatism, W.T. Welford 267 292
VOLUME 14 (1976)
1 The statistics of speckle patterns, J.C. Dainty 1 46
2 High-resolution techniques in optical astronomy, A. Labeyrie 47 87
3 Relaxation phenomena in rare-earth luminescence, L.A. Riseberg, M.J. Weber 89 159
4 The ultrafast optical Kerr shutter, M.A. Duguay 161 193
5 Holographic diffraction gratings, G. Schmahl, D. Rudolph 195 244
6 Photoemission, P.J. Vernier 245 325
7 Optical fibre waveguides a review, P.J.B. Clarricoats 327 402
VOLUME 15 (1977)
1 Theory of optical parametric amplification and oscillation, W. Brunner,
H. Paul 1 75
2 Optical properties of thin metal films, P. Rouard, A. Meessen 77 137
3 Projection-type holography, T. Okoshi 139 185
4 Quasi-optical techniques of radio astronomy, T.W. Cole 187 244
5 Foundations of the macroscopic electromagnetic theory of dielectric media,
J. Van Kranendonk, J.E. Sipe 245 350
VOLUME 16 (1978)
1 Laser selective photophysics and photochemistry, V.S. Letokhov 1 69
2 Recent advances in phase profiles generation, J.J. Clair, C.I. Abitbol 71 117
3 Computer-generated holograms: techniques and applications, W.-H. Lee 119 232
4 Speckle interferometry, A.E. Ennos 233 288
5 Deformation invariant, space-variant optical pattern recognition,
D. Casasent, D. Psaltis 289 356
6 Light emission from high-current surface-spark discharges, R.E. Beverly III 357 411
7 Semiclassical radiation theory within a quantum-mechanical framework,
I.R. Senitzky 413 448
VOLUME 17 (1980)
1 Heterodyne holographic interferometry, R. Dandliker 1 84
2 Doppler-free multiphoton spectroscopy, E. Giacobino, B. Cagnac 85 161
3 The mutual dependence between coherence properties of light and
nonlinear optical processes, M. Schubert, B. Wilhelmi 163 238
4 Michelson stellar interferometry, W.J. Tango, R.Q. Twiss 239 277
5 Self-focusing media with variable index of refraction, A.L. Mikaelian 279 345
VOLUME 18 (1980)
1 Graded index optical waveguides: a review, A. Ghatak, K. Thyagarajan 1 126
2 Photocount statistics of radiation propagating through random and
nonlinear media, J. Perina 127 203
3 Strong fluctuations in light propagation in a randomly inhomogeneous
medium, V.I. Tatarskii, V.U. Zavorotnyi 204 256
4 Catastrophe optics: morphologies of caustics and their diffraction patterns,
M.V. Berry, C. Upstill 257 346
VOLUME 19 (1981)
1 Theory of intensity dependent resonance light scattering and resonance
fluorescence, B.R. Mollow 1 43
2 Surface and size effects on the light scattering spectra of solids, D.L. Mills,
K.R. Subbaswamy 45 137
3 Light scattering spectroscopy of surface electromagnetic waves in solids,
S. Ushioda 139 210
4 Principles of optical data-processing, H.J. Butterweck 211 280
5 The effects of atmospheric turbulence in optical astronomy, F. Roddier 281 376
VOLUME 20 (1983)
1 Some new optical designs for ultra-violet bidimensional detection of
astronomical objects, G. Courte`s, P. Cruvellier, M. Detaille 1 61
2 Shaping and analysis of picosecond light pulses, C. Froehly, B. Colombeau,
M. Vampouille 63 153
3 Multi-photon scattering molecular spectroscopy, S. Kielich 155 261
4 Colour holography, P. Hariharan 263 324
5 Generation of tunable coherent vacuum-ultraviolet radiation, W. Jamroz,
B.P. Stoicheff 325 380
VOLUME 21 (1984)
1 Rigorous vector theories of diffraction gratings, D. Maystre 1 67
2 Theory of optical bistability, L.A. Lugiato 69 216
3 The Radon transform and its applications, H.H. Barrett 217 286
4 Zone plate coded imaging: theory and applications, N.M. Ceglio,
D.W. Sweeney 287 354
5 Fluctuations, instabilities and chaos in the laser-driven nonlinear ring cavity,
J.C. Englund, R.R. Snapp, W.C. Schieve 355 428
VOLUME 22 (1985)
1 Optical and electronic processing of medical images, D. Malacara 1 76
2 Quantum fluctuations in vision, M.A. Bouman, W.A. Van De Grind,
P. Zuidema 77 144
3 Spectral and temporal fluctuations of broad band laser radiation,
A.V. Masalov 145 196
4 Holographic methods of plasma diagnostics, G.V. Ostrovskaya,
Yu.I. Ostrovsky 197 270
5 Fringe formations in deformation and vibration measurements using laser
light, I. Yamaguchi 271 340
6 Wave propagation in random media: a systems approach, R.L. Fante 341 398
VOLUME 23 (1986)
1 Analytical techniques for multiple scattering from rough surfaces,
J.A. DeSanto, G.S. Brown 1 62
2 Paraxial theory in optical design in terms of Gaussian brackets, K. Tanaka 63 111
3 Optical films produced by ion-based techniques, P.J. Martin, R.P. Netterfield 113 182
4 Electron holography, A. Tonomura 183 220
5 Principles of optical processing with partially coherent light, F.T.S. Yu 221 275
VOLUME 24 (1987)
1 Micro Fresnel lenses, H. Nishihara, T. Suhara 1 37
2 Dephasing-induced coherent phenomena, L. Rothberg 39 101
3 Interferometry with lasers, P. Hariharan 103 164
4 Unstable resonator modes, K.E. Oughstun 165 387
5 Information processing with spatially incoherent light, I. Glaser 389 509
VOLUME 25 (1988)
1 Dynamical instabilities and pulsations in lasers, N.B. Abraham, P. Mandel,
L.M. Narducci 1 190
2 Coherence in semiconductor lasers, M. Ohtsu, T. Tako 191 278
3 Principles and design of optical arrays, Wang Shaomin, L. Ronchi 279 348
4 Aspheric surfaces, G. Schulz 349 415
VOLUME 26 (1988)
1 Photon bunching and antibunching, M.C. Teich, B.E.A. Saleh 1 104
2 Nonlinear optics of liquid crystals, I.C. Khoo 105 161
3 Single-longitudinal-mode semiconductor lasers, G.P. Agrawal 163 225
4 Rays and caustics as physical objects, Yu.A. Kravtsov 227 348
5 Phase-measurement interferometry techniques, K. Creath 349 393
VOLUME 27 (1989)
1 The self-imaging phenomenon and its applications, K. Patorski 1 108
2 Axicons and meso-optical imaging devices, L.M. Soroko 109 160
3 Nonimaging optics for flux concentration, I.M. Bassett, W.T. Welford,
R. Winston 161 226
4 Nonlinear wave propagation in planar structures, D. Mihalache,
M. Bertolotti, C. Sibilia 227 313
5 Generalized holography with application to inverse scattering and
inverse source problems, R.P. Porter 315 397
VOLUME 28 (1990)
1 Digital holography computer-generated holograms, O. Bryngdahl,
F. Wyrowski 1 86
2 Quantum mechanical limit in optical precision measurement and
communication, Y. Yamamoto, S. Machida, S. Saito, N. Imoto,
T. Yanagawa, M. Kitagawa 87 179
3 The quantum coherence properties of stimulated Raman scattering,
M.G. Raymer, I.A. Walmsley 181 270
4 Advanced evaluation techniques in interferometry, J. Schwider 271 359
5 Quantum jumps, R.J. Cook 361 416
VOLUME 29 (1991)
1 Optical waveguide diffraction gratings: coupling between guided modes,
D.G. Hall 1 63
2 Enhanced backscattering in optics, Yu.N. Barabanenkov, Yu.A. Kravtsov,
V.D. Ozrin, A.I. Saichev 65 197
3 Generation and propagation of ultrashort optical pulses, I.P. Christov 199 291
4 Triple-correlation imaging in optical astronomy, G. Weigelt 293 319
5 Nonlinear optics in composite materials. 1. Semiconductor and metal
crystallites in dielectrics, C. Flytzanis, F. Hache, M.C. Klein, D. Ricard,
Ph. Roussignol 321 411
VOLUME 30 (1992)
1 Quantum fluctuations in optical systems, S. Reynaud, A. Heidmann,
E. Giacobino, C. Fabre 1 85
2 Correlation holographic and speckle interferometry, Yu.I. Ostrovsky,
V.P. Shchepinov 87 135
3 Localization of waves in media with one-dimensional disorder,
V.D. Freilikher, S.A. Gredeskul 137 203
4 Theoretical foundation of optical soliton concept in fibers, Y. Kodama,
A. Hasegawa 205 259
5 Cavity quantum optics and the quantum measurement process, P. Meystre 261 355
VOLUME 31 (1993)
1 Atoms in strong fields: photoionization and chaos, P.W. Milonni,
B. Sundaram 1 137
2 Light diffraction by relief gratings: a macroscopic and microscopic view,
E. Popov 139 187
3 Optical amplifiers, N.K. Dutta, J.R. Simpson 189 226
4 Adaptive multilayer optical networks, D. Psaltis, Y. Qiao 227 261
5 Optical atoms, R.J.C. Spreeuw, J.P. Woerdman 263 319
6 Theory of Compton free electron lasers, G. Dattoli, L. Giannessi,
A. Renieri, A. Torre 321 412
VOLUME 32 (1993)
1 Guided-wave optics on silicon: physics, technology and status, B.P. Pal 1 59
2 Optical neural networks: architecture, design and models, F.T.S. Yu 61 144
3 The theory of optimal methods for localization of objects in pictures,
L.P. Yaroslavsky 145 201
4 Wave propagation theories in random media based on the path-integral
approach, M.I. Charnotskii, J. Gozani, V.I. Tatarskii, V.U. Zavorotny 203 266
5 Radiation by uniformly moving sources. Vavilov-Cherenkov effect,
Doppler effect in a medium, transition radiation and associated
phenomena, V.L. Ginzburg 267 312
6 Nonlinear processes in atoms and in weakly relativistic plasmas,
G. Mainfray, C. Manus 313 361
VOLUME 33 (1994)
1 The imbedding method in statistical boundary-value wave problems,
V.I. Klyatskin 1 127
2 Quantum statistics of dissipative nonlinear oscillators, V. Perinova,
A. Luks 129 202
3 Gap solitons, C.M. De Sterke, J.E. Sipe 203 260
4 Direct spatial reconstruction of optical phase from phase-modulated
images, V.I. Vlad, D. Malacara 261 317
5 Imaging through turbulence in the atmosphere, M.J. Beran,
I. Oz-Vogt 319 388
6 Digital halftoning: synthesis of binary images, O. Bryngdahl,
T. Scheermesser, F. Wyrowski 389 463
VOLUME 34 (1995)
1 Quantum interference, superposition states of light, and nonclassical effects,
V. Buzrek, P.L. Knight 1 158
2 Wave propagation in inhomogeneous media: phase-shift approach,
L.P. Presnyakov 159 181
3 The statistics of dynamic speckles, T. Okamoto, T. Asakura 183 248
4 Scattering of light from multilayer systems with rough boundaries,
I. Ohldal, K. Navratil, M. Ohldal 249 331
5 Random walk and diffusion-like models of photon migration in turbid
media, A.H. Gandjbakhche, G.H. Weiss 333 402
VOLUME 35 (1996)
1 Transverse patterns in wide-aperture nonlinear optical systems,
N.N. Rosanov 1 60
2 Optical spectroscopy of single molecules in solids, M. Orrit, J. Bernard,
R. Brown, B. Lounis 61 144
3 Interferometric multispectral imaging, K. Itoh 145 196
4 Interferometric methods for artwork diagnostics, D. Paoletti, G. Schirripa
Spagnolo 197 255
5 Coherent population trapping in laser spectroscopy, E. Arimondo 257 354
6 Quantum phase properties of nonlinear optical phenomena, R. Tanas,
A. Miranowicz, Ts. Gantsog 355 446
VOLUME 36 (1996)
1 Nonlinear propagation of strong laser pulses in chalcogenide glass films,
V. Chumash, I. Cojocaru, E. Fazio, F. Michelotti, M. Bertolotti 1 47
2 Quantum phenomena in optical interferometry, P. Hariharan,
B.C. Sanders 49 128
3 Super-resolution by data inversion, M. Bertero, C. De Mol 129 178
4 Radiative transfer: new aspects of the old theory, Yu.A. Kravtsov,
L.A. Apresyan 179 244
5 Photon wave function, I. Bialynicki-Birula 245 294
VOLUME 37 (1997)
1 The Wigner distribution function in optics and optoelectronics,
D. Dragoman 1 56
2 Dispersion relations and phase retrieval in optical spectroscopy,
K.-E. Peiponen, E.M. Vartiainen, T. Asakura 57 94
3 Spectra of molecular scattering of light, I.L. Fabelinskii 95 184
4 Soliton communication systems, R.-J. Essiambre, G.P. Agrawal 185 256
5 Local fields in linear and nonlinear optics of mesoscopic systems,
O. Keller 257 343
6 Tunneling times and superluminality, R.Y. Chiao, A.M. Steinberg 345 405
VOLUME 38 (1998)
1 Nonlinear optics of stratified media, S. Dutta Gupta 1 84
2 Optical aspects of interferometric gravitational-wave detectors,
P. Hello 85 164
3 Thermal properties of vertical-cavity surface-emitting semiconductor
lasers, W. Nakwaski, M. Osi nski 165 262
4 Fractional transformations in optics, A.W. Lohmann, D. Mendlovic,
Z. Zalevsky 263 342
5 Pattern recognition with nonlinear techniques in the Fourier domain,
B. Javidi, J.L. Horner 343 418
6 Free-space optical digital computing and interconnection, J. Jahns 419 513
VOLUME 39 (1999)
1 Theory and applications of complex rays, Yu.A. Kravtsov, G.W. Forbes,
A.A. Asatryan 1 62
2 Homodyne detection and quantum-state reconstruction, D.-G. Welsch,
W. Vogel, T. Opatrny 63 211
3 Scattering of light in the eikonal approximation, S.K. Sharma,
D.J. Somerford 213 290
4 The orbital angular momentum of light, L. Allen, M.J. Padgett,
M. Babiker 291 372
5 The optical Kerr effect and quantum optics in fibers, A. Sizmann,
G. Leuchs 373 469
VOLUME 40 (2000)
1 Polarimetric optical fibers and sensors, T.R. Woli
nski 1 75
2 Digital optical computing, J. Tanida, Y. Ichioka 77 114
3 Continuous measurements in quantum optics, V. Perinova,
A. Luks 115 269
4 Optical systems with improved resolving power, Z. Zalevsky,
D. Mendlovic, A.W. Lohmann 271 341
5 Diffractive optics: electromagnetic approach, J. Turunen, M. Kuittinen,
F. Wyrowski 343 388
6 Spectroscopy in polychromatic fields, Z. Ficek, H.S. Freedhoff 389 441
VOLUME 41 (2000)
1 Nonlinear optics in microspheres, M.H. Fields, J. Popp, R.K. Chang 1 95
2 Principles of optical disk data storage, J. Carriere, R. Narayan, W.-H. Yeh,
C. Peng, P. Khulbe, L. Li, R. Anderson, J. Choi, M. Mansuripur 97 179
3 Ellipsometry of thin film systems, I. Ohldal, D. Franta 181 282
4 Optical true-time delay control systems for wideband phased array
antennas, R.T. Chen, Z. Fu 283 358
5 Quantum statistics of nonlinear optical couplers, J. Perina Jr, J. Perina 359 417
6 Quantum phase difference, phase measurements and Stokes operators,
A. Luis, L.L. Sanchez-Soto 419 479
7 Optical solitons in media with a quadratic nonlinearity, C. Etrich,
F. Lederer, B.A. Malomed, T. Peschel, U. Peschel 483 567
VOLUME 42 (2001)
1 Quanta and information, S. Ya. Kilin 1 91
2 Optical solitons in periodic media with resonant and off-resonant
nonlinearities, G. Kurizki, A.E. Kozhekin, T. Opatrny, B.A. Malomed 93 146
3 Quantum Zeno and inverse quantum Zeno effects, P. Facchi, S. Pascazio 147 217
4 Singular optics, M.S. Soskin, M.V. Vasnetsov 219 276
5 Multi-photon quantum interferometry, G. Jaeger, A.V Sergienko 277 324
6 Transverse mode shaping and selection in laser resonators, R. Oron,
N. Davidson, A.A. Friesem, E. Hasman 325 386
VOLUME 43 (2002)
1 Active optics in modern large optical telescopes, L. Noethe 1 69
2 Variational methods in nonlinear fiber optics and related fields,
B.A. Malomed 71 193
3 Optical works of L.V. Lorenz, O. Keller 195 294
4 Canonical quantum description of light propagation in dielectric media,
A. Luks, V. Perinova 295 431
5 Phase space correspondence between classical optics and quantum
mechanics, D. Dragoman 433 496
6 Slow and fast light, R.W. Boyd, D.J. Gauthier 497 530
7 The fractional Fourier transform and some of its applications to optics,
A. Torre 531 596
VOLUME 44 (2002)
1 Chaotic dynamics in semiconductor lasers with optical feedback, J. Ohtsubo 1 84
2 Femtosecond pulses in optical fibers, F.G. Omenetto 85 141
3 Instantaneous optics of ultrashort broadband pulses and rapidly varying
media, A.B. Shvartsburg, G. Petite 143 214
4 Optical coherence tomography, A.F. Fercher, C.K. Hitzenberger 215 301
5 Modulational instability of electromagnetic waves in inhomogeneous
and in discrete media, F.Kh. Abdullaev, S.A. Darmanyan, J. Garnier 303 366
VOLUME 45 (2003)
1 Anamorphic beam shaping for laser and diffuse light, N. Davidson,
N. Bokor 1 51
2 Ultra-fast all-optical switching in optical networks, I. Glesk, B.C. Wang,
L. Xu, V. Baby, P.R. Prucnal 53 117
3 Generation of dark hollow beams and their applications, J. Yin,
W. Gao, Y. Zhu 119 204
4 Two-photon lasers, D.J. Gauthier 205 272
5 Nonradiating sources and other invisible objects, G. Gbur 273 315
6 Lasing in disordered media, H. Cao 317 370
VOLUME 46 (2004)
1 Ultrafast solid-state lasers, U. Keller 1 115
2 Multiple scattering of light from randomly rough surfaces, A.V. Shchegrov,
A.A. Maradudin, E.R. Mendez 117 241
3 Laser-diode interferometry, Y. Ishii 243 309
4 Optical realizations of quantum teleportation, J. Gea-Banacloche 311 353
5 Intensity-field correlations of non-classical light, H.J. Carmichael,
G.T. Foster, L.A. Orozco, J.E. Reiner, P.R. Rice 355 404
VOLUME 47 (2005)
1 Multistep parametric processes in nonlinear optics, S.M. Saltiel,
A.A. Sukhorukov, Y.S. Kivshar 1 73
2 Modes of wave-chaotic dielectric resonators, H.E. T ureci, H.G.L. Schwefel,
Ph. Jacquod, A.D. Stone 75 137
3 Nonlinear and quantum optics of atomic and molecular fields, C.P. Search,
P. Meystre 139 214
4 Space-variant polarization manipulation, E. Hasman, G. Biener, A. Niv,
V. Kleiner 215 289
5 Optical vortices and vortex solitons, A.S. Desyatnikov, Y.S. Kivshar,
L.L. Torner 291 391
6 Phase imaging and refractive index tomography for X-rays and
visible rays, K. Iwata 393 432
VOLUME 48 (2005)
1 Laboratory post-engineering of microstructured optical fibers,
B.J. Eggleton, P. Domachuk, C. Grillet, E.C. Magi, H.C. Nguyen,
P. Steinvurzel, M.J. Steel 1 34
2 Optical solitons in random media, F. Abdullaev, J. Garnier 35 106
3 Curved diffractive optical elements: Design and applications, N. Bokor,
N. Davidson 107 148
4 The geometric phase, P. Hariharan 149 201
5 Synchronization and communication with chaotic laser systems,
A. Uchida, F. Rogister, J. Garca-Ojalvo, R. Roy 203 341
VOLUME 49 (2006)
1 Gaussian apodization and beam propagation, V.N. Mahajan 1 96
2 Controlling nonlinear optical processes in multi-level atomic systems,
A. Joshi, M. Xiao 97 175
3 Photonic crystals, H. Benisty, C. Weisbuch 177 313
4 Symmetry properties and polarization descriptors for an arbitrary
electromagnetic wavefield, C. Brosseau, A. Dogariu 315 380
5 Quantum cryptography, M. Dusek, N. L utkenhaus, M. Hendrych 381 454
6 Optical quantum cloning, N.J. Cerf , J. Fiurasek 455 545
VOLUME 50 (2007)
1 From millisecond to attosecond laser pulses, N. Bloembergen 1 12
2 Conical diffraction: Hamiltons diabolical point at the heart of crystal optics,
M.V. Berry, M.R. Jeffrey 13 50
3 Historical papers on the particle concept of light, O. Keller 51 95
4 Field quantization in optics, P.W. Milonni 97 135
5 The history of near-field optics, L. Novotny 137 184
6 Light tunneling, H.M. Nussenzveig 185 250
7 The influence of Youngs interference experiment on the development
of statistical optics, E. Wolf 251 273
8 Planck, photon statistics, and Bose-Einstein condensation,
D.M. Greenberger, N. Erez, M.O. Scully, A.A. Svidzinsky, M.S. Zubairy 275 330
VOLUME 51 (2008)
1 Negative refractive index metamaterials in optics, N.M. Litchinitser,
I.R. Gabitov, A.I. Maimistov, V.M. Shalaev 1 67
2 Polarization techniques for surface nonlinear optics, M. Kauranen,
S. Cattaneo 69 120
3 Electromagnetic fields in linear bianisotropic mediums, T.G. Mackay
A. Lakhtakia 121 209
4 Ultrafast optical pulses, C.R. Pollock 211 249
5 Quantum imaging, A. Gatti, E. Brambilla, L. Lugiato 251 348
6 Assessment of optical systems by means of point-spread functions,
J.J.M. Braat, S. van Haver, A.J.E.M. Janssen, P. Dirksen 349 468
7 The discrete Wigner function, G. Bj ork, A.B. Klimov, L.L. Sanchez-Soto 469 516
VOLUME 52 (2009)
1 Coherent backscattering and Anderson localization of light, C.M. Aegerter,
G. Maret 1 62
2 Soliton shape and mobility control in optical lattices, Y.V. Kartashov,
V.A. Vysloukh, L. Torner 63 148
3 Signal and quantum noise in optical communications and cryptography,
P. Gallion, F. Mendieta, S. Jiang 149 259
4 Invisibility cloaking by coordinate transformation, M. Yan, W. Yan, M. Qiu 261 304
VOLUME 53 (2009)
1 The resolution challenge in 3D optical microscopy, M. Martnez-Corral,
G. Saavedra 1 67
2 Transformation optics and the geometry of light, U. Leonhardt,
T.G. Philbin 69 152
3 Photorefractive solitons and their underlying nonlocal physics, E. DelRe,
B. Crosignani, P. Di Porto 153 200
4 Stimulated scattering effects of intense coherent light, G.S. He 201 292
5 Singular optics: Optical vortices and polarization singularities,
M.R. Dennis, K. OHolleran, M.J. Padgett 293 363
6 Quantum feed-forward control of light, U.L. Andersen, R. Filip 365 414
VOLUME 54 (2009)
1 Propagation-Invariant Optical Fields, J.Turunen, A.T. Friberg 1 88
2 Theoretical Tools for Quantum Optics in Structured Media,
D. Mogilevtsev, S. Kilin 89 148
3 Polarization and Coherence Optics: Historical Perspective, Status,
and Future Directions, C. Brosseau 149 208
4 Optical Quantum Computation, T.C. Ralph, G.J. Pryde 209 269
5 Science and Engineering of Photonic Crystals, K. Sakoda, J.W. Haus 271 317
6 The Mathematical Theory of Laser Beam-Splitting Gratings,
L.A. Romero, F.M. Dickey 319 386
VOLUME 55 (2010)
1 Optical Pulse Propagation in Biological Media: Theory and Numerical
Methods, M. Premaratne 1 40
2 Manipulating Light States by Single-Photon Addition and Subtraction,
M. Bellini, A. Zavatta 41 84
3 Vacuum-Induced Processes in Multilevel Atoms, M. Kiffner, M. Macovei,
J. Evers, C.H. Keitel 85 198
4 Adaptive Lens, Guoqiang Li 199 284
5 The Structure of Partially Coherent Fields, G. Gbur, T.D. Visser 285 342
VOLUME 56 (2011)
1 Space-Time Analogies in Optics V. Torres-Company, J. Lancis, P. Andres 1 80
2 Unconventional Polarization States: Beam Propagation, Focusing, and
Imaging, T.G. Brown 81 129
3 Quantum Scissors Finite-Dimensional States Engineering,
W. Leonski, A. Kowalewska-Kudaszyk 131 186
4 Polarization and Spectral Properties of Biphotons, M. Chekhova 187 226
5 Engineering Nonlinear Optic Sources of Photonic Entanglement,
J.P. Torres, K. Banaszek, I.A. Walmsley 227 331
VOLUME 57 (2012)
1 The Microscope in a Computer: Image Synthesis from Three-Dimensional
Full-Vector Solutions of Maxwells Equations at the Nanometer Scale,
Ilker R. C
apoglu, Jeremy D. Rogers, Allen Taflove, Vadim Backman 1 91
2 Microstructures and Nanostructures in Nature, Dora Kroisova 93 132
3 Quantitative Phase Imaging, Mustafa Mir, Basanta Bhaduri, Ru Wang,
Ruoyu Zhu, Gabriel Popescu 133 217
4 Tight Focusing of Light Beams: Effect of Polarization, Phase, and
Coherence, Ziyang Chen, Limin Hua, Jixiong Pu 219 260
5 Direct and Inverse Problems in the Theory of Light Scattering,
Daomu Zhao, Tao Wang 261 308
VOLUME 58 (2013)
1 Dynamic Photonic Materials Based on Liquid Crystals,
Luciano De Sio, Nelson Tabiryan, Timothy Bunning, Brian R. Kimball,
and Cesare Umeton 1 64
2 Invisibility Physics: Past, Present, and Future, Greg Gbur 65 114
3 Phase Anomalies in Micro-Optics,
Myun-Sik Kim, Toralf Schaf, Carsten Rockstuhl, and Hans Peter Herzig 115 197
4 Subwavelength Atom Localization, Kishore T. Kapale 199 250
5 Low-Dimensional Silicon Structures for Use in Photonic Circuits,
Tyler Roschuk, Iain F. Crowe, Andrew P. Knights, and Matthew P. Halsall 251 315
VOLUME 59 (2014)
1 Active Optical Metamaterials,
Sebastian Wuestner and Ortwin Hess 1 88
2 Spontaneous Parametric Down-Conversion in Nonlinear Layered
Structures, Jan Perina Jr. 89 158
3 Spatial Heterodyne Fourier-Transform Waveguide Spectrometers,
Aitor V. Velasco, Pavel Cheben, Mirosaw Florjanczyk, and Maria L. Calvo 159 208
4 Precursors and Dispersive Pulse Dynamics, A Century after the Sommerfeld-
Brillouin Theory: Part I. The Original Theory, Natalie A. Cartwright and
Kurt E. Oughstun 209 266
5 The Role of Coherence in Image Formation in Holographic Microscopy,
Radim Chmelik, Michala Slaba, Vera Kollarova, Tomas Slaby, Martin Lostak,
Jana Collakova, and Zbynek Dostal 267 336
CUMULATIVE INDEX VOLUMES 160
Abdullaev, F. and J. Garnier: Optical solitons in random media 48, 35

Abdullaev, F.Kh., S.A. Darmanyan and J. Garnier: Modulational instability
of electromagnetic waves in inhomogeneous and in discrete media 44, 303
Abele`s, F.: Methods for determining optical parameters of thin films 2, 249
Abella, I.D.: Echoes at optical frequencies 7, 139
Abitbol, C.I., see Clair, J.J. 16, 71
Abraham, N.B., P. Mandel and L.M. Narducci: Dynamical instabilities and
pulsations in lasers 25, 1
Aegerter, C.M. and G. Maret: Coherent backscattering and Anderson
localization of light 52, 1
Agarwal, G.S.: Master equation methods in quantum optics 11, 1
Agranovich, V.M. and V.L. Ginzburg: Crystal optics with spatial dispersion 9, 235
Agrawal, G.P.: Single-longitudinal-mode semiconductor lasers 26, 163
Agrawal, G.P., see Essiambre, R.-J. 37, 185
Allen, L. and D.G.C. Jones: Mode locking in gas lasers 9, 179
Allen, L., M.J. Padgett and M. Babiker: The orbital angular momentum of light 39, 291
Alfalou, A. and C. Brosseau: Recent advances in optical image processing 60, 119
Ammann, E.O.: Synthesis of optical birefringent networks 9, 123
Andersen, U.L. Filip, R.: Quantum feed-forward control of light 53, 365
Anderson, R., see Carriere, J. 41, 97
Apresyan, L.A., see Kravtsov,Yu.A. 36, 179
Arimondo, E.: Coherent population trapping in laser spectroscopy 35, 257
Armstrong, J.A. and A.W. Smith: Experimental studies of intensity fluctuations
in lasers 6, 211
Arnaud, J.A.: Hamiltonian theory of beam mode propagation 11, 247
Asakura, T., see Okamoto, T. 34, 183
Asakura, T., see Peiponen, K.-E. 37, 57
Asatryan, A.A., see Kravtsov, Yu.A. 39, 1
Babiker, M., see Allen, L. 39, 291

Baby,V., see Glesk, I. 45, 53
Backman,V., see C apoglu, I.R. 57, 1
Baltes, H.P.: On the validity of Kirchhoffs law of heat radiation for a body in a
nonequilibrium environment 13, 1
Banaszek, K., see Juan P. Torres 56, 227
Barabanenkov, Yu.N., Yu.A. Kravtsov, V.D. Ozrin and A.I. Saichev:
Enhanced backscattering in optics 29, 65
Barakat, R.: The intensity distribution and total illumination of aberration-free
diffraction images 1, 67
Barrett, H.H.: The Radon transform and its applications 21, 217
Bashkin, S.: Beam-foil spectroscopy 12, 287
Volumes IXL were previously distinguished by roman rather than by arabic numerals.
615
616 Cumulative Index Volumes 160
Bassett, I.M., W.T. Welford and R. Winston: Nonimaging optics for

flux concentration 27, 161
Beckmann, P.: Scattering of light by rough surfaces 6, 53
Bellini, M. and Zavatta, A.: Manipulating light states by single-photon addition
and subtraction 55, 41
Benisty, H. and C. Weisbuch: Photonic crystals 49, 177
Beran, M.J. and J. Oz-Vogt: Imaging through turbulence in the atmosphere 33, 319
Bernard, J., see Orrit, M. 35, 61
Berry, M.V. and C. Upstill: Catastrophe optics: morphologies of caustics and
their diffraction patterns 18, 257
Bertero, M. and C. De Mol: Super-resolution by data inversion 36, 129
Bertolotti, M., see Chumash, V. 36, 1
Bertolotti, M., see Mihalache, D. 27, 227
Bertolotti, M., F. Bovino and C. Sibilia: Quantum state engineering:
generation of single and pairs of photons 60, 1
Beverly III, R.E.: Light emission from high-current surface-spark discharges 16, 357
Bhaduri, B., see Mir, M. 57, 133
Bialynicki-Birula, I.: Photon wave function 36, 245
Biener, G., see Hasman, E. 47, 215
ork, G., A.B. Klimov and L.L. Sanchez-Soto: The discrete Wigner function
Bj 51, 469
Bloembergen, N.: From millisecond to attosecond laser pulses 50, 1
Bloom, A.L.: Gas lasers and their application to precise length measurements 9, 1
Bokor, N. and N. Davidson: Curved diffractive optical elements: Design and
applications 48, 107
Bokor, N., see Davidson, N. 45, 1
Bouman, M.A., W.A. Van De Grind and P. Zuidema: Quantum fluctuations in
vision 22, 77
Bousquet, P., see Rouard, P. 4, 145
Bovino, F., see Bertolotti, M. 60, 1
Boyd, R.W. and D.J. Gauthier: Slow and fast light 43, 497
Braat, J.J.M., S. van Haver, A.J.E.M. Janssen and P. Dirksen: Assessment
of optical systems by means of point-spread functions 51, 349
Brambilla, E., see Gatti, A. 51, 251
Brosseau, C., see Alfalou, A. 60, 119
Brosseau, C. and A. Dogariu: Symmetry properties and polarization descriptors
for an arbitrary electromagnetic wavefield 49, 315
Brosseau, C.: Polarization and coherence optics: Historical perspective, status,
and future directions 54, 149
Brown, G.S., see DeSanto, J.A. 23, 1
Brown, R., see Orrit, M. 35, 61
Brown, T. G.: Unconventional Polarization States: Beam Propagation,
Focusing, and Imaging 56, 81
Brunner, W. and H. Paul: Theory of optical parametric amplification and
oscillation 15, 1
Bryngdahl, O.: Applications of shearing interferometry 4, 37
Bryngdahl, O.: Evanescent waves in optical imaging 11, 167
Bryngdahl, O., T. Scheermesser and F. Wyrowski: Digital halftoning: synthesis
of binary images 33, 389
Cumulative Index Volumes 160 617
Bryngdahl, O. and F. Wyrowski: Digital holography - computer-generated

holograms 28, 1
Bunning, T., see De Sio L. 58, 1
Burch, J.M.: The metrological applications of diffraction gratings 2, 73
Butterweck, H.J.: Principles of optical data-processing 19, 211
Buzek, V. and P.L. Knight: Quantum interference, superposition states of light,
and nonclassical effects 34, 1
Cagnac, B., see Giacobino, E. 17, 85

Calvo, M.L., see Velasco, A.V. 59, 159
Cao, H.: Lasing in disordered media 45, 317
apoglu, I.R., J.D. Rogers, A. Taflove and V. Backman: The Microscope in a
C
Computer: Image Synthesis from Three-Dimensional Full-Vector
Solutions of Maxwells Equations at the Nanometer Scale 57, 1
Carmichael, H.J., G.T. Foster, L.A. Orozco, J.E. Reiner and P.R. Rice:
Intensity-field correlations of non-classical light 46, 355
Carriere, J., R. Narayan, W.-H. Yeh, C. Peng, P. Khulbe, L. Li, R. Anderson,
J. Choi and M. Mansuripur: Principles of optical disk data storage 41, 97
Cartwright, N.A. and K.E. Oughstun: Precursors and dispersive pulse
dynamics, a century after the Sommerfeld-Brillouin Theory: Part I. The
Original Theory 59, 209
Cartwright, N.A. and K.E. Oughstun: Precursors and dispersive pulse
dynamics: a century after the SommerfeldBrillouin theory: Part II. the
modern asymptotic theory 60, 263
Casasent, D. and D. Psaltis: Deformation invariant, space-variant optical
pattern recognition 16, 289
Cattaneo, S., see Kauranen, M. 51, 69
Ceglio, N.M. and D.W. Sweeney: Zone plate coded imaging: theory and
Cerf, N.J. and J. Fiurasek: Optical quantum cloning 49, 455
Chang, R.K., see Fields, M.H. 41, 1
Charnotskii, M.I., J. Gozani, V.I. Tatarskii and V.U. Zavorotny: Wave
propagation theories in random media based on the path-integral approach 32, 203
Cheben, P., see Velasco, A.V. 59, 159
Chen, R.T. and Z. Fu: Optical true-time delay control systems for wideband
phased array antennas 41, 283
Chen, Z., L. Hua and J. Pu: Tight Focusing of Light Beams: Effect of
Polarization, Phase, and Coherence 57, 219
Chiao, R.Y. and A.M. Steinberg: Tunneling times and superluminality 37, 345
Chmelik, R., M. Slaba, V. Kollarova, T. Slaby, M. Lostak, J. Collakova, and
Z. Dostal: The role of coherence in image formation in holographic
microscopy 59, 267
Choi, J., see Carriere, J. 41, 97
Christensen, J.L., see Rosenblum, W.M. 13, 69
Christov, I.P.: Generation and propagation of ultrashort optical pulses 29, 199
Chumash, V., I. Cojocaru, E. Fazio, F. Michelotti and M. Bertolotti:
Nonlinear propagation of strong laser pulses in chalcogenide glass films 36, 1
Clair, J.J. and C.I. Abitbol: Recent advances in phase profiles generation 16, 71
Clarricoats, P.J.B.: Optical fibre waveguides - a review 14, 327

Cohen-Tannoudji, C. and A. Kastler: Optical pumping 5, 1
Cojocaru, I., see Chumash, V. 36, 1
Cole, T.W.: Quasi-optical techniques of radio astronomy 15, 187
Collakova, J., see Chmelik, R. 59, 267
Colombeau, B., see Froehly, C. 20, 63
Cook, R.J.: Quantum jumps 28, 361
Courte`s, G., P. Cruvellier and M. Detaille: Some new optical designs for ultra-
violet bidimensional detection of astronomical objects 20, 1
Creath, K.: Phase-measurement interferometry techniques 26, 349
Crewe, A.V.: Production of electron probes using a field emission source 11, 223
Crosignani, B., see DelRe, E. 53, 153
Crowe, I.F., see Roschuck, T. 58, 251
Cruvellier, P., see Courte`s, G. 20, 1
Cummins, H.Z. and H.L. Swinney: Light beating spectroscopy 8, 133
Dainty, J.C.: The statistics of speckle patterns 14, 1

Dandliker, R.: Heterodyne holographic interferometry 17, 1
Darmanyan, S.A., see Abdullaev, F.Kh. 44, 303
Dattoli, G., L. Giannessi, A. Renieri and A. Torre: Theory of Compton free
electron lasers 31, 321
Davidson, N. and N. Bokor: Anamorphic beam shaping for laser and diffuse
light 45, 1
Davidson, N., see Bokor, N. 48, 107
Davidson, N., see Oron, R. 42, 325
De Mol, C., see Bertero, M. 36, 129
De Sterke, C.M. and J.E. Sipe: Gap solitons 33, 203
De Sio, L., N. Tabiryan, T. Bunning, B.R. Kimball, and C. Umeton: Dynamic
Photonic Materials Based on Liquid crystals 58, 1
Decker Jr, J.A., see Harwit, M. 12, 101
Delano, E. and R.J. Pegis: Methods of synthesis for dielectric multilayer filters 7, 67
DelRe, E., Crosignani, B. and Di Porto, P.: Photorefractive solitons and their
underlying nonlocal physics 53, 153
Demaria, A.J.: Picosecond laser pulses 9, 31
Demkowicz-Dobrza nski, R., M. Jarzyna and J. Koody
nski: Quantum limits in
optical interferometry 60, 345
DeSanto, J.A. and G.S. Brown: Analytical techniques for multiple scattering
from rough surfaces 23, 1
Dennis, M.R., OHolleran, K. and Padgett, M.J.: Singular optics: Optical
vortices and polarization singularities 53, 293
Desyatnikov, A.S., Y.S. Kivshar and L.L. Torner: Optical vortices and vortex
solitons 47, 291
Detaille, M., see Courte`s, G. 20, 1
Dexter, D.L., see Smith, D.Y. 10, 165
Di Porto, P., see DelRe, E. 53, 153
Dickey, F.M., see Romero, L.A. 54, 319
Dirksen, P., see Braat, J.J.M. 51, 349
Dogariu, A., see Brosseau, C. 49, 315
Domachuk, P., see Eggleton, B.J. 48, 1

Dostal, Z., see Chmelik, R. 59, 267
Dragoman, D.: The Wigner distribution function in optics and optoelectronics 37, 1
Dragoman, D.: Phase space correspondence between classical optics and
quantum mechanics 43, 433
Drexhage, K.H.: Interaction of light with monomolecular dye layers 12, 163
Duguay, M.A.: The ultrafast optical Kerr shutter 14, 161
Dusek M., N. L utkenhaus and M. Hendrych: Quantum cryptography 49, 381
Dutta, N.K. and J.R. Simpson: Optical amplifiers 31, 189
Dutta Gupta, S.: Nonlinear optics of stratified media 38, 1
Eberly, J.H.: Interaction of very intense light with free electrons 7, 359
Eggleton, B.J., P. Domachuk, C. Grillet, E.C. Magi, H.C. Nguyen, P.
Steinvurzel and M.J. Steel: Laboratory post-engineering of microstructured
optical fibers 48, 1
Englund, J.C., R.R. Snapp and W.C. Schieve: Fluctuations, instabilities and
chaos in the laser-driven nonlinear ring cavity 21, 355
Ennos, A.E.: Speckle interferometry 16, 233
Erez, N., see Greenberger, D.M. 50, 275
Essiambre, R.-J. and G.P. Agrawal: Soliton communication systems 37, 185
Etrich, C., F. Lederer, B.A. Malomed, T. Peschel and U. Peschel: Optical
solitons in media with a quadratic nonlinearity 41, 483
Evers, J., see Kiffner, M. 55, 85
Fabelinskii, I.L.: Spectra of molecular scattering of light 37, 95

Fabre, C., see Reynaud, S. 30, 1
Facchi, P. and S. Pascazio: Quantum Zeno and inverse quantum Zeno effects 42, 147
Fante, R.L.: Wave propagation in random media: a systems approach 22, 341
Fazio, E., see Chumash, V. 36, 1
Fercher, A.F. and C.K. Hitzenberger: Optical coherence tomography 44, 215
Ficek, Z. and H.S. Freedhoff: Spectroscopy in polychromatic fields 40, 389
Fields, M.H., J. Popp and R.K. Chang: Nonlinear optics in microspheres 41, 1
Filip, R.: see Andersen, U.L. 53, 365
Fiorentini, A.: Dynamic characteristics of visual processes 1, 253
Fiurasek, J., see Cerf, N.J. 49, 455
Florja nczyk, M., see Velasco, A.V. 59, 159
Flytzanis, C., F. Hache, M.C. Klein, D. Ricard and Ph. Roussignol: Nonlinear
optics in composite materials. 1. Semiconductor and metal crystallites in
dielectrics 29, 321
Focke, J.: Higher order aberration theory 4, 1
Forbes, G.W., see Kravtsov, Yu.A. 39, 1
Foster, G.T., see Carmichael, H.J. 46, 355
Francon, M. and S. Mallick: Measurement of the second order degree of
coherence 6, 71
Franta, D., see Ohldal, I. 41, 181
Freedhoff, H.S., see Ficek, Z. 40, 389
Freilikher, V.D. and S.A. Gredeskul: Localization of waves in media with one-
dimensional disorder 30, 137
Friberg, A.T., see Turunen, J. 54, 1

Frieden, B.R.: Evaluation, design and extrapolation methods for optical signals,
based on use of the prolate functions 9, 311
Friesem, A.A., see Oron, R. 42, 325
Froehly, C., B. Colombeau and M. Vampouille: Shaping and analysis of
picosecond light pulses 20, 63
Fry, G.A.: The optical performance of the human eye 8, 51
Fu, Z., see Chen, R.T. 41, 283
Gabitov, I.R., see Litchinitser, N.M. 51, 1

Gabor, D.: Light and information 1, 109
Gallion, P., F. Mendieta and S. Jiang: Signal and quantum noise in optical
communications and cryptography 52, 149
Gamo, H.: Matrix treatment of partial coherence 3, 187
Gandjbakhche, A.H. and G.H. Weiss: Random walk and diffusion-like models
of photon migration in turbid media 34, 333
Gantsog, Ts., see Tanas, R. 35, 355
Gao, W., see Yin,J. 45, 119
Garcia-Ojalvo, J., see Uchida, A. 48, 203
Garnier, J., see Abdullaev, F. 48, 35
Garnier, J., see Abdullaev, F.Kh. 44, 303
Gatti, A., E. Brambilla and L. Lugiato: Quantum imaging 51, 251
Gauthier, D.J.: Two-photon lasers 45, 205
Gauthier, D.J., see Boyd, R.W. 43, 497
Gbur, G.: Nonradiating sources and other invisible objects 45, 273
Gbur, G. and Visser, T.D.: The structure of partially coherent fields 55, 285
Gbur, G.: Invisibility Physics: Past, Present, and Future 58, 65
Gea-Banacloche, J.: Optical realizations of quantum teleportation 46, 311
Ghatak, A. and K. Thyagarajan: Graded index optical waveguides: a review 18, 1
Ghatak, A.K., see Sodha, M.S. 13, 169
Giacobino, E. and B. Cagnac: Doppler-free multiphoton spectroscopy 17, 85
Giacobino, E., see Reynaud, S. 30, 1
Giannessi, L., see Dattoli, G. 31, 321
Ginzburg, V.L.: Radiation by uniformly moving sources. Vavilov-Cherenkov
effect, Doppler effect in a medium, transition radiation and associated
phenomena 32, 267
Ginzburg,V.L., see Agranovich, V.M. 9, 235
Giovanelli, R.G.: Diffusion through non-uniform media 2, 109
Glaser, I.: Information processing with spatially incoherent light 24, 389
Glesk, I., B.C. Wang, L. Xu, V. Baby and P.R. Prucnal: Ultra-fast all-optical
switching in optical networks 45, 53
Gniadek, K. and J. Petykiewicz: Applications of optical methods in the
diffraction theory of elastic waves 9, 281
Goodman, J.W.: Synthetic-aperture optics 8, 1
Gozani, J., see Charnotskii, M.I. 32, 203
Graham, R.: The phase transition concept and coherence in atomic emission 12, 233
Gredeskul, S.A., see Freilikher, V.D. 30, 137
Greenberger, D.M., N. Erez, M.O. Scully, A.A. Svidzinsky and M.S. Zubairy:
Planck, photon statistics, and Bose-Einstein condensation 50, 275
Grillet, C., see Eggleton, B.J. 48, 1
Hache, F., see Flytzanis, C. 29, 321

Hall, D.G.: Optical waveguide diffraction gratings: coupling between guided
modes 29, 1
Halsall, M.P., see Roschuk T. 58, 251
Hariharan, P.: Colour holography 20, 263
Hariharan, P.: Interferometry with lasers 24, 103
Hariharan, P.: The geometric phase 48, 149
Hariharan, P. and B.C. Sanders: Quantum phenomena in optical
interferometry 36, 49
Harwit, M. and J.A. Decker Jr: Modulation techniques in spectrometry 12, 101
Hasegawa, A., see Kodama, Y. 30, 205
Hasman, E., G. Biener, A. Niv and V. Kleiner: Space-variant polarization
manipulation 47, 215
Hasman, E., see Oron, R. 42, 325
Haus, J.W., see Sakoda, K. 54, 271
He, G.S., Stimulated scattering effects of intense coherent light 53, 201
Heidmann, A., see Reynaud, S. 30, 1
Hello, P.: Optical aspects of interferometric gravitational-wave detectors 38, 85
Helstrom, C.W.: Quantum detection theory 10, 289
Hendrych, M., see Dusek, M. 49, 381
Herriot, D.R.: Some applications of lasers to interferometry 6, 171
Herzig, H.P., see Kim, M.S. 58, 115
Hess, O., see Wuestner, S. 59, 1
Hitzenberger, C.K., see Fercher, A.F. 44, 215
Horner, J.L., see Javidi, B. 38, 343
Hua, L., see Chen, Z. 57, 219
Huang, T.S.: Bandwidth compression of optical images 10, 1
Ichioka,Y., see Tanida, J. 40, 77

Imoto, N., see Yamamoto,Y. 28, 87
Ishii, Y.: Laser-diode interferometry 46, 243
Itoh, K.: Interferometric multispectral imaging 35, 145
Iwata, K.: Phase imaging and refractive index tomography for X-rays and
visiblerays 47, 393
Jacobsson, R.: Light reflection from films of continuously varying refractive

index 5, 247
Jacquinot, P. and B. Roizen-Dossier: Apodisation 3, 29
Jacquod, Ph., see Tureci, H.E. 47, 75
Jaeger, G. and A.V. Sergienko: Multi-photon quantum interferometry 42, 277
Jahns, J.: Free-space optical digital computing and interconnection 38, 419
Jamroz, W. and B.P. Stoicheff: Generation of tunable coherent vacuum-
ultraviolet radiation 20, 325
Janssen, A.J.E.M., see Braat, J.J.M. 51, 349
Jarzyna, M., see Demkowicz-Dobrza nski, R. 60, 345

Javidi, B. and J.L. Horner: Pattern recognition with nonlinear techniques in the
Fourier domain 38, 343
Jesus Lancis, see Victor Torres-Company 56, 1
Jiang, S., see Gallion, P. 52, 149
Jones, D.G.C., see Allen, L. 9, 179
Joshi, A. and M. Xiao: Controlling nonlinear optical processes in multi-level
atomic systems 49, 97
Juan P. Torres, K. Banaszek, and I. A. Walmsley: Engineering nonlinear optic
sources of photonic entanglement 56, 227
Kapale, K.T.: Subwavelength Atom Localization 58, 199

Kartashov, Y.V., V.A. Vysloukh and L. Torner: Soliton shape and mobility
control in optical lattices 52, 63
Kastler, A., see Cohen-Tannoudji, C. 5, 1
Kauranen, M. and S. Cattaneo: Polarization techniques for surface nonlinear
optics 51, 69
Keitel, C.H., see Kiffner, M. 55, 85
Keller, O.: Local fields in linear and nonlinear optics of mesoscopic systems 37, 257
Keller, O.: Optical works of L.V. Lorenz 43, 195
Keller, O.: Historical papers on the particle concept of light 50, 51
Keller, U.: Ultrafast solid-state lasers 46, 1
Kemp, B.A.: Macroscopic theory of optical momentum 60, 437
Khoo, I.C.: Nonlinear optics of liquid crystals 26, 105
Khulbe, P., see Carriere, J. 41, 97
Kielich, S.: Multi-photon scattering molecular spectroscopy 20, 155
Kiffner, M., Macovei, M., Evers, J. and Keitel, C.H.: Vacuum-induced
processes in multilevel atoms 55, 85
Kilin, S., see Mogilevtsev, D. 54, 89
Kilin, S.Ya.: Quanta and information 42, 1
Kimball, B.R., see De Sio L. 58, 1
Kinosita, K.: Surface deterioration of optical glasses 4, 85
Kim, M.S., T. Scharf, C. Rockstuhl, H.P. Herzig: Phase Anomalies in Micro-
Optics 58, 115
Kitagawa, M., see Yamamoto,Y. 28, 87
Kivshar,Y.S., see Desyatnikov, A.S. 47, 291
Kivshar,Y.S., see Saltiel, S.M. 47, 1
Klein, M.C., see Flytzanis, C. 29, 321
Kleiner,V., see Hasman, E. 47, 215
Klimov, A.B., see Bj ork, G. 51, 469
Klyatskin, V.I.: The imbedding method in statistical boundary-value wave
problems 33, 1
Knight, P.L., see Buzek, V. 34, 1
Knights, A.P., see Roschuk, T. 58, 251
Kodama, Y. and A. Hasegawa: Theoretical foundation of optical-soliton
concept in fibers 30, 205
Kollarova,V., see Chmelik, R. 59, 267
Koody nski, J., see Demkowicz-Dobrza nski, R. 60, 345
Koppelman, G.: Multiple-beam interference and natural modes in open

resonators 7, 1
Kottler, F.: The elements of radiative transfer 3, 1
Kottler, F.: Diffraction at a black screen, Part I: Kirchhoffs theory 4, 281
Kottler, F.: Diffraction at a black screen, Part II: electromagnetic theory 6, 331
Kowalewska-Kudlaszyk, A. see W. Leo nski 56, 131
Kozhekin, A.E., see Kurizki, G. 42, 93
Kravtsov, Yu.A.: Rays and caustics as physical objects 26, 227
Kravtsov, Yu.A. and L.A. Apresyan: Radiative transfer: new aspects of the old
theory 36, 179
Kravtsov, Yu.A., G.W. Forbes and A.A. Asatryan: Theory and applications of
complex rays 39, 1
Kravtsov,Yu.A., see Barabanenkov, Yu.N. 29, 65
Kroisova, D.: Microstructures and Nanostructures in Nature 57, 93
Kubota, H.: Interference color 1, 211
Kuittinen, M., see Turunen, J. 40, 343
Kurizki, G., A.E. Kozhekin, T. Opatrny and B.A. Malomed: Optical solitons in
periodic media with resonant and off-resonant nonlinearities 42, 93
Labeyrie, A.: High-resolution techniques in optical astronomy 14, 47

Lakhtakia, A., see Mackay, T.G. 51, 121
Lean, E.G.: Interaction of light and acoustic surface waves 11, 123
Lederer, F., see Etrich, C. 41, 483
Lee, W.-H.: Computer-generated holograms: techniques and applications 16, 119
Leith, E.N. and Upatnieks, J.: Recent advances in holography 6, 1
Leonhardt, U. and Philbin, T.G.: Transformation optics and the geometry of
light 53, 69
Leo nski, W., and A. Kowalewska-Kudlaszyk: Quantum scissors
finite-dimensional states engineering 56, 131
Letokhov, V.S.: Laser selective photophysics and photochemistry 16, 1
Leuchs, G., see Sizmann, A. 39, 373
Levi, L.: Vision in communication 8, 343
Li, G.: Adaptive lens 55, 199
Li, L., see Carriere, J. 41, 97
Lipson, H. and C.A. Taylor: X-ray crystal-structure determination as a branch
of physical optics 5, 287
Litchinitser, N.M., I.R. Gabitov, A.I. Maimistov and V.M. Shalaev: Negative
refractive index metamaterials in optics 51, 1
Lohmann, A.W., D. Mendlovic and Z. Zalevsky: Fractional transformations in
optics 38, 263
Lohmann, A.W., see Zalevsky, Z. 40, 271
Lostak, M., see Chmelik, R. 59, 267
Lounis, B., see Orrit, M. 35, 61
Lugiato, L., see Gatti, A. 51, 251
Lugiato, L.A.: Theory of optical bistability 21, 69
Luis, A. and L.L. Sanchez-Soto: Quantum phase difference, phase
measurements and Stokes operators 41, 419
Luks, A. and V. Perinova: Canonical quantum description of light propagation

in dielectric media 43, 295
Lukrs, A., see Perinova, V. 33, 129
Lukrs, A., see Perinova, V. 40, 115
utkenhaus, N., see Dusek, M.
L 49, 381
Machida, S., see Yamamoto, Y. 28, 87

Mackay, T.G. and A. Lakhtakia: Electromagnetic fields in linear bianisotropic
mediums 51, 121
Macovei, M., see Kiffner, M. 55, 85
Magi, E.C., see Eggleton, B.J. 48, 1
Mahajan, V.N.: Gaussian apodization and beam propagation 49, 1
Maimistov, A.I., see Litchinitser, N.M. 51, 1
Mainfray, G. and C. Manus: Nonlinear processes in atoms and in weakly
relativistic plasmas 32, 313
Malacara, D.: Optical and electronic processing of medical images 22, 1
Malacara, D., see Vlad, V.I. 33, 261
Mallick, S., see Francon, M. 6, 71
Malomed, B.A.: Variational methods in nonlinear fiber optics and related fields 43, 71
Malomed, B.A., see Etrich, C. 41, 483
Malomed, B.A., see Kurizki, G. 42, 93
Mandel, L.: Fluctuations of light beams 2, 181
Mandel, L.: The case for and against semiclassical radiation theory 13, 27
Mandel, P., see Abraham, N.B. 25, 1
Mansuripur, M., see Carriere, J. 41, 97
Manus, C., see Mainfray, G. 32, 313
Maradudin, A.A., see Shchegrov, A.V. 46, 117
Marchand, E.W.: Gradient index lenses 11, 305
Maret, G., see Aegerter, C.M. 52, 1
Maria Chekhova: Polarization and Spectral Properties of Biphotons 56, 187
Martin, P.J. and R.P. Netterfield: Optical films produced by ion-based
techniques 23, 113
Martnez-Corral, M. and Saavedra, G.: The resolution challenge in 3D optical
microscopy 53, 1
Masalov, A.V.: Spectral and temporal fluctuations of broad-band laser radiation 22, 145
Maystre, D.: Rigorous vector theories of diffraction gratings 21, 1
Meessen, A., see Rouard, P. 15, 77
Mehta, C.L.: Theory of photoelectron counting 8, 373
Mendez, E.R., see Shchegrov, A.V. 46, 117
Mendieta, F., see Gallion, P. 52, 149
Mendlovic, D., see Lohmann, A.W. 38, 263
Mendlovic, D., see Zalevsky, Z. 40, 271
Meystre, P.: Cavity quantum optics and the quantum measurement process 30, 261
Meystre, P., see Search, C.P. 47, 139
Michelotti, F., see Chumash, V. 36, 1
Mihalache, D., M. Bertolotti and C. Sibilia: Nonlinear wave propagation in
planar structures 27, 227
Mikaelian, A.L.: Self-focusing media with variable index of refraction 17, 279
Mikaelian, A.L. and M.L. Ter-Mikaelian: Quasi-classical theory of laser

radiation 7, 231
Mills, D.L. and K.R. Subbaswamy: Surface and size effects on the light
scattering spectra of solids 19, 45
Milonni, P.W.: Field quantization in optics 50, 97
Milonni, P.W. and B. Sundaram: Atoms in strong fields: photoionization and
chaos 31, 1
Mir, M.: Quantitative Phase Imaging 57, 133
Miranowicz, A., see Tanas, R. 35, 355
Miyamoto, K.: Wave optics and geometrical optics in optical design 1, 31
Mogilevtsev, D. and Kilin, S.: Theoretical tools for quantum optics in
structured media 54, 89
Mollow, B.R.: Theory of intensity dependent resonance light scattering and
resonance fluorescence 19, 1
Murata, K.: Instruments for the measuring of optical transfer functions 5, 199
Musset, A. and A. Thelen: Multilayer antireflection coatings 8, 201
Nakwaski, W. and M. Osi nski: Thermal properties of vertical-cavity surface-

emitting semiconductor lasers 38, 165
Narayan, R., see Carriere, J. 41, 97
Narducci, L.M., see Abraham, N.B. 25, 1
Navratil, K., see Ohldal, I. 34, 249
Netterfield, R.P., see Martin, P.J. 23, 113
Nguyen, H.C., see Eggleton, B.J. 48, 1
Nishihara, H. and T. Suhara: Micro Fresnel lenses 24, 1
Niv, A., see Hasman, E. 47, 215
Noethe, L.: Active optics in modern large optical telescopes 43, 1
Novotny, L.: The history of near-field optics 50, 137
Nussenzveig, H.M.: Light tunneling 50, 185
Obod, Yu.A., see Shvartsburg, A.B. 60, 489

Ohldal, I. and D. Franta: Ellipsometry of thin film systems 41, 181
Ohldal, I., K. Navratil and M. Ohldal: Scattering of light from multilayer
systems with rough boundaries 34, 249
Ohldal, M., see Ohldal, I. 34, 249
OHolleran, K., see Dennis, M.R. 53, 293
Ohtsu, M. and T. Tako: Coherence in semiconductor lasers 25, 191
Ohtsubo, J.: Chaotic dynamics in semiconductor lasers with optical feedback 44, 1
Okamoto, T. and T. Asakura: The statistics of dynamic speckles 34, 183
Okoshi, T.: Projection-type holography 15, 139
Omenetto, F.G.: Femtosecond pulses in optical fibers 44, 85
Ooue, S.: The photographic image 7, 299
Opatrny, T., see Kurizki, G. 42, 93
Opatrny, T., see Welsch, D.-G. 39, 63
Oron, R., N. Davidson, A.A. Friesem and E. Hasman: Transverse mode
shaping and selection in laser resonators 42, 325
Orozco, L.A., see Carmichael, H.J. 46, 355
Orrit, M., J. Bernard, R. Brown and B. Lounis: Optical spectroscopy of single

molecules in solids 35, 61
nski, M., see Nakwaski, W.
Osi 38, 165
Ostrovskaya, G.V. and Yu.I. Ostrovsky: Holographic methods of plasma
diagnostics 22, 197
Ostrovsky, Yu.I. and V.P. Shchepinov: Correlation holographic and speckle
interferometry 30, 87
Ostrovsky,Yu.I., see Ostrovskaya, G.V. 22, 197
Oughstun, K.E.: Unstable resonator modes 24, 165
Oughstun, K.E., see Cartwright, N.A. 59, 209
Oughstun, K.E., see Cartwright, N.A. 60, 263
Oz-Vogt, J., see Beran, M.J. 33, 319
Ozrin,V.D., see Barabanenkov, Yu.N. 29, 65
Padgett, M.J., see Allen, L. 39, 291

Padgett, M.J., see Dennis, M.R. 53, 293
Pal, B.P.: Guided-wave optics on silicon: physics, technology and status 32, 1
Paoletti, D. and G. Schirripa Spagnolo: Interferometric methods for artwork
diagnostics 35, 197
Pascazio, S., see Facchi, P. 42, 147
Patorski, K.: The self-imaging phenomenon and its applications 27, 1
Paul, H., see Brunner, W. 15, 1
Pedro Andres, see Vctor Torres-Company 56, 1
Pegis, R.J.: The modern development of Hamiltonian optics 1, 1
Pegis, R.J., see Delano, E. 7, 67
Peiponen, K.-E., E.M. Vartiainen and T. Asakura: Dispersion relations and
phase retrieval in optical spectroscopy 37, 57
Peng, C., see Carriere, J. 41, 97
Perina Jr, J.: Spontaneous parametric down-conversion in nonlinear layered
structures 59, 89
Perina Jr, J. and J. Perina: Quantum statistics of nonlinear optical couplers 41, 359
Perina, J.: Photocount statistics of radiation propagating through random and
nonlinear media 18, 127
Perina, J., see Perina Jr, J. 41, 359
Perinova, V. and A. Luks: Quantum statistics of dissipative nonlinear oscillators 33, 129
Perinova, V. and A. Luks: Continuous measurements in quantum optics 40, 115
Perinova,V., see Luks, A. 43, 295
Pershan, P.S.: Non-linear optics 5, 83
Peschel, T., see Etrich, C. 41, 483
Peschel, U., see Etrich, C. 41, 483
Petite, G., see Shvartsburg, A.B. 44, 143
Petykiewicz, J., see Gniadek, K. 9, 281
Philbin, T.G., see Leonhardt, U. 53, 69
Picht, J.: The wave of a moving classical electron 5, 351
Pollock, C.R.: Ultrafast optical pulses 51, 211
Popescu, G., see Mir, M. 57, 133
Popov, E.: Light diffraction by relief gratings: a macroscopic and microscopic view 31, 139
Popp, J., see Fields, M.H. 41, 1
Porter, R.P.: Generalized holography with application to inverse scattering and

inverse source problems 27, 315
Premaratne, M.: Optical pulse propagation in biological media: theory and
numerical methods 55, 1
Presnyakov, L.P.: Wave propagation in inhomogeneous media: phase-shift
approach 34, 159
Prucnal, P.R., see Glesk, I. 45, 53
Pryde, G.J., see Ralph, T.C. 54, 209
Psaltis, D. and Y. Qiao: Adaptive multilayer optical networks 31, 227
Psaltis, D., see Casasent, D. 16, 289
Pu, J., see Chen, Z. 57, 219
Qiao,Y., see Psaltis, D. 31, 227

Qiu, M., see Yan, M. 52, 261
Ralph, T.C. and Pryde, G.J.: Optical quantum computation 54, 209
Raymer, M.G. and I.A. Walmsley: The quantum coherence properties of
stimulated Raman scattering 28, 181
Reiner, J.E., see Carmichael, H.J. 46, 355
Renieri, A., see Dattoli, G. 31, 321
Reynaud, S., A. Heidmann, E. Giacobino and C. Fabre: Quantum fluctuations
in optical systems 30, 1
Ricard, D., see Flytzanis, C. 29, 321
Rice, P.R., see Carmichael, H.J. 46, 355
Riseberg, L.A. and M.J. Weber: Relaxation phenomena in rare-earth
luminescence 14, 89
Risken, H.: Statistical properties of laser light 8, 239
Rockstuhl, C., see Kim, M.S. 58, 115
Roddier, F.: The effects of atmospheric turbulence in optical astronomy 19, 281
Rogers, J.D., see C apoglu, I.R. 57, 1
Rogister, F., see Uchida, A. 48, 203
Roizen-Dossier, B., see Jacquinot, P. 3, 29
Romero, L.A. and Dickey, F.M.: The mathematical theory of laser beam-
splitting gratings 54, 319
Ronchi, L., see Wang Shaomin 25, 279
Rosanov, N.N.: Transverse patterns in wide-aperture nonlinear optical systems 35, 1
Roschuk, T., I.F. Crowe, A.P. Knights, M.P. Halsall: Low-Dimensional
Silicon Structures for Use in Photonic Circuits 58, 251
Rosenblum, W.M. and J.L. Christensen: Objective and subjective spherical
aberration measurements of the human eye 13, 69
Rothberg, L.: Dephasing-induced coherent phenomena 24, 39
Rouard, P. and P. Bousquet: Optical constants of thin films 4, 145
Rouard, P. and A. Meessen: Optical properties of thin metal films 15, 77
Roussignol, Ph., see Flytzanis, C. 29, 321
Roy, R., see Uchida, A. 48, 203
Rubinowicz, A.: The Miyamoto-Wolf diffraction wave 4, 199
Rudolph, D., see Schmahl, G. 14, 195
Saavedra, G., see Martnez-Corral, M. 53, 1

Saichev, A.I., see Barabanenkov, Yu.N. 29, 65
Saito, S., see Yamamoto,Y. 28, 87
Sakai, H., see Vanasse, G.A. 6, 259
Sakoda, K. and Haus, J.W.: Science and engineering of photonic crystals 54, 271
Saleh, B.E.A., see Teich, M.C. 26, 1
Saltiel, S.M., A.A. Sukhorukov and Y.S. Kivshar: Multistep parametric
processes in nonlinear optics 47, 1
Sanchez-Soto, L.L., see Bj ork, G. 51, 469
Sanchez-Soto, L.L., see Luis, A. 41, 419
Sanders, B.C., see Hariharan, P. 36, 49
Scharf, T., see Kim, M.S., 58, 115
Scheermesser, T., see Bryngdahl, O. 33, 389
Schieve, W.C., see Englund, J.C. 21, 355
Schirripa Spagnolo, G., see Paoletti, D. 35, 197
Schmahl, G. and D. Rudolph: Holographic diffraction gratings 14, 195
Schubert, M. and B. Wilhelmi: The mutual dependence between coherence
properties of light and nonlinear optical processes 17, 163
Schulz, G.: Aspheric surfaces 25, 349
Schulz, G. and J. Schwider: Interferometric testing of smooth surfaces 13, 93
Schwefel, H.G.L., see T ureci, H.E. 47, 75
Schwider, J.: Advanced evaluation techniques in interferometry 28, 271
Schwider, J., see Schulz, G. 13, 93
Scully, M.O. and K.G. Whitney: Tools of theoretical quantum optics 10, 89
Scully, M.O., see Greenberger, D.M. 50, 275
Search, C.P. and P. Meystre: Nonlinear and quantum optics of atomic and
molecular fields 47, 139
Senitzky, I.R.: Semiclassical radiation theory within a quantum-mechanical
framework 16, 413
Sergienko, A.V., see Jaeger, G. 42, 277
Shalaev,V.M., see Litchinitser, N.M. 51, 1
Sharma, S.K. and D.J. Somerford: Scattering of light in the eikonal
approximation 39, 213
Shchegrov, A.V., A.A. Maradudin and E.R. Mendez: Multiple scattering of
light from randomly rough surfaces 46, 117
Shchepinov, V.P., see Ostrovsky, Yu.I. 30, 87
Shvartsburg, A.B. and G. Petite: Instantaneous optics of ultrashort broadband
pulses and rapidly varying media 44, 143
Shvartsburg, A.B., Yu.A. Obod and O.D. Volpian: Tunneling of
electromagnetic waves in all-dielectric gradient metamaterials 60, 489
Sibilia, C., see Bertolotti, M. 60, 1
Sibilia, C., see Mihalache, D. 27, 227
Simpson, J.R., see Dutta, N.K. 31, 189
Sipe, J.E., see De Sterke, C.M. 33, 203
Sipe, J.E., see Van Kranendonk, J. 15, 245
Sittig, E.K.: Elastooptic light modulation and deflection 10, 229
Sizmann, A. and G. Leuchs: The optical Kerr effect and quantum optics in
fibers 39, 373
Slaba, M., see Chmelik, R. 59, 267

Slaby, T., see Chmelik, R. 59, 267
Slusher, R.E.: Self-induced transparency 12, 53
Smith, A.W., see Armstrong, J.A. 6, 211
Smith, D.Y. and D.L. Dexter: Optical absorption strength of defects in
insulators 10, 165
Smith, R.W.: The use of image tubes as shutters 10, 45
Snapp, R.R., see Englund, J.C. 21, 355
Sodha, M.S., A.K. Ghatak and V.K. Tripathi: Self-focusing of laser beams in
plasmas and semiconductors 13, 169
Somerford, D.J., see Sharma, S.K. 39, 213
Soroko, L.M.: Axicons and meso-optical imaging devices 27, 109
Soskin, M.S. and M.V. Vasnetsov: Singular optics 42, 219
Spreeuw, R.J.C. and J.P. Woerdman: Optical atoms 31, 263
Steel, M.J., see Eggleton, B.J. 48, 1
Steel, W.H.: Two-beam interferometry 5, 145
Steinberg, A.M., see Chiao, R.Y. 37, 345
Steinvurzel, P., see Eggleton, B.J. 48, 1
Stoicheff, B.P., see Jamroz, W. 20, 325
Stone, A.D., see T ureci, H.E. 47, 75
Strohbehn, J.W.: Optical propagation through the turbulent atmosphere 9, 73
Stroke, G.W.: Ruling, testing and use of optical gratings for high-resolution
spectroscopy 2, 1
Subbaswamy, K.R., see Mills, D.L. 19, 45
Suhara, T., see Nishihara, H. 24, 1
Sukhorukov, A.A., see Saltiel, S.M. 47, 1
Sundaram, B., see Milonni, P.W. 31, 1
Svelto, O.: Self-focusing, self-trapping, and self-phase modulation of laser
beams 12, 1
Svidzinsky, A.A., see Greenberger, D.M. 50, 275
Sweeney, D.W., see Ceglio, N.M. 21, 287
Swinney, H.L., see Cummins, H.Z. 8, 133
Tabiryan, N., see De Sio L

Taflove, A., see C apoglu, I.R. 57, 1
Tako, T., see Ohtsu, M. 25, 191
Tanaka, K.: Paraxial theory in optical design in terms of Gaussian brackets 23, 63
Tanas, R., A. Miranowicz and Ts. Gantsog: Quantum phase properties of
nonlinear optical phenomena 35, 355
Tango, W.J. and R.Q. Twiss: Michelson stellar interferometry 17, 239
Tanida, J. and Y. Ichioka: Digital optical computing 40, 77
Tatarskii, V.I. and V.U. Zavorotnyi: Strong fluctuations in light propagation in
a randomly inhomogeneous medium 18, 204
Tatarskii,V.I., see Charnotskii, M.I. 32, 203
Taylor, C.A., see Lipson, H. 5, 287
Teich, M.C. and B.E.A. Saleh: Photon bunching and antibunching 26, 1
Ter-Mikaelian, M.L., see Mikaelian, A.L. 7, 231
Thelen, A., see Musset, A. 8, 201
Thompson, B.J.: Image formation with partially coherent light 7, 169

Thyagarajan, K., see Ghatak, A. 18, 1
Tonomura, A.: Electron holography 23, 183
Torner, L., see Kartashov, Y.V. 52, 63
Torner, L.L., see Desyatnikov, A.S. 47, 291
Torre, A.: The fractional Fourier transform and some of its applications to
optics 43, 531
Torre, A., see Dattoli, G. 31, 321
Tripathi,V.K., see Sodha, M.S. 13, 169
Tsujiuchi, J.: Correction of optical images by compensation of aberrations and
by spatial frequency filtering 2, 131
T
ureci, H.E., H.G.L. Schwefel, Ph. Jacquod and A.D. Stone: Modes of wave-
chaotic dielectric resonators 47, 75
Turunen, J., M. Kuittinen and F. Wyrowski: Diffractive optics:
electromagnetic approach 40, 343
Turunen, J. and Friberg, A.T.: Propagation-invariant optical fields 54, 1
Twiss, R.Q., see Tango, W.J. 17, 239
Uchida, A., F. Rogister, J. Garca-Ojalvo and R. Roy: Synchronization and

communication with chaotic laser systems 48, 203
Umeton, C., see De Sio L. 58, 1
Upatnieks, J., see Leith, E.N. 6, 1
Upstill, C., see Berry, M.V. 18, 257
Ushioda, S.: Light scattering spectroscopy of surface electromagnetic waves in
solids 19, 139
Vampouille, M., see Froehly, C. 20, 63

Van De Grind, W.A., see Bouman, M.A. 22, 77
van Haver, S., see Braat, J.J.M. 51, 349
Van Heel, A.C.S.: Modern alignment devices 1, 289
Van Kranendonk, J. and J.E. Sipe: Foundations of the macroscopic
electromagnetic theory of dielectric media 15, 245
Vanasse, G.A. and H. Sakai: Fourier spectroscopy 6, 259
Vartiainen, E.M., see Peiponen, K.-E. 37, 57
Vasnetsov, M.V., see Soskin, M.S. 42, 219
Velasco, A.V., P. Cheben, M. Florjanczyk, and M.L. Calvo: Spatial heterodyne
fourier-transform waveguide spectrometers 59, 159
Vernier, P.J.: Photoemission 14, 245
Vctor Torres-Company, Jesus Lancis, and Pedro Andres: Space-time analogies
in optics 56, 1
Visser, T.D., see Gbur, G. 55, 285
Vlad, V.I. and D. Malacara: Direct spatial reconstruction of optical phase from
phase-modulated images 33, 261
Vogel, W., see Welsch, D.-G. 39, 63
Volpian, O.D., see Shvartsburg, A.B. 60, 489
Vysloukh,V.A., see Kartashov, Y.V. 52, 63
Walmsley, I.A., see Raymer, M.G. 28, 181

Walmsley, I. A., see Juan P. Torres 56, 227

Wang Shaomin, and L. Ronchi: Principles and design of optical arrays 25, 279
Wang, B.C., see Glesk, I. 45, 53
Wang, T., see Zhao, D. 57, 261
Weber, M.J., see Riseberg, L.A. 14, 89
Weigelt, G.: Triple-correlation imaging in optical astronomy 29, 293
Weisbuch, C., see Benisty, H. 49, 177
Weiss, G.H., see Gandjbakhche, A.H. 34, 333
Welford, W.T.: Aberration theory of gratings and grating mountings 4, 241
Welford, W.T.: Aplanatism and isoplanatism 13, 267
Welford, W.T., see Bassett, I.M. 27, 161
Welsch, D.-G., W. Vogel and T. Opatrny: Homodyne detection and
quantum-state reconstruction 39, 63
Whitney, K.G., see Scully, M.O. 10, 89
Wilhelmi, B., see Schubert, M. 17, 163
Winston, R., see Bassett, I.M. 27, 161
Woerdman, J.P., see Spreeuw, R.J.C. 31, 263
Wolf, E.: The influence of Youngs interference experiment on the
development of statistical optics 50, 251
Woli
nski, T.R.: Polarimetric optical fibers and sensors 40, 1
Wolter, H.: On basic analogies and principal differences between optical and
electronic information 1, 155
Wuestner, S. and O. Hess: Active optical metamaterials 59, 1
Wynne, C.G.: Field correctors for astronomical telescopes 10, 137
Wyrowski, F., see Bryngdahl, O. 28, 1
Wyrowski, F., see Bryngdahl, O. 33, 389
Wyrowski, F., see Turunen, J. 40, 343
Xiao, M., see Joshi, A. 49, 97

Xu, L., see Glesk, I. 45, 53
Yan, M., W. Yan and M. Qiu: Invisibility cloaking by coordinate

transformation 52, 261
Yan, W., see Yan, M. 52, 261
Yamaguchi, I.: Fringe formations in deformation and vibration measurements
using laser light 22, 271
Yamaji, K.: Design of zoom lenses 6, 105
Yamamoto, T.: Coherence theory of source-size compensation in interference
microscopy 8, 295
Yamamoto, Y., S. Machida, S. Saito, N. Imoto, T. Yanagawa and M.
Kitagawa: Quantum mechanical limit in optical precision measurement and
communication 28, 87
Yanagawa, T., see Yamamoto, Y. 28, 87
Yaroslavsky, L.P.: The theory of optimal methods for localization of objects in
pictures 32, 145
Yeh, W.-H., see Carriere, J. 41, 97
Yin, J., W. Gao and Y. Zhu: Generation of dark hollow beams and their
Yoshinaga, H.: Recent developments in far infrared spectroscopic techniques 11, 77

Yu, F.T.S.: Principles of optical processing with partially coherent light 23, 221
Yu, F.T.S.: Optical neural networks: architecture, design and models 32, 61
Zalevsky, Z., D. Mendlovic and A.W. Lohmann: Optical systems with

improved resolving power 40, 271
Zalevsky, Z., see Lohmann, A.W. 38, 263
Zavatta, A., see Bellini, M. 55, 41
Zavorotny,V.U., see Charnotskii, M.I. 32, 203
Zavorotnyi,V.U., see Tatarskii, V.I. 18, 204
Zhao, D. and T. Wang: Direct and Inverse Problems in the Theory of Light
Scattering 57, 261
Zhu, R., see Mir, M. 57, 133
Zhu, Y., see Yin, J. 45, 119
Zubairy, M.S., see Greenberger, D.M. 50, 275
Zuidema, P., see Bouman, M.A. 22, 77

Prog Opt Vol 60

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Prog Opt Vol 60

Uploaded by

Copyright:

Available Formats

VOLUME SIXTY

G.S. Agarwal Stillwater, USA

Amsterdam Boston Heidelberg London New York Oxford

First edition 2015

Copyright 2015, Elsevier B.V. All rights reserved.

For information on all Elsevier publications

1. Quantum State Engineering: Generation of Single

2. Recent Advances in Optical Image Processing 119

3. Precursors and Dispersive Pulse Dynamics: A Century

3. Uniform Asymptotic Expansions 288

4. Quantum Limits in Optical Interferometry 345

5. Macroscopic Theory of Optical Momentum 437

6. Tunneling of Electromagnetic Waves in All-Dielectric

4. Tunneling of Microwaves Through Gradient TLs 529

Author Index for Volume 60 565

Quantum State Engineering:

Quantum Technologies Laboratory, SELEX-ES, Genoa, Italy

Progress in Optics, Volume 60 # 2015 Elsevier B.V. 1

Fourier-transformed linewidth. Much progress has been made in the years

where is h/2 with h Plancks constant, is the frequency (pulsation) of

~ajni njn 1i (6)

jni ~a{n =n!j0i (8)

The wave function (q) of the state j0i is

where o is the vacuum dielectric constant and V is the volume.

Eo 2=o V 1=2 n + 1=21=2 (15)

The phase of the wave is of course completely uncertain.

ji is an eigenstate of the annihilation operator a with eigenvalue , which

Vice versa n states may be expressed as a superposition of coherent states

3. THE PROBLEM OF LOCALIZING PHOTONS

relativistically invariant. Two observers in relative motion would not quite

4. ANTIBUNCHING OF SINGLE-PHOTON STATES

ptjt + =pt hI tI t + i=hI ti2 g2 (19)

g2 hntnt + i=hni2 pt=t + =pt (20)

In Equation (20), n(t) is the number of photons counted at time t.

The Hanbury Brown and Twiss experimental arrangement is shown in

An example of the result obtained when antibunching is present is shown

5. PHOTON STATISTICS AND SPECTRAL PURITY

where PN(W) is the probability distribution of the integrated intensity W

is the photodetection efficiency and I(t) is the intensity of light incident

n photons in a time interval t, t + T. Quantum mechanics arrives at an

Trf : W n =n! expW : PN W dW g

fhAigav TrA (25)

where the trace Tr of an operator O is defined as

the j is forming a set of states verifying the completeness relation

pm T hmim =1 + hmim + 1 (28)

Observation of sub-Poissonian photon statistics was first made by Short

6. THE PURCELL EFFECT AND THE CONTROL OF

7. PREPARATION OF SINGLE-PHOTON STATES:

previously specified quantum state by a series of elementary operations on

t cos o t=2je, 0i + sin o t=2jg, 1i (40)

The group at Max-Planck-Institute, Garching, had great experience

Quality factors Q of the order of 1010 at frequencies in the microwave region

In a second paper, the control, via an electric accelerating field, of the

Holland, Walls, and Zoller (1991) proposed a quantum nondemolition

A new method was proposed by Kuhn, Hennrich, Bondo, and Rempe

the one-photon quantum state can be reached. Later, Leonski (1996)

8. REALIZATION OF SINGLE-PHOTON SOURCES

is a necessary but not sufficient condition for a single-photon turnstile

8.1 Strongly Attenuated Sources

would be modified. Therefore, Taylor performed an experiment taking

pm hnim ehni =m! (42)

8.2 Single-Photon Sources Emitting one Photon in a

Single photons have been obtained at least in the following systems.