Quantum Aspects of Light Propagation

Antonı́n Lukš · Vlasta Peřinová

Quantum Aspects of Light


Antonı́n Lukš Vlasta Peřinová
Joint Laboratory of Optics Joint Laboratory of Optics
Palacký University and Palacký University and
Institute of Physics of the Czech Institute of Physics of the Czech
Academy of Sciences Academy of Sciences
772 07, Olomouc 772 07, Olomouc
Czech Republic Czech Republic
luks@prfnw.upol.cz perinova@prfnw.upol.cz

Consulting Editor
D.R. Vij
Kurukshetra University
E-5 University Campus
Kurukshetra 136119

ISBN 978-0-387-85589-9 e-ISBN 978-0-387-85590-5
DOI 10.1007/b101766
Springer Dordrecht Heidelberg London New York

Library of Congress Control Number: 2009930842 

c Springer Science+Business Media, LLC 2009
All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer
software, or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if
they are not identified as such, is not to be taken as an expression of opinion as to whether or not
they are subject to proprietary rights.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)


Quantum descriptions of light propagation frequently exhibit a replacement of time
by propagation distance. It seems to be natural since a propagation lasts some
amount of time. The primary intention was to inform more fundamentally inclined,
open-minded readers on this approach by this book. We have included also spatio-
temporal descriptions of the electromagnetic field in linear and nonlinear optical
media. We call some of these formalisms one dimensional (more exactly 1 + 1-
dimensional), even though they comprise the time variable along with the position
coordinate. These descriptions, however, are 3 + 1-dimensional in principle. The
rapid development of applications of photonic band-gap structures and experiments
on lasing in a disordered medium has directed us to pay attention even to these
topics, which has influenced the style of the book, which becomes a very review of
these streams.
This book has the following features. It reviews both macroscopic and micro-
scopic theories of the electromagnetic field in dielectrics. It takes into account para-
metric down-conversion experiments. It covers results on nonlinear optical couplers.
It includes optical imaging with nonclassical light. It expounds basics of quasimode
theory. It respects success of the Green-function approach in describing optical
field at dielectric devices, left-handed materials and the Casimir effect for some
geometries. It refers to quantization in waveguides, photonic crystals, disordered
media, and propagation in strongly scattering media, incoherent and coherent ran-
dom lasers, and important problems in optical resonators including chaotic cavities.
In our opinion it is appropriate to do something more than only formal comparison
of various approaches in the future, even though the reader will already have formed
an idea of their scope.
The simplest approach with one variable (time or propagation distance) and
with several frequencies has proven its vitality in the development of the quan-
tum information theory and the quantum computation. At present there exist even
books devoted to these fields: Alber, G., Beth, T., Horodecki, M., Horodecki, P.,
Horodecki, R., Rötteler, M., Weinfurter, H., Werner, R., and Zeilinger, A. (2001),
Quantum Information: An Introduction to Basic Theoretical Concepts and Experi-
ments, Springer-Verlag, Berlin; Nielsen, Michael A. and Chuang, Isaac L. (2000),
Quantum Computation and Quantum Information, Cambridge University Press,


vi Preface

The fundamental problem of light propagation in dielectric media is connected
with the role of nonclassical light in applications and has been pursued intensively
in quantum optics since about 1984. In the present book we review spatio-temporal
descriptions of the electromagnetic field in linear and nonlinear dielectric media
applying macroscopic and microscopic theories. We mainly pay attention to canoni-
cal quantum descriptions of light propagation in a nonlinear dispersionless dielectric
medium and linear and nonlinear dispersive dielectric media. These descriptions are
regularly simplified by a transition to the one-dimensional propagation, which is
illustrated also by descriptions of some optical processes.
Quantum theories of light propagation in optical media are generalized from
dielectric media to magnetodielectrics. Classical and nonclassical properties of radi-
ation propagating through left-handed media will be presented. The theory is uti-
lized for the quantum electrodynamical effects to be determined in periodic dielec-
tric structures which are known to be a basis of new schemes for lasing and a control
of light field state. Quantum descriptions of random lasers are provided.
It is an interesting question, to what extent the topic of this book overlaps
with the condensed-matter theory. Restricting ourselves to optical devices, we can-
not exclude such overlap in principle, because many of them are made of condensed
matters. The condensed-matter theory, however, is devoted mainly to problems of
conductors and semi-conductors. Photonic crystals can be studied similarly as ordi-
nary electronic crystals, even though for instance the conductivity is replaced by the
transmissivity. This does not mean any thematic overlap.
Texts on quantum optics have so far based the spatio-temporal description on the
quantization of the electromagnetic field in a free space in the hope that differences
from the field in a medium are negligible or can be easily included in other ways.
A rare exception was for instance the text Vogel, W. and Welsch, D.-G. (1994),
Lectures on Quantum Optics, Akademie Verlag, Berlin, where a choice of a suitable
approach, albeit a selection of one of possibilities, was declared.
The book will be useful to research workers in the field of general optics,
quantum optics and electronics, optoelectronics, and nonlinear optics, as well as
to students of physics, optics, optoelectronics, photonics, and optical engineering.

Olomouc Vlasta Peřinová
Olomouc Antonı́n Lukš


We have pleasure in thanking Dr. J. Peřina, Jr., Ph.D., for communicating files to
the publisher, graphics, and word processing and Ing. J. Křepelka, Ph.D., for the
careful preparation of figures. This book has arisen under the financial support by
the Ministry of Education of the Czech Republic in the framework of the project
No. 1M06002 “Optical structures, detection systems, and related technologies for
few-photon applications”.



1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Origin of Macroscopic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Lossless Nonlinear Dielectric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Nondispersive Lossless Linear Dielectric . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Quantization in Terms of a Dual Potential . . . . . . . . . . . . . . . 11
2.2.2 Momentum Operator as Translation Operator . . . . . . . . . . . . 13
2.2.3 Wave Functional Description of Gaussian States . . . . . . . . . 20
2.2.4 Source-Field Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.5 Continuum Frequency-Space Description . . . . . . . . . . . . . . . 31
2.3 Quantum Description of Experiments with Stationary Fields . . . . . . 36
2.3.1 Spatio-temporal Descriptions of Parametric
Down-Conversion Experiments . . . . . . . . . . . . . . . . . . . . . . . 38
2.3.2 From Coupled Quantum Harmonic Oscillators Back
to Interacting Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3 Macroscopic Theories and Their Applications . . . . . . . . . . . . . . . . . . . . . 85
3.1 Momentum-Operator Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.1.1 Temporal Modes and Their Application . . . . . . . . . . . . . . . . 86
3.1.2 Slowly Varying Amplitude Momentum Operator . . . . . . . . . 88
3.1.3 Space–Time Displacement Operators . . . . . . . . . . . . . . . . . . 102
3.1.4 Generator of Spatial Progression . . . . . . . . . . . . . . . . . . . . . . 104
3.1.5 Nonlinear Optical Couplers . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.2 Dispersive Nonlinear Dielectric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.2.1 Lagrangian of Narrow-Band Fields . . . . . . . . . . . . . . . . . . . . 117
3.2.2 Propagation in One Dimension and Applications . . . . . . . . . 126
3.3 Modes of Universe and Paraxial Quantum
Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.3.1 Quasimode Description of Spectrum of Squeezing . . . . . . . 133
3.3.2 Steady-State Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.3.3 Approximation of Slowly Varying Envelope . . . . . . . . . . . . . 143
3.3.4 Optical Imaging with Nonclassical Light . . . . . . . . . . . . . . . 152


x Contents

3.4 Optical Nonlinearity and Renormalization . . . . . . . . . . . . . . . . . . . . . 173
3.5 Quasimode Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
3.5.1 Relation to Quantum Scattering Theory . . . . . . . . . . . . . . . . 193
3.5.2 Mode Functions for Fabry–Perot Cavity . . . . . . . . . . . . . . . . 200
3.5.3 Atom–Field Interaction Within Cavity . . . . . . . . . . . . . . . . . . 207
3.5.4 Several Sets of Quasimodes . . . . . . . . . . . . . . . . . . . . . . . . . . 213

4 Microscopic Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
4.1 Method of Continua of Harmonic Oscillators . . . . . . . . . . . . . . . . . . . 223
4.1.1 Dispersive Lossy Homogeneous Linear Dielectric . . . . . . . . 224
4.1.2 Correlation of Ground-State Fluctuations . . . . . . . . . . . . . . . 235
4.2 Green-Function Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4.2.1 Dispersive Lossy Linear Inhomogeneous Dielectric . . . . . . 239
4.2.2 Dispersive Lossy Nonlinear Inhomogeneous
Dielectric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
4.2.3 Elaboration of Linear Theory . . . . . . . . . . . . . . . . . . . . . . . . . 245
4.2.4 Optical Field at Dielectric Devices . . . . . . . . . . . . . . . . . . . . . 253
4.2.5 Modification of Spontaneous Emission by
Dielectric Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
4.2.6 Left-Handed Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.2.7 Application to Casimir Effect . . . . . . . . . . . . . . . . . . . . . . . . . 280

5 Microscopic Models as Related to Macroscopic Descriptions . . . . . . . . 303
5.1 Quantum Optics in Oscillator Media . . . . . . . . . . . . . . . . . . . . . . . . . . 303
5.2 Problem of Macroscopic Averages . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
5.2.1 Conservative Oscillator Medium . . . . . . . . . . . . . . . . . . . . . . 305
5.2.2 Kramers–Kronig Dielectric . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.2.3 Dissipative Oscillator Medium . . . . . . . . . . . . . . . . . . . . . . . . 312
5.3 Single-Photon Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317

6 Periodic and Disordered Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
6.1 Quantization in Periodic Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
6.1.1 Classical Description of Electromagnetic Field . . . . . . . . . . 322
6.1.2 Modal Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
6.1.3 Method of Coupled Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
6.1.4 Normalized Modes of the Electromagnetic Field . . . . . . . . . 334
6.1.5 Quantization in Linear Nonhomogeneous
Nonconducting Medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
6.2 Corrugated Waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
6.2.1 Lossless Propagation in a Waveguide Structure . . . . . . . . . . 351
6.2.2 Coupled-Mode Theory Including Gain or Losses . . . . . . . . . 359
6.3 Photonic Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
6.4 Quantization in Disordered Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
6.4.1 Quantization in Chaotic Cavity . . . . . . . . . . . . . . . . . . . . . . . . 394
6.4.2 Open Systems Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398

. . . . . . . . . . . . . . . . . . . .3 Modal Decomposition in Optical Resonators . . . . . . . . . 445 7 Conclusions . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . . . . .3 Semiclassical Approach . . . . . . . . . . . 414 6. .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . 457 Index . . . . . . . . . . . . 408 6. 453 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . 475 . . . . . . . .2 Incoherent and Coherent Random Lasers . . . . . . . . . . . . . . . . . . . . . . . .4 Chaotic Resonators . . . . . . . . . . . . . . 444 6. . . . . . . . . . . . . . . . . . . . .5 Propagation in Amplifying Random Media .4. . . . . . . . . . . . . . . . . . . . 404 6. . . . 408 6. . . . . . . . . . . . . . . . . . . . . . .Contents xi 6. . . . . . . . . . . . . . . . . . . . .1 Strongly Scattering Media . . . . . . . . . . . . . .

In quantum optics. In the first. Quantization is accomplished by imposing the standard equal-time commutation relations.Chapter 1 Introduction The importance of quantum optics has been recognized by both specialists and pub- lic since Roy J. In the second approach. Peřinová. the microscopic. the normal-mode expansion approach is used that is well suited for systems in optical cavities. the Maxwell equations com- pleted by the constitutive relations are solved with the method of slowly varying envelope approximation and the resultant equations are sometimes simplified on the assumption of parametric approximation. The quantization of the electromagnetic field in the presence of a dielectric is possible.1007/b101766 1. After a Lagrangian which produces the macroscopic Maxwell equations for the field in a nonlinear medium is found. the problem of quantization of the electromag- netic field in vacuo was solved by Dirac (1927) long ago and the quantization of a nonlinear theory is due to Born and Infeld (1934. 1935). such as an optical parametric oscil- lator. First it resembled some dissatisfaction with the situation following the advent of laser in 1958. the macroscopic approach. The result is a A. the canonical momenta and the Hamiltonian are derived. 1 DOI 10. From the historical viewpoint. Quantum Aspects of Light Propagation. Shen 1984). With respect to the propagation in linear dielectric media it is appropriate to refer first to Jauch and Watson (1948).  C Springer Science+Business Media. In nonlinear optics (Bloembergen 1965. but is not appropriate for open systems such as a parametric amplifier. It has become standard that the phe- nomenological Hamiltonians of quantum optics are frequently introduced without a quantitative connection to the classical equations describing the nonlinear optical effects. Quantum informatics is closely connected with this field. Ingenious. Glauber was awarded the Nobel Prize in Physics 2005. Lukš. the medium is com- pletely described by its linear and nonlinear susceptibilities. but simple solutions are preferred to intricacies of the quantized field theory with the hope that experimenters realize the simple proposals with appropriate means. The new optical effects are analyzed both by the methods of non- linear optics which belong to classical physics and by those of quantum optics (Shen 1969). A revived interest in this problem can be perceived since the 1990s. No matter degrees of freedom appear explicitly in this treatment. a model for the medium is constructed and both the field and the matter degrees of freedom appear in the theory. LLC 2009 . This can be done in two ways which are called the macroscopic and microscopic approaches. Both are quantized. V.

The formalism of the macroscopic approach to the quantization has been developed (Abram and Cohen 1991). The optical solitons have been studied in the nonlinear optics and their quan- tum properties have been calculated using spatio-temporal descriptions by erudite authors. in which the momentum operator is used along with the Hamiltonian. The vacuum propagation and low-order perturbation theory have sufficed for spatio-temporal descriptions of para- metric down-conversion experiments (Casado et al. which are coupled by a nonlinear interaction. shifts. 2003). The macroscopic quantization of the electromagnetic field was applied to inhomogeneous media (Glauber and Lewenstein 1991). shifts. In the past. They have pointed out that there is a difficulty in including the dispersion in the quantized macroscopic theory. The problem of a proper quantum mechanical description of the operation of optical devices has been addressed (Knöll et al. Besides this. 1987). The experiment on the “induced coherence without induced emission” has been described on restriction to spatial behaviour of fields and the multimode description has been restored too (Peřinová et al. Jr. have been developed (Toren and Ben- Aryeh 1994). we will use the term space progression instead of space evolution. Accordingly. The space–time displacement operators have been related to the elements of the energy–momentum tensor (Serulnik and Ben-Aryeh 1991). and translations of the elec- tromagnetic field along with the time displacements. and translations. where the restriction to merely spatial behaviour of interesting optical fields is accepted. 1997a. Hillery and Mlodinow (1984) have used the electric displacement field as the canonical variable for nonlinear quantization and they have explored the macro- scopic approach to the quantization of homogeneous nondispersive media. He has based the for- malism on the momentum operator for the radiation field and investigated in this way not only the spatial progression of the electromagnetic wave but also refraction and reflection. The excellent review of linear and nonlinear couplers (Peřina. 1990).2 1 Introduction theory of mixed matter-field (polariton) modes. The dispersion has been treated on the assumption of a narrow-frequency interval (Drummond 1990). Abram (1987) intended to overcome the difficulties of the con- ventional quantum optics by reformulating its assumptions. Casado et al. 1997b). and Peřina 2000). The theoretical methods for investigating propagation in quantum optics. In the following. an attempt at a formula- tion of quantum theory of propagation of the optical wave in a lossless dispersive dielectric material has been made (Blow et al. many authors that dealt with macroscopic quantum theories of light propagation wrote also on space displacements. 1990). they used the term “space evolution” in the former case. Drummond (1994) has presented a review of his theory . or sim- ply on the (time) evolution. The importance of a proper space–time description of squeezing has been recognized (Bialynicka-Birula and Bialynicki-Birula 1987). 1986. has used a similar approach. The applications have used the fact that the nonlinear processes of quantum optics are described quantum optically in the parametric approximation with linear mathematical tools so that quantization procedures and solutions of the dynamics need not face immense difficulties as for the really nonlinear formalism (Huttner et al.

c). This scheme is based on the Hopfield model of such a dielectric. Gruner and Welsch (1995) have calculated the ground-state correlation of the quantum-mechanical fluctuations of the intensity. The input and output operators that are related via scattering operator are directly linked to multi-time quantum corre- lation functions (Dalton et al. 1998) devoted to the concept of quasinormal modes. 1999b. the modes of the universe have been used in the treatment of the spectrum of squeezing (Gea-Banacloche et al. In the case of a . Gruner and Welsch (1996a) have realized the expansion of the field operators which is based on the Green function of the classical Maxwell equations and preserves the equal- time canonical commutation relations of the field. A quantum scattering theory approach to quantum-optical measurements has been expounded (Dalton et al. In addition to (Lang et al. 1999b). (1998) have developed a quantiza- tion scheme for the electromagnetic field in a spatially varying three-dimensional linear dielectric which causes both dispersion and absorption. the microscopic approach to the quantum theory of light propagation has been extended to nonlinear media and the generalized nonlinear Schrödinger equa- tion well known from the description of quantum solitons has been derived for a dielectric with a Kerr nonlinearity. Following (Huttner and Barnett 1992a. 1998). 1999a).1 Introduction 3 and its applications. In (Schmidt et al. where the matter is represented by a harmonic polarization field (Hopfield 1958).b). Barnett. 1999c). It has taken into account the existence of small photodetectors or pixels (Kolobov 1999).b). In addition to previous work (Lang et al. 1973) and along with an independent work (Ho et al. 1990a. Abram and Cohen (1994) have developed a travelling-wave formulation of the theory of quan- tum optics and have applied it to quantum propagation of light in a Kerr medium. Brown and Dalton (2001a) have generalized the quasimode theory of macroscopic quantization in quantum optics and cavity quantum electrodynamics developed by Dalton. One-dimensional description of beam propa- gation has been completed with transverse position coordinates. Huttner and Barnett (1992a. The relationship between the pure mode and quasi- mode annihilation and creation operators is determined (Dalton et al. and Knight (1999a. They have found that the spatial progression can be derived on the assumption of weak absorption. 1999d). quasimode theory of macroscopic canonical quantization has been invented and applied (Dalton et al. based on expanding the vector potential in terms of quasimode functions has been carried out (Dalton et al.b) have presented a fully canonical quantization scheme for the electromagnetic field in dispersive and lossy linear dielectrics. The generalized form of quasimode theory has beeen applied to provide a fully quantum-theoretical derivation of the laws of reflection and refraction at a boundary (Brown and Dalton 2001b). This generalization admits the case where two or more quasipermittivities are introduced. A macroscopic canonical quantization of the electro- magnetic field and radiating atom system involving classical. 1973) devoted to the concept of quasinormal modes. Dung et al. linear optical devices. A quantum theory of the lossless beam splitter is given in terms of the quasimode theory of macroscopic canonical quantization. An original approach to the description of a degenerate parametric amplifier (Deutsch and Garrison 1991a) has been related to the theory of paraxial quantum propaga- tion (Deutsch and Garrison 1991b).b).

Spontaneous parametric down conversion in a finite-length mul- tilayer structure has been considered (Centini et al. Both localization and laser theory. Sakoda (2002) has formulated quantization of the electromagnetic field in photonic crystals. Nonlinear optics in a photonic band-gap structure has been studied (Tricca et al. A microscopic theory of an optical field in a lossy linear optical medium has been developed (Knöll and Leonhardt 1992). which were developed in the 1960s. quantum theory is needed. Yablonovitch (1987) suggested that three-dimensional periodic dielectric struc- tures could have a photonic band gap in analogy to electronic band gaps in semicon- ductor crystals. (1996) have dealt with the quantization of a field in dielectrics and have applied it to the theory of atomic radiation in one-dimensional Fabry–Pérot resonator. The idea of one-dimensional propagation may be compared with results concerning a mirror waveguide. As soon as quantization in nonhomogeneous dielectric media is solved. Jr. This idea and its subsequent experimental proof in the macrowave domain have led to extensive activity aimed at the optimization of photonic band-gap structures for the visible domain and the exploration of their potential applications (Journal of Modern Optics 1994. 2007). 2006). Peřina. 2000). The Green function has also been used in the more complicated case of two dielectric bod- ies with a common planar interface. a band of frequencies for which an electromagnetic wave cannot propagate in any direction. 2004. Lasing in disordered media has been a subject of intense theoret- ical and experimental studies. et al. Journal of the Optical Soci- ety of America B 2002.). Spontaneous decay of an excited atom in the presence of dispersing and absorbing bodies has been investigated using an exten- sion of this formalism (Dung et al. although the field theory in linear dielectrics should have such a property. Random lasers have been classified into incoherent and coherent random lasers. 2004. They were used in strongly scatter- ing gain media. not only the case of a finite dielectric medium is worth a treatment but also the case of an infinite periodic medium (Caticha and Caticha 1992. the well-known Green function has been used and it has been shown that the indicated quantization scheme exactly preserves the funda- mental equal-time commutation relations of quantum electrodynamics. Peřina. 2005. the character of lasing modes depends on the amount of disorder. They have shown that macroscopic averaging of the dynamical variables can lead to a macroscopic description.4 1 Introduction homogeneous dielectric. 2005. In a random medium. Dutra and Furuya (1997) have considered a single-mode cavity filled with a medium consisting of two-level atoms that are approximated by harmonic oscillators. namely. Dalton et al. were jointly applied in the study of random laser. Weak disorder leads to a poor . et al. Peřina.b) have observed that the (full) Huttner–Barnett model of a dielectric medium does not comprise all the dielectric permittivities of the medium which can be expected from the classical electrody- namics. Jr. Tip 1997). In order to understand quantum-statistical properties of random lasers. Research works on both types of random lasers have been summarized in the monographic chapter (Cao 2003). etc. Dutra and Furuya (1998a. et al. Standard quantum theory for lasers applies only to quasidiscrete modes and cannot account for lasing in the presence of overlapping modes. Kweon and Lawandy 1995. Jr.

1 Introduction 5 confinement of light and to strongly overlapping modes. (3.1)–(2. 2000.322) are in the Gaussian units.323)–(3.25)–(2. which is restricted to linear media.2). and the relations (2. Statistics naturally belongs to the theory of amplifying random media (Beenakker 1998.14)–(3.494)– (3. Cheng and Siegman (2003) have derived a generalized formalism of radiation-field quantization.392) are in the rationalized cgs units. so some of the relations (2.578). (3.109)–(3. True eigenmodes of a system will be non-orthogonal and the method is intended for quantization of an open system. .108). and (2. (2002) have developed a quantization scheme for optical resonators with overlapping (nonorthogonal) modes. there are exceptions. in which a gain or loss medium is involved. Patra and Beenakker 1999.69) and (3.24) are in the Heaviside–Lorentz units. the relations (3. (3. 2001).15)–(2. We will use units following original papers and although the system of interna- tional (SI) units prevails. Hackenbroich et al. Mishchenko et al.276)–(3.125). which need not rely on a set of orthogonal eigenmodes.

1969). Macroscopic approaches to quantization of the electromagnetic field are not funda- mental theories and modify the free-space electromagnetic-field theory. A similar travelling-wave description for the electromagnetic field–matter interaction was considered to be possible in terms of a virtual cavity and a momentum operator of the field.1 Lossless Nonlinear Dielectric An approach to the quantum theory of light propagation was considered stan- dard until the critique by Hillery and Mlodinow (1984) and is still.1007/b101766 2. Concerning this approach. 2. 7 DOI 10. Especially. With the progress in (classical) optics interest in the quantization of the field power in quantum optics has increased. Efforts emerged to formulate so simple a quantum theory of the electromagnetic field that it allows one to recognize the role of the momentum operator. Quantum Aspects of Light Propagation. This approach to quantization was rather distant from the quantum theory of the electromagnetic field. Shen (1967) studied A. LLC 2009 . Although the virtual cavity has been beaten. V. with which the radiation is transformed. Lukš. quantization of the field power has been assumed. to the contrary. both passive and active. Peřinová. On a fundamental level the theory of the electromagnetic field in the free space does not differ from the theory of this field in the matter. Not always is it necessary to utilize the formalism of the electromagnetic field in the matter.  C Springer Science+Business Media. let us consider papers by Shen (1967. the momentum operator has still enabled one to study quantum aspects of nonlinear optical processes. did not consider the momentum operator. Quantization restrictions of any kind such as the frequency dispersion of the refractive index were apparent on published work. Formalisms were presented which. For description of experiments with correlated photons it suffices to describe the electromagnetic field between optical devices and to know the input– output relations for the optical elements.Chapter 2 Origin of Macroscopic Approach With the birth of quantum optics in the 1960s it became clear that it would be easy to describe the interaction between the electromagnetic field and the matter in a cavity even on elimination of matter degrees of freedom.

Quantum statistics has been determined using descriptions suited to the case of a cavity. the cavity treatment of the problems of light propagation in media seems to be valid on the following assumption. For investigation of properties of a medium. (2. Weak nonlinearity has a significant effect on light only after a longer interaction distance. the same treatment can be applied to problems of light propagation in media (Shen 1967). which moves in the z direction with a light velocity c ( √c in Shen 1969). Here the dependence on the time seems to be more fundamental. This case should be treated by the method of many-body transport theory (Ter Haar 1961). k = (ωk ) is the value of the dielectric function  at ωk . He contributed to the contemporary research (Glauber 1965). In quantum optics. Then a cavity problem can be converted to a corresponding steady-state propagation problem by replacing t by − cz in the field amplitudes and the latter problem can be changed to the former one by replacing z by −ct when ez is the direction of propagation. and associating the discrete values of the wave vector k with d (instead of L). The finite medium can be extended to infinity.c. Light can cover a longer distance easily when contained in a cavity res- onator. t) = c b̂k (z) exp[−i(ωk t − kz)] + H. it becomes difficult. An appropriate compo- nent of the vector-potential operator has the expansion of the form      Â(z. It raises expectations that the same√ is true in the quantum treatment. where T is the counting time of photodetectors. A partial interaction of the light with the medium can be approximated with no interaction and a complete interaction. The operators will be space dependent (localized) instead of time dependent. In nonlinear optics a number of classical descriptions have been developed both as a cavity problem and as a steady-state propagation problem. and H. incoherent scattering has been a useful tool. One is advised to imagine a box of length cT . It is evident that one is interested in a conversion of a cavity problem to a corresponding steady-state propagation problem. Quantum theory of radiation had long been formulated (Heitler 1954). .c. coherent scattering has been interesting as well or more. which lasts for a time t. denotes the term Hermitian .8 2 Origin of Macroscopic Approach quantum statistics of nonlinear optics. the localized annihilation † and creation operators b̂k (z) and b̂k (z) have been proposed. But for coherent scattering. For nonlinear optics. On quantizing in a volume L 3 and assuming that the field does not vary appre- ciably over a distance d large compared with the wavelength. Shen (1969) pays attention to replacing t by − z c  and to replacing z by √ct . The resultant change of statistical properties of fields in the box can now be calculated using the cavity treatment (Shen 1967). Transformations will be generated with a localized momentum operator instead of the Hamiltonian operator. Photon fields may be quantized in a box of finite volume.  is the Planck constant divided by 2π .1) k 2ωk k L 3 where ωk is the frequency. In principle.

respectively. (2. There is a difficulty. (2. The localized photon-number operator is realized as a configuration- space photon-number operator (Mandel 1966) Ad  † n̂(z) = b̂ (z)b̂k (z). The Hamiltonian is Ĥ(t) = L 2 Ĥ (z  . b̂k  (z) = δkk  1̂. A Hamiltonian density Ĥ (z. satisfy the equal-space commutation relation  †  b̂k (z). the localized momentum operator is  . with c ≡ c0 . the free-space speed of light. not by using an integration with respect to time. Here it is replaced by the phase velocity √ck .2.2) The smallvariation of the field has been formulated as that of the normally ordered †m moments b̂k (z)b̂kn (z) . It has been assumed that k = 2πnd . t) dz  . where n is an integer. For free fields. t) is considered.4) L A third difficulty is that the localized momentum operator is defined as Ĥ (z. The annihilation and creation operators b̂k (z) and † b̂k (z). It is also specified that k = 2πd n .3) L3 k k where A is the cross-sectional area of the beam. (2. The above picture of a moving box requires a light velocity c independent of the frequency ωk . There is another difficulty in view of this picture that d has been used instead of cT .t) c essen- tially.1 Lossless Nonlinear Dielectric 9 conjugate to the previous one. Shen (1969) utilizes the notation √c for this velocity. not that ωk = 2πT n .

t).7) c ∂t . but in fact one does not utilize this. P̂(z) . but with b̂k (z) and b̂k (z) replacing âk (t) and âk (t). (2. (2. the localized momentum operator has the form of a Hamilto- † † nian.6) dz  The electric strength vector is derived from the vector potential according to the relation 1 ∂ Ê(z. The momentum operator generates translations d i   b̂k (z) = b̂k (z).5) k For interacting fields. be a vector. A momentum operator should have the form ez P̂(z). t) = − Â(z. † P̂(z) = k b̂k (z)b̂k (z) + 12 1̂ . (2.

the unitary translation operator is z  i Û (z. t)Û (z. t1 ) .13) The equation of motion for a statistical operator ρ̂(z) is ∂ i   ρ̂(z) = P̂(z). Ê (+) (z. t) = Ê (z. the calculations for steady-state propa- gation in a medium become the same as the corresponding calculations for a cavity . t1 ) . (2. Then the correlation function of fields at different times is expressed in two forms:   Ê (−) (z. (2. tn ) Ê (+) (z. . tn ) . . . Ê (−) (0. ρ̂(z) . . . . t)) contains the functions exp(−iωk t) (exp(iωk t)). In Shen (1967) the opposite convention is used. Ê (−) (z. t) + Ê (−) (z. z 0 ). t) ( Ê (−) (z. Such an indication is the fact that the localized momentum operator has been derived from the Hamiltonian density. . (2. z 0 ) Ê(z 0 . The space-ordered product has a similar definition as the time-ordered product. (2.11) There are indications that any “alternative” quantum theory is avoided. tn ) . t1 ) . a localized density matrix (statistical operator) progresses: ρ̂(z) = Û (z. t) = Û † (z. t).9). . Field operators at different spatial points z. Ê (+) (0. Ê (−) (z. (2. . (2. we pass from the “spatial Heisenberg picture” to a spatial Schrödinger picture. tn ) Ê (+) (0. According to equations (2.9) dz  Something is more suitable for propagation problem: We define all the quantities at a given plane z = z 0 for all times and try to obtain the propagation towards z ≥ z 0 . P̂(z) . Then d (+) i  (+)  Ê (z. . Ê (+) (z. .10 2 Origin of Macroscopic Approach We decompose this operator as Ê(z. . t1 )   = Tr ρ̂(z) Ê (−) (0. With this in mind.8) where Ê (+) (z.6) and (2. t1 )   = Tr ρ̂(0) Ê (−) (z. t1 ) . tn ) . 0). (2. t) = Ê (+) (z.12) Here ρ̂(0) is a given statistical operator.10)  z0 where S is the space-ordering operation. tn ) Ê (+) (z. z 0 ) = S exp P̂(z  ) dz  . t). 0)ρ̂(0)Û † (z.14) ∂z  With the help of these localized operators. In the latter picture. z 0 are connected by this unitary operator: Ê(z.

2 Nondispersive Lossless Linear Dielectric The study of nonlinear optical phenomena and their inclusion in an effective non- linear theory of the electromagnetic field has utilized the asymmetry of most optical media. The canonical momentum should be the magnetic induction in place of the more usual electric-field strength. this macroscopic classical theory is quantized. The problem of beam splitting was mentioned. 2. Hillery and Mlodinow (1984) have pointed out some problems with the standard theory. (2. Such a theory may not be capable of describing the Bohm–Aharonov effect. a macroscopic theory of the quantized electromagnetic field in a medium can be very close to the usual theory of this field in free space.15) .1 Quantization in Terms of a Dual Potential According to a pioneering paper of Hillery and Mlodinow (1984). which include the dispersion and the nonlinearity at least approximately. but linear relative to the magnetic field. the same proposal has been included in Shen (1969). and to keep interpreting the electric-field (up to the sign) as the canonically conjugated variable to the vector potential. the standard macroscopic quantum theory of electrodynamics in a nonlinear medium is due to Shen (1967) and has been elaborated upon by Tucker and Walls (1969). On neglecting dispersion and nonlinearity. A description of the field distribution in space must be completed with a quantum state of the field in quantum physics.2. Then. solutions have been disseminated. which are nonlinear with respect to the electric-field. The Hamiltonian formulation of the theory consists in the noncanonical Hamiltonian Ĥnoncan = ĤEM + ĤInoncan . Besides such a theory we expound a simple quantization connected to considerations of the role of the Poynting vector operator and the momentum operator.2.2 Nondispersive Lossless Linear Dielectric 11 √ with t replaced by − cz (Shen 1967) and by z c  (Shen 1969). The other approach is to take the expression for the energy of the radiation in nonlinear medium. 2. above all that it is not consistent with the macroscopic Maxwell equations. which differs from the free-field Hamiltonian in part. One approach to the derivation of a macroscopic quantum theory would be to begin from a quantum microscopic theory as explored in the linear case by Hopfield (1958). In contrast to this. Essentially. A renewed interest in the spatio-temporal description leads to the study of the wave functional of the electromagnetic field des- pite the doubts of the pioneers of theoretical physics of the photonic wave function. (Let us note that it differs from Shen (1969)).

18) is replaced by ∂Λ D = ∇ × Λ. (2. A Lagrangian is considered which gives proper equations of motion. The momentum canonical to A is Π = (Π0 . nondispersive. It has been shown how to utilize the Dirac quantization procedure for constrained Hamiltonian systems (Dirac 1964).21) ∂t It can be shown that the canonical momentum is Π × = B. The vanishing of Π0 indicates that the system is constrained. (2. A).20) 0 In order to simplify the quantization of the macroscopic Maxwell theory.17) 2 with Ê being the electric field strength operator and P̂ being the polarization of the medium. The canonical Hamiltonian has the form H = HEM + HI . The electric and magnetic fields are expressed in terms of the vector potential A and the scalar poten- tial A0 : ∂A E=− − ∇ A0 . we obtain an improper expression for the time derivative of the magnetic-induction field B̂. (2. The relation (2. It is assumed that the medium is lossless. the dual potential Λ has been introduced along with Λ and Λ0 .12 2 Origin of Macroscopic Approach where 1 ĤEM = (Ê2 + B̂2 ) d3 x. (2. which we call the dual vector and scalar potentials.18) ∂t The appropriate Lagrangian density depends on the first partial derivatives of the four-vector A = ( A0 . This theory may be called standard. B = + ∇Λ0 . as an undesirable “quantum effect”. The polarization is a function of the electric field which may be written as a power series. (2. B = ∇ × A. It can easily be seen that. It can be derived that the canonical momentum is Π = −D. and the Heaviside–Lorentz units having been used. Upon expressing the canonical Hamiltonian functional in terms of the electric-displacement and . Π). where Π0 = 0.19) where 1  HI = E· P− P(λE) dλ d3 x. (2. and homogeneous.16) 2 1 ĤInoncan = Ê · P̂ d3 x.

the usual Hamiltonian theory for the electromagnetic field in a nonlinear dielectric medium and the alternative have been quantized in the ordinary way. mainly to a change of the frequencies of the modes which does not occur. (2. and V is a quantization volume.23) with   Λ̂i (x.2 Momentum Operator as Translation Operator In the late 1980s. we expound the main ideas of Abram (1987). Abram criticized the modal Hamiltonian formalism. suggests . 2. t) = iδi⊥j (x − x )1̂. (2. the problem of propagation did not seem to be typical of quantum optics.24) The transverse δ function has been used and made a reference to Bjorken and Drell (1965). It is added that different diagonalizations indicated by the quadratic part of the total Hamiltonian generate different kinds of normal ordering. (2. Dispersion is also considered a reason for a microscopic theory to be contemplated. Π̂×j (x . χ is the (linear) susceptibility of the material. Although this model can be an appropriate limit of the Huttner–Barnett model. t) = iδi⊥j (x − x )1̂.2 Nondispersive Lossless Linear Dielectric 13 magnetic-induction fields. being aware that they require the consideration of propagation. Hillery and Mlodinow (1984) do not mention propagation except a paragraph on the interpretation problems. t). He decided to extend the traditional theory of quantum optics to describe propagation phenomena without invoking the modal Hamilto- nian. t). A doubt is expressed that there is an appropriate kind and the microscopic approach is pro- pounded. We can compare   Âi (x.25) 8π V where E (H ) is the magnitude of the electrical (magnetic) field strength. (2. where they recommend to confine the medium to part of the quantization volume and to place the field source and the detector outside of the medium.2. especially the inclusion of the linear polarization term in the Hamiltonian: ? 1 H= (E 2 + H 2 + 4π χ E 2 ) dV. refraction. This would lead to an incorrect result. the results are the same: H = H ×. Π̂ j (x .2. Abram addressed the problem of light propagation through a linear nondis- persive lossless medium (Abram 1987). According to him one of the propagation phenomena.22) Then.

with the electric field polarized along the x-axis and the magnetic field along the y-axis. independent of frequency (no dispersion).28) ∇ · D = 0. We shall further assume that light is propagating in a linear dielectric.32) The change in the total energy which is given by the integrated energy flux (the Poynting vector) over the surface of a body or volume is proposed in Abram (1987) as the proper quantum-mechanical Hamiltonian.14 2 Origin of Macroscopic Approach the momentum as the concept appropriate for the description of these phenomena. The momentum is treated . (2. (2.27) c ∂t ∇ · B = 0. (2. The change in the total momentum is given as the integrated flux of the Maxwell stress tensor. (2. (2. √ n= . we shall consider only the case of plane waves propagating along the z-axis. (2.30) where we assume the susceptibility of the material for simplicity to be a scalar (neglecting its tensorial properties). Quantum mechanically. (2. so that B = H. B is the magnetic induction. We assume that there are no free charges or currents and that we are dealing with nonmagnetic materials. respectively.26) c ∂t 1 ∂B ∇ ×E=− .31) and the refractive index n. where the induced polarization is at all times proportional to the incident electric field: P = χ E.29) where D = E + 4π P is the electric displacement. Propagation of the electromagnetic field is described by the Maxwell equations: 1 ∂D ∇ ×H= . For simplicity. the directions of all vectors being implicit. It is convenient to define also the dielectric function  of the material  = 1 + 4π χ . E and H are the electric and magnetic field strengths. and c is the speed of light. These are a good antidote against the idea that “space and momentum are canonically conjugate variables like time and energy”. Let us remark that microscopic models demonstrate that a Hamiltonian including light–matter interaction can be considered. P is the (linear and non- linear) polarization induced in the medium. space and momentum are canonically conjugate variables. This reduces the Maxwell equations to scalar differential equa- tions.

In volume V  the wavelengths of the modes become λ = λn . (2.25) is solved. The total energy is conserved. t) = − Â(z. It is interesting that in the absence of reflection.   2π  2  †  1 Â(z. It is convenient to rearrange equation (2.37) j V . t) is usually written as ( = 1)   2π  2  †  1 Â(z.33) n √ HT = n HI . the electromagnetic vector-potential operator  ≡ Â(z. t) = c â j eiω j t + â− j e−iω j t e−ik j z . (2.2. In propagation along the z-axis the Maxwell stress tensor is replaced by the energy density. but the oscillator frequencies remain unchanged.  = 1). t) = ê j c ∂t j   2π ω j  2  1 †  = −i b̂ j − b̂− j . When the propagation along the ±z-axis in free space is considered with the elec- tric field polarized along the x-axis and the magnetic field along the y-axis (χ = 0. the enigma of the Hamiltonian (2. t) = c â j eiω j t−ik j z + â j e−iω j t+ik j z . respectively. To simplify the notation. we omit unit vectors. because the volume V reduces to V  = Vn . the components of which are always proportional to the wave vectors of the excited modes. (2.36) j V ωj The electric and magnetic field operators may be obtained as 1 ∂  Ê(z. â j are the creation. annihilation operators. (2. (2. the electric and magnetic fields of the transmitted (T ) waves in the dielectric are related to the corresponding incident (I ) fields in free space by 1 ET = √ E I .35) in a manner that is familiar to solid-state physicists. but the energy density is increased by a factor of n. However. for a photon in the jth mode of the wave vector k j (with k− j = −k j ) and the frequency ω j = c|k j | fulfilling the Bose commutation relations.2 Nondispersive Lossless Linear Dielectric 15 on the same footing as the Hamiltonian.35) j V ωj † where â j . We may consider a square pulse which enters a dielectric.34) This change in the energy density implies a similar increase for the total momentum of the pulse.

42) 8π  = û j (2.45) V j 2 The energy fluxes due to the forward (backward) waves alone can be expressed uniquely:  ωj  † 1   ωj  † 1  û + = b̂ j b̂ j + 1̂ . t) = Â(z. (2.43) j 1   = ê j ê− j + ĥ j ĥ − j 8π j 1   † †  = ω j b̂ j b̂ j + b̂− j b̂− j + 1̂ (2. t) = ĥ j ∂z j    12 2π ω j  †  = −i sj b̂ j + b̂− j . The Hermiticity of the operators Ê ≡ Ê(z. (2.40) † ĥ j = ĥ − j .38) j V where s j ≡ sgn j and b̂ j = â j e−iω j t+ik j z . t) can be verified using the relations † ê j = ê− j . (2. (2.46) j(>0) V 2 j(<0) V 2 . we suppose that they are sym- metrized.39) When products of these operators are encountered. (2.44) 2V j   1  † 1 = ω j b̂ j b̂ j + 1̂ .16 2 Origin of Macroscopic Approach and ∂  Ĥ (z. (2. û − = b̂ j b̂ j + 1̂ .41) The energy density operator can be written as 1  2  û = Ê + Ĥ 2 (2. t) and Ĥ ≡ Ĥ (z.

51) as appropriate for any operator Q̂(z. fades.9) and (3. The simi- larity with equation (3. Let us consider Q̂ ≡ Q̂(z. We may calculate the Poynting vector operator as c c  Ŝ = Ê Ĥ = ê j ĥ − j (2. t) = Q[ Ê(z. Ŝ− = − b̂ j b̂ j + 1̂ .52) 4π 4π j  cω j  † 1  = sj b̂ j b̂ j + 1̂ .10) in Abram (1987) well. •].49) ∂z where Q̂ is any operator.54) j(>0) V 2 j(<0) V 2 .50) ∂ where Q[•.49) for Q̂ = Ê. t)]. the Heisenberg equation of motion.47) c j 2 It is important to understand the relations (3. it suffices to verify the relation (2. Q̂]. 0) = â j (2. (2. (2. (2. we perceive that the operators do not obey our definition of the operator Q̂.48) ∂z where |z are position coordinate states and |ψ is an arbitrary pure state. (2. (2.2 Nondispersive Lossless Linear Dielectric 17 The total momentum operator Ĝ is then V   † 1  Ĝ = (û + − û − ) = k j b̂ j b̂ j + 1̂ . Although the operators b̂ j ≡ b̂ j (z. •] is a formal series in Ê and Ĥ . We interpret (3. Since the differential operator ∂z is just as differentiation as the superoperator −i[Ĝ. It is true at least in the situations treated in Abram (1987).10) from Abram (1987) ∂ Q̂ = −i[Ĝ.49).53) j V 2 The Poynting vector operators due to the forward (backward) waves alone can be expressed uniquely:  cω j  † 1   cω j  † 1  Ŝ+ = b̂ j b̂ j + 1̂ . Ĥ (z. t). t). We would prefer a definition of the operator Q̂.9) concerning elementary quantum mechanics as ∂ z| p̂z |ψ = −i z|ψ . Ĥ . t) are studied using (2.2. and the initial condition b̂ j (0. (2.

37) and (2. ĥ l ] = s− j δ− j.42) becoming 1  2  û =  Ê + Ĥ 2 8π 1   =  ê j ê− j + ĥ j ĥ − j . (2.60) j and introduce the operators † B̂ j = e−γ R̂ b̂ j eγ R̂ = (cosh γ )b̂ j − (sinh γ )b̂− j .59) 2V The energy density operator û refr may be diagonalized through a Bogoliubov trans- formation.42) through (2.l is the Kronecker δ function. (2.38) satisfy a modified operator algebra with respect to that of the harmonic oscillator: [ê j . (2. (2.55) c j 2 The investigation of the case χ = 0. The knowledge of these commutators and of the generalized total momentum operator Ĝ.  = 1 does not lead to any new expansions of the field operators Ê and Ĥ . the derivation of (2.g.43) we can set û j = û j refr .56)   4π ω j [ê j .47).57) V where δ j. ωj  † †  †  †  û j refr = b̂ j b̂ j + b̂− j b̂− j − 2π χ b̂ j − b̂− j b̂− j − b̂ j .18 2 Origin of Macroscopic Approach The total energy operator of the free field inside the volume of quantization is thus   V    † 1 Ĥ = Û = Ŝ+ − Ŝ− = ω j b̂ j b̂ j + 1̂ . The energy density operator (2. ĥ l ] = 0̂.58) can be generalized. (2. should have been generalized accordingly. The individual components of the rearranged elec- tric and magnetic field operators according to (2.61) where 1 1 γ = ln  = ln n.58) 8π j which enables us to derive the Maxwell equations both via the temporal derivatives and via the spatial derivatives. To this end we introduce an anti-Hermitian operator R̂ of the form  † †  R̂ = b̂ j b̂− j − b̂ j b̂− j (2. In the expansion (2. (2.62) 4 2 . êl ] = [ĥ j . the relation (2. (2.l 1̂. e.

64) j and the energy density operator has the diagonal form nω j  † †  û j refr = B̂ j B̂ j + B̂− j B̂− j . t) = −i sj ( B̂ j + B̂− j ).38).65) 2V The momentum operator is then given by    V † 1 Ĝ refr = (û + − û − ) = Kj B̂ j B̂ j + 1̂ . (2.63) the operator R̂ takes the form  † †  R̂ = B̂ j B̂− j − B̂ j B̂− j .2 Nondispersive Lossless Linear Dielectric 19 On substitution † b̂ j = eγ R̂ B̂ j e−γ R̂ = (cosh γ ) B̂ j + (sinh γ ) B̂− j . (2.69) j V Similarly as above. respectively. (2. this relation can be interpreted as a result of the replacement b̂ j . (2.2.68) j nV and    12 2π nω j † Ĥ (z.63) into (2.66) c j 2 with K j = nk j and the Hamiltonian can be calculated as    † 1 Ĥrefr = ωj B̂ j B̂ j + 1̂ . we can obtain the electric and magnetic field operators inside the dielectric:   2π ω j  2 1 † Ê(z. t) = −i ( B̂ j − B̂− j ) (2. (2.37) and (2.67) j 2 By inserting (2. (2.

. both reflection and diffraction occur.34). We will not treat this more general case according to Abram (1987).33) and (2.→ B̂ j and a consequence of the quantized classical equations (2. For normal incidence on a sharp vacuum–dielectric interface.

Let us recall the quadrature representation in quantum optics. the exposition is confined to pure Gaussian states while it is possible to generalize it also to mixed Gaussian states of the electromagnetic field. The Hamiltonian in this representation has the form   1 2 δ 2 1 H= − + [∇ × A(r)]2 d3 r. Only the dispersion of the medium has not been considered.e. t) B(r.e. as is appropriate with SI units. as well as that of the “most general”. the wave functional of the vac- uum state. It is shown that the expectation values B̂ = B and D̂ = D (equivalently. the argument A(r) of the wave functional does not depend on t. but the wave functional does depend on t. by two real matrix kernels. the simplest Gaussian state of the electromagnetic field. The wave function and the corresponding Wigner function become then functionals of the field variables. (2. can be found. t).70) is then useful. The Riemann–Silberstein–Kramers complex vector has been introduced  1 D(r. This approach allows that the medium under investigation is inhomogeneous and time dependent. It is not clear whether the complex vector (2. It has been suggested that the periodicity of the electric per- mittivity tensor (r. Mrówczyński and Müller (1994) have . t) F(r. t) = √ √ +i √ .2. t) or the magnetic permeability μ(r. Thus. (2. t).70) 2 0 μ0 √ √ where we have divided by 0 . It has been shown how the Green function method can be used for solving linear equations for the field operator F̂(r. It is tempting to generalize the concept of a Gaussian state of the finite- dimensional harmonic oscillator to the case of an infinite oscillator. i. i.20 2 Origin of Macroscopic Approach 2.71) 2 0 δA(r)2 μ0 In Białynicka-Birula and Białynicki-Birula (1987). The whole electromagnetic field is treated as a huge infinite-dimensional har- monic oscillator. The pure Gaussian state is determined by a complex matrix kernel. Let us observe that contrary to the operator Â(r. Ê = E) evolve according to the free-field Maxwell equations and also the equations which the complex matrix kernel obeys can be found there. μ0 . It has been derived that photon pair production is a necessary condition for squeezing. Białynicka- Birula and Białynicki-Birula (1987) treat the time development of the Gaussian states in the free-field case.3 Wave Functional Description of Gaussian States Białynicka-Birula and Białynicki-Birula (1987) have tried first to define the squeez- ing that is a generalization of the standard definition for one mode of radia- tion. t) can be important for the generation of squeezed states. t]. This definition can be reformulated with respect to Białynicki-Birula (2000). There the Schrödinger picture is adopted and an ana- logue of the Schrödinger representation in quantum mechanics has been introduced. This representation is a wave functional Ψ[A.

the wave func- tional and the Wigner functional have been introduced. The Wigner functional for the thermal state of the elec- tromagnetic field has been presented.75) also plays the role of a norm for the photon wave function (Białynicki-Birula 1996a. The norm (2. This state is mixed and it even has infinitely many photons in the whole field. which characterize the state. has no wave functional. D(r) are the vector potential and the electric displacement vector. The exposition continues with the Wigner functionals for the states of the electromagnetic field that describe a definite number of photons. The analogy with the one-dimensional harmonic oscilla- tor leads to other notions. −D] = B(r) · B(r ) 4π  2 μ0 |r − r |2   μ0 1 + D(r) · D(r ) d3 r d3 r . −D has been presented.72) 4π  μ0 |r − r |2 and from the wave functional (we change à → −D)   1 μ0 1  3 3  Ψ̃0 [−D] = C exp − 2 D(r) · D(r ) d r d r .74) where  1 0 1 N [A. A and A . −D] = exp{−2N [A.  (2. Białynicki-Birula (2000) starts from the wave func- tional of the vacuum state (Misner et al. An example of the functional for the one-photon state with the photon mode function f(r) has been included. The exposi- tion is related to the hot topic of the superpositions of coherent states of the electro- magnetic field.75) 0 |r − r |2 The expression (2.2 Nondispersive Lossless Linear Dielectric 21 considered only the scalar field. (2. the Wigner functional for the coherent state of the electromagnetic field |A. (2.2. where A(r).b). 1970)   1 0 1  3 3  Ψ0 [A] = C exp − 2 B(r) · B(r ) d r d r (2. the mixed state. In contrast to Białynicka-Birula and . Let us remark that for (the statistical operator of) such a state the matrix element can be considered which is a functional of two arguments. respectively.75) has not been related to any inner product of the photon wave functions. but these notions are connected. −D]}. The Wigner functional of the electromagnetic field in the ground state is W0 [A. In each of the subsequent cases. In particular.73) 4π  0 |r − r |2 The normalization constant C is an issue and it has not been completely solved in Białynicki-Birula (2000). The exception.

(2.80) Let us note that the right-hand sides of (2. 1984). and KBD are real matrix kernels. X̂ 2 [B]] = i f(r) · [∇ × g(r)] d3 r.78) 4π 2  0 |r − r |2  1 0 1 g(r) = B(r ) d3 r . we introduce quadrature operators as X̂ 1 [D] = D̂(r. (2.76) and (2. (2. we obtain that i [ X̂ 1 [D].79) 4π 2  μ0 |r − r |2 The commutator of the X̂ 1 and X̂ 2 operators is [ X̂ 1 [D]. −D] = exp − B · KBB · B  μ0    μ0 + D · KDD · D + B · KBD · D d3 r d3 r .76) X̂ 2 [B] = B̂(r. The kernel KBD is not inde- pendent of KBB and KDD .77).81) 4 We see easily that the usual commutator − 12 i1̂ is yielded by the field (D. 0) · g(r) d3 r. (2. X̂ 2 [B]] = D(r1 ) · A(r1 ) d3 r1 1̂. (2. Milburn et al. but it must obey the condition that is reminiscent . 0) · f(r) d3 r. (2. Without resorting to this notation. Białynicki-Birula (2000) presents the Wigner functional for the squeezed vacuum state:  1 0 Wsq [A. (2.22 2 Origin of Macroscopic Approach Białynicki-Birula (1987). −D)) with the property [−D(r1 )] · A(r1 ) d3 r1 = 2.82) We have not deepened the contrast by introducing the notation X̂ 1 [−D] and X̂ 2 [A] on the left-hand sides of (2. B) (or (A. (2.83) 0 where KBB .79) comprise the operator |∇|−1 up to a certain factor (cf.77) where  1 μ0 1 f(r) = D(r ) d3 r .78) and (2. KDD .

Mendonça et al.84) where n(r) is the space-dependent refractive index. Another open question is how the projection of this Wigner functional onto Wigner functions of any orthogonal (complete or incomplete) modal system looks out. (2000) have quantized the linearized equations for an electromagnetic field in a plasma.4 Source-Field Operator Knöll et al. The problem of the time evolution is also discussed. also Gea-Banacloche et al. He has derived formulae for the mass of photon in resting and moving dielectric and the velocity of the photon as a particle. (2.2. (1987) have elaborated on the approach developed on the basis of quantum field theory and applied to the problem of spectral filtering of light (Knöll et al. Nilsson et al. the symmetric ordering is vexed.b). (1987) have compared the problem of quantum-mechanical treatment of action of optical devices with the input–output formalism (Collett and Gardiner 1984. Yamamoto and Imoto 1986. He has extended the concept of the vector potential to relativistic velocities of the medium. The only assumptions are that the interaction between sources and light is linear in the vector potential and the optical system is lossless and that the condition of sufficiently small dispersion is fulfilled. (2. It is appropriate to mention here work concerning the photon wave function (Inagaki 1998. Knöll et al. 1986). Just as in the field theory. Zaleśny (2001) has found that the influence of a medium on a photon can be described by some scalar and vector potentials. First. cf. The classical Maxwell equations comprise the relative permittivity (r) = n 2 (r).2. Apart from the fact that only a very particular setup is considered in the input–output formalism. the classical Maxwell equations with sources and optical devices are formulated and solved by the proce- dure of mode expansion and the quantized version is derived. although it is relevant mainly to the electromagnetic field in vacuo. An extension of the quantization pro- cedure leads to the definition of a photon charge operator. They have determined an effective mass for the transverse photons.85) c2 . Gardiner and Collett 1985. 2. 1986. The mode functions Aλ (r) are introduced as the solutions of equation ωλ2 ∇ × (∇ × Aλ (r)) − (r) Aλ (r) = 0. Using a straightforward procedure. the theory does not take into account the full space–time structure of the field.2 Nondispersive Lossless Linear Dielectric 23 of the Schrödinger–Robertson uncertainty relation (Białynicki-Birula 1998). Hawton 1999. Kobe 1999). 1990a. It has been conceded that the Wigner function is not a very powerful tool for making detailed calculations.

87) In terms of these functions. âλ ] = δλλ 1̂. âλ ].88) † On inserting the operators âλ and âλ into the decomposition of the vector potential. âλ ] = 0̂ = [âλ . the operator of the vector potential Â(r. (2. which have the properties † [âλ . † † [âλ . the vector potential can be decomposed. t) is defined:  . from which the gauge condition can be derived ∇ · [(r)Aλ (r)] = 0.24 2 Origin of Macroscopic Approach where ωλ2 is the separation constant for each λ. (2.86) It is assumed that these solutions are normalized and orthogonal in the sense of equation (r)Aλ (r) · Aλ (r) d3 r = δλλ 1̂. (2. In the standard † manner the destruction and creation operators âλ and âλ are defined.

p̂k  a  ] = iδaa  δkk  1̂.90) and the commutation relations † [r̂ka . † Â(r. âλ ].91) The operator Â(r.92) ∂t and for the derivation of the magnetic field strength operator B̂(r.93) . p̂k  a  ]. r̂k  a  ] = 0̂ = [ p̂ka . (2.89) λ The source quantities ra and pa are considered as the operators r̂a and p̂a . (2. t) = ∇ × Â(r. (2. âλ ] = 0̂ = [ p̂ka . which obey the standard commutation relations [r̂ka . † [ p̂ka . âλ ]. t) = Aλ (r) âλ (t) + âλ (t) . t) = − Â(r. âλ ] = 0̂ = [r̂ka . (2. t). [r̂ka . t) (2. t) can be used for the derivation of the electric-field strength operator which is associated with the radiation field by the relation ∂ Ê(r.

the mode functions are redefined so that they obey the normalization condition  (r)Aλ (r) · Aλ (r) d3 r = δλλ .94) 20 ωλ The form of the normalization conditions (2. (2.94) is tailored to real- mode functions and the necessity of modification of some fundamental relations is commented on by Knöll et al.87) and (2.2.2 Nondispersive Lossless Linear Dielectric 25 Nevertheless. (1987). All of these field operators may be written in the form  † .

(2.99) where (+)  F̂kfree (r. t  ) d3 r dt  . .97) λ F̂(−) (r.96) where  F̂(+) (r. (2. (2. It is often convenient to decompose a given field operator F̂(r. t. t)]† . t). (2. t  ) Ĵk  (r .95) λ In dependence on the choice of the operator F̂(r. (2. It is typical of the approach of Knöll et al. r . t). t) = θ (t − t  )K kk  (r. t) + F̂(−) (r. t). t) into two parts by the relation F̂(r. so that the field operators can be expressed in terms of free-field and source-field operators. t) = F̂(+) (r. t) = Fλ (r)âλ (t) + F∗λ (r)âλ (t) . t) = Fλ (r)âλ (t). (1987) that any field oper- ator F̂k(+) is decomposed into a free-field operator and a source-field operator as follows: F̂k(+) (r.98) Further. (2. t) = F̂kfree (+) (r.100) λ F̂ks (r. the functions Fλ (r) can be derived from the mode functions of the vector potential Aλ (r). (2. F̂(r.101) Here vector components are labelled by the index k and repeated indices k  mean summation. the Heisenberg equations of motion for the field operators are derived. t) + F̂ks (r. t) = Fkλ (r)âλfree (t). t) = [F̂(+) (r.

r. t} (2. t) (2. respectively. (2. The following abbreviations of the notation are used: x = {r. t ) = ∓K k  k (r . t.105)  λ So the symmetry relations ∗     K kk  (r. (2.104) i λ Analogously. so that we may only guess that âλfree (t)|t=t0 = âλ (t0 ) for t = t0 . The information on the action of the opti- cal instruments on the source field is contained in the space–time structure of the kernel K kk  . the kernel K kk  is defined by the relation 1  K kk  (r. t . the operator âλfree (t) was not defined. t  ) d3 r dt  + F̂kfree (+) (r. t. the commutation relations for various combinations of field operators at different times are studied and relationships between field commutators and source- quantity commutators are derived. Further. r .101). t  ) = − Akλ (r)A∗k  λ (r ) exp[−iωλ (t − t  )]. r . (2. t  ) Ĵk  (r .102) i λ Inserting equation (2. r .106) are valid for Â(+) (+) k and Ê k . the appropriate form of the kernel K kk  is 1 K kk  (r. (1987). t. t. r . if F̂k(+) is identified with the vector potential Â(+) k . which may be regarded as the apparatus function also used in classical optics.99) yields the following representation of F̂k(+) : F̂k(+) (r. it holds that Fkλ = Akλ and the kernel K kk  takes the form 1  K kk  (r. if one is interested in the electric-field strength of the radiation Ê k(+) . t  ) = − ωλ Akλ (r)A∗k  λ (r ) exp[−iωλ (t − t  )]. t. r .26 2 Origin of Macroscopic Approach Unfortunately.103) In particular. (2. and the dynamics for t ≥ t0 can be found in Knöll et al. t  ) = − Fkλ (r)A∗k  λ (r ) exp[−iωλ (t − t  )].107) . t). In equation (2. t) = θ(t − t  )K kk  (r.

Ân (tn ) = Âi1 (ti1 ) Âi2 (ti2 ). (2. −.j ) (j .116) . by which the superscripts +.. x2 ) − D̂k22k1 1 (x2 . j = +. Ân (tn ).112) where (j . ( j) ( j) ( j) (2.114) The commutators in (2. x2 )[ Ĵk  1 (x1 ).j ) D̂k11k2 2 (x1 . x1 )K k2 2k  (x2 . (j ) (j ) (j ) (j ) (2. it holds that ( j) ( j) ( j) F̂k (x) = F̂kfree (x) + F̂ks (x).j ) + D̂k11k2 2 (x1 ..2 Nondispersive Lossless Linear Dielectric 27 and others. Let us consider any operator product Â1 (t1 ) Â2 (t2 ).113) 1 2 1 2 From an inspection of equation (2. x2 )1̂. Âin (tin ) with ti1 < ti2 < · · · < tin . we readily learn that (j .. The symbol T+ introduces the time ordering of the operators Âi (ti ) with the latest time to the far left: T+ Â1 (t1 ) Â2 (t2 ). x2 ) = 0̂ if t1 > t2 .109) When appropriate. the time ordering symbols T+ and T− are used.. (2. −. F̂k2 free (x2 )] (j . (2. F̂k(−) 2 free (x2 )] = Fk1 k2 (x1 ..2.. (2. x2 ) =− θ (t2 − t2 )θ (t2 − t1 )θ (t1 − t1 ) ⊗K k1 1k  (x1 .115) [ F̂k(+) 1 free (x1 ). Ân (tn ) = Âi1 (ti1 ) Âi2 (ti2 ).91) it follows that (j ) (j ) (j ) (j ) [ F̂k1 1 (x1 ).108) F̂ks (x) = θ (t − t  )K kk  (x..113). F̂k2 free (x2 )] = 0̂. F̂k2 2 (x2 )] = [ F̂k1 free 1 2 (x1 ).112) are ( j) ( j) [ F̂k1 free (x1 ).. (2. (2. Ĵk  2 (x2 )] dx1 dx2 . x  ) Ĵk  (x  ) dx  .110) and the symbol T− introduces time ordering of the operators Âi (ti ) with the latest time to the far right: T− Â1 (t1 ) Â2 (t2 ). x1 ). (2. − are introduced also for Ĵk (x) and K kk  (x. x  ). j = +.j ) D̂k11k2 2 (x1 . With these generalizations. Âin (tin ) with ti1 > ti2 > · · · > tin ..111) From (2..

.. x2 ) = Fk1 λ (r1 )Fk∗2 λ (r2 ) exp[−iωλ (t1 − t2 )].117) λ It would be interesting to find the particular forms of the commutators between  and Ê or between Â(+) and Ê (−) ... The method of transformation of normal and time orderings is demonstrated for the following important class of correlation functions: G (m..km+n (x 1 .118) j=1 j=m+1 This transformation is understood in the relation T+ F̂k(+) (x1 ) F̂k(+) (x2 ) 1 . these commutation relations are used to express field correlation func- tions of free-field operators and source-field operators and to describe the effect of the optical system on the quantum properties of light fields.n) k1 . x m+n ) ⎡ ⎤⎡ ⎤"  m  m+n = ⎣T− F̂k(−) j (x j )⎦ ⎣T+ F̂k(+) j (x j )⎦ . Further. (2. (2..28 2 Origin of Macroscopic Approach where  Fk1 k2 (x1 .

2 .

109) for the operators F̂k(+) is (xi ) and T+ time ordering of the source-quantity operators Ĵki (xi ) in the resulting source-quantity operator products before performing the integrations with respect to ti . = O+ F̂k(+) 1 free (x 1 ) + F̂ (+) k1 s (x 1 ) F̂ (+) k2 free (x 2 ) + F̂ (+) k2 s (x 2 ) . (ii) Substitution of equation (2. F̂k(−) j free (x j ): (i) Ordering of the operators F̂k(−) is (xi ). (ii) Substitution of equation (2.109) for the operators F̂k(−) is (xi ) and T− time ordering †  of the source-quantity operators Ĵk  (xi ) in the resulting source-quantity operator i products before performing the integrations with respect to ti . . (+) F̂k j free (x j ): (i) Ordering of the operators F̂k(+) is (xi ). The symbol O+ introduces the following ordering of operators F̂k(+) is (xi ).119) and the following ones the ordering symbols O+ and O− are used. F̂k(+) j free (x j ) with the operators F̂k(+) j free (x j ) to the right of the operators F̂k(+) is (xi ).119) In equation (2. F̂k(−) j free (x j ) with the operators F̂k(−) j free (x j ) to the left of the operators F̂k(−) is (xi ). The symbol O− introduces the following operator ordering in products of operators F̂k(−) is (xi ). (2.

119) may now be generalized:  n n  .2.2 Nondispersive Lossless Linear Dielectric 29 Equation (2.

(2. T+ F̂k(+) j (x j ) = O+ F̂k(+) j free (x j ) + F̂k (+) j s (x j ) .120) j=1 j=1  n n  .

n) k1 .. T− F̂k(−) j (x j ) = O− F̂k(−) j free (x j ) + F̂ (−) kjs (x j ) .. (2..121) j=1 j=1 Using relations (2.121). (2. x m+n ) ⎧ ⎫ ⎨ m  .km+n (x 1 .118).120). we may represent the correlation func- tions as G (m.. and (2.. .

⎬ = O− F̂k(−) (x j ) + F̂k(−) (x j ) ⎩ j free js ⎭ j=1 ⎧ ⎫" ⎨  m+n .

. . the complex kernels K k j k j (x j . F̂kfree = 0 = F̂kfree ..122) can be simplified: G (m.125) .. (2. x m+n )  ⎡ ⎤⎡ ⎤"  m  m+n = ⎣O− F̂k(−) js (x j )⎦ ⎣O+ F̂k(+) js (x j )⎦ . (2.. In Knöll et al.. into the relation (2.km+n (x 1 .. .124) j=1 j=m+1 When written in more detail.123) then the relation (2.122) is specialized to a multimode coherent free field |{αλ } .124).n) k1 . (+) F̂kfree (x)|{αλ } = Fk (x)|{αλ } . It is noted that the effect of the beam splitter that is used for mixing of source light with the reference beam in the case of homodyne detection is described by the assumption that the reference light beam is a free field. x j ) are introduced. (1987) the relation (2. (2.⎬ ⊗ O+ F̂k(+) (x j ) + F̂k(+) (x j ) . . (2.122) ⎩ j free js ⎭ j=m+1 When at the points of observation the following conditions are fulfilled  (+)   (−)  .. .

30 2 Origin of Macroscopic Approach that is to say G (m.... x m+n ) ⎧ ⎫ ⎨ m  .. .n) k1 .km+n (x 1 ..

⎬ = O− Fk∗j (x j )1̂ + F̂k(−) (x j ) ⎩ js ⎭ j=1 ⎧ ⎫" ⎨  m+n .

the probability of observing precisely n events in a counting time interval [t. Let us note that one usually assumes that S(t1 − t2 ) = ηδ(t1 − t2 ). Δt)] .126) ⎩ js ⎭ j=m+1 Finally.128) i t t may be interpreted as the operator of the integrated intensity.130) n! .⎬ ⊗ O+ Fk j (x j )1̂ + F̂k(+) (x j ) . the theory is applied to the photocount statistics. (ii) T+ ordering of the operators Ê k(+) (x) and T− ordering of the operators Ê k(−) (x).127) n! where  t+Δt t+Δt Γ̂(t. In analogy with (2. Kelley and Kleiner 1964). Δt) exp[−Γ̂(t. Following Glauber’s theory of photon detection (Glauber 1965.122). Δt) = S(t1 − t2 ) Ê k(−) (ri . Δt) = Ω Γ̂(t. Here ri are position vectors of the detector atoms and S(t) is a response function. In relation (2. Δt) = O Γ̂(t. the ordering symbol Ω introduces the following operator ordering: (i) The normal ordering of the operators Ê k(−) (x). Δt) . t1 ) Ê k(+) (ri . Ê k(+) (x) with the operators Ê k(−) (x) to the left of the operators Ê k(+) (x). Δt) exp −Γ̂(t. t + Δt) is given by the relation ) * 1  n   pn (t.127) becomes ) * 1  n pn (t. (2. (2. t2 ) dt1 dt2 (2. (2. (2.129) with some η.127). relation (2.

128) as follows:  t+Δt t+Δt (−) (+) Γ̂(t. there is an analogy with the relation (2. t2 ) dt1 dt2 .126):  t+Δt t+Δt Γ̂(t. Ê kfree (x) with (−) (−) (+) (+) the operators Ê ks (x).2. Δt) = S(t1 − t2 ) i t t . Ê kfree (x) and O− ordering of the opera- (−) (−) tors Ê ks (x). Ê kfree (x). (2. Δt) = S(t1 − t2 ) Ê ks (ri . The fulfilling of the conditions (2.131) i t t In the case of mixing the source field light with a coherent free-field reference beam. Ê kfree (x). Ê kfree (x) to the left of the operators Ê ks (x).2 Nondispersive Lossless Linear Dielectric 31 where the Ω ordering is simply replaced by the O ordering defined as follows: (−) (−) (+) (+) (i) The normal ordering of the operators Ê ks (x). t1 ) Ê ks (ri . (+) (+) (ii) O+ ordering of the operators Ê ks (x). Ê kfree (x).123) causes a modification of relation (2. Ê ks (x).


(1991) a continuous-mode quantum theory of electromagnetic field has been developed. As usual in the quan- tum field theory.5 Continuum Frequency-Space Description Blow et al. and absorptionless dielectric with space- dependent refractive index has been applied to the description of the action of a resonator-like cavity with input–output coupling and filled with an active medium (Knöll and Welsch 1992).132) A generalization of the Wick theorem on transforming a time-ordered product onto a sum of normally ordered terms was performed by Agarwal and Wolf (1970). â j ] = δi j 1̂. (2. The mode spectrum becomes .2. 2. have frequencies given by dif- ferent integer multiples of the mode spacing Δω. t1 ) Ek (ri . t2 ) dt1 dt2 .133) Different modes of the cavity. dispersionless. t2 )1̂ + Ê ks (ri . (1990) have formulated the quantum theory of optical wave propagation without recourse to cavity quantization. the box-related modes are considered whose creation and destruc- tion operators satisfy the usual independent boson commutation relations: † [âi . × Ek∗ (ri . This approach avoids the introduction of a box-related mode spacing and enables one to use a continuum frequency-space description. (2. The quantum theory of the radiation field interacting with atomic sources in the presence of a linear. t1 )1̂ + Ê ks (−) (+) (ri . labelled by i and j. In this chapter and in that by Blow et al.

Further specific states of the field have been treated such as coherent states.140) λ=1. (2. (2. t) has the diagonal form: ∞ Ĥfree = A Û (z.2 . t) = Â (z. B̂ (+) (z. λ)â(k. noise and squeezed states. Such operators have all the usual properties of the operators of the monochromatic mode. With the use of noncontinuous operators. Field quan- tization in a dielectric has been treated including the material dispersion and the theory has been applied to the pulse propagation in an optical fibre. (2. λ) exp[−i(ωt − kz)] dk. Let us consider the fields in a lossless dielectric material with the real relative permittivity (ω) and the refractive index n(ω) related by (ω) = [n(ω)]2 . A comparison with results by Drummond (1990. num- ber states.139) ∂t ∂z and with the expansion of the vector-potential operator  ∞ vG (ω) Â (+) (z.138) −∞ The field operators are obtained in accordance with the relation ∂ (+) ∂ (+) Ê (+) (z. t) dz = ωâ † (ω)â(ω) dω.135) Let us recall the definition of the phase velocity ω c vF (ω) = = (2. (2. t) = − Â (z. 1994) would be in order.136) k n(ω) and that of the group velocity vG (ω) 1 ∂k 1 ∂ = = [ωn(ω)]. t) = −∞ 4π 0 cωn(ω)A  × (k. a generalization of the single-mode normal ordering theorem was proved. t). t) (2.137) vG (ω) ∂ω c ∂ω The normalization of the field operators is fixed by requirement that the normally ordered total energy density operator Û (z.134) A complete orthonormal set of functions was considered which may describe states of finite energy. (2. The set is numerable infinite and to each function in it a destruction operator is assigned.32 2 Origin of Macroscopic Approach continuous as Δω → 0 and in this limit the transformation to continuous-mode operators is convenient: √ âi → Δω â(ω).

139) that the field operators are  ω Ê (+) (z. The advantage of this treat- ment is that the functional dependence on z and t is contained in the c-number functions rather than the operators â(z. (2. t) + â(z. λ) = vG (ω)â(ω). but they seem to be defined otherwise. t) are a complete orthonormal set of functions on z and ĉ j are destruc- tion operators obeying the usual commutation relations.142) c and  ωn(ω) B̂ (+) (z. t) = 0̂. Assuming a narrow bandwidth. (1990) that the solution of . t) = i 4π 0 cAn(ω)   n(ω)z × â(ω) exp −iω t − dω (2. it follows from (2. t) = i 4π 0 c3 A   n(ω)z × â(ω) exp −iω t − dω.144) for example. It is not emphasized by Blow et al.145) j where φ j (z. which obeys the equation ∂ k  ∂ 2 i â(z.2. Drummond 1990).2 Nondispersive Lossless Linear Dielectric 33 Noting that dω + dk = . The equation has been simplified using the transformation of envelope into a frame moving with the group velocity. (2. (2. The treatment of this problem in the noncontinuous basis proceeds from the replacement  â(z.141) vG (ω) and taking the polarization to be parallel to the x-axis. evaluated at the central frequency.143) c Alternatively.144) ∂z 2 ∂t 2 where k  is the second derivative with respect to the frequency of the propagation constant. the slowly varying field envelope can be represented by the operator â(z. t) = φ j (z. the propagation constant can be expanded to the second order in frequency and a partial differential equation can be obtained (cf. In the classical nonlinear optics the stationary fields have also envelopes. t) as in the propagation equation (2. â(k. (2. t). This is necessary for the envelope to be slowly varying. t)ĉ j .

(2. properties. (2. (2. (2. (2. t) = 0. the field of the local oscillator being in the coherent state |{αL (t)} . commutators. the process of photodetection in free space is considered and the results applied to homodyne detection with both local oscillator and signal fields pulsed. The propagation equation (2. For homodyne detection of pulsed signals it is advantageous to use a pulsed local oscillator. t) enjoys the orthonormality and completeness only as the equal-space.147) τ Here τ is the start time of the measurements.149) τ † where âL (t) and âL (t) are the continuum creation and destruction operators of the local oscillator field and â † (t) and â(t) correspondingly for the signal field.150) with NL the mean total number of photons in the pulse and θL the externally con- trolled local oscillator phase. Similarly.34 2 Origin of Macroscopic Approach equation (2. (2. 1987) τ +T † Ô = i [â † (t)âL (t) − âL (t)â(t)] dt. the detector is placed at z = 0.151) with   D̂({α(t)}) = exp [α(t)â † (t) − α ∗ (t)â(t)] .144) preserves the equal-space.148) 2π Let us further consider a balanced homodyne detector in which the light beam under study is superposed on a local oscillator by combining them at a 50:50 beam splitter.152) . The pulsed signal is described by the noncontinuous basis function φ0 (t) and the local oscillator is described by a normalized function φL (t). The measured quantity is the difference in the photocurrents of two detectors placed in the output arms of the beam splitter and it can be represented by the operator (Collett et al. the set of functions φ j (z. (2.144) now yields the following equations for the noncontinuous basis functions: ∂ k  ∂ 2 i φ j (z. not equal-time.146) ∂z 2 ∂t 2 Finally. and 1 â(t) = √ â(ω) exp(−iωt) dω. The results of sets of measurements in which the photocurrent is integrated over periods T can be predicted by the use of an operator τ +T M̂ = â † (t)â(t) dt. where + αL (t) = NL exp(iθL )φL (t). but not equal-time. t) + φ j (z. Let us recall the definition of a coherent state: |{α(t)} = D̂({α(t)})|0 .

i > 0. (2. It is shown how the formulation of the quantum field theory is modified for the one-dimensional optical system.149) it is necessary to substitute  â(t) = φi (t)d̂i . (2. but of finite cross-sectional area A of the rectangular form with sides parallel to the x. be removed and putting k z = k. λ). we consider  âL (t) = φiL (t)ĉiL .k x y On the assumption that the modes with k x = 0 or k y = 0 are vacuum ones. It is assumed that the signal field is described by a set of noncontinuous operators d̂i at the output of a nonlinear system and the signal field at the input to the system is described by a similar set of operators ĉi .154) i In analogy. a reduced Hilbert (namely Fock) space can be exploited. the other conversions are A δ (3) (k − k ) → δ(k − k  ).158) 2π The vector-potential operator has been modified for the dispersive lossless medium and compared with Drummond (1990) and Loudon (1963). the positive-frequency part is .2 Nondispersive Lossless Linear Dielectric 35 which is close enough to that by Blow et al. The summation in (2.155) i where the subscript L only modifies the familiar meanings and φ0L (t) = φL (t). (2.2.and y-axes.156) A k .156) can. The fields are defined in an infinite waveguide parallel to the z-axis. therefore. (2. (2. (2. d̂i = ĉi .157) (2π)2 √ A â(k.153) In the relation (2. (1990) except for the exchange of space for time. λ) → â(k. The action of the nonlinear system is defined by the relations † d̂0 = μĉ0 + ν ĉ0 . The x and y wave-vector components are thus restricted to discrete values and any three-dimensional integral over this spatial region is converted according to (2π )2  d3 k → dk z .

cf. This process may occur as frequency down conversions. λ)â(k. A second-order interference has been . Laboratory techniques for doing experi- ments with single photons also have advanced. e. (1997a). (2. In a travelling-wave setting.3 Quantum Description of Experiments with Stationary Fields Burnham and Weinberg (1970) found that the measured value of the correlation time between the two optical photons produced in a parametric process was very small.36 2 Origin of Macroscopic Approach  vG (ω) Â (+) (r. (1990). λ) exp[−i(ωt − k · r)] d3 k. and the uniformity has been achieved also by the use of the Wigner (or Weyl) representation of the field operators. 1990).140)). The photon pairs (biphotons) produced in parametric down-conversion are useful in experiments concerning fundamental questions of quantum theory. signal. an effect of a practical interest.159) λ=1. The process of optical parametric three-wave mixing in a second-order nonlinear medium consists of the coherent interaction between pump. A similar treatment of the famous experiment and of another one has been presented by Casado et al. Since 1985.159) can be converted to the one-dimensional form easily (as indicated above. A unified treatment of the experiment on the interference of a “biphoton with itself” and of other three experiments has been provided by Casado et al.g. t) = 16π 3 0 cωn(ω)  × (k. McDonald (2001) has considered a variation of the physical situation of “slow light” to show that the group velocity can be negative at central frequency. 2. and idler waves. (2. (1994). in Joobeur et al. The description of experiments has been facilitated by studies of Campos et al. specifically as an optical parametric oscillation and an optical parametric amplification. such photon pairs have become familiar for the study of nonclassical aspects of light (Horne et al. The autocorrelation and cross-correlation properties of the signal and idler beams produced in the parametric down-conversion have been studied.2 c ω= |k|. the antipulse propagating within the slab so as to annihilate the incident pulse at the near side. the optical parametric generation is called a spontaneous parametric down- conversion. (2. (1997b).160) n(ω) The expression (2. A Gaus- sian pulse can emerge from the far side of a slab earlier than it hits the near side and the pulse emission at the far side is accompanied by an antipulse emission. A fourth-order interference has been obtained in the four cases.

The studies of the fundamentals of quantum mechanics underlie such interest- ing applications as quantum cryptography and quantum computing (Bowmeester et al. which is assumed to be monochromatic however. Coincidences are not studied and a second-order interference has been obtained in the two cases. Here we return from the Wigner to the Hilbert-space formalism as in Peřinová and Lukš (2003). We may refer to Casado et al. This experiment was proposed in order to test a Bell inequality for energy and time. which was also used to test Bell’s inequality using phase and momentum. . in the case of the experiment of Franson (1989) we are allowed to return to the simple description. but only on the paths.2. but since two paths belong to each crystal. Next we deal with induced coherence and indistinguishability in two-photon interference (Zou et al. Two experiments are analysed: frustrated two-photon creation by interference. the number of paths is greater. 2000).3 Quantum Description of Experiments with Stationary Fields 37 treated in the two cases and the stochastic properties of the pump beam have been respected. Last we mention the frustrated two-photon creation via interference (Herzog et al. The experiment on the interference of signal and idler photons (Ghosh et al. but with the response of the output fields of a nonlinear crystal to the input fields (Casado et al. 2005). the biphoton arrives at the two detectors) is still held for the reason of interference. although we cannot define them everywhere. and induced coherence and indistinguishability. even though the schematic is more complicated. 1994) restricting ourselves to the second-order interference and the monochromatic pump. The lack of induced emission made it a “mind- boggling” experiment (Greenberger et al. 1993). we do not reproduce the well-known result. As in the schematics of the experiments the field is restricted to paths leading to detec- tors. In the use of two detectors we consider four paths and modify (double) the description. In this case the schematic comprises two nonlinear crystals. 1997a) when two paths cross such a crystal. as essentially two paths are involved. we introduce one-dimensional expansions of the electric-field operator. We attempt to consider orthogonal modal functions. (1997b). (1997a). 1991). where one cannot evaluate the orthogonality property for the lack of a complete definition. Design of experiments for undergraduate students has become feasible (Galvez et al. In con- trast. The experiments have become very popular (Shih 2003). Such a response depends also on stochastic properties of the pump beam. when the lack of the second-order interference is derived. First we consider the three-dimensional expansion of the operator of a chosen component of the electric vector after Casado et al. where stochastic properties of the pump beam are taken into account. but the indistinguishability of the paths along which the signal photon arrives at the detector (in fact. In this approach we do not start with the description of the process of parametric down-conversion from a Hamiltonian. Similarly we proceed in the case of the experiment of Rarity and Tapster (1990). the simple description is appropriate. We are aware of the dangerous position. Nevertheless. 1986) can do with the simple description.

162) k.λ (t) is the annihilation operator for a photon whose wave vector is k and whose polarization vector is k. The interaction lasting only for a short time and being spatially confined to the medium suggests to us an appropriate modification of the linear dynamics. which are often related to the device. We suppose that such an approach can be interesting also in our study.38 2 Origin of Macroscopic Approach 2. (1997a) use the subscripts on the (Wigner representations of) field opera- tors they indicate that the light beam contains frequencies within a range and that “transverse” components of wave vectors are limited by small upper values. in general.λ (t) and âk. t). (2. In this picture the state of the field is represented by a time-independent statistical operator ρ̂.162) correspond to the Heisenberg picture. We refer to any of our figures for a sketch of the . The action of the scattering operator on the initial field can be “guessed”.λ (t). where all time dependence of the averages comes from the creation and annihilation oper- † ators âk. the free evolution of operators âk0 (zeroth-order solution) is transformed into a kind of linear dynamics of the “relevant” component Ê ss(+)0 (r. When Casado et al.λ 2L where L 3 is the normalization volume.λ . the theory of resonators connect the quantum field with the annihilation operators not via the complex exponentials.3.161) and (2. t) = i (+)  â (t)eik·r . which differs from the free-field Hamiltonian only by the meaning of the creation and annihilation operators. t) + Ê(−) (r. (i) The process of parametric down-conversion We are going to study the process of parametric down-conversion of light in the Hilbert-space representation. 3 k. t) = Ê(+) (r. We believe that such subscripts indicate which part of the field is considered. Equations (2.λ (2. we find it conve- nient to use a scalar approximation well known in classical optics. the electric vector is represented as a sum of two mutually conjugate operators (Casado et al. after we find modal functions that are connected to the linear devices used and to the mirrors. Obviously. with ss being any subscript. We suppose that one or two nonlinear crystals involved in the experiment are described in terms of interaction Hamiltonians. and ωk = c|k|. but via more general modal functions. âk.161)   ωk Ê (r. 1997a) Ê(r. The laser theory and.λ k. We modify also the notation for the resulting field by omitting the initial subscript 0.1 Spatio-temporal Descriptions of Parametric Down-Conversion Experiments In the Hilbert-space representation of the light field. This process can be formalized by a quadratic Hamiltonian. We restrict ourselves to the operator ρ̂ that represents a vacuum state. As we do not study experiments involving polarizing devices. t) of the electric vector via the appro- priate modal functions.

1997a). or just the detectors.. We treat the pump beam as an intense monochromatic wave represented. The interaction does not change the field just in front of the crystal. t).164) The response Ê s(+) (r. t)] for . In order to determine the detection probabilities in the Hilbert-space representa- tion. we adopt the correlation properties (Casado et al. (2.3 Quantum Description of Experiments with Stationary Fields 39 setup used for parametric down-conversion.c. t) = V ei(k0 ·r−ω0 t) + c. t) + e−iω0 t gV Ĝ 0i (−) ˆ † (+) ˆ Ê (r. called “signal” Ê s and “idler” Ê i . t). respectively. 1̂ˆ is the identity superoperator. A nonlinear crystal is pumped by a laser beam V . ks + ki = k0 . it may be interest- ing to imagine Equations (2.163) where V is a complex amplitude of a pump beam.165) i 0s 0i where g is an effective coupling constant. j = 1. Ê 0i(−)j (r. t).c. with average frequencies ωs . Since a pointlike crystal is considered (Casado et al. t) = (1̂ˆ + g 2 |V |2 Ĵˆ ) Ê 0s (+) ˆ Ê (−) (r. t)]† . t). respectively. t) of a nonlinear crystal to the input fields (+) Ê 0s (r. filters. As it is almost at the centre of the crystal. so we can interpret the initial field as the “in” resulting field. ωi and wave vectors ks . 0 ≡ 01 . In a product with the identity operator it may be added to the electric-field operator. ki . are selected by means of apertures. t) and Ê i(−) (r. (r. Ê i(−) j (+) (r. 2. means the complex conjugate term to the previous one.165) at r = 0 without the subscript 0 on the right-hand side. Now. t) is as Ê s(+) (r. It can occur. and Ĝ Ê 0s (r. which substitute expan- sions in annihilation (creation) operators for annihilation (creation) operators ( Ĵˆ yields an expansion in the annihilation operators in the first equation and Ĝ ˆ yields ˆ † (+) an expansion in the creation operators in the same equation).2. which becomes the “out” resulting field. and c. but at the cost of other notation. fulfilling the match- ing conditions ωs + ωi = ω0 . t) and Ê 0i(−) (r. In such a work it has proved convenient to substitute slowly varying amplitudes F̂J(+) (r. called signal and idler. t) + (1̂ + g |V | Ĵ ) Ê (r. j = 1. let us consider two narrow correlated beams. k0 is an appropriate wave vector. In experimental practice two narrow correlated beams. t). in the scalar approximation. by V (r. The relation (2. (2. ω0 is a frequency of the pump beam. t) [ F̂J(−) (r. Let us take the origin of the coordinate system. it differs negligibly from the initial field just behind the crystal.165) can be modified (doubled) so that it relates out- 0s put fields Ê s(+) j (r. producing a continuum of coloured cones around the axis defined by the pump. t) = e gV Ĝ Ê (r. iω0 t 2 2ˆ (−) (2. 2. to input fields Ê 0s j (r. at the centre of the crystal. t). 1997b). t) ≡ [Ĝˆ Ê (−) (r. Ĝ ˆ and Ĵˆ are antilinear and linear superoperators.

with ea being the unit vector in the direction of propagation. which is still an approximation:  rab   r  ab F̂ (+) (rb . t) F̂J(+) (r. (2. i.167) c c where ωa is some frequency appropriate to a light beam and r ab = ea ·(rb −ra ). J = s.172) in the Heisenberg picture. The well-known property of Gaussian random variables A. t). B.169) The relation holds at any point of the outgoing beam. t) F̂i(+) (r. we prefer to characterize the signal and idler field operators just behind the crystal at different times (Casado et al. the relation between them being F̂J(+) (r. the usual theory of detection (by photon absorp- tion) is based on the normal ordering. t) Ê (−) (rb . (2. t  ) Ê (+) (ra . With respect to the cross correlation.168) The following autocorrelations.171) In the Hilbert-space formalism. (2. t ) = gV ν(t − t). Since the vectors whose dot product is taken are usually of the same direction. t  ) = F̂s(−) (r. t. t − exp iωa . 1997a):   F̂s(+) (+) out (0.170) It is useful to know that F̂s(+) (r. and their complex conjugates. The joint detection rate is given by Pab (ra .166) According to Casado et al.40 2 Origin of Macroscopic Approach the amplitudes Ê (+) (−) J (r. t)|0 (2. i. we can use the autocorrelations (Casado et al. (2. vanish: F̂J(+) (r. J = s. where K  = K a K b and K a (K b ) is a constant related to the efficiency of the detector and the energy of a single photon. ABC D = AB C D + AC B D + AD BC . t) F̂i out (0. (1997a) it is essential to use the following relation. (2. t) F̂J(+) (r. (2. (2. i. t) = eiω J t Ê (+) J (r. t)]. If we consider the signal beam emerging from the crystal at different times t and t  . t) F̂i(−) (r. the magnitude of displacement vector may be evoked. most interestingly just behind the crystal. t) [ Ê J (r. t) = F̂ (+) ra . t  ) = K  0| Ê (−) (ra . t  ) = g 2 |V |2 μ J (t  − t). J = s. and D. t  ) Ê (+) (rb . t  ) = 0. t  ) = 0. C. rb . 1997a): F̂J(−) (r.173) .

(2.176) − w2 − w2 with w the coincidence window which we choose to be w = 13 × 10−9 s. Let us remember that for σ −1  w. = erf √ + w + erf √ w − . σ2 h − d 2 = exp − 2 c      1 σ d +h σ d +h × erf √ + w ∓ erf ∓ √ w − .        d d 1 σ 2d σ 2d M . ν(τ ) is a Gaussian.178) π 0 d. t  ) = K  | Ê (+) (ra .175) Rabmax + Rabmin where w w 2 2 Rabmax = Pabmax (τ  ) dτ  .   1   1 2 4 √ −σ 2 τ 2 ν(τ ) = |ν(τ )| = σe . = K . (2. Rabmin = Pabmin (τ  ) dτ  . c = 2. (2. −w -2   . 1997a) and finally obtain Pab (ra . 2 2 c 2 c (2. . . M . Particularly.2. while the last term is second order in g. (2. 2 erf(x) = √ (2. c c 2 2 c 2 c   σ M(0. d h 2 .180) 2 ..177) Here x 2 e−t dt. 0) = erf √ w . t. t) Ê (+) (rb . . defined in terms of the integral   w . (2.3 Quantum Description of Experiments with Stationary Fields 41 applies not only in the Weyl (Casado et al.ν τ  + d . c .174) We will determine the visibility V  of the intensity interference: Rabmax − Rabmin V = . we have erf(±∞) = ±1. The first two terms are fourth order in g.179) K π where σ = 1012 s−1 . 1997a) but also in the normal ordering and entails that the joint detection rate is written in three terms. c . t  ) |2 . We may discard the first two terms (Casado et al.998 × 108 ms−1 is the speed of light.  . rb . . . dτ  c c . h are parameters of an experimental setup.  .ν τ  + h .

2. j = 1. 0. A report on the experiment was brief (Ghosh and Mandel 1987).42 2 Origin of Macroscopic Approach (ii) Experiment on the interference of signal and idler photons Let us start with an experiment demonstrating the coherence properties of the parametric down-conversion photon pairs as proposed in Ghosh et al. Fig. (2. z Ms j ). When two detectors are put on the screen one can show a fourth-order. j = 1. 0Mi j = (−b. We consider the initial electric field in the form 2  . the tracing of the beams is not so evident. It is assumed that ωs = ωi = ω20 and the signal and idler beams are directed to a screen by means of two mirrors. and we modify the well-known result (Casado et al. As seen from Fig. 1986). z Mi j ). 2. we will specify modal functions and nonlinear dynamics of field operators.181) 2b − x j 2b + x j with d being the distance from the centre of the crystal to the screen and b being the distance from the axis of the pumping to the mirrors.1.1 Experimental setup on interference on a screen In what follows. Ghosh et al. We introduce the notation for the points of reflection 0Ms j = (b. where bd bd z Ms j = . There is no second-order interference between the two beams. z Mi j = . or intensity–intensity. 0. 2. interference. 1997a. 2. (1986).

t) = vs j k (r)âs j k0 (t).184) k∈[k]i j . Ê 0(+) (r.183) k∈[k]s j  Ê i(+) j0 (r. k z ). z). t) + Ê (+) ij0 (r. t) = V (+) (r. (2. k = (k x . t)1̂ + Ê s(+) j0 (r. (2. k y . t) = vi j k (r)âi j k0 (t). t) . (2.182) j=1 where r = (x. y. and  Ê s(+) j0 (r.

187) with fields we denote as Ê (+) J 0 (r. is of the form  ωk v J k (r) = i exp(ik · r) for r  k. t) = Ê s(+) j0 (r. i j . 0) for J = s1 . After switching on the nonlinear interaction. J = s j . 2. where G j and J j are appropriate linear .3 Quantum Description of Experiments with Stationary Fields 43 with the orthonormal systems of functions vs j k (r). (1997a). t) + e−iω0 t gV Ĝ j j0 j sj0 + g 2 |V |2 Ĵˆ j Ê i(+) j0 (r. 2. j = 1. k ∈ [k]s j . 0J = (−2b. respectively. t)]† . k ∈ [k]i j . j = 1.188) whereas for z > 0 provided that g|V |  1. Compare Casado et al. k z ). (r. k ∈ [k]i j . t)]† . 0. t) for z < 0. the perturbative approximation of the solution of the Heisenberg equations of motion that retains terms up to g 2 can be written as Ê s(+) (r. k = (−k x . vi j k (r). we associate the signal and idler modal functions (2. 2. vs j k (r)2 = |vs j k (r)|2 d3 r = ωk . s2 . k y . (2. t) = [ Ê s(+) j0 (r. AL  ωk v J k (r) = −i exp[ik · (r − 0J )] for (r − 0J )  k . t). 2. 0. 2. t). J = s j . t). k ∈ [k] J . 0J = (2b. i j .186) The [k]s j is a set of integer multiples of the vector 2π e . 0) for J = i1 . (2. t) = Ê s(+) ˆ Ê (−) (r. part of the field is not influenced: Ê s(+) j (r. Similarly for [k]i j . es j being a unit vector of L sj the signal beam at the origin. j = 1. t) + g 2 |V |2 Ĵˆ Ê (+) (r. which substitute and Ĝ j j the expansions in annihilation operators ( Ĵˆ j yields an expansion in the annihilation operators) for annihilation (creation) operators (Ĝ ˆ yields an expansion in creation j operators). (2.2. k ∈ [k] J . Ê i(−) j0 (r. k ∈ [k]s j . A formal expression for v J k (r). j = 1.185) vi j k (r)2 = |vi j k (r)|2 d3 r = ωk .191) ˆ and Ĵˆ are antilinear and linear superoperators. A is the effective transverse area of the beam. (2. (2. In a standard fashion.189) j j0 j ij0 j sj0 Ê i(+) (r. AL z > z M J . t) (r.190) where Ê s(−) j0 (r. i2 .187) where J = s j . t) = Ê i(+) ˆ Ê (−) (r. z < z M J . i j . Ê i(+) j (r. t) + e−iω0 t gV Ĝ (2. t) = [ Ê i(+) j0 (r. t). t) = Ê i(+) j0 (r. (2. j = 1.

195) Supposing that in the sense of classical nonlinear optics.198) [k ]s j [k ]i j which is a great unexpected simplification. t − . k )u (ω0 − ωk − ωk  ) [k ] 2 sj † × exp [it(ω0 − ωk − ωk )] â j0k (t) for k ∈ [k]i j .197) j j0k [k ]s j ⎡ ⎤   Ĵˆ j â j0k (t) = ⎣ f (k.196) we easily obtain that  † ˆ â (t) = Ĝ f (k. t).44 2 Origin of Macroscopic Approach operators acting on functions of the argument r for complex-valued functions.192)  Ĵˆ j â j0k (t) = f (k.194) x â j0k (t) = â j0k (0)e−iωk t . (2.199) we express the field at a point r j . j = 1. (2. (2. (2. (2. k )â j0k (t) for k ∈ [k]i j . k ) f ∗ (k .193) respectively. t exp i c 2c  ri j   ω0 r i  (+) + F̂i j out 0. (2. t) = exp (iωs t) Ê s(+) j (r. 2. Introducing F̂s(+) j (r. (2. j = 1. k ) is a distribution with a support determined by the condition ω0 − ωk − ωk = 0. k )⎦ â j0k (t) for k ∈ [k]s j . Further we will express the intensity correlations that have been determined in the experiment. (2. on the screen as  rs j   ω0 r s  F̂ (+) (r j . f (k. k ) [k ]i j [k ]s j   Δt Δt ×u (ωk + ωk − ω0 ) u   (ωk − ωk ) exp [it(ωk − ωk )]  2 2 × â j0k (t) for k ∈ [k]s j . with sin x ix u(x) = e . k ) f ∗ (k . t) = F̂s(+) − j j out 0. The ˆ and Ĵˆ have the properties superoperators Ĝ j j   ˆ â (t) =  Δt Ĝ j j0k f (k. 2.200) c 2c . j exp i (2.

we can obtain the relations (2. t  ) F̂s(+) 2 out (0. t1 ) F̂ (+) (r2 . By taking into account the correlation relations F̂s(+) 1 out (0. t) = gV ν(t  − t).3 Quantum Description of Experiments with Stationary Fields 45 where / / rs j = rs (x j ) = 2 zM s + b2 + (d − z Ms j )2 + (b − x j )2 . Hence. 1997a) F̂ (+) (r1 . Casado et al. t) F̂i(+) 2 out (0.202) we get (cf.209) below. t  ) = F̂i(+) 1 out (0. j = 1.2. (2. 2. j / / ri j = ri (x j ) = 2 zM i + b2 + (d − z Mi j )2 + (b + x j )2 . (2. t2 )   rs ri  ω 0 .201) j and the subscript out indicates that the field behind the nonlinear crystal is consid- ered.

= gV ν t2 − t1 + 1 − 2 exp i (ri2 + rs1 ) c c 2c  rs2 ri1  ω .

 0 + ν t1 − t2 + − exp i (ri1 + rs2 ) . + . r2 ..  .ν τ2 − τ1 + 1 − 2 . t + τ2 ) ≈ K  g 2 |V |2 . t + τ1 . rs ri .2 × . (2. we finally get (cf..203) c c 2c Assuming that the beams with different subscripts j are mutually uncorrelated.2 .  rs ri . Casado et al.. 1997a) P12 (r1 .ν τ1 − τ2 + 2 − 1 . c c c c  rs ri   rs ri  + 2Re ν τ2 − τ1 + 1 − 2 ν ∗ τ1 − τ2 + 2 − 1 c c c c ω .

(2. J = 1. 2η1 2η2 K  = K1 K2. (2. K2 = .  0 × exp i (ri2 + rs1 − ri1 − rs2 ) . 2. .204) 2c where K  is a constant related to the efficiency of the detectors. K1 = .205) ω0 ω0 η J . is the efficiency of the detector D J .

whether it is associated only with the extremes of the detection rate. The short oscillation period corresponds to the change in the signed distance Fig.206) c c c c   rs1 − ri2 ri1 − rs2 Rsimax − Rsimin = 4g 2 |V |2 M .e. The cosine C of the phase depends only on the signed distance of the detectors .1. 2. +M . j = 1. An equal phase is assumed on any of the lines y2 ≡ x2 −x1 = constant. (2. (2. ri j (instead of the relation (2. 2.2 The dependence of the cosine C of the phase on the position coordinates x1 and x2 for d = 1 and b = 0. . / ri j = ri (x j ) = (2b + x j )2 + d 2 . x1 and x2 for which Rsi = Rsimax or Rsi = Rsimin .201)) were as (Ghosh et al.179) we are interested in C = ±1.209) In Fig.207) c c We may ask whether the visibility has its proper meaning for all x1 .175). 2.204) and the choice (2. The analysis of the setup in Fig. i.1 after Ghosh et al. where   ω0 ri2 + rs1 ri + rs2 C = cos − 1 .46 2 Origin of Macroscopic Approach The visibility is expressed by the formula (2. where Rsimax + Rsimin = 2g 2 |V |2     rs1 − ri2 rs1 − ri2 ri1 − rs2 ri1 − rs2 × M . (1986).2 a short period of the oscillations of the cosine is depicted after Ghosh et al. x2 . (2. . The original formulae for rs j . 1986) / rs j = rs (x j ) = (2b − x j )2 + d 2 . (2. 2. (1986).208) 2 c c with ω0 = 2πc351 × 109 Hz. By relation (2.

3 Quantum Description of Experiments with Stationary Fields 47 y2 ≡ x2 − x1 .162).0015 the visibility almost vanishes. and (2.161). (2. . k ∈ [k]i j .1. On the paths to the mirror M2 they increased the phase of the signal by ϕs and that of the idler by ϕi . (1986) (iii) The experiment of Rarity and Tapster Rarity and Tapster (1990) demonstrated a violation of Bell’s inequality using phase and momentum of photon pairs instead of polarization as in previous exper- iments.4 Experimental setup of Rarity and Tapster Now we will determine nonlinear dynamics of field operators. The distinction is in the sets [k]s j and [k]i j . The analysis of the setup in Figure 2. versus the position coordinates x1 and x2 for d = 1 and b = 0. 2. 2.165).3 The complement of the visibility.3. which are appropriate to the experimental setup. with another orthonormal system of functions vs j k (r). We assume the electric field in the form (2.2.4). j = 1. They directed one of the signal beams and one of the idler beams to a mirror M1 and another mirror M2 (see Fig. As obvious from Fig.1 after Ghosh et al. Fig. vi j k (r). depends only on the coordinate of the point amid the detectors D1 and D2. They selected two signal beams of the same colour (the frequency ωs ) and two idler beams also of the same colour (frequency ωi = ωs ). Fig. 1 − V  . the complement of the visibility. They coherently mixed the two signals and idlers. k ∈ [k]s j . 1 − V  . 2. An equal visibility V  is assumed on the lines y1 = 12 (x1 + x2 ) = constant. For y1 = 0. 2. 2.

We have located the spot of reflection of the idler beam and the spot of reflection of the signal beam. (2. i1 . z Ms2 ). z < z PS J1 . 0. z Ms1 . (2. z BSi ).216)  ωk v J k (r) = −i exp{i[k · (r − 0J ) + ϕ J1 ]} for (r − 0J )  k .  ωk v J k (r) = i exp(ik · r) for r  k. J = s j . AL c z M J < z < z BS J1 . We introduce the notation z BSi . z Ms2 . 0. J = s1 . 0) in the mirror M2 . 0. 2. We specify that  ωk v J k (r) = i exp(ik · r) for r  k. k ∈ [k] J . AL z M J < z < z BS J1 .217) where J = s2 . z PSs . (b. 0. i2 . respectively. . k ∈ [k] J . which is assumed to be simple so far. (−b. J1 = s. z Mi2 ).211) z Mi2 z Ms2 The origin 0 has its image 0s2 = 0i2 = (−2b. i j . 0BSs = (0. 0. J1 = s. (2. z Ms1 ). k ∈ [k] J . 0. k = (−k x . k ∈ [k] J . k z ). AL z PS J1 < z < z M J . (2. 0. the phase shifter for the signal beam. (2. j = 1. i.210) We respect the phase shifters and the spots of reflection on the mirror M2 with the notation z PSi . (2. −b . 0.48 2 Origin of Macroscopic Approach We will give a specification of v J k (r). We respect the spots of reflection on the mirror M1 with the notation z Mi1 . respectively. and the spot of reflection of the signal beam. z Mi1 ).214) where δx is a path-length difference. z < z M J . the spot of reflection of the idler beam. i. (−b. at (b. (2. z BSs such that 0BSi = (0. 0) in the mirror M1 . (2. 0BSs with the beam splitters. 0. z Mi2 . We associate points 0BSi . Then we assume that the mirror M1 is simple. z PSs . at 0 1 0 1 z PSi z PSs −b . 0. k y . We have located the phase shifter for the idler beam. z PSi .215) AL  ωk v J k (r) = i exp[i(k · r + ϕ J1 )] for r  k.213) AL   ωk   ω J1 δx v J k (r) = −i exp ik · (r − 0 J ) + for (r − 0J )  k .212) The origin 0 has its image 0s1 = 0i1 = (2b. k ∈ [k] J . z BSs ).

 ωk vs1 k (r) = −i exp[ik · (r − 0s1 )]ts AL  ωk −i exp{i[k · (r − 0BSs ) + k · (0BSs − 0s2 ) + ϕs ]}rs . e. ti and reflectivities ri . t). i2 .  ωk vs2 k (r) = −i exp{i[k · (r − 0BSs ) + k · (0BSs − 0s1 )]}rs AL  ωk −i exp{i[k · (r − 0s2 ) + ϕs ]}ts .3 Quantum Description of Experiments with Stationary Fields 49 Beam splitter BSs with the transmissivities ts . we associate the modal functions.222) AL and for (r − 0s2 )  k .218) AL c  ωk vs2 kin (0BSs ) = −i exp{i[k · (0BSs − 0s2 ) + ϕs ]}. with fields we denote Ê (+)j0 (r. (2.223).221). (2.189). k ∈ [k]s2 .221) AL Performing the exchange k ↔ k in (2. (2. z > z BSs .219) AL Under the assumption δx = 0 we can perform the exchange k ↔ k and write the output values for k ∈ [k]s1  ωk vs1 kout (0BSs ) = −i exp{ik · (0BSs − 0s1 )}ts AL  ωk −i exp{i[k · (0BSs − 0s2 ) + ϕs ]}rs .223) we perform the replacement s ↔ i. (2.213). Nonlinear dynamics is described in the same way as for the experiment on the interference of signal and idler photons by relations (2.217).216). (2.223) AL Beam splitter BSi with the transmissivities ti . z > z BSs . (2. (2. k ∈ [k]s1 .214).220) AL  ωk vs2 k out (0BSs ) = −i exp{ik · (0BSs − 0s1 )}rs AL  ωk −i exp{i[k · (0BSs − 0s2 ) + ϕs ]}ts .222) and (2.215). (2.g. k ∈ [k]s2 . and (2. k ∈ [k]s1 .2. (2. (2. we have for (r − 0s1 )  k . ri can be described analogously: In (2.188). J = s1 . i1 . ts and reflectivities rs . (2. s2 .190). rs : Input values are   ωk   ωs δx vs1 kin (0BSs ) = −i exp ik · (0BSs − 0s1 ) + .222). (2. . In a standard fashion. (2. (2.

t).192).224) where J = s1 . t) and also F̂s2 (r. In the following. t − + exp i + iϕi c ωi c  ri   ωi r i  + ri F̂i(+) 1 out 0. since the upper and the lower mirrors were not at exactly the same distance from the pumping beam axis (Casado et al. J1 = i. If we consider that the time intervals rcs − rci − ωϕss .226) c c with 0 the centre of the coordinate system. are    ωr   (+) rs ϕs s s F̂s(+) (rs .50 2 Origin of Macroscopic Approach (2. we assume δx = 0. The mirror above the pumping beam axis is not simple. t) is correlated with F̂i(+) (+) (+) 1 (r. t) = t F̂ s s2 out 0. rs and ri the path lengths of the lower signal and idler beams. t − exp i . (2. t). and (2. 1997a) . t  − exp i . s2 .193). for expressing the intensity correlations.225) c c    ωr  (+)   (+)  ri ϕi i i F̂i2 (ri . t) is correlated with the idler field F̂i2 (r. t ) = ti F̂i2 out 0. respectively. respectively. We will take into account that the signal field F̂s(+) 1 (r. (2. to the appropriate detector. t − + exp i + iϕs 2 c ωs c  rs   ωs r s  + rs F̂s(+) 1 out 0. we obtain   r . We introduce the slowly varying field operators F̂J(+) (r. t) = exp (iω J1 t) Ê (+) J (r. 2. J1 = s and J = i1 . j = 1. In the experiment under consideration both upper paths were modified by δx. It allows the fields with z < 0 to stay initial and those with z > 0 at least to obey the same rules we have used to calculate the modal functions. these pairs not fulfilling matching conditions. The field operators at the signal and idler detectors placed at rs . t) is uncorrelated with F̂i(+) j (r. t). ri . but F̂s(+) j (r. rcs − rci + ωϕii are small in comparison with the coherence time of signal and idler given by the function ν(τ ). but a mirror assembly which enables one to change δx (Rarity and Tapster 1990). i2 . (2.

t  ) ≈ ri ts exp i ωs + ωi + ϕs s F̂s(+) (rs . ri (ri . t) F̂i(+) 2 2 c c  r ri .

t + τ  ) = K  g 2 |V |2 |ν(τ  − τ )|2  . t + τ . (2.227) s c c From this we have Psi (rs .  + rs ti exp i ωs + ωi + ϕi ν(t  − t). ri .

(2.228) . × |ri ts |2 + |rs ti |2 + 2Re{r∗s ts ri ti∗ exp[i(ϕs − ϕi )]} .

where   Rsimax + Rsimin = 2g 2 |V |2 M (0. (2.6).229) Rsimax − Rsimin = 8g 2 |V |2 M (0. |ti | = 0 and |ts | = 1.230) The dependence of the visibility on the beam splitters + with the transmissivities + |ts | ∈ [0. |ti | = 1. 2. (2. 1] and the reflectivities |rs | = 1 − |ts |2 .5 The visibility V  versus moduli of the amplitude transmissivities ts . where ΔL s (ΔL i ) is the length difference between the long (short) route of the signal (idler) beam through the corresponding inter- ferometer. Fig. |ti | ∈ [0. He arranged two Mach–Zehnder interferometers and let a signal and an idler beam each pass through an interferometer (see Fig. The surface plotted has a boundary condi- tion zero. The experiment was originally proposed for an atom and free-space propagation. . 0) |ts ri ||rs ti |. |ri |= 1 − |ti |2 is plotted in Fig. 2. The maximum is attained on the line segment connecting the points |ts | = 0. 2.5. In Tapster et al. (1994) Franson’s experiment has been adapted to parametric down-conversion and fibres. 0) |ts ri |2 + |rs ti |2 .2.175).3 Quantum Description of Experiments with Stationary Fields 51 The visibility is expressed by the formula (2. It may be equal to unity in the sense of the equality |ts ri | = |ti rs |. ti from the “lower side” of the beam splitters for signal and idler beams for δx = 0 Unfortunately. The coincidence detection shows a fourth-order interference as a cosine depen- dence on 1c (ωs ΔL s + ωi ΔL i ). 1]. (iv) The experiment of Franson Franson (1989) proposed a test of “Bell’s inequality for energy and time”. the quantity under consideration is not dependent on the charac- teristic of the nonlinear optical process. In the past few years several groups have performed experiments of that type. The interference manifests itself as a cosine variation of the coincidence rate with ϕs − ϕi .

i. J = s. t) + Ê s(+) BS 0 (r. (2. sBS1i . k ∈ [k]iBS1s .231) 1i where  Ê (+) J 0 (r. z BS1s < z < z BS2s . k ∈ [k]s .232) k∈[k] J The modal functions as restricted to linear segments are  ωk ik·r vsk (r) = i e for r  k. t)1̂ + Ê s0 (+) (r.235) . k ∈ [k]s .234) AL Here 0BS1s is the centre of the beam splitter BS1s .233) AL  ωk ik·r viBS1s k (r) = i e for (r − 0BS1s )  k. and [k]iBS1s is the set of wave vectors k of the beam corresponding to the unused input port of this beam splitter.52 2 Origin of Macroscopic Approach Fig. t) 1s + Ê i0(+) (r. z BS1s is the corresponding z-coordinate. we consider the initial electric-field in the form Ê 0(+) (r.6 Experimental setup of Franson’s type. The modal functions at the output of this beam splitter are   ωk ik·r ωk ik·(r−0BS1s iBS ) vsk (r) = i e ts + i e 1s r s AL AL for r  k. t) + Ê i(+) BS 0 (r. z < z BS1s . (2. (2. For simplicity Ê iBS 0 ≡ Ê iBS1s 0 and Ê sBS 0 ≡ Ê sBS1i 0 For the description of the nonlinear dynamics of field operators. t). (2. t) = V (+) (r. 2. z > z BS1s . (2. iBS1s . t) = v J k (r)â J k0 (t).

236) where 0BS1s iBS .239) where z M2s is the z-coordinate of the centre of the second mirror and 0M1s BS1s s and 0M1s are chosen such that k · (r − 0M1s BS1s s ) = k · (r − 0M1s ) + k · (0M1s − 0BS1s s ). k ∈ [k]s . 0BS1s s are chosen such that 1s k · (r − 0BS1s iBS ) = k · (r − 0BS1s ) + k · 0BS1s .238) Here z M1s is the z-coordinate of the centre of the signal mirror and z BS2s is the z-coordinate of 0BS2s . AL z M1s < z < z M2s . (2. k ∈ [k]iBS1s . k ∈ [k]iBS1s . z M1s < z < z BS1s . k ∈ [k]iBS1s .244) . (2. (2. (2. k ∈ [k]iBS1s . (2.241) After the reflection from the second mirror.237) 1s k · (r − 0BS1s s ) = k · (r − 0BS1s ) + k · 0BS1s . k ∈ [k]iBS1s . AL z M2s < z < z BS2s .242) where 0M2s M1s BS1s s and 0M2s M1s are chosen such that −k · (r − 0M2s M1s BS1s s ) = −k · (r − 0M2s ) +k · (0M2s − 0M1s BS1s s ). (2.240)  k · (r − 0M1s )  = k · (r − 0M1s ) + k · 0M1s .243) −k · (r − 0M2s M1s ) = −k · (r − 0M2s ) +k · (0M2s − 0M1s ). After the reflection from the first mirror. k ∈ [k]iBS1s . k ∈ [k]iBS1s .3 Quantum Description of Experiments with Stationary Fields 53   ωk ik·(r−0BS s ) ωk ik·r viBS1s k (r) = i e 1s r + i s e ts AL AL for (r − 0BS1s )  k. (2. k ∈ [k]iBS1s . the modal function is  ωk −ik·(r−0M M BS s ) viBS1s k (r) = i e 2s 1s 1s r s AL  ωk −ik·(r−0M M ) +i e 2s 1s t for (r − 0 s M2s )  −k. (2. (2. the centre of the beam splitter BS2s .2. the modal function is  ωk ik ·(r−0M BS s ) viBS1s k (r) = −i e 1s 1s r s AL  ωk ik ·(r−0M )  −i e 1s t for (r − 0 s M1s )  k .

Ê i(+) BS1s 0 (r. k ∈ [k]s . we associate the modal functions which travel to the above detector. (2.  ωk ik·(r−0BS1s iBS ) ωk ik·(r−0BS M M ) + i e 1s + i e 2s 2s 1s ts rs AL AL  ωk ik·(r−0BS M M BS s ) 2 +i e 2s 2s 1s 1s r . Switching on the nonlinear interaction. (2. Ê s(+) BS1i 0 (r. (2. (2. t). (2. . J = s.245). (2. and (2.233). Counter to propagation the field stays initial and along with propagation it at least obeys the rules we have used to generate the modal functions.245) AL s where 0BS2s M2s M1s and 0BS2s M2s M1s BS1s s are chosen such that k · (r − 0BS2s M2s M1s ) = k · (r − 0BS2s ) − k · (0BS2s − 0M2s M1s ). AL s z > z BS2s . we find the field to obey the relations (2.234).  ωk −ik·(r−0BS s ) ωk −ik·(r−0M M BS ) + i e 2s +i e 2s 1s 1s rs ts AL AL  ωk −ik·(r−0M M ) 2 +i e 2s 1s t for (r − 0 BS1s )  −k. k ∈ [k]iBS1s . For simplicity.189) and (2. k ∈ [k]iBS1s . (2. which travel to the lower detector.235). Exchanging s ↔ i. i.249) 1s −k · (r − 0BS2s s )  = −k · (r − 0BS2s ) + k · 0BS2s .239).190).248).54 2 Origin of Macroscopic Approach The output modal functions for the beam to be detected are  ωk ik·r 2 vsk (r) = i e ts - AL . (2. (2. (2.246) k · (r − 0BS2s M2s M1s BS1s s ) = k · (r − 0BS2s ) −k · (0BS2s − 0M2s M1s BS1s s ).248) where 0BS2s BS1s iBS and 0BS2s s are chosen such that 1s −k · (r − 0BS2s BS1s iBS ) = −k · (r − 0BS2s ) 1s +k · (0BS2s − 0BS1s iBS ).236). z > z BS2s . it is assumed that t J = tJ = r J = rJ = √12 .247) The output modal functions for the second beam are  ωk −ik·(r−0BS2s BS1s iBS ) 2 viBS2s k (r) = i e 1s r - AL s .242). k ∈ [k]iBS1s . We relate them with fields we denote as Ê i0(+) (r. t). t).250) In a standard fashion. t). k ∈ [k]s . (2. (2. k ∈ [k]s . (+) with fields we denote as Ê s0 (r. (2. we introduce modal functions. for r  k.

t − − exp iωs 1s c c c     (+) |rs − rBS2s | L s. t) = exp (iωs t) Ê s(+) (r. t) = exp (iωi t) Ê i(+) (r. t − − − exp iωs c c c c     |rs − rBS2s | L s. s↔i Let us denote L s. (2.253) .long L s.short |rBS1s | |rBS1s | + F̂sout 0. (2. t − − − exp iωs 2 c c c c     (+) |rs − rBS2s | L s. (2. . ri .251) The field operators at the signal and idler detectors placed at rs .. t)      1 (+) |rs − rBS2s | L s. Supposing that ΔL s ≡ L s. t − − exp iω s c c c   |rs − rBS2s | × exp iωs .short (ΔL i ≡ L i.short + i F̂i(+) BS1s in 0BS1s .long − L i.long + |ri − rBS2i | c   ΔL i − ΔL s i  + ν τ − τ + exp ωs |rBS1s | + L s. we introduce the slowly varying field operators F̂s(+) (r. t).long − i F̂iBS in 0BS1s . we get (Casado et al. t + τ  ) = gV 4  i   × ν(τ  − τ ) exp ωs |rBS1s | + L s.2. F̂i(+) (r.short c c   i   + |rs − rBS2s | + ωi |rBS1i | + L i. are F̂s(+) (rs .short (L i.long |rBS1s | |rBS1s | = F̂sout 0. t) = F̂s(+) (rs .long + |rs − rBS2s | c  i   + ωi |rBS1i | + L i. t). t).long − L s.254) c . t + τ ) F̂i(+) (ri .short ) is much greater than the coherence length of the signal (idler) in order to avoid the second-order interference.3 Quantum Description of Experiments with Stationary Fields 55 For the calculation of intensity correlations determined in the experiment. respectively.short ) a length of the short arm of the interferometer for the signal (idler) beam.short L s.short + |ri − rBS2i | .252) c F̂i(+) (ri . (2. 1997a) 1 F̂s(+) (rs .

(2.. where Rsimax + Rsimin   1 ΔL i − ΔL s ΔL i − ΔL s = g 2 |V |2 M (0. (2. Fig.7. We finally obtain  1  2 2 Psi (rs . but the function erf does not contribute to it.175). t + τ . ΔL i − ΔL s .256) 8 c c   1 2 2 ΔL i − ΔL s Rsimax − Rsimin = g |V | M 0. t + τ  ) = K g |V | |ν(τ  − τ )|2 16 . . 0) + M . The interference manifests itself as a cosine variation of the coincidence rate with 1c (ωs ΔL s + ωi ΔL i ).257) 4 c The dependence of the visibility on the difference (ΔL i −ΔL  s) is plotted in Fig. hc . . 2.2 + . . The variation of the visibility is due to the function M dc . ri . (2.. The distance between the points of inflection is 2 σc = 6×10−4 m. 10−3 ] measured in metres .  .7 The visibility V  versus the difference (ΔL i − ΔL s ) ∈ [−10−3 .56 2 Origin of Macroscopic Approach provided that |rBS1J | + L J. c     ∗  ΔL i − ΔL s − 2Re ν(τ − τ )ν τ − τ + c   i × exp (ωs ΔL s + ωi ΔL i ) .ν τ  − τ + .short + |r J − rBS2J | is the same for J = s and J = i.255) c The visibility is given in (2. 2.

t) + Ê i0(+) (r. i. (2. For simplicity. (2. z < z BSs . 2. j = 1. but we consider more general ones. i2 are aligned and the path differ- ence between the two signals is varied slightly. The experimental setup is outlined in Fig. we will specify modal functions and the nonlinear dynamics of field operators. z < z Ms . t) = v J k (r)â J k0 (t).258) We assume that V1 = V2 exp(ik0 · 02 ) = V . s2 . We are interested in the joint detection rate of the detectors Ds and Di when the trajectories of the two idlers i1 . 2. sBSi . 2. J = s1 . We consider the initial electric field in the form  2 Ê 0(+) (r. when the paths of the idler photons are aligned. t) + Ê s(+) BS 0 (r.8. t).2. Fourth-order interference disappears when the idlers are misaligned or separated by a beam stop. k ∈ [k]s . Fig. t) = V j exp[i(k0 · r − ω0 t)].262) AL . (1991) performed an experiment in which fourth-order interference is observed in the superposition of signal photons from two coherently pumped para- metric down-conversion crystals.260) k∈[k] J The modal functions as restricted to linear segments are  ωk ik·r vs1 k (r) = i e for r  k. (2. The parametric down-conversion occurs at both crystals. each with the emission of a signal photon and an idler photon. Ê sBS 0 ≡ Ê sBSi 0 In what follows. there were simi- lar crystals in the experiment.261) AL  ωk ik·r vs2 k (r) = i e for (r − 02 )  k. (2. classical pump waves of complex amplitudes V j (r.259) i j=1 where  Ê (+) J 0 (r. t)1̂ + Ê s(+) j0 (r. k ∈ [k]s .3 Quantum Description of Experiments with Stationary Fields 57 (v) Induced coherence and indistinguishability in two-photon interference Zou et al. On the contrary.8 Experimental setup on induced coherence without induced emission. in which two nonlinear crystals NL1 and NL2 are optically pumped by two mutually coherent. t) =V (+) (r. (2.

z < z BSi . k ∈ [k]i . k ∈ [k]sBSi .270) AL   ωk ik·r ωk ik·(r−0BS s )  vik (r) = i e t+i e i BSi r AL AL for r  k.266) where 0BSs and 0BSs Ms have been chosen so that. k ∈ [k]s . z > z BSi . k ∈ [k]i . 2 r = r = √i 2 . k ∈ [k]s . k · (r − 0BSs ) = k · (r − 0BSs ) + k · 0BSs . (2. (2. z > z BSs . (2.58 2 Origin of Macroscopic Approach  ωk ik ·(r−0M )  vs1 k (r) = −i e s for (r − 0 Ms )  k . as in the definitions related to Fig. (2. (2. k ∈ [k]s . 2. (2.222) and (2. respectively.273) k · (r − 0BSi i )  = k · (r − 0BSi ) + k · 0BSi for k ∈ [k]sBSi .264) Similarly as in (2. k ∈ [k]s . respectively.272) where k is defined relative to the beam splitter BSi and 0BSi i and 0BSi sBSi have been chosen so that.274) .263) Here 0Ms and k are used as in the definition of modal functions related to Fig.1 and k has the same meaning. z > z BSi .223) for t = t = √1 . (2. k · (r − 0BSi sBSi ) = k · (r − 0BSi ) + k · 0BSi i for k ∈ [k]i . (2. Here. k ∈ [k]sBSi .265)   ωk ik·(r−0 ) i ωk ik·r 1 vs2 k (r) = −i e BS s M s √ +i e √ AL 2 AL 2 for (r − 02 )  k.268)  ωk ik·r vik (r) = i e for r  k.271)   ωk ik·(r−0BS i ) ωk ik·r  vsBSi k (r) = i e i r + i e t AL AL for (r − 0BSi )  k.4. (2. we specify that 0Ms has been chosen so that k · (r − 0Ms ) = k · (r − 0Ms ) + k · 0Ms . (2. z > z BSs . we have   ωk ik ·(r−0M ) 1 ωk ik ·(r−0BS ) i vs1 k (r) = −i e s √ +i e s √ AL 2 AL 2  for (r − 0Ms )  k . (2. 2. (2. k ∈ [k]s . k ∈ [k]s .269) AL  ωk ik·r vsBSi k (r) = i e for (r − 0BSi )  k.267) k · (r − 0BSs Ms )  = k · (r − 0BSs ) + k · 0Ms . z < z BSi . AL z Ms < z < z BSs .

190). The field operators at the signal and idler detectors placed at rs . which would lead to the use of the subscript 1. 2 2 2 It also depends on the pump beam at the same crystal. are     1 d d F̂s(+) = √ − i F̂s1 (rs . t) (+) 01 .191) so that they confess the first-stage operators on the left-hand side. (2. Now we would like to interpret the subscript 0 not as the order of solution but as a number of the initial stage. (2.272) travel to the lower detector.276) c c We still assume different crystals and derive a slight generalization of the well- known experiment (Casado et al. t). J1 = s. t). we introduce the operators (2.190).271). (2.189). Further we will express the intensity correlations that have been determined in the experiment. Again.270).2. t  ) F̂s(+) 2 (02 . and (2. (2. we find the field to obey the relations (2.278) .275) 2 c c     (+)  (+)  l l F̂i (ri . i. t ) = tgV1 ν1 t − t − exp iωi . and (2. i. t − exp iωi . We relate them with fields we denote as Ê i0 (r.269). Counter to propagation the field remains initial and along with propagation it at least obeys the rules we have used to generate the modal functions. where J = s1 . s. (2. and (2.277) 1 c c F̂i(+) (02 . Switching on the second nonlinear interaction. (2. t − exp iωs . By taking into account the correlation relations     (+)   f f F̂s(+) (01 .3 Quantum Description of Experiments with Stationary Fields 59 The modal functions (2. s2 . t) = gV2 ν2 (t  − t). The action of the operators Ĝ ˆ and Ĵˆ depends on the nonlinear crystal located at 0 . Switching on the first nonlinear interaction. Ê sBSi 0 (r. Counter to propagation the field stays first stage and along with propagation it still at least obeys the rules to generate the modal functions.189).190). t ) = F̂i 02 . (2. t − exp iωs 2 2 c c     h h + F̂s(+) 02 . respectively. and (2. t) F̂i (02 . (2.191) for j = 1 with i1 → i. we find the field to obey the relations (2.191) for j = 2 with i2 → i. 1997a). ri . (2.189). but the first-stage field operators to have been substituted for the operators on the right-hand side. Since the stage is followed by the first stage. we would like to modify the relations (2.224).

The visibility is expressed by the formula (2. In both cases the idler photon may arrive at the detector. As the two possibilities are indistinguishable . and put a detector into the reflected idler beams (see Fig. as can be seen from Fig.175). it depends on the transmis- sivity of the beam splitter. 1997a) gV F̂s(+) 2 (rs . The interference manifests itself as a cosine variation of the coincidence rate with ωc0 f .2 . l f d .10). signal. t + τ  ) = K  g 2 |V |2 2 .  . c c .  . l h . NL. 0) |t|. (vi) Frustrated two-photon creation via interference Herzog et al. where ν1 (τ ) = ν2 (τ ) ≡ ν(τ ). that emerge from a nonlinear crystal. 0) |t|2 + 1 .2 . In the standard quantum interpretation a pair of correlated photons can be created either by the laser beam travelling from left to right or when the reflected laser beam travels from right to left. laser. ri .281) Rsimax − Rsimin = 2g |V | M (0.282) The maximum visibility is equal to unity and.ν2 τ − τ − + c c c . t) F̂i(+) (ri . l = h   Rsimax + Rsimin = g 2 |V |2 M (0. They placed three mirrors in the three beams..60 2 Origin of Macroscopic Approach we get (Casado et al. (2. t  ) = √ 2      l f d i × − itν1 τ − τ − − + exp [ωs d + ωi (l + f )] c c c c     l h i + ν2 τ  − τ − + exp [ωs h + ωil] ..280) c We have hopefully corrected the factor.279) c c c We finally obtain 1 Psi (rs . 2 2 (2. (2. and changed signs of the argument of ν(τ ) relying on the identity ν(τ ) = ν(−τ ). in general. and idler. (2. changed a sign with respect to the reflection from the mirror. . 2.tν1 τ − τ − − +  + . × . 2.     l f d l h + 2Im tν1 τ  − τ − − + ν2∗ τ  − τ − + c c c c c    i × exp [ωs (d − h) + ωi f ] . (1994) performed a simple experiment interpreted as showing inter- ference of two processes. where for d = l + f .9. . t + τ ..

ϕ0 = 2|k0 |l0 = 2 ω0cl0 . k ∈ [k]s . It is assumed that d =l + f.3 Quantum Description of Experiments with Stationary Fields 61 Fig. Accordingly the description of the pump beam is given by V (+) (r. t). we must consider that (+) (+) (+) Ê s0 (r. (2. 1] of the beam splitter BSi . The modal functions are   ωk ik·r ωk iϕs −ik·r vsk (r) = i e −i e e 2AL 2AL for r  k. Associating the modal functions with quantum fields.285) .283) where e0 is the direction vector of the forward-propagating pump beam.l = h Fig. 2. z Ms = e0 · 0Ms . t) = Ê sF0 (r.10 Experimental setup on frustrated two-photon creation via interference they interfere and the counting rate oscillates depending on the position of a chosen mirror.284) where ϕs = 2 ωcs ls . t) + Ê sB0 (r.2. (2. 2. t) = V ei(k0 ·r−ω0 t) − V ei[−k0 ·(r−2l0 e0 )−ω0 t] = V ei(k0 ·r−ω0 t) − V eiϕ0 ei(−k0 ·r−ω0 t) for x = 0. z < z Ms . The modal functions vik (r) are expressed similarly. z < l0 . y = 0. (2.9 The visibility V  versus the modulus of the amplitude transmissivity |t| ∈ [0.

287) where e j . t − c   2ˆ (+) 2ls = −(1 + g |V | Ĵ ) Ê sF0 0. (0. (2. idler beam. t). t) = Ê sF0 (0. (2. t) + (1 + g |V | Ĵ ) Ê (+) (0. (+) (+) Ê iF (0 − (r = 0)ei . t) are expressed similarly. (2. t).288)   (+) (+) 2ls Ê sB (0 + (r = 0)es . t − c   = −ei(ϕs +ϕi ) e−iω0 t gV Ĝ ˆ Ê (−) 0. t). t) = −i e e âsk0 (t) (2. t) = Ê iF0 (0. t) sBout sB s +e −iω0 t gV e Ĝ ˆ Ê (−) (0 + (r = 0)e . iϕ0 iB i (+) −iω0 t iϕ0 ˆ (−) Ê iBout (0. t) = (1 + g 2 |V |2 Ĵˆ ) Ê sF0 (+) ˆ Ê (−) (0. The field operators Ê i0(+) (r. (−) 2 iF i sF0 iF0 (2.289) c   (+) (+) 2li Ê iB (0 + (r = 0)ei . and Ê iB0 (r. i. t) = e −iω0 t ˆ 2ˆ gV Ĝ Ê (0. is the direction vector of the forward-propagating signal.62 2 Origin of Macroscopic Approach where  (+) ωk ik·r Ê sF0 (r. t) = e gV e Ĝ Ê (0 + (r = 0)e . t). Nonlinear dynamics is described by the relations (+) (+) Ê sF (0 − (r = 0)es . t − . Ê iF0 (+) (+) (r. t) =i e âsk0 (t). k∈[k]s 2AL  (+)  ωk iϕs −ik·r Ê sB0 (r. t) = − Ê iF 0 + (r = 0)ei . t) + e−iω0 t gV Ĝ iF0 (+) Ê (0 + (r = 0)e . t) = − Ê sF 0 + (r = 0)es . t). (2.286) k∈[k] 2AL s are the forward-propagating component and the backward-propagating component. respectively: (+) Ê sF (0 + (r = 0)es . t). t) sB s + (1 + g |V | Ĵˆ ) Ê iB 2 (+) 2 (0 + (r = 0)ei . t). t − .290) c Ê (+) (0. t − 2 c   i(ϕs +ϕi ) −iω0 t ˆ (−) 2ls −e e gV Ĝ Ê iF0 0. j = s. respectively.291) . t) = (1 + g 2 |V |2 Ĵˆ ) Ê (+) (0 + (r = 0)e . t − 2li sF0 c   2li − (1 + g 2 |V |2 Ĵˆ ) Ê iF0 (+) 0. t).

F̂J(−) (−) F (r. F̂J B (r. we have mostly dealt with the fourth-order interference in para- metric down-conversion experiments.297) c c The photodetection rate in the detector Di is expressed similarly (Casado et al.294)  . J = s. t). The 1986.    2li 2ls Ps (rs . t) . Coincidence measurements in the various setups are essentially (or suffi- ciently well) described in terms of the cross correlation between the signal and the idler. t) . Ê (−) (rs . we express the field operators at the signal and idler detectors placed at rs . Considering the forward propagation. Introducing the slowly varying field operators F̂J(+) (r. ri . (2. 0 (2. t) = F̂ (+) J Bout 0. t) = 2K g 2 |V |2 μs (0) + μs − cos(ϕs + ϕi − ϕ0 ) . t). t). . . t) = e iω J t (+) Ê J B (r. J = s. reflections. t) Ê (+) (rs . F̂sB (rs . In conclusion. (2. F̂J(+) B (r. 1997b). t − exp i . 1990. i. t).296) c c where we have relied on the identity μs (τ ) = μs (−τ ). From this. where only paths through nonlinear and linear optical elements and the free space (with possible reflections from perfect mirrors) to detectors are drawn. We have “promoted” the schemes of the experiments. t) = e iω J t (+) Ê J F (r. i. F̂J(+) F (r. Ps (rs . t) F̂sB (rs . t −   c c 2l i 2l s = 2g 2 |V |2 μs (0) + μs − cos(ϕs + ϕi − ϕ0 ) . (2.3 Quantum Description of Experiments with Stationary Fields 63 Further we will calculate the quantum mean intensities that have been determined in the experiment.293) c c The quantum mean intensity or single photodetection rate in the detector Ds is  . t − F̂sB 0. (−) (+) . respectively. t). = K 0 . t) = F̂sB 0. 1994 (adapted back to free space). . t) = K 0 . t). we obtain that    rs  (+)  rs  (−) (+) (−) F̂sB (rs . t) = eiω J t Ê (+) J (r. . 0 .295) where K is a constant related to the efficiency of the detector and the energy of a single photon. . to a reason of a certain neglect of the . and 1991 experiments were chosen according to a review article of other authors. (2. and the backward propagation.2. (2.292) and F̂J(−) (r. t) F̂sB (rs . as  rJ  ω r  J J F̂J(+) B (r J .

It has been found that the distribution of phase difference derived from the Q function narrows when the connection of both the down-converters via the idler mode closes up. The assumption of the strong idler field is implicit in work contributing to the monochromatic theory. The monochromatic treatment associates each travelling wave with a quantum harmonic oscillator. We have performed con- ventional quantization by introducing annihilation operators in place of the classical complex amplitudes of the modes. the second nonlinear crystal is proven to “ignore” the idler photons. even though completed by optimal scaling of its results. by the hypothesis that there exists an incomplete or relatively complete system of more complicated modal functions. which have still been specified only on the paths. is far from being persuasive here. We have replaced the usual assumption that the electric field is expanded in terms of an incomplete set of plane waves. the exper- iment of Zou et al. When the idler field is weak per frequency unit. A nonclassical distribution of photon-number sums restricted to even sums of photon numbers cannot occur. The monochromatic description. about multiples of the latter. which is relatively com- plete with respect to the expected direction of propagation.2 From Coupled Quantum Harmonic Oscillators Back to Interacting Fields One of the interference experiments we have described in Section 2. the polychromatic theory yields results similar to those obtained by the monochro- matic treatment. (1991) which has been analysed in Wang et al. The spontaneous emission from the first nonlinear crystal in the idler serves as a stimulating idler input to the second nonlinear crystal that acts as an optical amplifier.64 2 Origin of Macroscopic Approach beams’ divergence. has attracted much attention.3. The latter distribution occurs when the idler beam is blocked.e. i. In Řeháček and Peřina (1996) it has been shown that the distribution of photon- number sum in signal modes interpolates between a Bose–Einstein distribution and a convolution of two Bose–Einstein distributions. We tried to choose sufficiently realistic values of the parameters for all the four experiments and to find visibilities of the intensity interference. In this situation. 2.1.3. The simple formalism of several coupled harmonic oscillators is useful for an analysis of the travelling-wave setup of interference experiment due to . In general. The arrangement of two down-converters is pumped by mutually coherent beams and the two down-converters are connected by the idler beam.b). the photon-number sum is distributed as if it corresponded to the number of signal degrees of freedom which varied between 1 and 2. A beam splitter placed between the two nonlinear crystals in the idler beam can change the strength of their connection since it attenuates the emerging field. (1991a. The parametric down-conversion in the second nonlinear crystal is stimulated by idler photons when the idler field is strong “per frequency unit”. The interference of signal beams from both the crystals can be observed. because the correlation between the photon numbers of the two signal beams is not complete.

2. The experimental arrangement consists of two parametric down-conversion crystals with . (1991) up to suppression of the induced emission. If the field used to lock the idler of one down-converter (crystal NL2 in Fig. the interference was lost when one could tell which crystal had emitted each signal photon. (i) Formalism of several modes We turn to the quantum analysis of the Zou–Wang–Mandel experiment. Using multimode analysis of the experiment. 2. To the contrary. The induced emission can also be used in parametric down-conversion to lock the phase of the idler and.11.3 Quantum Description of Experiments with Stationary Fields 65 Zou et al. and Peřinová et al. (1991) and Wang et al. in principle. they will have. (1991a).2. Fig. 2. Wiseman and Mølmer 2000). the two signal fields will also be locked in phase. Such simple models have been published. (2000) and provide what is a continuation of Peřinová et al. (1991a) had a negligible probability of both crystals producing a down-converted photon pair and used a quantum-mechanical explanation based on indistinguishability of paths to explain the interference they observed in the experiment. the two signals will be incoherent and there will be no interference. 1991a. and hence no induced emission. perfect first-order coherence and so will interfere at a final beam splitter not included in Fig. they derived that there could be no induced emission in their experiment. here we try to come near their analysis.11) is itself the idler output of another down-converter (crystal NL1 in Fig. the signal fields from NL1 and NL2 show perfect interference. When the induced emission occurs. Nonetheless. it can be utilized. from this. we may refer to Řeháček and Peřina (1996). that of the signal (since the phase sum of the signal and idler is locked to the pump phase). they found that for perfect matching of idler modes. Wiseman and Mølmer (2000). We try to formalize here a comparison between the two approaches. The phenomenon of induced emission makes the phase of an amplified field adopt the same phase as the incident locking field (Wang et al. Thus. (1991a) yields an explanation involving a sufficient number of realistic parameters. there exists a sim- ple quantal description that may claim that it conforms to results of the multimode analysis.11). However. 2. Even though in the foregoing section the formalism yielded results similar enough to those of Wang et al. The multimode approach to the analysis used by Wang et al. Concerning this. If there is no connection between the two down-converters.11 Scheme of two parametric processes with aligned idler beams with the spatial Heisenberg picture made explicit Zou et al. (2000). The more complicated approach used originally for the analysis seems to be unsuitable for treating the phenomenon of induced emission.

âs1 (1) = âs1 (2) = âs1 (3). respectively. in the Hilbert space of these four modes. j = 0. 1. Of course.298) Having prescribed equations of motion of operators. while in the modified Heisenberg picture it changes like that of the free field. 2. 1. we consider also the appendage input annihilation operators â0 (0). 2000). Here s1 . the change of the state cannot be specified satisfactorily. âi (2). We will write the equation giving the transformation of M̂( j) from its value M̂(0) before the interaction to its value M̂(1) after the action of the first down converter. â0 (2) = â0 (3). j = 0. 2. We consider the input annihilation operators âs1 (0). 2. an arbitrary operator M̂( j). and to its value M̂(3) after the action of the second down converter. the output annihilation operators âs1 (1). i for the idler mode. s1 (the signal mode for crystal NL1). 3. âi (3). but fortunately. i (the idler modes.298) Û j+1 ( j) for j = 0. âs2 (2). â0 (3). and 0 for the “escape” mode of the beam splitter. âs2 (0). âs1 (3). which are partially connected due to the presence of a beam splitter in between. Let us consider.11. to its value M̂(2) after the action of the beam splitter. a jth-stage operator. The appropriate relations read † M̂( j + 1) = Û j+1 ( j) M̂( j)Û j+1 ( j). s2 (the signal mode for crystal NL2). s2 stand for the signal mode of crystal 1 and that of crystal 2. 2. â0 (1). 3. we can describe the system by four modes. which are identified). In our case of the “discrete” space (cf. and is illustrated in Fig. and 0 (the escape mode for the beam splitter) (Peřinová et al. in the description of the dynamics below. and the intermediate annihilation operators âi (1). âs2 (1) and the appendage output annihilation operators âs1 (2). (2. To obtain four-mode unitary transformations between the stages 0. The crystals are assumed to be identical or distinct and pumps are assumed to be identical so that  .66 2 Origin of Macroscopic Approach aligned idler beams. In the Heisenberg picture the input state does not change. â0 (2). we will have to be consistent with the identities â0 (0) = â0 (1). In (2. âs2 (3). âs2 (0) = âs2 (1) = âs2 (2). 3). 2. 1. âi (0). 1. we have adopted a spatial mod- ified Heisenberg picture of the dynamics. 2 describe the down conversion in the undepleted pump approximation. j = 0. Restricting ourselves to the quasimonochromatic light beams (or quasimonochro- matic components of these). we will not need it.

j = 0.c.c. . respectively. 2. (2.300) . Û j+1 ( j) = exp iκ j +1 âs j +1 ( j)âi ( j) + H. In between the down converters the idler from crystal NL1 is put through a beam splitter BS and becomes the idler for crystal NL2. This process is described by   † † † Û2 (1) = exp i[ω̄0 â0 (1)â0 (1) + ω̄i âi (1)âi (1) + (γ ∗ â0 (1)âi (1) + H. l1 and l2 are the lengths of the first crystal and the second crystal.299) 2 2 χ where κ j +1 = v c p j +1 l j . (2. χ is the quadratic susceptibility of the matter of which 2 2 2 +1 both the nonlinear crystals are made. vp j +1 are classical complex amplitudes of 2 pumping beams.)] . c is the speed of light.

t+t 2 .3 Quantum Description of Experiments with Stationary Fields 67 where    ω̄0 t + t (t − t ) = arg ∓i f (t. respectively. t ). t ). Hence.2 |t + t | 1 − . (2.309) . t ) = / .304) and the unitarity of the transformation matrix implies that ∗ ∗ |t|2 + |r|2 = 1. .303)  . Then Arccos(Re t) f (t. (2. Arccos . (2. we obtain that † âs1 (1) = âs1 (0) cosh(κ1 ) + iâi (0) sinh(κ1 ). t ).306) 1 − (Re t)2  ω̄0 = ±(Im t) f (t. (2. (2. and that Re t > 0. for the idler mode and t and r are those for the “escape” mode. The modulus of the transmission amplitude coefficient |t| can vary between zero (where the second emission is spontaneous) and unity (where the second down conversion is stimulated in the highest degree). † â0 (2) = Û2 (1)â0 (1)Û2 (1) = râi (1) + t â0 (1).2. (2. (t + t )∗ f (t. . 2 Here t and r are the transmission and reflection amplitude coefficients. † âi (2) = Û2 (1)âi (1)Û2 (1) = tâi (1) + r â0 (1). âi (3) = iâs2 (2) sinh(κ2 ) + âi (2) cosh(κ2 ). and from this r = −r ∗ (Řeháček and Peřina 1996). .307) ω̄i On applying the relation (2. (2. âi (1) = iâs†1 (0) sinh(κ1 ) + âi (0) cosh(κ1 ).302) with .308) and † âs2 (3) = âs2 (2) cosh(κ2 ) + iâi (2) sinh(κ2 ). t ) = + . .298) at the input ( j = 0) and at the stage 2 ( j = 2).301) ω̄i 2 2 γ = −ir f (t. |r |2 + |t |2 = 1. (2.305) It is advantageous to assume that t = t∗ . (2. tr + rt = 0. t+t .

The equivalence of both the pictures can be proved and the statistical properties can be expressed in similar terms as in (2. it is easy to obtain the expec- tation values âs†1 (1)âs1 (1) = sinh2 (κ1 ).314) with Û j ≡ Û j (0) given in (2. ⎛ ⎞ ⎛ ⎞⎛ ⎞ âs1 (3) m s1 s1 m s1 i m s1 s2 m s1 0 âs1 (0) ⎜ â † (3) ⎟ ⎜ m is1 m ii m is2 m i0 ⎟ ⎜ † ⎟ ⎜ i ⎟ ⎜ ⎟ ⎜ âi (0) ⎟ .300).i.0 We may introduce also the abbreviations M̂ ≡ M̂(0) and we consider the Schrödinger picture. we easily obtain the following relations: † âs1 (1) = cosh(κ1 )âs1 (0) + i sinh(κ1 )âi (0). 1.312) Here.315) Since all the initial fields are in the vacuum states. the statistical operator is a tensor product of separate vacuum sta- tistical operators  ρ̂ = |0 j j 0|. “permanent” statistical operator of the system is given as ρ̂ ≡ ρ̂(0) and when we average M̂( j) = Tr{ρ̂ M̂( j)}. (2. 2. (2. (2. (2. concretely.299) and (2. 2 2 2 (2.317) âs†1 (1)âs2 (3) ∗ = t sinh(κ1 ) cosh(κ1 ) sinh(κ2 ).s2 .316) âs†2 (3)âs2 (3) = sinh (κ2 )[1 + |t| sinh (κ1 )]. ⎝ âs2 (3) ⎠ = ⎝ m s2 s1 m s2 i m s2 s2 m s2 0 ⎠ ⎝ âs2 (0) ⎠ (2. in fact. (2. In fact.312) M̂ ( j) = Tr{ρ̂( j) M̂} = M̂( j) .310) ∗ ∗ † âs2 (3) = t sinh(κ1 ) sinh(κ2 )âs1 (0) + it cosh(κ1 ) sinh(κ2 )âi (0) † + it∗ cosh(κ2 )âs2 (2) + ir∗ sinh(κ2 )â0 (1).318) We will show in the Heisenberg picture that the input–output relation is con- nected to the SU(2. (2. (2.2) group. j = 0.313) j=s1 .298) is replaced by the evolution relations † ρ̂( j + 1) = Û j+1 ρ̂( j)Û j+1 .68 2 Origin of Macroscopic Approach Using Equations (2. where the relation (2. (2.304).311) The statistical properties of the system in the Heisenberg picture can be obtained when we take into account that the initial.319) † † â0 (3) m 0s1 m 0i m 0s2 m 00 â0 (0) .

s2 . s2 . The coeffi- cients of the transformation (2. m 0s2 = 0. âs2 (3)âs†1 (3) = âs1 (3)âs†2 (3) ∗ . † † â0 (3)âi (3) = âi (3)â0 (3) ∗ . From the form of the relation (2.326) † âi (3)â0 (3) = m is1 m ∗0s1 + m is2 m ∗0s2 . (2. m s1 i = i sinh(κ1 ). (2.319) it is evident that the operator N̂ ( j) = n̂ s1 ( j) + n̂ s2 ( j) − n̂ i ( j) − n̂ 0 ( j). (2.325) More generally. j = 0. (2. m is1 = −it∗ sinh(κ1 ) cosh(κ2 ). (2. 0. 0.319) verify the pseudoorthogonality relations m js1 m ∗ks1 + m js2 m ∗ks2 − m ji m ∗ki − m j0 m ∗k0 = g jk .327) Further nonvanishing moments are â j (3)âk (3) = m js1 m ∗ks1 + m js2 m ∗ks2 . gs1 s1 = gs2 s2 = 1. j.322) where g jk = g j j δ jk . (2. k = s1 .323) We observe that the antinormally ordered moments have the expression † â j (3)â j (3) = |m js1 |2 + |m js2 |2 . j = s1 . s2 .2) group. j = s1 .329) . m 0s1 = ir sinh(κ1 ). m s2 i = it∗ cosh(κ1 ) sinh(κ2 ).3 Quantum Description of Experiments with Stationary Fields 69 where m s1 s1 = cosh(κ1 ).321) is independent of j. m 00 = t. m s2 0 = ir∗ sinh(κ2 ). (2. m is2 = −i sinh(κ2 ). j = i. m 0i = −r cosh(κ1 ). m ii = t∗ cosh(κ1 ) cosh(κ2 ). 3. i. m i0 = r∗ cosh(κ2 ). m s2 s2 = cosh(κ2 ).2. âs1 (3)âs†2 (3) = m s1 s1 m ∗s2 s1 + m s1 s2 m ∗s2 s2 . k = i. (2. m s1 s2 = m s1 0 = 0. This conservation law suggests the SU(2.320) m s2 s1 = t∗ sinh(κ1 ) sinh(κ2 ). gii = g00 = −1.328) and † † â j (3)âk (3) = â j (3)âk (3) ∗ . (2. (2.324) and the normally ordered moments † â j (3)â j (3) = |m js1 |2 + |m js2 |2 . 0.

330) † † â j (3)âk (3) = â j (3)âk (3) = 0. (2. 3) D̂s2 (βs2 . βs2 . β0 . βi . 0.70 2 Origin of Macroscopic Approach The rest second-order moments vanish: † † â j (3)âk (3) = â j (3)âk (3) = 0. j = s1 . j = s1 . (2. β0 . k = i. βs2 . 0) D̂i (βi . s2 . k = s1 . i.332) where the displacement operators are given by † D̂ j (β j .336) ⎩ 2 ⎭ j=s1 . 0. j = s1 . 0)} = Tr{ρ̂ D̂s1 (βs1 .i. βi (3) = −βs∗1 m s1 i − βs∗2 m s2 i + βi m ii + β0 m 0i . β0 (3) = −βs∗1 m s1 0 − βs∗2 m s2 0 + βi m i0 + β0 m 00 . D̂ j (β j ) ≡ D̂ j (β j .333) By the remark above.0 we derive that   CS (βs1 . 3) D̂0 (β0 . we obtain that CS (βs1 . k = i.i.331) Quantum statistics of radiation in the process under study is that of a four-mode Gaussian state.335) From the known quantum characteristic function for the initial vacuum state ⎧ ⎫ ⎨ 1  ⎬ CS (βs1 . βs2 .0 ⎡ ⎤    + ⎣−βs1 βs∗2 Bs∗1 s2 − βi β0∗ Bi0∗ + β j βk C ∗jk + c. β0 . 0). k) = exp[β j â j (k) − β ∗j â j (k)]. (2.337) j=s1 . βi . (2. s2 . (2. k = 0. 3) = Tr{ρ̂(3) D̂s1 (βs1 . (2. βi .334) where − βs∗1 (3) = −βs∗1 m s1 s1 − βs∗2 m s2 s1 + βi m is1 + β0 m 0s1 .s2 k=i. s2 .s2 .c. 3) = Tr{ρ̂(0) D̂s1 (βs1 (3)) D̂s2 (βs2 (3)) D̂i (βi (3)) D̂0 (β0 (3))}. On substituting into the relation (2. (2. −βs∗2 (3) = −βs∗1 m s1 s2 − βs∗2 m s2 s2 + βi m is2 + β0 m 0s2 .0 . s2 . starting with the quantum characteristic function: CS (βs1 . βs2 . 0) = exp − |β j |2 . 0. (2. 3)}. ⎦ . 0. 0) D̂0 (β0 . and j = i. 0) D̂s2 (βs2 .s2 . βi . 3) = exp − |β j |2 B jS j=s1 .319). 3) D̂i (βi . 3.332) according to (2. β0 .

(2. k = i. 0. s2 for the signal modes to 1.s2 In the following we simplify the notation s1 . We confine ourselves to the study of the signal beams in what follows. (2. 1. From the characteristic function s . s2 .2. 3) = Tr ρ̂(0) D̂s1 (βs1 . † B jN = Δâ j (3)Δâ j (3) .3 Quantum Description of Experiments with Stationary Fields 71 Here the coefficients B jS . j = s1 .340) where Tri and Tr0 are partial traces over the idler and escape modes. s2 . 3).341) In Peřinová et al. (2. s2 . i. and with the coefficients Bs1 s2 . respectively.340) can easily be obtained: CS (βs1 . C jk . Taking into account that â j (3) = 0. βs2 . 0. 2. 0. 0. we see that we are consistent with the more general notation † B jA = Δâ j (3)Δâ j (3) . which are described by the reduced statistical operator ρ̂signal (3) = Tri Tr0 {ρ̂(3)}. s2 . B jk . ⎦ . j = s1 .338) † Bi0 = âi (3)â0 (3) . j = i. Bi0 . 1) D̂s2 (βs2 . C jk = â j (3)âk (3) . βs2 . j = s1 . s2 . 2 † 1 B jS = â j (3)â j (3) + . (2003). 3) = CS (βs1 . 2 Bs1 s2 = âs1 (3)âs†2 (3) . βs2 ) ≡ CS (βs1 . C jk can be expressed in the form † 1 B jS = â j (3)â j (3) − . j = i. j = s1 . (2. 3) ⎡ ⎤    = exp ⎣− |β j |2 B jS + −βs1 βs∗2 Bs∗1 s2 + c.339) where Δâ = â− â . k = i. 0. 0. (2. Quantum characteristic function in the state described by the statistical operator (2. the same function has been introduced as   CS (βs1 .342) j=s1 . 0. j = s1 . after similar replacement.c. respectively. βs2 .

β2 ) = exp (|β1 |2 + |β2 |2 ) CS (β1 . and antinormal orderings of field operators. β2 ). Cs (β1 . −1 in the subscript and also s = N . we can establish the Φs . A denote the normal. symmetrical. 0.343) 2 where s = 1. S. (2.

350) .)] . 2 2 (2. B1S = B1A − .c. which yields the expansion ∞ ∞  1 ΦA (α1 . n 2 |ρ̂signal (3)|m 1 . (2. α2 |ρ̂signal (3)|α1 . for s = −1 it holds that 1 ΦA (α1 . α2 is the two-mode coherent state. 2 B2A = cosh2 (κ2 ) + |t| sinh2 (κ2 ) sinh2 (κ1 ). m 2 ) = n 1 .72 2 Origin of Macroscopic Approach quasidistribution related to the respective ordering of field operators 1 Φs (α1 . n 2 . m 1 . n 2 + q) √ (2. we obtain 1 Φs (α1 . (2.349) n 2 =max(0.348) π2 where |α1 .345) K 12s where 1 B1A = cosh2 (κ1 ). α2 ) = α1 . (2.347) Especially. m 2 (2. α1∗ α2∗ . m 1 . B2N = B2A − 1. 2 ∗ B12 = t∗ sinh(κ1 ) cosh(κ1 ) sinh(κ2 ). α2 ) = π2K  12s  1 ∗ ∗ × exp [−B2s |α1 | − B1s |α2 | + (B12 α1 α2 + c.−q) (m 1 + q)!m 1 !n 2 !(n 2 + q)! for any ΦA quasidistribution that does not depend on α1 α2 .346) and K 12s = B1s B2s − |B12 |2 .344) After integrating. α2 ) = exp(−|α 1 | 2 − |α 2 | 2 ) π2 q=−∞ m 1 =max(0. 1 B2S = B2A − . B1N = B1A − 1. α2 ) = π4   × Cs (β1 .−q) ∞  ∗(m +q) n +q α1 1 α1m 1 α2∗n 2 α2 2 × ρ(m 1 + q. (2. β2 ) exp α1 β1∗ − α1∗ β1 + α2 β2∗ − α2∗ β2 d2 β1 d2 β2 . n 2 . α2 . Here ρ(n 1 .

(2. the joint photon-number distribution is approximately the product of its marginal photon-number distributions.2. 10]. m 1 . It differs from the product of pertinent marginal distributions by larger “diagonal” probabilities.351) K 12A while obviously ρ(m 1 + q1 . t. n 2 − q2 ) = 0 for q1 = −q2 . n 2 . 2 As for the experimental arrangement under study.12 Joint photon-number distribution for |t| = 1. (2. for the sake of illustrations. Fig. l2 . The joint photon-number distribution p(n 1 . Equating the expansion coefficients for (2.3 Quantum Description of Experiments with Stationary Fields 73 are the usual matrix elements. n 2 ) = ρ(n 1 . A substitution into (2.353) This distribution can be seen in Fig. 2. Consequently. n 2 ). the mean photon number B1N in the first sig- nal mode is constant and this convenient behaviour is.349). B1N = B2N = 3. it depends on l1 . n 2 + q) = p=max(0. n j ∈ [0. B1N = B2N = 3. To the contrary for |t| small.351) leads to slightly more complicated expression: p(n 1 . n 1 .352) (ii) Photon-number statistics Numbers of photons in signal modes complete the picture of the quantum cor- relation between these beams.345) with s = −1 and those of (2. . m 1 .−q) (m 1 − p)!(n 2 − p)! p!( p + q)! ∗p (K 12A − B2A )m 1 − p (K 12A − B1A )n 2 − p B12 B12 p+q × m +n 2 +q+1 1 . n 2 . j = 1. n 2 ) can be expressed in the concise form (Řeháček and Peřina 1996).n 2 ) min(m √ (m 1 + q)!m 1 !n 2 !(n 2 + q)! ρ(m 1 + q. (2.12 for |t| = 1. n 2 . we arrive at the expression 1 . 2. whereas the numerical demonstration is restricted to the case when the length of the first crystal is kept fixed. |B12 | = 3.

13 Quantum correlation measure RN versus the modulus of the transmission amplitude coefficient |t| ∈ [0.74 2 Origin of Macroscopic Approach extended also to the second one as a relationship between |t| and κ2 . respectively) 1991). of the first signal beam from NL1 to Ds .13. |γ12 | = RN = 1. (2. see Fig. the interference manifests itself as oscillations in the counting rate Is (see (2. The maximum obtainable visibility .357) B1N + B2N √ As B1N B2N ≤ 12 (B1N +B2N ). respectively. This maximum correlation does not correspond to a weaker correlation between the signal photon numbers. 2. 10. 1. 1]. e. c. (2. (2. In the multimode analysis (Wang et al. Fig.356) (2) The limit case |t| = 1 is interesting. δτ2 . 104 (the curves a. the visibility of the interference between the signal fields has been expressed. b. d.355) B1N B2N In the numerical illustration of the quantum correlation measures. Deriving a simplified visibility Vsimple for the formalism of several modes provides √ 2RN B1N B2N Vsimple = . and of the second signal beam from NL2 to Ds are incremented by δτ0 . B2N sinh2 (κ2 ) = . the visibility cannot exceed the correlation measure RN and the equality is attained for B1N = B2N . 100.354) 1 + |t|2 B1N (2) The Glauber degree of coherence (Peřina 1991) γ12 is the complex-valued quan- tum correlation measure related to the normal ordering: ∗ (2) B12 γ12 =√ . we assume κ1 = κ2 = κ and find κ1 by the inversion of the formula B1N = sinh2 (κ1 ) = âs†1 (1)âs1 (1) = n 1 (1).412) below) when propagation times of the idler beam from NL1 to NL2. (2. it is assumed that n 1 (1) = 10−2 . 1991a) of the experiment (Zou et al. δτ1 . 2.

360) where Û j+1 (0) for j = 1. that form broad-band signal fields. In the spatial interaction picture. and (2. κ2 of the crystals.0 .2. (2. (2. The single-photon regime applies in the multimode analysis of Wang et al. n s2 . 3 are given by relations (2. or it is negligible.0 ≡ |0 s1 |0 i |0 s2 |0 0 and.359) 1 + |t|2 n 1 (1) Wiseman and Mølmer (2000) considered the relevant limits in this form.318) into (2. we simply obtain RN = |t|. we can rewrite (2. (1991) and Wang et al. which is the modified Schrödinger picture. An analogue of relation (2. before it enters the beam splitter. they change like the Heisenberg picture free-field operators. |n s1 .317). Up to the first order in the rescaled lengths κ1 . The same applies to idler modes and idler fields.300). 2. n 0 ≡ |n s1 s1 |n i i |n s2 s2 |n 0 0 .298) for a discrete space is not used .3 Quantum Description of Experiments with Stationary Fields 75 between two fields in an experiment is given by the correlation measure RN . because each of the narrow-bandwidth signal modes (ks1 .355).s2 . 2 2 We will drop the argument (0) at the annihilation operators in what follows. Here |0 s1 . On substituting (2. (2. (1991a. we find that |t| cosh(κ1 ) RN = + . ωs2 ).316).358) 1 + |t|2 sinh2 (κ1 ) Noting that the idler beam.s2 . j = 1. or it is small. (2. (2. (1991a). has the same statistics as the output signal 1.i. in general. with directions characterized by ks j and with the frequencies ωs j . n i .299) and (2. The signal fields s1 and s2 from the two down converters are allowed to come together and interfere at the detector Ds . The probability to have down conversions at both crystals over inter- † † action time is less than or equal to âs1 (1)âs1 (1)âs2 (3)âs2 (3) = B1N B2N + |B12 |2 .i.b) occurs for n 1 (1)  1. the operators do not change and in the interaction picture. receives only a small part of the pumping photons over interaction time. â0 ( j) → â0 (0). the state of the field produced by the crystals is given by |ψ(3) ≡ Û3 (0)Û2 (0)Û1 (0)|0 s1 . (ks2 . âi ( j) → âi (0). The probability of a down conversion at crystal NL1 over interaction time is less than or equal to n 1 (1).358) in terms of the mean photon number n̄ 1 (1) = sinh2 (κ1 ) as  1 + n 1 (1) RN = |t| .338) with (2.361) In the Schrödinger picture. 2. The single- photon regime which is the regime of experiment and theory in Zou et al. cf. where the annihilation operators âs j +1 ( j) → âs j +1 (0). ωs1 ). Wiseman and Mølmer (2000).

1.0.0 (3) = |κ2 |2 (2.363) Let us assume that one infers a conversion at crystal NL2 after no photocounts in the escape mode. 0 + r|0. j = 0.365) On a photocount in the escape mode it is certain that the conversion has happened at NL1. 2.369) Hence âs†1 âs2  κ1 κ2 t∗ . Fortu- nately.i. We introduce the probability of the detection p1. âs†2 âs2  κ22 . (2. 1. 1. Let us consider the annihilation operators âs j +1 . In general.76 2 Origin of Macroscopic Approach in quantum optics (the free-field propagation is absorbed in the interaction).1 (3) = |κ1 |2 |r|2 .1. we will use just the interaction-picture annihilation operators.s2 . 0 + r|1. 0 + iκ1 (t|1. Expanding the operators Û j+1 (0) according to κ j +1 . 1991a). 0.364) and that of such a wrong inference p1. (2. On no photocounts in this mode. 0 . the posterior probabilities are |κ2 |2 Prob(n s1 = 0 ∩ n s2 = 1|n i = 1 ∩ n 0 = 0) = . (2. 1. 1. 1 ). 0. For |t| = 1 we have a single-photon state in the idler mode 2 and in the collection of the signal modes. 0. 0. (2. This may be regarded as reflecting the intrinsic impossibility of knowing whether the detected photon comes from NL1 or NL2 (Wang et al. one can infer a conversion at crystal NL1 after a photocount in the escape mode. The multiphoton conditional states can be found in Luis and Peřina (1996a). (2. (2.0 + iκ2 |0.0. 1 ).0.368) âs2 |ψ(3)  iκ2 |0. 0.362) when κ j +1 are small. We introduce the probability of a correct inference of the conversion at crystal NL2 p0. (2. (2.1. 0. 0.371) .367) |κ1 | |t|2 + |κ2 |2 2 Here the underlining means a random variable. j = 0. 2. we obtain that 2 |ψ(3)  |0 s1 .1. The counting rate registered by Ds exhibits perfect interference when the idler fields are perfectly aligned.0 (3) = |κ1 |2 |t|2 . (2. The action of the two 2 operators on the state |ψ(3) is asymptotically for small κ j +1 expressed as 2 âs1 |ψ(3)  iκ1 (t|0.366) |κ1 |2 |t|2 + |κ2 |2 |κ1 |2 |t|2 Prob(n s1 = 1 ∩ n s2 = 0|n i = 1 ∩ n 0 = 0) = .370) âs†1 âs1  κ12 .

t0 ). In this case. respectively. are unitary operators: Û j+1 (0. (2.0 . but it can be verified that it is valid up to the first order in κ1 . m = s1 . Ê s j (t)≡ Ê s j (rs . and “escape” modes (ks j . 2. 2. (ki . The multimode formalism views the electric fields as temporal-interaction-picture operators  δω  Ê m (r. t0 ). t) = (+) âm (ωm ) exp[i(km · r − ωm t)]. t. t. 2. (2. at the appropriate detector. i. m = s1 . The multimode formalism enables one to respect that the two crystals are centred at 01 and 02 . That is.i. ωi ). j = 0. i.t=t0 = 1̂. (2. we must present another appropriate description of the dynamics of the same down-conversion experiment. s2 . t). κ2 . 2. idler (m = i). In order to compare the sophisticated multimode formalism with the simple for- malism of several modes. Û j+1 (0. We can find that γ12(2) = t∗ in the limit of small κ j +1 . The opposite regime is that where n 1 (1)  1. j = 1. t) ≡ Û3 (0. κ2 . t)Û2 (0. whose vacuum states may be designated as |0 m (ωm ). 2). j = 1. The Hilbert space for the mul- timode analysis is a tensor product of those Hilbert spaces of separate modes.2. 0. We adopt the temporal interac- tion picture and combine it with the spatial interaction picture. 2. 2). where τ j . the equation 2 holds up to the zeroth order. ωm ) at the frequency ωm . Obviously.373) Here |0 s1 . t. t − τ j ). (k0 . ω0 ). We con- sider photon-flux amplitude operators Ê s(+) j (t).372) 2π ω m where δω is the mode spacing and âm (ωm ) is the photon annihilation operator for narrow-bandwidth signal (m = s j . j = 1. s2 . j = 1. j = 0. the state produced by the crystal is given by |ψ(3.3 Quantum Description of Experiments with Stationary Fields 77 It can be verified that these relations hold up to the second order in κ1 .s2 . t) = lim Û j+1 (0. is the propagation time (+) (+) of s j from NL j to Ds .i. idler. t)|0 s1 . and “escape” (m = 0) modes (km . t) = Ê s j (0 j .0 is the vacuum state of all the narrow-bandwidth signal. (1991a) the pump beams at each crystal are represented by com- plex analytic signals V1 (t) and V2 (t) such that |V j (t)|2 is in units of photons per second ( j = 1. Û j+1 (0. 0. t)Û1 (0. t0 ) are unitary operators that obey the initial condition . ωs j ). (iii) Multimode formalism In Wang et al. (2.s2 . Here there are many photons on average in all of the down-converted beams. the phase of the stage-1 or stage-2 idler mode should lock the phase of the output signal-2 mode for any nonzero transmission amplitude coefficient t.374) t0 →−∞ where Û j+1 (0.375) .

t). t). t) = exp i ω0 â0 (ω )â0 (ω ) + ωi âi (ω )âi (ω ) ω   . 2 describe the parametric down conversion and are expressed. t0 ) for j = 0. indirectly. Û2 (0. in terms of Ê s(+) j (r. t.78 2 Origin of Macroscopic Approach Û j+1 (0. Ê i(+) (r. t) describes the beam 2 +1 splitter and is expressed as   † † Û2 (0.

ω ) t−t1 ω ω   . Let T denote the time ordering.376) In this point we differ from the paper by Wang et al. † + γ ∗ â0 (ω )âi (ω ) + H.c. t − t1 ) = T exp ν2 V2 (0)δω φ(ω0 − ω . t. We will introduce the unitary operator  t  Û3 (0. . (2. who used the initial condition at t  = t − t1 and they did not write down the decomposition into stages. (1991a).

(2.378) From this we obtain the vector |ψ(3. is a constant such that |ν j |2 gives the fraction of incident pump photons that is spontaneously down converted in the steady state. t − t1 ) = Û3 (0. ω ) ω ω 1    −i(ks2 +k )·02 sin 2 (ω0 − ω − ω )t1 × e 1 2 (ω0 − ω − ω )      t1 †  †  × exp −i(ω0 − ω − ω ) t − âs2 (ω )âi (ω ) − H. 2. t − t1 )  = |ψ(2. t − t1 )|ψ(2. j = 1. ω ) ω ω 1  −i(ks2 +k )·02 sin (ω0 − ω − ω )t1  ×e 2 1 (ω0 − ω − ω )   2   t1 × exp −i(ω0 − ω − ω ) t − |ω s2 |ω i |0 s1 . t.c. dt  . ks2 (k ) is a wave vector that is determined by the frequency ωs 2 (ω ) and the direction of the second signal beam (the idler beam).377) where ν j .0 . t. t. t.379) . ω0 is the frequency of the monochromatic pump beam. t. 2 (2. To the first order in the processes the unitary operator may be expressed as   Û3 (0.c. t − t1 ) + ν2 V2 (0)δω φ(ω0 − ω . t − t1 ) = 1̂ + ν2 V2 (0)δω φ(ω0 − ω . . 2 (2.  )·02 −i(ω0 −ω −ω )t  † † ×e−i(ks2 +k e âs2 (ω )âi (ω ) − H.

0 . (2. t − t1 )|0 s1 . ω ) 2π ω  ω  1  −ik ·02 sin (ω0 − ω − ω )t1 ×e 2 1 (ω0 − ω − ω ) 2     t1 × exp −i(ω0 − ω − ω ) τ2 − exp[−i(ω0 − ω )(t − τ2 )] 2 × |ω i |0 s1 . t.384) φ(ω̃ . ω) = φ2 (ω̃.0 † = Û2 (0.0 (2. t)Û1 (0. t. t. ω .382) = ν2 V2 (t − τ2 )|1(02 . (2.s2 . t − t1 ) = Û2 (0. (2.0 is the vacuum state of the first signal and escape modes. j = 1.i. ω) = φ1 (ω̃.i. t − t1 ) to a vector  δω   Ê s2 (t)|ψ(3. 2. ω ) 2π ω ∞   −ik ·02 sin 12 (ω0 − ω − ω )t1 ×e −∞ 1 (ω0 − ω − ω ) 2   t1 × exp −i(ω0 − ω − ω ) τ2 − dω exp[−i(ω0 − ω )(t − τ2 )] 2 × |ω i |0 s1 . t − τ2 ) i |0 s1 .s2 . ω). t) i = 2π δω φ(ω0 − ω . t)Û1 (0. (2. ω ) is connected with spectral functions φ j (ω . ω .383) where the single-photon state of the idler beam √  |1(r.s2 . t)|0 s1 . (2. t.380) We transform the vector |ψ(3.386) . t − t1 )Û2 (0.0 . t − t1 ) = ν2 V2 (0) δω φ(ω0 − ω . ω) characterizing the sig- nal and idler fields at any crystal NL j. respectively. ω0 ). φ j (ω̃ . t.3 Quantum Description of Experiments with Stationary Fields 79 where |ω s2 and |ω i are frequency eigenstates of the second signal and the idler beam. φ(ω̃.385) The frequency eigenstates are single-photon states: |ω m = âm† (ω )|0 m . and |ψ(2. ω ) ω −i(k ·r−ω t) ×e |ω i .s2 .381)  δω   ν2 V2 (0) φ(ω0 − ω .s2 . |0 s1 .2. ω ) = φ j (ω̃ .0 (2.

t)Û1 (0. Wang et al. Should t1 mean the interaction time. (2. We further calculate Ê s1 (t)|ψ(3.80 2 Origin of Macroscopic Approach where 8 |0 m = |0 m (ω ). (2.389) for K 2 t1 < τ2 < (K 2 + 1)t1 (2. for instance.388) To resolve this. t. (1991a) let t1 → ∞. t − t1 ) . t]. t − t1 ) = Ê s1 (t)|ψ(2. [t − (K 2 + 1)t1 . t. 0. it is better to change the integration limits. namely not to consider the integration interval [t − t1 . t. t − t1 )Û2 (0. t − t1 ) by the action of the unitary operator  t † Û2 (0.391) We obtain the appropriate component of the vector |ψ(2. t) = T exp ν1 V1 (0)δω t−t1  −i(ks1 +k )·01 −i(ω0 −ω −ω )t    × φ(ω0 − ω . ω )e e ω ω . (2. but. the nonvanishing result is obtained only for 0 < τ2 < t1 . (2. m = i.390) to hold.387) ω Unfortunately. t − K 2 t1 ]. t.

s2 ≡ |ψvac s1 .s2 stand for vacuum states.s2 .c.0 . so that  Ê s1 (t)|ψ(3. |1(r.s2 .s2 . |0 s1 . t) 0 ≡ 2π δω φ(ω0 − ω . ω ) exp[−i(k · r − ω t)]|ω 0 . t) i is defined in (2. but we replace âi (ω ) by † † tâi (ω ) + râ0 (ω ). |ω i by t|ω i + r|ω 0 and we change all the other subscripts that underlie to a change. dt .392) † The calculation proceeds similarly as in the case of NL2.0  + r|1(01 .393) where |0 s1 .i.  × âs†1 (ω )[t∗ âi (ω ) ∗  † + r â0 (ω )] − H.384). (2.  (2.394) ω .i. (2. t − t1 ) = ν1 V1 (01 .0 ≡ |ψvac s1 .i. t. and |1(r. t − τ1 ) i |0 s1 . t) 0 stands for single-photon state of the “escape” beam √  |1(r. t − τ1 ) 0 |0 s1 . t − τ1 ) t|1(01 .s2 .

s2 .0 . t − τ ) i |0 s1 . t − K 1 t1 ]. t − τ ) i |0 s1 .2. t)|] M̂  f [|ψ(3. The operator  f is not unitary and is even “slightly” nonlinear. The operator  f compensates for the difference we have caused with the initial condition at t  = t0 → −∞ instead of the Wang–Zou–Mandel shortening of the integration interval. no difficulties arise if we change the limits independently of NL2.396) for K 1 t1 < τ1 < (K 1 + 1)t1 (2.s2 ] = |1(01 . t) . Using such an operator we can describe. but.401) The angular brackets will mean averages and.395) Considering a change of the integration limits as above. (2. (2. Âcτ [|1(01 . t − τ ) 0 |0 s1 . the relations (2. for instance. . We do not consider the integration interval [t − t1 . t − (K 1 + 1)t1 ) .i.i. when operators are involved.393) can be generalized to provide Âcτ1 [ Ê s(+) 1 (t)|ψ(3. (2. t − τ ) 0 |0 s1 . t) .s2 .398) Âc(τ0 +τ2 ) [ Ê s(+) 2 (t)|ψ(3. (2. t) ] = Ê s2 (t)|ψ(3.s2 . Its consideration depends on a neglect of the coherence length in comparison with the interaction length. t − K 1 t1 .403) Hence. (2. one may omit the unusual notation when no ambiguities arise. M̂ = ψ(3. t].400) Âcτ [|1(01 . t − K 2 t1 .383) and (2.  f denotes an attenuation of the field down to the vacuum state outside an interaction length centred at the distance f from NL1 in the direction of propagation of the beam. the brackets are supposed to average in the state |ψ(3. it also holds that M̂ =  f [ ψ(3. where (within which interaction length) the single-photon states are localized at the time t. to the first order of the processes. we see that. [t − (K 1 + 1)t1 .3 Quantum Description of Experiments with Stationary Fields 81 Here a nonvanishing result is obtained only for 0 < τ1 < t1 . (2. (2. t − (K 2 + 1)t1 ) .399) where τ0 is the propagation time of the idler from NL1 to NL2.402) with M̂ being an operator.0 ] = |1(01 .397) to hold. t)| M̂|ψ(3. When the operator is situated inside an interaction length centred in the propagation distance f from NL1. (2. t) ] = Ê s1 (t)|ψ(3. t) ]. In other words.

we have ωs + ωi = ω0 . t − τ1 )|1(r2 . (2.407) . ω)|2 e−iωτ dω. t − τ2 ) i = μ(τ0 + τ2 − τ1 ) exp[−iωi (τ0 + τ2 − τ1 )]. (2. Ê s(−) 2 (t) Ê s(+) 2 (t)  |ν2 |2 |V2 (t − τ2 )|2 .82 2 Origin of Macroscopic Approach Letting ωs and ωi denote the centre frequency of the signal beam and the idler beam. respectively.404) where ω0 e−iωi τ μ(τ ) = 2π |φ(ω̃.405) 0 we obtain that the relations (2. (2. Introducing the normalized correlation function μ(τ ) of the down-converted light i 1(r1 .406) Ê s(−) 1 (t) Ê s(+) 1 (t)  |ν1 |2 |V1 (t − τ1 )|2 . (2.371) ought to read Ê s(−) 1 (t) Ê s(+) 2 (t)  ν1∗ ν2 t∗ V1∗ (t − τ1 )V2 (t − τ2 ) ×μ(τ0 + τ2 − τ1 ) exp[−iωi (τ0 + τ2 − τ1 )].370) and (2.

359). Wang et al.410) . Hence the modulus of the normalized correlation function is | Ê s(−) (t) Ê s(+) (t) | / 1 2 Ê s(−) (+) (−) (+) 1 (t) Ê s1 (t) Ê s2 (t) Ê s2 (t) V1∗ (t − τ1 )V2 (t − τ2 ) =+ |μ(τ0 + τ2 − τ1 )||t|. (1991a) comprises the parameters t∗ .† where we introduce Ê s(−) j (t) ≡ Ê (+) sj (t) . A linear dependence of visibility on |t|. On the contrary. j = 1. Nevertheless. considering the photon-flux amplitude operators Ê s(+) (t) at the detector Ds with a quantum efficiency ηs (Wang et al. Especially.408) |V1 (t − τ1 )|2 |V2 (t − τ2 )|2 The maximum value is equal to |t|. the statistical properties in Peřinová et al.409) 2 substituting into the formula for the average rate of photon counting Is = ηs ψ(t)| Ê s(−) (t) Ê s(+) (t)|ψ(t) . (1991a). which is predicted also by equation (2. (2. because the differences under discussion resemble distinct. r ∗ . (2003) coincide with those in Wang et al. as seen convincingly in the original work (Zou et al. As concerns Ê s(+) j (t)|ψ(t) . they are not explicitly presented in Wang et al. 1991a). is the true signature of induced coherence without induced emission. (1991a). (2. 1991. 2. yet equivalent pictures. It emerges that the parameters of the beam splitter do not enter the relation for Ê s(+) 1 (t)|ψ(t) . 1   Ê s(+) (t) = √ i Ê s(+) 1 (t) + Ê s(+) 2 (t) . (2. but they may be derived. Ê s(+) 1 (t)|ψ(t) in Wang et al. 1991a).

the joint number-sum and phase-difference distribution has been considered. They have addressed the number sum and the quantum phase difference as simultaneously measurable observables and the number difference and the quantum phase difference as canonically conjugate observables. it concerned the variances of number sum and num- ber difference and the dispersions of quantum phase differences according to vari- ous definitions. The paper (Peřinová et al. but for the canonical quantum phase difference and the Luis– Sánchez-Soto phase difference only. whereas the system of canonical phase related to the antinormal ordering of exponential phase operators ranges between the symmetrical and antinormal order- ings.c. (2.2. especially the entropic or information-based measure with the modulus of the usual degree of coherence in the dependence on absolute value of the transmission amplitude coefficient of the beam splitter inserted as an attenuator of the perfect alignment. t) 0 uniquely results in the relation 1 Is = ηs {|ν1 |2 |V1 (t − τ1 )|2 + |ν2 |2 |V2 (t − τ2 )|2 2 +[−iν1∗ ν2 V1∗ (t − τ1 )V2 (t − τ2 ) t∗ μ(τ0 + τ2 − τ1 )e−iωi (τ0 +τ2 −τ1 ) + c.412) Peřinová et al. They have compared different measures of correlation. 2000) reveals that the correlated chaotic state is the mixed partial phase-difference state. whose correlation is not maximum for the perfect alignment. (2000) have studied quantum statistics of radiation in signal modes of the two-mode parametric processes with aligned idler beams. (2. through the symmetrical and antinormal order- ings. photon-number difference.411) and taking into account the orthogonality of single-photon states |1(r1 . The situation with the photon numbers in the signal modes. t) i and |1(r1 .]}. but by no means exactly. The quasidistribution of number difference and phase difference has been defined with the properties that the marginal distribution of the phase difference is the canonical one. In contrast to the normal ordering. They have found that the signal beams are in the correlated chaotic state. Some other measures have been introduced taking into account the symmetrical and antinormal orderings of field operators. and quantum phase- difference statistics. The strength of correla- tion depends on the degree to which the paths of the idler beams are superposed and aligned. In addition to the marginal distributions.3 Quantum Description of Experiments with Stationary Fields 83 where Ê s(−) (t) = [ Ê s(+) (t)]† . They have taken into account that the quantum correlation has a significant effect on the photon-number sum. these orderings do not indicate the maximum correlation for the perfect alignment. A comparison of distributions of quantum phase difference derived from the phase-space distributions has shown that the phase-difference uncertainty increases from the normal ordering. The theory of canonical correlation has been applied to the quasidis- tribution of complex amplitudes related to the symmetrical ordering of field oper- ators. serves as motivation for such a more general consideration. . Essentially.

the pair time duration. The utility of a simple single-mode theory has been clarified in the case where single spatial mode filters and narrow-band optical filters are used to filter the output state of parametric down-conversion Li et al. (2003) have compared the simple formalism of several coupled harmonic oscillators with multimode formalism in the analysis of an interference experiment.84 2 Origin of Macroscopic Approach Peřinová et al. Peřinová et al. and angular decomposition of the pump-beam frequency and their effect on characteristics of a photon pair. the previous authors did not try to include time delays between optical elements into the analysis. Peřina and Křepelka (2006) have provided the generalization of this description to stimulated parametric down-conversion. pump-beam transverse width. He considers the pump-pulse duration. They have indicated that. (2007) have reported on a measurement of the joint signal–idler photoelectron distribution of twin beams. Parameters of the previously published model (Peřina and Křepelka 2005) have been estimated. for instance. Peřina (2008) has shown that a nonlinear planar waveguide pumped by a beam orthogonal to its surface may serve as a versatile source of photon pairs. (2003) have also formally expressed. Peřina et al. Peřina and Křepelka (2005) have derived joint photon-number distributions in signal and idler modes and have illustrated related concepts taking into account experimental data. The specific result that the joint signal–idler quasidis- tribution of integrated intensities can be approximated by a well-behaved function even in the case where the quasidistribution is not an ordinary function has been comprised. . Then they have assumed the single-photon regime as also previous authors did. 2005). assuming several coupled harmonic oscillators. On focusing on several modes they have been able to study phase prop- erties of “correlated chaotic beams”. that one works with a single-photon state of some signal modes in the several-mode formalism whenever one describes the experiment with a superposition of single-photon states of modes that form the signal beam. such as the spectral widths of signal and idler fields. and the degree of entanglement between the two fields.

It has been noted that the nonlinearity of the process may lead up to a need of a renormal- ization. 3.  C Springer Science+Business Media.1 Momentum-Operator Approach Several papers devoted to macroscopic approaches to quantization of the elec- tromagnetic field advocate the momentum operator. An attempt has been made to take into account the frequency dispersion of a medium at least up to inclusion of the group velocity and to preserve the traditional descriptions of nonlinear processes by introducing narrow-band fields. We present applications of the traditional approach in quantum optics. V. Nevertheless. Lukš. LLC 2009 . such an operator A. Peřinová. Finally the application of one of the macroscopic approaches has led to the description of several linear optical devices and to the study of radiating atoms in a linear medium. A mention of the quasimode description of the spectrum of squeezing will be restricted to an analysis of coupling of the cavity modes and propagating modes. In optical imaging with nonclassical light one wants to investigate the quantum fluctuations of light at different spatial points in the plane perpendicular to the prop- agation direction of the light beam. whose description includes the renormalization.Chapter 3 Macroscopic Theories and Their Applications There were several attempts at a justification of the momentum-operator approach. The spatio-temporal approach has been developed with respect to quantum soli- tons. It is appropriate that they include quantization of the electromagnetic field at least in the one-dimensional case. which is a recurrent theme by the way. The paraxial propagation of a light beam with the parabolic approximation and the asymptotic expansion of the beam has been completed with a quantized version. A general approach to quantization of the electromagnetic field in a nonlinear medium enables one to compare properties of the momentum operator with those of the space–time displacement operator. In general. Quantum Aspects of Light Propagation. A complete analysis could be provided only for the para- metric processes. in which the momentum operator is effectively quadratic. there is a modicum of papers on this theme in quantum optics. In such spatial points very small photodetec- tors or pixels may be located.1007/b101766 3. As an example the nonlinear process has been presented. 85 DOI 10.

t0 +T Ĝ(z) ≡ ĝ(z. cannot substitute the spatial modes. In Huttner et al.e. they assume a similar situation as Abram (1987). Instead of immediately reducing the unsymmetrical Maxwell stress tensor to a single component. In comparison with Abram (1987). They remark that the change in the quantization volume pointed out by Abram (1987) is not defined when the medium is dispersive. the momentum density is first reduced. Specifying the state at a given point (e. that relation (2. (1990). ω j )] = δi j 1̂ (3. To the contrary. t) = c[ D̂ (−) (z. “The generalization” of the relation for the momentum-flux operator ĝ(z.c. The exposition is concluded with applications of the traditional approach to the nonlinear optical couplers. i. Its integration over T gives the momentum operator Ĝ(z). we see the following changes. t) Ê (+) (z. where ωm = 2mπ T . a quantity whose square is proportional to the photon flux.3) is not founded well.e. 3. t) dt. â † (z. the classical analysis of nonlinear optical processes shows that in order to obtain simple equations of propagation it is useful to introduce photon-flux amplitude. means the Hermitian conjugate to the previous term. (1990) have developed a formalism that describes in a fully quantum- mechanical way the propagation of light in a linear and nonlinear lossless dispersive medium.]. they consider only the one-dimensional case restricting themselves even to fields propagating in the +z-direction only.86 3 Macroscopic Theories and Their Applications should be one of the space–time displacement operators.1.4) t0 .49) holds for any operator. (3. The normal ordering is used where necessary. The equal-space commutation relations [â(z.e.2) where H. t) B̂ (+) (z. 1992). but they develop Abram’s idea of the use of the energy flux not being dependent on the medium (cf. ωi ). i. the MKSA (SI) system of units is used. In their opinion.c. t) + H.1 Temporal Modes and Their Application Huttner et al. i. t) = [ D̂ (−) (z.] (3. to the form ĝ(z. however Ben-Aryeh et al.1) cannot substitute the usual equal-time commutation relations.e. z = 0) and within a time period T cannot substitute specifying the state at an initial time (t = 0) and within a quantization length L.c. They take for granted that in quantum field theory there is a generator for spatial progression. The notation ceases to express the dependence on both z and t and states the dependence on z only. almost paradoxical properties of these operators have been derived. At first. Temporal modes of discrete frequencies ωm .. t) + H. (3. i.g. when the refractive index depends on the frequency. At present we hesitate to accept the consequences of their approach (cf. Caves and Crouch 1987).

(3. In Abram (1987) there is not pure Heisenberg picture. ω0 + m )â † (z.8). the momentum operator is obtained  Ĝ lin (z) = (km )â † (z. ωm )â(z. (3. t) = |E|e−i(ωp t−kp z) is the positive-frequency part of the pump field. (1990) have concentrated on the propagation of light in a multimode degenerate paramet- ric amplifier. the electric-field operator Ê(z. (3. the equal-time commutation relations can be derived   Â(z. ω0 = (3. t). Huttner et al. so that the equivalence of the two theories (for n independent of ω) is not excluded.11) 0 c n(ω0 + m ) n(ω0 − m ) is the coupling constant between the different modes. .5) m 20 cT n(ωm ) while in Abram (1987) the operator is independent of the medium.1 Momentum-Operator Approach 87 In the case of a linear dielectric medium in contrast with Abram (1987). t) + H. From relations (3. the momentum operator is obtained    Ĝ nonlin (z) = λ(m ) â † (z.c.9) 4 m where ωp m ≡ ωm − ω0 . The equal-space com- mutation relations are conserved. ω0 − m )eikp z + H. From relation (3. .10) 2 and  χ (2) |E| ω0 + m ω0 − m λ(m ) ≡ (3. ωm ). . (3.8) where E (+) (z.3. t) Ê (−) (z. The postulated relation (3. − D̂(z  .7) Attempting at the quantization in a nonlinear medium. with the pump frequency ωp . (3. For such a medium. t) = iδ(z − z  )1̂. t) = χ (2) E (+) (z. t) = â(z.6) m where km = n(ωmc )ωm is the wave vector in the (linear) medium.4) and (3. .c.3). t) is dependent on the refractive index n(ωm )   ωm   Ê(z.c.3) then leads to the nonlinear part of the momentum-flux operator   2  ĝnonlin (z. ωm )e−iωm t + H.

For exam- ple. but. 1985. a wave vector mismatch appears as energy (frequency) nonconservation. In the case |Δk(m )| > λ(m ).88 3 Macroscopic Theories and Their Applications It is assumed that the phase-matching condition at ω0 . for which the modal approach is very convenient. a standard two-port homodyne detection scheme is assumed. Collett and Gardiner 1984. it mixes effects related to spatial pro- gression of the beam with the spectral manifestations of the nonlinearity. In relation to the experi- ment. It is the cavity quantum electrodynamics. Another technique considers the spatial progression of the temporal Fourier components of the local electric field (Yurke et al. For the correlation function g S (τ ) of the photocurrent difference and its Fourier transform y(η) = g S (τ )e−iητ dτ. the Bogoliubov transformation for squeezing emerges and amplifying behaviour can be recognized. Let us mention its use in the development of the input–output formalism for nonlinear interactions in cav- ity (Yurke 1984. where the usual interpretation of homodyne detection in terms of the field quadratures is used. Caves and Crouch 1987). Carmichael 1987). (1990).2 Slowly Varying Amplitude Momentum Operator Nevertheless. It is found that the phase-mismatch Δk(m ) is proportional to m2 . the evolution is not essentially different from that in a linear medium. there is a class of problems. The result is comparable with Crouch (1988).7) can be recovered only approximately. t) of the frequency ω0 . for the case of the travelling-wave parametric generation (Tucker and Walls 1969).12) 2 and relation (3. t). t) ≈ δ(z − z  )1̂ (3. For the equality |Δk(m )| = λ(m ). the squeezing effect is band lim- ited. n(ωp ) = n(ω0 ). Several authors have tried to return the quantum-mechanical propagation to direct space. The modal approach can describe many of the features of travelling-wave phenomena. For the nonlinear medium. − D̂ (+) (z  . It has been shown that the values of y(η) can be minimized uniformly enough by an adequate choice of the local oscillator phase. 1987.1. the equal-time commutation relations are   i Â(−) (z. The propagation of light in a magnetic (dielectric) medium is not considered in quantum optics. As far as |Δk(m )| < λ(m ). It is an alternative of the conventional reciprocal space approach to . (3.13) we refer to Huttner et al. the amplifying is present. where the light is mixed at a beam splitter with a strong local oscillator ε(z. One technique (Drummond and Carter 1987) involves the partition of the box of quantization into finite cells. in principal. Gardiner and Collett 1985. is satisfied. but the increase is only linear not exponential. 3. We proceed with the field inside an effective (linear or nonlinear) medium and the direct-space formulation of the theory of quantum optics as presented by Abram and Cohen (1991).

P = χ (1) E + χ (2) E 2 + · · · + χ (n) E n + · · · . To impose the canonical structure on the field. Abram and Cohen (1991) simplify the geometry for the electromagnetic field so that the electric field E is polarized along the x-axis. the vector potential is polarized along the x-axis and is related . but it can be introduced phenomenologically (Drummond and Carter 1987). (3. in terms of an effective linear or nonlinear polarization. They use the Heaviside– Lorentz units and take  = c = 1. Drummond and Carter 1987). In this simple geometry the Maxwell equations reduce to two scalar differential equations ∂E ∂B =− . Glauber and Lewenstein 1991. which can be expressed as a converg- ing power series in the electric field E. while propagation occurs along the z-axis.15) ∂z ∂t where the electric displacement field D is defined by D = E + P. In propagation problems. The dispersion cannot be taken into account rigorously within a quantum-mechanical theory based on the effective (macroscopic) Hamiltonian formulation (Hillery and Mlodinow 1984).17) where χ (n) is the nth-order susceptibility of the medium. Hillery and Mlodinow 1984. The theory is applied to squeezed light generation by the parametric down-conversion of a short laser pulse as an illustration. In the assumed geometry. such an effective theory still permits a quantum-mechanical description of the field (Jauch and Watson 1948. they introduce the vector potential A and adopt the Coulomb gauge in which the scalar potential vanishes and A is transverse. (3. An important simplification of quantum optics results when the microscopic description of the material is replaced by a macroscopic description. This approach does not use the conventional modal description of the field. Glauber and Lewenstein 1989. (3. In spite of the phenomenological treatment of the medium. In classical nonlinear optics the slowly varying amplitude approximation of the electromagnetic wave equation has arisen. the magnetic field B along the y-axis.14) ∂z ∂t ∂B ∂D =− . They have derived an operatorial wave equation that relates the temporal evolution of an electromagnetic pulse to its spatial progression. Their approach relies on the electromagnetic momentum operator as well as on the Hamiltonian and is restricted to the dispersionless lossless nonmag- netic dielectric medium. The appeal of the classical theory of optics may consist in its considering material as a continuous dielectric characterized by a set of phenomenological constants. Shen 1967.3. (3. the interactions undergone by a short pulse of light are examined.1 Momentum-Operator Approach 89 quantum optics.16) with P being the polarization of the medium.

90 3 Macroscopic Theories and Their Applications to the electric and magnetic fields by ∂A E =− (3. 1989).24) ∂z In setting up the Hamiltonian functional. 1 2 1 1 1 L= (E − B 2 ) + χ (1) E 2 + χ (2) E 3 + χ (3) E 4 + · · · . the electric field E is to be expressed in terms of the electric displacement.17) (Hillery and Mlodinow 1984). (3. (3.20) is not renormalizable (Power and Zienau 1959.22) ∂t 1 2 1 2 3 = (B + A2 ) + χ (1) E 2 + χ (2) E 3 + χ (3) E 4 + · · · .21) ∂ Ȧ The Lagrangian density is then transformed to some components of the energy– momentum tensor of the electromagnetic field inside a nonlinear medium Θμν .18) ∂t and ∂A B= . (3. (3.25) where the β coefficients may be expressed in terms of the susceptibilities χ (n) through definition (3. Let us note that the theory with the effective Lagrangian density (3. Babiker and Loudon 1983. namely the energy density ∂A Θtt = Π −L (3.20) is the electric displacement ∂L Π= = −D. Cohen-Tannoudji et al.23) 2 2 3 4 and the momentum density ∂A Θt z = −Π = D B.21). . It is assumed that E = β (1) D + β (2) D 2 + β (3) D 3 + · · · . Drummond and Carter 1987). (3. The canonically conjugated momentum of A with respect to the Lagrangian density (3. Woolley 1971.20) 2 2 3 4 which is known to be the most general density dependent only on the electric field and having the gauge invariance. (3. which is the canonically conjugated momentum of A according to relation (3.16) and relation (3.19) ∂z The effective Lagrangian density has been chosen (Hillery and Mlodinow 1984.

. Not caring for divergencies.28) and using relation (3. respectively. such as B̂ D̂ 2 + D̂ B̂ D̂ + D̂ 2 B̂ B D 2 .27) V V where the integration is over the cavity obeying periodic boundary conditions and lower and upper limits are supposed to converge to −∞ and ∞. t).19) in the simple geometry. The field can now be quantized by replacing each field variable by the corre- sponding operator and by replacing the Poisson bracket between the displacement D and the vector potential A by the equal-time commutator [ D̂(r. Â(r . Abram and Cohen (1991) consider any product of noncommuting operators that appear in an expression as fully symmetrized. . t). but rather in terms of its spatial derivative B̂. d3 r.29) where d δ  (z − z  ) = δ(z − z  ) (3.e.26) and momentum (3.27) operators. i. (3.30) dz is the derivative of the δ function.26) V 2 2 3 4 while the momentum has the form G= Θt z d r = 3 B D d3 r. (3. B̂(z  . (3. The vector potential A replaced by the operator  does not appear explicitly in the Hamiltonian (3. we obtain that [ D̂(z.3.28) exactly by its (−i) multiple. (3. including all possible permutations of the individual field operators. t)] = −iδ  (z − z  )1̂. t)] = iδT (r − r )1̂.1 Momentum-Operator Approach 91 The Hamiltonian functional is then written as H= Θtt d3 r V   1 2 1 (1) 2 1 (2) 3 1 (3) 4 = B + β D + β D + β D + . Taking the curl according to r of both the sides of the canonical commutation relation (3. where the transverse δT (r − r ) reduces to the ordinary δ function and where the three-dimensional position vector r can be replaced by the coordinate z.

→ .31) 3 In contrast. i. the normal order- ing and an elimination of divergencies.e. more recently they carried out the renormalization. (3. .

(3.28) corre- spond to the requirement that the field be specified over all space at one instant of time (e. (3. In analogy with the classical equations (3.34) The Heisenberg-like equation involving the momentum can be considered ∂ Q̂ = −iĜ × Q̂.32) has the solution Q̂(t) = exp(it Ĥ × ) Q̂(0) (3. The superoperator assigns opera- tors to operators. (3. the Heisenberg equation can be written as follows ∂ Q̂ = i[ Ĥ .14).92 3 Macroscopic Theories and Their Applications The description of propagative optical phenomena has been discussed within the framework of a direct-space formulation of quantum optics and the operatorial (or better commutator) equivalent of the Maxwell equations and the electromagnetic wave equation (Abram and Cohen 1991).32) ∂t where Q̂ is any field operator and the superscript × denotes the superoperator. ∂z ∂t . Equation (3. namely the commutation of the operator it loads with another operator which fol- lows. ∂z ∂t (3. at t = 0).27) and the equal-time commutator (3.35).36) Apart from the obvious similarity of equations (3. (3. there is also a difference.15) rather than with the classical equations ∂(β (1) D + β (2) D 2 + β (3) D 3 + · · · ) ∂B =− . The Hamiltonian of the electromagnetic field relates the desired spatial distribution of the field at the instant t + dt to its spatial distribution at t.33) −i Ĥ t =e i Ĥ t Q̂(0)e .35) ∂z This equation has the solution Q̂(z) = exp[−i(z − z 0 )Ĝ × ] Q̂(z 0 ). The integrals in (3. Abram and Cohen (1991) use the Kubo (1962) notation for the commutator or more exactly for a corresponding superoperator. Q̂] = i Ĥ × Q̂. It is emphasized that in the Hamiltonian formulation of mechanics the time variable plays a particular role.37) ∂B ∂D =− . but the momentum operator Ĝ relates the translated and nontranslated fields only (at the same instant of time).g. Respecting this.32) and (3. (3.26) and (3.

without resorting to a modal decomposition of a propagating pulse) is illustrated by examining the propagation of light (of the short light pulse) through a linear medium and through a vacuum–dielectric interface. 0) − iv sinh(−ivt Ĝ × ) B̂(z.38) and (3.41) Let us note again that it is an analogue of the wave equation ∂2 E ∂2 D = .41) reduces to (Ĝ × Ĝ × −  Ĥ × Ĥ × ) Ê = 0̂.45) i B̂(z.46) v Equations (3. the velocity of an electromagnetic wave in a refractive medium.42) ∂z 2 ∂t 2 not of the more complicated equation ∂ 2 (β (1) D + β (2) D 2 + β (3) D 3 + · · · ) ∂2 D = . 0). that is Ĝ × Ĥ = 0̂. (3. It is also convenient to define v = √1 .47) . For a linear medium the commutator wave equation (3. (3. Wave equation (3.40) Equations (3.44) enables one to rewrite equations (4.44) where  = 1 + χ (1) is the dielectric function of the medium.43) ∂z 2 ∂t 2 In Abram and Cohen (1991) the direct-space description of propagation (i. (3. t) = cosh(−ivt Ĝ × ) Ê(z.38) Ĝ × B̂ = Ĥ × D̂.45) and (3. 0). so that the Hamiltonian and momen- tum operators commute with each other.2b) of Abram and Cohen (1991) in the form Ê(z. 0) + cosh(−ivt Ĝ × ) B̂(z. (3. t) = Ê(z.1 Momentum-Operator Approach 93 two commutator equations can be derived Ĝ × Ê = Ĥ × B̂. (3.3.39) may be combined into the commutator equivalent of the electromagnetic wave equation Ĝ × Ĝ × Ê = Ĥ × Ĥ × D̂. In the following exposition. (3. (3. the convention c = 1 will be dissolved or not used. t) (3.2a) and (4.e.39) On assuming that the medium is homogeneous. (3.46) indicate that the linear combination Ŵv+ (z. t) = − sinh(−ivt Ĝ × ) Ê(z. t) + v B̂(z. (3.

as well as the displacement field D̂0 . incident.48) Similarly. (3. and the relations they obey. In classical nonlinear optics. (3. on which the classical theory of nonlinear optics is based (Abram and Cohen 1991). For simplicity.94 3 Macroscopic Theories and Their Applications evolves in time as Ŵv+ (z. a single nonlinear susceptibility χ (n) is considered. t) − v B̂(z. Not even in classical optics. t) = exp(−ivt Ĝ × )Ŵv− (z. In the absence of the nonlinearity. t). the linear combination Ŵv− (z. 0). Ŵv+ (ζ. We then consider three waves.52) 2  0 and the momentum operator Ĝ 0 = B̂0 D̂0 d3 r. reflected.53) which are of zeroth order in the nonlinear susceptibility χ (n) . In Abram and Cohen (1991) further a perturbative treatment of the time evolution of the field in a nonlinear medium is examined that corresponds to propagation within the slowly varying amplitude approximation. the assumption of a weak nonlinearity makes the slowly vary- ing amplitude (SVA) approximation of the electromagnetic wave equation possible (Shen 1984). In the perturbative treatment of the nonlinear propagation it is assumed that the optical nonlinearity of the medium is absent at t = −∞ and turned on adiabatically. the electric and magnetic fields in the medium. we now consider two half-spaces such that the z ∈ (−∞. Ŵc− (z. t) = Ê(z. +∞) half-space consists of a transparent linear dielectric. and transmitted. t) (3. Following the stan- dard perturbation theory (Itzykson and Zuber 1980).49) evolves as Ŵv− (z. Abram and Cohen derive the commutator equivalent of the slowly varying ampli- tude wave equation. 0) half-space is empty. t). the problem of propagation of a short pulse in a nonlinear medium can be solved in the general case. Ê 0 and B̂0 . t) = exp(ivt Ĝ × )Ŵv+ (z. D̂0 =  Ê 0 (3. while the z ∈ (0. the exact field operators in the . 0) = Ŵv− (z + vt.50) To examine the problem of the interface. 0) = Ŵv+ (z − vt. 0). (3.51) propagate under the energy operator   1 1 2 Ĥ0 = B̂02 + D̂ d3 r (3. t). Ŵc+ (z.

56) ∂t ˆ (t) the nonlinear interaction part of the Hamiltonian. Û (t) = −iλ H̃1 (3. a dimensionless parameter. The exact Hamiltonian (3.1 Momentum-Operator Approach 95 nonlinear medium. (3.3.26) can then be expressed perturbatively up to the first order in λ as Ĥ = Ĥ0 + λ Ĥ1S + O(λ2 ). t)Û (t). which is the solution to the differential equation ∂ ˆ (t)Û (t).60) where Ĥ1S is the “diagonal part” of Ĥ1 . which has been introduced for the bookkeeping of this and higher powers of χ (n) .61) n+1 with [ n2 ]  −n+1 n! Ŝn = 2  −m B̂02m Ê 0n−2m . (3.54) and B̂(z. H̃ (3. (3. (n) χ (n) Ĥ1S ≡ Ĥ1S =− Ŝn+1 d3 r. S stands for stationary.59) t→−∞ The Hamiltonian Ĥ1 is first order in the nonlinear susceptibility χ (n) . D̂ and B̂. with H̃1 ˆ (t) ≡ 1 1 H̃1 β (n) D̂0n+1 d3 r = − χ (n) Ê 0n+1 d3 r (3.58) 1 0 1 and obeys the initial condition . can be derived from the zeroth-order fields by the unitary transformation D̂(z. (3. which is expressed also by λ.62) m=0 (n − 2m)!(2m)! . t)Û (t) (3. = 1̂. (3. . t) = Û −1 (t) D̂0 (z. Û (t). which commutes with the linear Hamiltonian Ĥ0 . t) = Û −1 (t) B̂0 (z.57) n+1 n+1 or ˆ (τ ) = exp(iτ Ĥ × ) Ĥ .55) Here Û (t) is the unitary operator.

t) = i ˆ (τ ) dτ H̃ D̂0 (z.68) namely P̂W ≡ P̂W(n) = χ (n) Ŝn . Relying on the relation 0̂ = − Ĥ0× Ĥ0× D̂1 + Ĝ × × × × 0 Ĝ 0 D̂1 − Ĝ 0 Ĝ 0 ( P̂NL − P̂W ). (3. (3. t) + λ D̂1 (z.33). The notion of the diagonal part belongs to the perturbation theory which was treated in Sczaniecki (1983).71) .63) can be compared with the multiplication of the fast-varying (“carrier”) wave by a slowly varying envelope function.62).64) 1 −∞ is the first-order correction to the displacement field. (3. P̂W and its complement.96 3 Macroscopic Theories and Their Applications which leads to the decoupling of opposite-going fields.67) The nonlinear polarization P̂NL consists of two parts. the time evolution of the displacement operator D̂ can be explained and cast in the form × D̂(z. t).61) and (3. a connection with the modal approach has been mentioned in Abram and Cohen (1991). t) ≈ exp(itλ Ĥ1S ) D̂0 (z.41) can be written up to order λ0 as (Ĝ × × × × 0 Ĝ 0 −  Ĥ0 Ĥ0 ) D̂0 = 0̂ (3.69) This partition again eliminates all terms that couple opposite-going waves in P̂NL . and P̂W obeys the zeroth-order wave equation (Ĝ × × × × 0 Ĝ 0 −  Ĥ0 Ĥ0 ) P̂W = 0̂. In the context of (3. The action of the superoperator on D̂0 in relation (3.65) it can be shown that the exact commutator wave equation (3.70) we can derive the commutator equation 2 Ĥ0× Ĥ1S × D̂0 = −Ĝ × × 0 Ĝ 0 P̂W . (3. According to (3.63) where λ ≡ 1 and t × D̂1 (z. (3. On introducing the nonlinear polarization P̂NL = −β (n) D̂0n = χ (n) Ê 0n . t) (3. (3. (3.66) and that to order λ1 as 2 Ĥ0× Ĥ1S × D̂0 = − Ĥ0× Ĥ0× D̂1 + Ĝ × × × × 0 Ĝ 0 D̂1 − Ĝ 0 Ĝ 0 P̂NL .

73) ∂z 2  where Ẽ is the envelope function of the electric field. (3. the commutator SVA equation relates directly the slow component of the temporal evolution of a short pulse of the displacement field D̂0 to the long- scale modulation of its spatial progression. the commutator wave equation (3. which is written as Abram and Cohen (1991) ∂ ∂2 2ik Ẽ = 2 PW .1 Momentum-Operator Approach 97 which has been compared to the classical slowly varying amplitude (SVA) wave equation. in the context of (3. more often.78) In this form.75) 2 namely χ (n) Ĝ SVA ≡ Ĝ (n) = R̂n+1 d3 r. This problem can be remedied by defining an effective “SVA” momentum operator such that it obeys 1 × Ĝ × SVA D̂0 = Ĝ P̂W . Also concerning P̂W . (3. With definition (3. .76) and (3.74). the connection to the modal approach has been shown in Abram and Cohen (1991).71) whose right-hand side does not contain D̂0 in contrast to the left-hand side. 1 Ĝ 1 = B̂0 P̂NL d3 r.3. (3. (3.72) ∂z ∂t or. The commutator equivalent of the slowly varying amplitude wave equation will be applied to the quantum-mechanical treatment of the propagation in a nonlinear medium.76) SVA n+1 with 2 ]−1 n [ −n+1 n! R̂n = 2  −m B̂02m+1 Ê 0n−2m−1 . (3. in terms of the temporal Fourier components of Ẽ and PW as ∂ iω Ẽ(ω) = √ PW (ω). the connection to the modal approach can be shown. Let us consider equation (3.74) 2 0 The solution Ĝ SVA is the stationary part of the effective “interaction” momentum operator Ĝ 1 .71) can be written as (Ĝ × × × × SVA Ĝ 0 +  Ĥ1S Ĥ0 ) D̂0 = 0̂.77) m=0 (n − 2m − 1)!(2m + 1)! Also.77) for Ĝ SVA . (3. (3.

relations (3.87) In most practical situations. (3.80) where in accordance with the perturbation theory the forward and backward elec- tromagnetic waves are defined as Ŵv± = Ê 0 ± v B̂0 . (3. t) ≈ exp(it Ĥ1S ) Ŵv+ (z. t) =  exp(ivt Ĝ × − − SVA ) Ŵv (z + vt. t). the first-order terms V̂1± may be neglected and Ŵv± can be introduced also on the left-hand sides of equations (3. t). (3. respectively. 0) + V̂1 (z. (3. (1987) and Caves and Crouch (1987) treated this problem by using spatial differential equations for appropriately defined creation and annihilation operators.84) √ 1S ×  Ĥ1S Ŵv− = Ĝ × − SVA Ŵv . t) + V̂1+ (z. For the case of a classical pump. More recently. These equations provide a simple rule for converting the temporal evolution of the modulation enve- lope to the spatial progression. Yurke et al. (3. Equation (3.86) and (3. 0) + V̂1 (z. this problem was treated through a modal anal- ysis by Tucker and Walls (1969). V̂1± play an important role in that they incorporate the coupling to the wave going in the opposite direction and do give rise to the nonlinear reflection. t) ≈ exp(it Ĥ1S ) Ŵv− (z.80). t).63) now becomes × V̂ + (z.86) V̂ − (z.49). the forward (+) and backward (−) polarization waves are defined in analogy with (3. (3. t) =  exp(−ivt Ĝ × + + SVA ) Ŵv (z − vt. (3.82) × V̂ − (z.78) simplifies to √  Ĥ × Ŵv+ = −Ĝ × + SVA Ŵv .79) which in the absence of the nonlinearity have the form V̂0± =  Ŵv± .98 3 Macroscopic Theories and Their Applications In order to clarify the role of Equation (3.82) and (3. .e.78). √ V̂ ± = D̂ ±  B̂. t) + V̂1− (z. (3. (3.87) using (3. the travelling-wave generation of squeezed light by the parametric down-conversion of a short pulse is examined. As an illustration of the above quantum treatment. t).85) for the forward-going and backward-going waves. i.81) Relation (3.47) and (3. Nevertheless.83) where V̂1± are the first-order corrections to V̂ given by equations analogous to (3.83) can be written as follows V̂ + (z.64).

equations similar to familiar classical first-order differential equations have been derived (see Shen 1984). For a medium that exhibits a second-order nonlinearity. we introduce also the notation Ê (±) . because it is not in definition (3. the right-hand side of relation (3. Similarly. the mod- ulated wave solution W ˆ can be separated into its positive. t).3. Ê (±) . relation (3. In Abram and Cohen (1991). (3.47). ω S = ω2P . the optical frequencies retain the same sign and relation (3. t).88) to (+) Eˆ P (z.95) .84) can be expressed as 1 χ (2) × 2 − Ĝ × Ŵ = − √ Ĥ Ŵ . (3. (3.and negative-frequency parts   Wˆ = 2 Eˆ (+) + Eˆ (−) . W (3. (3. (3. and we modify relation P S P S (3. The two fields involved in parametric down-conversion are introduced: the pump field with the central pump frequency ω P and the signal field that oscillates at approximately ω S . On this assumption.94) Ĝ × (−) SVA Ê S = κ Ê (−) (+) P Ê S . (3.93) and similar equations for the negative-frequency parts.88) SVA where the subscript v and the superscript + have been omitted. The pump field consists of a short pulse whose duration TP is much longer than the optical period ω2πP . t).89) for the signal field becomes Ĝ × (+) (+) (−) SVA Ê S = −κ Ê P Ê S .89) SVA 2  0 It is convenient to separate the field into its positive.1 Momentum-Operator Approach 99 ˆ can be written in terms of a carrier wave Ŵ as The full modulated wave W follows ˆ (z. t) = exp(−iz Ĝ × )Ŵ (z. (+) Eˆ S (z.88) can be modified to (+) Eˆ (z.92) with a similar equation holding for the negative-frequency part.90) where the factor of 2 arises. t) = exp(−iz Ĝ × (+) SVA ) Ê P (z. t). (3. Eˆ .91) In the first-order perturbative treatment. In (+) (+) fact. Eˆ . t) = exp(−iz Ĝ × (+) SVA ) Ê S (z.and negative-frequency parts:   Ŵ = 2 Ê (+) + Ê (−) . t) = exp(−iz Ĝ × SVA ) Ê (+) (z.

96).102) .95). t) + i sinh κz Î P (z.100) as |(P + S)(t) = exp(iĜ SVA vt)|P(t) . Ê (+) (−) P .101) We are now in a position to describe a travelling-wave experiment of paramet- ric down-conversion. (3. and (3. where κ =  Ĝ × (+) (+) (+) SVA Ê P = −κ Ê S Ê S (3. (3. which means that Ê P stands to the left from Ê (+) P .35) and comparison with (3. a pump pulse expressed by the state |P(t) P initially traverses a nonlinear crystal that extends from z = 0 to L and generates a signal pulse in the course of its propagation. |P(t) = |P(t) P ⊗ |0 S .99) is essentially the intensity operator for the pump field and the subscript N denotes the normal ordering of the operators Ê (±) (−) P . Using the previous approx- imations. (3.97).100) where |P(t) is the state of the field (pump and signal) in the remote past. t) (3. Ê P into relation (3. t) + Ê S(−) (z. we have an analogue of the classical description of the spatial progression. the solution of (3.94).100 3 Macroscopic Theories and Their Applications χ (2) √ωS .98) Î P (z. t) Ê (+) (z. Within the undepleted pump assumption. (3.97) Let us observe that on the substitution Q̂ = Ê S(+) . (3. t) S N P S 9 /  (+) : Ê P (z. Deviating slightly from Abram and Cohen (1991). we formulate the interaction picture as follows: |(P + S)(t) = Û (t)|P(t) . t) = cosh κz Î (z. t) = Ê (−) (+) P (z. t) Ê P (z. (3.96) and Ĝ × (−) (−) (−) SVA Ê P = κ Ê S Ê S . t) N where Î P (z. we rewrite (3. t). In such an experiment. (3.93) becomes /  ˆE (+) (z. Ê S(−) .

L . (3.3. L I S (t) = P P t − . because the vacuum expectation value of the operator product (+) (−) S 0| Ê S Ê S |0 S diverges. (3. a measurement of the intensity profile of the signal pulse can be expressed by the equal-time function I S (t) (−) (+) I S (t) = P(t)| Eˆ S (L . t) = Eˆ S (L − vt.104)   with Î P 0. L L 2 P t − Î 0.104) is simplified by omitting the terms without the antinormal ordering of signal field operators. t) Eˆ S (L .1 Momentum-Operator Approach 101 The dependence on z which has been introduced by the substitution t = vz in the previous exposition is not explicit here. .107) v .  " L .  *   L .. The interaction picture is not needed for calculation of expectation values of the observables.P t− v .108) v . 0).  . v . L . There exist results for the two-time correlation function g S (t2 . (3.   ..  . )  .sinh2N κ L Î P 0. t) = sinh2 κ L A P t − . relation (3. Let us assume that |P(t  ) is a coherent state with the property        Ê (+) P (0..U P t  . t − .106) Hence. A P t  ≥ 0. = 1. t1 ) and for the nth-order photon-coincidence rate for the signal pulse g S(n) . . For example.103) after a simplification yields   ... t )|P(t ) = A P t U P t |P(t ) .103) (+) (+) Since Eˆ S (L .105) where   . 0). v P × S 0| Ê S(+) Ê S(−) |0 S . t − Lv = Î P (L − vt. v . relation (3. P P P v P v so that (after a renormalization)   L I S (L . t − P t − = A t − . (1) The numerical results have been obtained for a laser pulse that has an amplitude profile A P (t  ) at z = 0. (3. (3. (3. but have been restricted to the intensity profile measured at the exit of the crystal z = L. This can be consid- ered as legitimate. t)|P(t) .

110) It is then consistent to assume ξ = 0 everywhere in a nonlinear medium and the dual scalar potential need not be taken into account. 1994). D1 .111) [ D̂1 (x3 . B2 . As a solution to this problem it is proposed to use vector potential ψ. The displacement operators are obtained from the energy–momentum tensor developed by Serulnik and Ben-Aryeh (1991) with an alternative definition for the vector potential. where we use Λ2 = −ψ2 according to Drummond (1990. B̂2 (y3 . This choice enables Serulnik and Ben-Aryeh (1991) to work in the new Coulomb gauge. ct). (3. The Lagrangian and Hamiltonian densities are derived from the Maxwell equa- tions by using nonconventional definitions for the scalar and vector potentials. it is shown that the conventional approach that uses the standard potentials A and V is not appropriate for treating the general case of nonlinear polarization when ∇ · P = 0. ct)] = icδ(x3 − y3 )1̂. The relation between this tensor and the space–time description of propagation is analysed.102 3 Macroscopic Theories and Their Applications 3. (3. since for such cases V does not vanish. The corresponding quantum fields obey the commutation relation [Λ̂2 (x3 . It has been possible to obtain explicit expressions for all the elements of the energy–momentum tensor and to discuss their physical meaning. ct)] = −icδ  (x3 − y3 )1̂. It is shown that the scalar potential can be neglected only by using alternative definitions. The general form of the energy–momentum tensor is derived and explicit expression for its elements is given.1.3 Space–Time Displacement Operators Serulnik and Ben-Aryeh (1991) have discussed a general problem of the electro- magnetic wave propagation through nonlinear nondispersive media. (3. Let us restrict ourselves to the usually treated one-dimensional case.109) which fulfils the relation ∇ · D = 0. where only the fields E 1 . They have used the four-dimensional formalism of the field theory in order to develop an extension of the formalism introduced by Hillery and Mlodinow (1984). Further the quantization is performed and the properties of space–time displacement operators are presented. The arguments of these fields are x3 and ct. D = −∇ × ψ. ct). The space–time is described by a Lie transform (Steinberg 1985).112) . In the following we will show that the relationship between the energy–momentum tensor and the space–time description of propagation is different from that derived by Serulnik and Ben-Aryeh (1991). The complications following from the common definitions for the vector and scalar potentials are indi- cated. First. so that from the condition ∇ · B = 0 it follows that the dual scalar potential ξ obeys the equation ∇ 2 ξ = 0. (3. B̂2 (y3 . where ∇ · ψ = 0. and Λ2 are significant.

123) .1 Momentum-Operator Approach 103 Considering for this case the Hamiltonian density ⎛ ⎞ ζ . ct). (3. (3. This failure of the application of the ordinary presentations of quantum field theory has been published by Ben-Aryeh and Serulnik (1991).114) explains the role of the dual vector potential and relation (3. We could obtain the first of them as the equation of motion for the quantum field D̂1 .117) ∂ x3 c  ∂ B̂2 i ∂ B̂2 = B̂2 . Considering. The same is obtained for the quantum field D̂1 .120) [ B̂2 (x3 .122) [ D̂1 (x3 . we obtain the equations of motion in the Heisenberg picture as follows  ∂ Λ̂2 i = − Λ̂2 . ct  )] = icδ(ct − ct  )1̂. ( D̂1 B̂2 )S dx3 = − D̂1 . ct  )] = 0̂. the equal-space commutators [Λ̂2 (x3 . (3. ct  )] = 0̂.115) is essentially the second of the Maxwell evolution equations. ct  )] = icδ  (ct − ct  )1̂. D̂1 (x3 . Abram and Cohen 1991).118) ∂ x3 c ∂ x3 Relation (3.116) c is a correct quantum density for generation of the displacement as Serulnik and Ben-Aryeh (1991) indicate.115) ∂t  ∂ x3 Relation (3.117) expresses the role of the dual vector potential and relation (3. B̂2 (x3 . ct). D̂1 (x3 . (3. B̂2 (x3 . (3. The presumable equations of the spatial progression are  ∂ Λ̂2 i = Λ̂2 .<=> 1⎜ ⎜ 1 ⎟ ⎟ Ĥ = ⎜ D̂12 + B̂22 ⎟ . (3. ct). (3.119) [Λ̂2 (x3 . (3. ct). in contrast.121) [ D̂1 (x3 . ct). ct  )] = icδ  (ct − ct  )1̂.114) ∂t     ∂ B̂2 i ∂ 1 D̂1 =− B̂2 .118) is a mere tautology. (3.113) 2⎝  ⎠ where the right-hand side is symmetrically ordered (cf. Ĥ dx3 = −c .3. It is a question whether the tensor element 1 T̂ 03 = ( D̂1 B̂2 )S (3. ( D̂1 B̂2 )S dx3 = . B̂2 (x3 . Ĥ dx3 = c B̂2 . (3.

1910). We could obtain the first of them as the equation of the spatial progression for the quantum field B̂2 . 2007). and (3. The force of the radiation pressure is obtained similarly using the Lorentz force density operator as using the momentum density according to Abraham (1909.125). a direct derivation from the first principles has been announced only. there exists no space–time description which in addition to the so-called time displacement operators suggests the use of their space-displacement analogues.125) ∂ x3  ∂t Relation (3.125) is essentially the second of the Maxwell evolution equations. by their definition. (3.115). the above example is a warning against excessive trust in the spatial progression technique. A colloquium has been devoted to the momentum of an electromagnetic wave in dielectric media (Pfeifer et al. 1985) by which the atoms or the bulk matter is considered to be at rest while the electromagnetic field is propagating.104 3 Macroscopic Theories and Their Applications we obtain peculiar equation of the spatial progression  ∂ Λ̂2 i = Λ̂2 .124).124) ∂ x3   ∂ D̂1 i ∂ B̂2 = D̂1 . The tensor is Abraham’s plus the energy– momentum of the medium characterized by a dielectric pressure and enthalpy density (Abraham 1909). Since we must often guess the commutators never known before. that is by a Lie transformation.1. While the consistency of this picture with the theory of dielectrics has been demonstrated. They show that it is always possible to relate the external field in front of the medium to that behind it by the use of the shift oper- ators. As we can see from relations (3. (3. Leonhardt (2000) has determined an energy–momentum tensor of the electro- magnetic fields in quantum dielectrics. the global nature of creation and annihilation operators is lost. A successful quantum-mechanical analysis has been given for various physical sys- tems which include amplification and coupling between electromagnetic modes . are based on the energy–momentum tensor. 3. (3. (3. ( D̂1 B̂2 )S dt = − D̂1 .4 Generator of Spatial Progression Theoretical methods for treating propagation in quantum optics have been devel- oped in which the momentum operator is used in addition to the Hamiltonian.114). Serulnik and Ben-Aryeh (1991) have introduced the shift operators which.124) expresses the role of the dual vector potential and relation (3. By consistently following this idea. The theory of the radiation pressure on dielectric surfaces (Loudon 2002) accepts the expression for the momentum density of an electromagnetic wave in a transpar- ent material medium due to Abraham (1909). They have followed in their treatment Peierls’ solution of the problem of momentum conservation in matter (Peierls 1976. For a medium with nonlinear polarization. ( D̂1 B̂2 )S dt = − .

(3.127) The energy p 00 is used as the Hamiltonian for the description of time evolution. ⎡ ⎤ H gx gy gz ⎢ Sx σx x σx y σx z ⎥ T jk =⎢ ⎣ Sy ⎥.σzy . z. Starting with Caves and Crouch (1987). (1992) did the same by using a slightly different notation.1 Momentum-Operator Approach 105 (Toren and Ben-Aryeh 1994). for example. Toren and Ben-Aryeh (1994) dissociate themselves from such an approach. 1990). in a four-dimensional form. Yariv 1989).128) can be used. but the overarching generalization of both successful analyses has not been developed.3. Distributed feedback lasers have been described. line 3. t) dx dy dz. Huttner et al.2.gz ). y. but the momentum component p 03 rather merely translates in the z-direction. but they are not very explicit about the point that the use of the . The energy and momentum properties of the electromagnetic field can be des- cribed. The vector (σzx . in which contradirectional beams are amplified by an active medium and are coupled by a small periodic perturbation of a refractive index. problems of propagation are treated by expand- ing the field operators in terms of mode operators associated with definite frequen- cies. The density of the vectorial momentum (proportional to D × B) is represented by the vector (gx . by the energy–momentum tensor T jk . the four-vector p 3k p 3k = T 3k (x.σzz ) refers to a flux of momentum in the propagation direction of z.g y . (3.3 (Roman 1969). where j. In Toren and Ben-Aryeh (1994). which are space dependent (cf. The authors have drawn attention to the distributed feedback lasers (Yariv and Yeh 1984.126) σ yx σ yy σ yz ⎦ Sz σzx σzy σzz The tensor element T 00 represents the energy density. but the energy p 30 rather merely causes trans- lation of the field in time. the approach has been associated with the conservation of commutation relations for creation and annihilation operators. Imoto (1989) has developed the basic equation of motion by using a modified procedure of canonical quantiza- tion in which time and space coordinates are interchanged in comparison with the conventional procedure. z. the four-vector p 0k is defined as p 0k = T 0k (x. The momentum component in the z-direction p33 can be used as the generator of the spatial progression. The tensor element T 30 is the component of the Poynting vector standing for the flux of energy in the z-direction. y. t)c dt dx dy (3.1. In the conventional approach (Roman 1969). Ben-Aryeh et al.k = 0. Ben- Aryeh and Serulnik (1991) have shown that for the description of the spatial pro- gression. Let us take further.

128) is compatible with the canonical quantization in which the time coordinate plays the usual role. According to Toren and Ben-Aryeh (1994). The Hamiltonian of this system is given by the relation    † Ĥ =  ωâ † â − ω jres b̂ j b̂ j (3.106 3 Macroscopic Theories and Their Applications integrals (3. Linear amplification is treated by the use of momentum for space-dependent amplification.129) j and the total momentum operator is   . Travelling-wave attenuators and amplifiers can be treated as continu- ous limits of an array of beam splitters (Jeffers et al. 1993. the propagating modes are coupled to a momentum reservoir. Ban 1994).

p. Ĝ] = −iκ ∗j â + iβ jres b̂ j .136) j .135) with ρ(βres ) the density function of the reservoir wave propagation constants β jres . (3. (3. The equations of motion obtained from the momentum operator (3. (3.132) dz  j By using the spatial Wigner–Weisskopf approximation. â and b̂ j represent (in the zeroth-order) modes which are propagating in the positive direction of the z-axis.130) are dâ i  † = [â. Ĝ] = iβ â + i κ j b̂ j . (3.134) −∞ βres − β γ = {2π|κ(βres )|2 ρ(βres )} |βres =β .130) j j where the subscript res stands for the reservoir and κ j are appropriate coupling constants. † † Ĝ =  β â † â − β jres b̂ j b̂ j + (κ j â † b̂ j + κ ∗j â b̂ j ) .131) dz  j † db̂ j i † † = [b̂ . dβres . the Heisenberg–Langevin equations can be obtained  dâ 1 = i(β − Δβ) + γ â + L̂ † . (3.133) dz 2 where ∞ |κ(βres )|2 ρ(βres ) Δβ = −V. (3. and  † L̂ † = iκ j b̂ j eiβ jres z . (3.

The total momentum oper- ator is † † † † Ĝ = [β1 â1 â1 + β2 â2 â2 + κ̃ â1 â2 + κ̃ ∗ â1 â2 ]. (3.142) into equations (3. (3.142) Substituting the operators (3. Ĝ] = iβ1 â1 + iκ̃ ∗ â2 . The Hamiltonian is given by † † Ĥ0 = Ĥ = ω[â1 â1 + â2 â2 ]. (1991) and Ben-Aryeh et al. (1992) are criticized that they do not take account of the spatial dependence (3. (3. It is assumed that two modes are propa- gating in the same direction and they are coupled by a periodic change in the refrac- tive index. (3.139) Λ with κ a coupling constant. For a classical description. The equations of motion obtained from the momentum operator (3. m an integer. (3. In this connection papers Peřinová et al.141). (3. we refer to Yariv and Yeh (1984) and Yariv (1989).140). Â2 (z) ≡ â2 (z)e−iβ2 z . and Λ the “wavelength” of the spatial periodic change in the index of refraction (a perturbation in the dielectric constant). Ĝ] = iκ̃ â1 + iβ2 â2 .144) dz where 2π Δβ ≡ β1 − β2 − m (3. we get d Â1 = iκ ∗ Â2 e−iΔβ z .3.139).138) where β1 and β2 are components of the wave vectors of the two modes in the prop- agation direction of z and   im2π κ̃ = κ exp − z .1 Momentum-Operator Approach 107 The codirectional coupling is analysed.143) dz d Â2 = iκ Â1 eiΔβ z .140) dz  dâ2 i = [â2 .141) dz  We define slowly varying operators of the form Â1 (z) ≡ â1 (z)e−iβ1 z . (3. (3.145) Λ .138) are dâ1 i = [â1 .137) where the classical relation ω1 = ω2 = ω has been used. (3.

108 3 Macroscopic Theories and Their Applications

is the mismatch. A “field” mismatch may be cancelled by a medium component.
For the input–output relations we refer to Yariv and Yeh (1984), Yariv (1989), and
Peřinová et al. (1991).
In Peřinová et al. (1991), m = 0 and 2δ = −Δβ has been introduced in an
application to the codirectional coupler. In general, the solution to equations (3.143),
(3.144) coincides with the classical solution, Â j ↔ A∗j , j = 1, 2, where A1 , A2 are
the amplitudes of the waves propagating in the +z-direction.
The counterdirectional coupling is analysed in Toren and Ben-Aryeh (1994). The
total Hamiltonian is given by

† †
Ĥ0 ≡ Ĥ = ω[â1 â1 − â2 â2 ]. (3.146)

The momentum operator is

† † † †
Ĝ = [β1 â1 â1 − β2 â2 â2 + κ̃ â1 â2 + κ̃ ∗ â1 â2 ]. (3.147)

It is reasonable that the Hamiltonian and the zeroth-order operator are related,
respectively, to the flux of energy and that of the component of momentum in the

z-direction. Compared to Toren and Ben-Aryeh (1994), the operators â2 and â2 have
† ?
been exchanged. They criticize our assumption [â2 , â2 ] = −1̂, which we obtained

in Peřinová et al. (1991) from the usual equal-space commutator [â2 , â2 ] = 1̂ by this
interchange. It is tempting to have the same alternation between the opposite-going
modes as can be seen in comparison of (3.524) with (3.526) (cf. Abram and Cohen
1994). The equations of motion obtained from operator (3.147) are given by

dâ1 i ,
= [â1 , Ĝ ,â2 ↔â † ] = iβ1 â1 + iκ̃ ∗ â2 , (3.148)
dz  2

dâ2 i ,
= [â2 , Ĝ ,â2 ↔â † ] = −iκ̃ â1 + iβ2 â2 . (3.149)
dz  2

Using the slowly varying operators (3.142), in contrast with Toren and Ben-Aryeh
(1994), we obtain that

d Â1
= iκ̃ ∗ Â2 e−iΔβ z ,
d Â2
= −iκ̃ Â1 eiΔβ z , (3.150)

where Δβ has been defined by equation (3.145).
In Peřinová et al. (1991) still 2δ = −Δβ holds in an application to the counter-
directional coupler. The solution to equations (3.150) coincides with the classical
solution, Â j ↔ A∗j , j = 1, 2, where A1 , A2 are the amplitudes of the waves prop-
agating in the +z- and −z-directions. In Yariv and Yeh (1984), the solution to the
corresponding classical equations has been obtained for the boundary conditions

3.1 Momentum-Operator Approach 109

A1 (z)|z=0 = A1 (0), A2 (z)|z=L = A2 (L). First, however, one obtains the solution for
the usual condition at z = 0.
In Peřinová et al. (1991), the output operators have been obtained in terms of
the input ones. While we simply determine the operators Â1 (L), Â2 (0) from two
equations for the operators Â1 (0), Â2 (0), Â1 (L), Â2 (L), we observe that in this
† ?
procedure the equal-space commutator [ Â2 , Â2 ] = −1̂ must depend on both z and

L in a complicated manner, and simplifies to [ Â2 , Â2 ] = 1̂. Since the commutators
correspond to the Poisson brackets, much is illustrated by the appropriate classical
theory (Luis and Peřina 1996b). One must be aware of the fact that in formulating
the theory, Luis and Peřina (1996b) have avoided the above considerations on the
z coordinate and the generator of spatial progression and they used in the bulk of
their paper the usual time dependence and the Hamiltonian function. Although still
obscure in the case of commutators, the situation is clear in the classical case, when
the input–output transformation is characterized by the usual Poisson brackets and
the solution for the usual boundary conditions at z = 0 requires the noncanonical
transformation α2 ↔ α2∗ , with the complex amplitude α2 . The richness of their
theory is due to nonlinearities, whereas it is shown that in the quantum case only a
poor linear theory is possible.
The difficulty lies in the formulation of an appropriate dynamical operator.
Tarasov (2001) has defined a map of a dynamical nonlinear operator into a dynam-
ical superoperator. He had in mind quantum dynamics of non-Hamiltonian and dis-
sipative systems.
A quantum-mechanical treatment of distributed feedback laser using the momen-
tum operator in addition to the Hamiltonian is developed in Toren and Ben-Aryeh
(1994). The authors start from the classical description based on two coupled
dA1 1
= γ A1 − iκ A2 eiΔβ z ,
dz 2
dA2 1
= iκ ∗ A1 e−iΔβ z − γ A2 , (3.151)
dz 2

where A1 , A2 are the amplitudes of the waves propagating in the +z- and −z-
directions, respectively, κ is the coupling constant, γ the amplification constant,
and Δβ is given by (3.145), with β1 = β, β2 = −β. The solution of the classical
equations is well known (Yariv and Yeh 1984, Yariv 1989) and it shows that on a
condition the amplification becomes extremely large. The classical theory does not
include the quantum noise which follows from the amplification process. Unfortu-
nately, Toren and Ben-Aryeh (1994) did not develop the overarching generalization
of the analysis of amplification and that of the counterdirectional coupling. To the
best of our knowledge, such a quantum-mechanical theory is not in hand.
The treatment of parametric down-conversion and parametric up-conversion by
Dechoum et al. (2000) is interesting with its use of the Wigner representation of
optical fields, but it starts just from the Maxwell equations for the field opera-
tors for the lossless neutral nonlinear dielectric medium. With the use of common

110 3 Macroscopic Theories and Their Applications

approximation of treating the laser pump as classical, classical equations of nonlin-
ear optics are obtained.

3.1.5 Nonlinear Optical Couplers

Optical couplers employing evanescent waves play an important role in optics, opto-
electronics, and photonics where they may be conveniently used in the switching of
light beams. They also provide a means for controlling light by light. Amplitude
and intensity behaviour of linear couplers has been investigated extensively (Yariv
and Yeh 1984, Solimeno et al. 1986, Saleh and Teich 1991). Substantial progress
in controlling light beams has been achieved after nonlinear waveguides with both
linear and nonlinear coupling have been taken into account (Finlayson et al. 1988,
Townsend et al. 1989, Leutheuser et al. 1990, Weinert-Raczka and Lederer 1993,
Assanto et al. 1994, Hatami-Hanza and Chu 1995, Hansen 1995, Weinert-Raczka
This gave new possibilities of fast all-optical switching, including digital switch-
ing, and reduction of switching power. New ways of controlling optical beams in
nonlinear couplers have been invented.
Nonlinear waveguide materials used in composing nonlinear couplers provide
new possibilities in constructing switching and memory elements for all-optical
devices. These elements are necessary for further development of optical process-
ing and computing. Classically, all-optical devices are analysed from the viewpoint
of their amplitude or intensity dependences. However, they can be treated fully in
quantum theory. Noise of light beams in nonlinear couplers is naturally included in
this quantum treatment.
Peřina, Jr., and Peřina (2000) in Section 2 indicate a consequential use of the
momentum operator we have mentioned in Section 3.1.4. In nonlinear quantum
systems where both directions of propagation are present this formalism confronts
difficulties. The generator of spatial progression is related to commutators which
do not lead to proper input (and output) commutators. The two ways of introducing
photon annihilation and creation operators obeying boson commutation relations are
not consistent mutually. Working with operators is not secure.
The authors work with the two algebras transparently. The commutators which
are preserved in spatial progression are exploited only for the derivation of evolution
equations for operators. The proper input commutators are used to the other goals.
In order to determine quantum-statistical properties of light beams, solutions of
nonlinear operator equations have to be found first. One can apply a short-length
approximation. Or pump modes can be assumed to be in strong coherent states and
a parametric approximation can be used.
The short-length approximation is used as a tool in the treatment of nonlinear
quantum systems. Calculations with operators are safe when one direction of prop-
agation is present, because the two algebras coincide. If both directions of propaga-
tion are present it is easy to use one of the two algebras in principle (cf. Peřinová
et al. 1991). Paradoxes may occur.

3.1 Momentum-Operator Approach 111

For instance, the operator equations of spatial evolution differ from the Heisen-
berg equations in interaction picture only by one or a few changes of sign. But these
changes entail, in the formalism of input commutators, that operator products on the
right-hand sides of the equations may have an incidental order. Elas, this problem is
not present in the formalism of the generator of spatial progression. We suppose it
should be better to consider boundary-value problems for equations for c-numbers
in the systems where both directions of propagation underlie to description and to
quantize at the end.
The parametric approximation is used as another instrument. It leads to linear
evolution equations of operators. Also here a description of a quantum system is
specific in which two directions of propagation are present. But linear equations
comprise no products on the right-hand sides.
A number of quantum descriptions are related to two modes and can be specified
† †
as two linear equations (and their conjugates) for the operators Âa , Âb , Âa , Âb . The
right-hand sides of these equations are often independent of time. We may let λ1,2,3,4
denote the eigenvalues of the matrix of the right-hand sides. Introducing operators
Âc , Âd as appropriate Bogoliubov transforms of the operators Âa and Âb , one can
express any quantum description (a regularity is assumed) in one of the six normal
forms (Williamson 1936)

d Âc d Âd †
= a †c , = b Âd (3.152)
dt dt
for λ1,2,3,4 = ±a, ±b;

d Âc d Âd †
= a †c + b Âd , = −b Âc + a Âd (3.153)
dt dt
for λ1,2,3,4 = ±a ± ib;

d Âc d Âd
= −iρa Âc , = −iσ b Âd , ρ, σ = ±1, (3.154)
dt dt
for λ1,2,3,4 = ±ia, ±ib;

d Âc d Âd
= a †c , = −iρb Âd , ρ = ±1, (3.155)
dt dt
for λ1,2,3,4 = ±a, ±ib;

d Âc 1 † d Âd 1 †
= a †c + ( Âd − Âd ), = ( †c + Âc ) + a Âd (3.156)
dt 2 dt 2
for λ1,2,3,4 = ±a, ±a;

d Âc ρ d Âd ρ †
= i ( †c − Âc ) − a Âd , = a Âc + i ( Âd − Âd ), ρ = ±1 (3.157)
dt 2 dt 2

112 3 Macroscopic Theories and Their Applications

for λ1,2,3,4 = ±ia, ±ia. The formulae for Âc and Âd are complicated and they are
not presented here.
Peřina and Peřina, Jr. (1995a) have studied a codirectional coupler composed
of one linear and one nonlinear waveguide. Second-subharmonic mode (b1 ) and
pump mode (b2 ) nonlinearly interact in the waveguide b. The second-subharmonic
mode b1 also interacts linearly with mode a in the waveguide a. The corresponding
momentum operator in interaction pictures is written in the form

† †
Ĝ int =  −κab1 Âa Âb1 − Γb Â2b1 Âb2 exp(iΔkb z) + H.c. , (3.158)

where κab1 denotes the linear coupling constant of modes a and b1 and Γb is the non-
linear coupling constant between modes b1 and b2 . The nonlinear phase mismatch
Δkb is defined as Δkb = 2kb1 − kb2 and kb1 (kb2 ) means the wave vector of mode b1
(b2 ). In (3.158), Âa , Âb1 , and Âb2 stand for optical-field operators of modes a, b1 ,
and b2 in interaction pictures. The conservation law
† †
Âa† (z) Âa (z) + Âb1 (z) Âb1 (z) + 2 Âb2 (z) Âb2 (z) = const. (3.159)

is fulfilled by the solution of Heisenberg equations in the interaction picture.
Peřina (1995a,b) and Peřina and Bajer (1995) have analysed squeezing of the
light in a short-length approximation. The assumption of a strong coherent field in
mode b2 with the amplitude ξb2 leads to the linearization of the operator equations
of motion. The analysis leads to at least one positive eigenvalue. Amplification may
occur dependent on the initial conditions. In the case where |Γb ξb | < |κab1 |, oscil-
lations also occur in the spatial development of quantities characterizing the fields.
Results on squeezing of the light have been obtained (Peřina and Peřina, Jr. 1995a).
The assumption of a strong coherent field in mode b2 and the linearization are
related here to three types of behaviour, which can be distinguished also using nor-
mal forms (3.152), (3.153), and (3.156).
Peřina and Peřina, Jr. (1995b) have treated a contradirectional coupler composed
again of one linear and one nonlinear waveguide. Mode a propagates against modes
b1 and b2 . The appropriate conservation law reads as
† †
− Âa† (z) Âa (z) + Âb1 (z) Âb1 (z) + 2 Âb2 (z) Âb2 (z) = const, z = 0, L; (3.160)

† †
Âa† (0) Âa (0) + Âb1 (L) Âb1 (L) + 2 Âb2 (L) Âb2 (L)
† †
= Âa† (L) Âa (L) + Âb1 (0) Âb1 (0) + 2 Âb2 (0) Âb2 (0). (3.161)

A phase matching (Δkb = 0) is assumed.
A formulation of short-length approximation seems to be obvious, but it uses
equal-space products of field operators. We can return to the boundary-value prob-
lem for classical equations which have the same solutions in the case of

3.1 Momentum-Operator Approach 113

contradirectional propagation as the initial-value problem for codirectional coupler
up to the second order in L provided we write Aa (L) in place of Aa (0). Quantization
up to the second order in L can be done in the case of the contradirectional propaga-
tion by using the quantum input–output relations of the codirectional coupler where
Âa (L) is written in place of Âa (0).
The assumption of a strong coherent field in mode b2 leads to linear operator
equations of motion as in the case of the codirectional coupler. Introducing sa = ±1,
sa = 1 when mode a propagates along with modes b1 and b2 and sa = −1 when it
propagates counter to the latter, we can write the eigenvalues as
λ1,2,3,4 = ± |Γb ξb2 | ± |Γb ξb2 |2 − sa |κab1 | . (3.162)

In the case of the contradirectional coupler oscillations cannot occur. Results on
squeezing of the light have been obtained (Peřina and Peřina, Jr. 1995b,c).
Let us note that one cannot assess input–output relations so easily in this case
using only the eigenvalues. In the case of the codirectional coupler the input–output
relations are just the solutions of the initial-value problem and their dependence on
exp(λ1,2,3,4 L) is linear. In the case of the contradirectional coupler the input–output
relations depend on exp(λ1,2,3,4 L) in a nonlinear way.
Peřina and Bajer (1995) have studied also a codirectional coupler with four
modes. A mode of frequency ω1 (b1 ) and a mode of frequency ω2 (b2 ) nonlinearly
interact in the waveguide b. The pump mode b1 of frequency ω1 interacts linearly
with mode a1 and the mode b2 of frequency ω2 is coupled linearly with mode c
(ω2 = 2ω1 holds). The momentum operator (3.158) is modified to the form
† † †
Ĝ int =  −κab1 Âa Âb1 − κcb2 Âc Âb2 + Γb Â2b1 Âb2 + H.c. , (3.163)

where κcb2 is the linear coupling constant of modes c and b2 , κab1 and Γb have the
original meaning. The conservation law

† †
Âa† (z) Âa (z) + Âb1 (z) Âb1 (z) + 2 Âb2 (z) Âb2 (z) + 2 †c (z) Âc (z) = const. (3.164)

is obeyed by the solutions of Heisenberg equations in the interaction picture. Peřina
and Bajer (1995) have treated squeezing of the light in the short-length approxima-
tion also in this case.
Mišta, Jr. and Peřina (1997) have investigated a codirectional coupler with five
modes. Second-subharmonic modes (a1 , c1 ) and pump modes (a2 , c2 ) nonlinearly
interact in the respective parts (a, c). The second-subharmonic mode a1 also inter-
acts linearly with mode c1 via mode b in the part b. On assuming linear and non-
linear phase matching, Peřina, Jr. and Peřina (2000) describe the coupler with the
following momentum operator:
Ĝ int =  Γa Âa21 Âa† 2 + Γc Â2c1 †c2 + κa1 b Âa† 1 Âb + κbc1 †c1 Âb + H.c. . (3.165)

114 3 Macroscopic Theories and Their Applications

The solutions of Heisenberg equations in the interaction picture obey the conserva-
tion law

Âa† 1 (z) Âa1 (z) + 2 Âa† 2 (z) Âa2 (z)

+ Âb (z) Âb (z) + †c1 (z) Âc1 (z) + 2 †c2 (z) Âc2 (z) = const. (3.166)

In a short-length approximation results on squeezing of the light have been obtained.
Peřina and Peřina, Jr. (1996) have studied a codirectional coupler composed of
two nonlinear waveguides. While they have used a parametric approximation from
the very beginning (Peřina, Jr. and Peřina 2000), here we present a generalization of
the momentum operator (3.158)
Ĝ int =  Γa Âa21 Âa† 2 exp(iΔka z) + Γb Â2b1 Âb2 exp(iΔkb z)
+ κab Âa1 Âb1 + H.c. , (3.167)

where a1 ≡ a, κab ≡ κab1 , and Γa is the nonlinear coupling constant between modes
a1 and a2 . The nonlinear phase mismatch Δka is defined as Δka = 2ka1 − ka2 and
ka1 (ka2 ) means the wave vector of mode a1 (a2 ). The parametric approximation
has consisted in replacements Âa2 → ξa2 exp(iΔlz), Âb2 → ξb2 exp(−iΔlz), where
Δl = 12 (kb2 − ka2 ), on assuming also some linear coupling between modes a2 and
b2 . Korolkova and Peřina (1997a) have obtained and discussed solutions on some
simplifying assumptions. Karpati et al. (2000) studied all-optical switching in this
system. Peřina and Peřina, Jr. (1996) and Korolkova and Peřina (1997a) have con-
sidered a contradirectional coupler as well.
Janszky et al. (1995) were first to investigate a coupler composed of two nonlin-
ear waveguides with nondegenerate optical parametric processes. Pump (aP ), signal
(aS ), and idler (aI ) modes in one waveguide are assumed to interact linearly with
their counterparts (bP , bS , bI ) in the other waveguide. The coupler is described by
the following momentum operator
  † † †
Ĝ =  ki âi âi + [Γa âaP âa†S âa†I + Γb âbP âbS âbI + H.c.]
i=aP ,aS ,aI ,bP ,bS ,bI

† † †
+ [κP âaP âbP + κS âaS âbS + κI âaI âbI + H.c.] , (3.168)

where ki denotes the wave vector of the ith mode along z-axis, Γa (Γb ) is the non-
linear coupling constant of modes aP , aS , aI (bP , bS , bI ) in waveguide a (b), and κP ,
κS , and κI stand for the linear coupling constants between the two pump, the two
signal, and the two idler modes.
Herec (1999) has used a short-length approximation to solve the Heisenberg
equations in the interaction picture and he has obtained results on squeezing of the
light. Mišta, Jr. (1999) has assumed strong coherent field in pump modes, κP = 0,
and phase matching, discerned three (of five) regimes in spatial evolution of the
coupler, and obtained various results on the squeezing.

3.1 Momentum-Operator Approach 115

Optical fibres and certain organic polymers with high third-order nonlinearities
may be used for the construction of couplers based on the Kerr effect. The nonlinear
directional couplers are interesting as they exchange energy periodically between
the guides like linear ones for low total intensities while they trap the energy in the
guide into which it has been launched initially for high intensities (Jensen 1982).
A coupler with two nonlinear waveguides (denoted as a and b) with Kerr nonlin-
earities has been considered by Chefles and Barnett (1996). It is described by the
momentum operator Ĝ int (ka = kb is assumed) (Korolkova and Peřina 1997b),

†2 †
Ĝ int =  g Âa†2 Âa2 + g Âb Â2b + gab Âa† Âa Âb Âb

(1999a. (3. The real constant gab describes nonlinear coupling of the modes and the real † † constant κab characterizes linear coupling of the modes. The behaviour of the coupler may exhibit a bifurcation in dependence on the parameter 1 η= |2g − gab | (|Aa |2 + | Ab |2 ) (3. The behaviour is more compli- cated in the quantum regime.c. Fiurášek et al.b) have obtained interesting results on assuming also initial coherent states. The threshold is at η = 1. This way. (3. An optimum energy exchange between modes a and b occurs if the difference of the phases of the complex amplitudes of modes a and b equals π2 or − π2 . We note that the solution is exact for gab = 2g. Korolkova and Peřina (1997b) have assumed that Âa and Âb are fast-oscillating operators due to the linear coupling and they have introduced the operators 1 B̂a = √ [ Âa exp(−iκab z) + Âb exp(iκab z)]. † + κab Âa Âb + H. In the interaction picture a numerical solution of the Schrödinger equation may use invariant subspaces defined as eigenspaces of the constant of motion (Chefles and Barnett 1996) and it is exact for any initial state from a finite direct sum of such subspaces.169) Here g means a Kerr nonlinear coupling coefficient which is the same in both waveg- uides. . Then the operators B̂a B̂a and † B̂b B̂b are conserved along z. The operator Âa Âa + Âb Âb is a constant of motion. 2 1 B̂b = √ [ Âa exp(−iκab z) − Âb exp(iκab z)].171) 2κab in the classical regime.170) 2 † On an approximation a solution has been obtained. The character of the evolution of mean photon numbers in the regions of revivals can be controlled by the z-dependent linear coupling constant κab (z) (Korolkova .

Fiurášek and Peřina (1998. Jr. 2003). Squeezing in a given waveguide also is preserved in such a way. Peřina. A codirectional coupler composed of two waveguides is described with the momentum operator Ĝ:   †  † † † .116 3 Macroscopic Theories and Their Applications and Peřina 1997c). and Peřina (1997) have paid attention to the couplers which are based on Raman and Brillouin scattering. a nonlinear optical switching matrix has been considered (Liu et al. In the classical regime. 2000a.b) have continued this work. 1999. Switching of energy between the waveguides is achieved by a suitable profile of the coupling functions.

S.A.b l=L. Ĝ =  k jl â jl â jl + g̃ jA â jL â jV â jA + g̃ jS â jL â jV â jS + H.c.V . j=a.

l = L (laser).174) 2 2 . j = a. and N denotes the number of surrounding waveguides. the momentum operator Ĝ n is  Ĝ n = Ĝ + [ξa exp(2ika z)âa†2 + H. and V (phonon) in waveguides j.c. It is described by the following momentum operator Ĝ ⎡ ⎤  N †  N † Ĝ =  ⎣ka âa† âa + kb j âb j âb j + gab j (âb j âa† + âb j âa )⎦ . b. Fiurášek and Peřina (2000a) have considered a Raman coupler with broad phonon spectra. † † +  κS âaS âbS + κA âaA âbA + H.172) where k jl are wave vectors of modes l. A parametric approximation consists in assuming strong coherent states in pump modes aL and bL . κS (κA ) is the linear coupling constant between the Stokes (anti-Stokes) modes in different waveguides. If the central waveguide a contains a second-order nonlinear medium. (1997) have treated one central waveguide (a) which interacts linearly with a greater number of mutually noninteracting waveguides (b j ) in its surroundings.173) j=1 j=1 where ka (kb j ) is the wave vector of mode a (b j ). S (Stokes).c.b). (3. The method utilized is based on linear operator corrections to a classical solution. They have described the phonon systems of the waveguides with multimode boson fields. A (anti-Stokes). Here g̃ jS (g̃ jA ) describes the Stokes (anti-Stokes) nonlinear coupling in waveguide j. Mogilevtsev et al. Fiurášek and Peřina (1999) have used another approximation in solving Heisen- berg equations in the interaction picture. (3. Then these fields have been eliminated from the description of the coupler using the Wigner–Weisskopf approximation (Peřina 1981a. Vectors characterizing phase mismatches are defined as follows: Δk jS = k jL − k jV − k jS . . ΔkS = kaS − kbS . and ΔkA = kaA − kbA . (3. Δk jA = k jL + k jV − k jA .]. gab j is the linear coupling constant between modes a and b j .

2 Dispersive Nonlinear Dielectric 117 where ξa2 stands for the amplitude of the pump field in the central waveguide. A reservoir spectrum has been considered which has a gap. The interaction momentum operator Ĝ int is .3. Only a nonlinear coupling is present. (1996) have considered a coupler composed of one waveguide with χ (2) medium and the other one with χ (3) medium. Mogilevtsev et al. The behaviour of the linear and nonlinear couplers agrees with the idea that the surrounding waveguides form a reservoir.

whereas it has an exponential character for 2ga > Nb gab . 3. The third type of behaviour could result using the superposition principle. gb stands for the Kerr constant. (3. and gab means the non- linear coupling constant between modes a and b. When an initial coherent state with the amplitude ξb in mode b is assumed. When an initial Fock state with Nb photons in mode b is assumed. But Abram and Cohen (1991) work with a single field. essentially the number required by the classical slowly varying amplitude approximation. the mean number of photons in mode a oscillates in z and the exponential terms can be neglected for 2ga  |ξb |2 gab . It increases exponentially in z and the oscillating terms do not contribute significantly for 2ga  |ξb |2 gab . the existence of the optical fields that do not change during propagation is conditioned by the frequency dispersion and the nonlinearity of the medium. As known. and Glauber and Lewenstein (1989) . Contrary to Abram and Cohen (1991). papers by Knöll (1987). Białynicka-Birula and Białynicki-Birula (1987). A macroscopic quantization was to take into account both the properties. 3. the solution for the operator Âb (z) oscillates in z for 2ga < Nb gab . one can distinguish two regimes essentially. Till 1990. The paradox of the validity of both the approaches can be resolved only by a detailed microscopic theory.175) where ga describes the process of second-subharmonic generation and includes the coherent pump amplitude. Drummond creates an arbitrary number of slightly varied copies of the vacuum electromagnetic field for the nonlinear dielectric medium. The nonlinearity would appear as an interaction of narrow-band fields.2 Dispersive Nonlinear Dielectric The spatio-temporal quantum description has been adopted in optics in spite of its complexity due to quantum solitons. Drummond (1990) generalizes the treatment of a linear homogeneous disper- sive medium (Schubert and Wilhelmi 1986).1 Lagrangian of Narrow-Band Fields Drummond (1990) has presented a technique of canonical quantization in a gen- eral dispersive nonlinear dielectric medium. Assuming the vacuum state in mode a and an incident pure state in mode b.2. †2 † Ĝ int =  ga ( Âa†2 + Âa2 ) + gb Âb Â2b + gab Âa† Âa Âb Âb .

e. Haus et al. Shelby et al. For simplicity. Let u. ∂t ∂D(x. i. 1987. . On using a tensorial product. etc. Hillery and Mlodinow (1984) were attractive with their use of the idea due to Born and Infeld (1934) for the quantization of homogeneous nonlinear nondis- persive medium. tensorial notation is used which will occur also elsewhere in this book.g. uvw.e. t) = − . be vectors. i. uv. 1989. . 1990. is constructed. t) ∇ × E(x. 1989). In addition to the quadrature squeezing in (Rosenbluh and Shelby 1991). is formed. . The scalar product denoted by the dot · is generalized to a contraction. mean contractions. a double sum and a triple sum of products of corresponding components. . The correspondence of compo- nents is achieved by using the same notation for the last subscript of the tensor to the left as for the first subscript of the tensor to the right. In description of a nonlinear dielectric medium. but intractable. The macroscopic quantization is a route to the simplest quantum theory compati- ble with known dielectric properties unlike the microscopic derivation of the nonlin- ear quantum theory of electromagnetic propagation in a real dielectric. Also the pieces of notation : and . and quantum chaos (Toda et al. t) = . pulsed squeezing (Slusher et al. 1987).. the dielectric of interest is regarded as having uniform linear magnetic susceptibility. An excellent example of this is the fibre optical soli- tons whose quantization is given in detail in Drummond (1994). the inverse expansion is necessary here. 1989. a tensor of rank 2. a tensor of rank 3.g. ∂t . In agreement with theoretical predictions (Carter et al. Drummond and Carter 1987. Drummond et al. t) ∇ × H(x. experiments (Rosenbluh and Shelby 1991) led to evidence of quantum solitons. a simple sum after the tensors are replaced by their components and products of corresponding components are formed. Contrary to the usual description (Bloembergen 1965).e. e. but unphysical. quantum properties of soliton collisions were mea- sured (Watanabe et al. The charges are assumed to occur only in the induced dipoles of polarization. w. Lai and Haus 1989. Drummond (1994) compares the quantum theory obtained via macroscopic quantization with the traditional quantum-field theory. The tractable model quantum field theory ceases to be unphysical when it is tested exper- imentally in quantum optics. 1992) demonstrate that solitons can be considered to be nonlinear bound states of a quantum field. which uses the dielectric permittivity tensors. More recent experiments (Friberg et al. or physical.118 3 Macroscopic Theories and Their Applications could be referred to as devoted to the theory of inhomogeneous nondispersive linear dielectric. 1989). microcavity quantum electrodynamics (Hinds 1990). v. Haus and Lai 1990). The field equations are therefore ∂B(x. He concedes that most model quantum field theories prove to be either tractable. It is advantageous to begin with the treatment of a classical dielectric introduc- ing the nonlinear response function in terms of the electric displacement field D. Similar nonlinearities are encoun- tered in photonic band-gap theory (Yablonovitch and Gmitter 1987).

τ1 . τ1 . τn ) = 2π × . we express the electric vector in the form ∞ E(x.. t − τ2 ) dτ1 dτ2 0 0 ∞ ∞ ∞ .. τ1 . the (n + 1)th-rank susceptibility tensor read 1 χ(x. t − τ1 )D(x. t − τ1 )E(x. t) = 0.176) where D(x.. t − τ3 ) dτ1 dτ2 dτ3 + ··· . t) = 0..2 Dispersive Nonlinear Dielectric 119 ∇ · D(x. τ2 .. τ2 ) : E(x. τ2 . dτn . After adding the vacuum electric displacement field 0 E(x.3.178).179) respectively. t) = μH(x. (3.E(x. in general. t) = χ (x. (3. t − τ ) dτ 0 ∞ ∞ + ζ (2) (x. t) = 0 E(x. t). t − τ2 ) dτ1 dτ2 0 0 ∞ ∞ ∞ . τ3 ). ∇ · B(x. t − τ ) dτ 0 ∞ ∞ + χ (2) (x. t − τ2 ) 0 0 0 × D(x. χ̃ (n) (x.D(x. t) + P(x. t − τ1 )E(x. t). t − τ2 ) 0 0 0 × E(x. 2π  n 1 χ (n) (x. τ ) · E(x. τ ) = χ̃ (x.. ω1 . (3. B(x. .177) Here ∞ P(x.. t) to both sides of (3. 1 n (3. τ1 . τ ) · D(x. . τ1 . + ζ (3) (x.. t) = ζ (x..178) where the tensor of rank 2.. (3. ω)e−iωτ dω.180) .. ωn )e−i(ω τ1 +···+ω τn ) dτ1 . t − τ1 )D(x.. t − τ3 ) dτ1 dτ2 dτ3 + ··· . + χ (3) (x. τ2 ) : D(x. τ3 ).

ω − ω1 ) dω1 . (3. ω − ω1 ) : D̃(x. ω1 . ω3 ) (2) +2χ̃ (2) (x... ω) (2) + ζ̃ (x. ω1 .120 3 Macroscopic Theories and Their Applications where 1 ζ (x. ω2 ) = 0(2) . . ω) = D(x.. ω1 . ω2 . respectively. ω2 ) (3) + × D̃(x. τ ) = ζ̃ (x.180). ω2 ) + χ̃ (2) (x. ω) the usual frequency-dependent tensor of permittivity. In particular. ζ̃ (x.183) Here 1 and 0(n) are the second-rank unit tensor and the (n + 1)th-rank zero tensor.. Introducing the Fourier components of the electric strength field and electric displacement field. t)eiωt dt. ω)]−1 .182) with (x.+ω τn ) dτ1 ..184) and performing the Fourier transform of both sides of (3. ω2 ) : ζ̃ (x.. ω) · ζ̃ (x.ζ̃ (x. ω1 . we obtain that Ẽ(x.. ω3 ).. ω2 . ω) = E(x. (3. Ẽ(x. t)eiωt dt. ωn )e−i(ω τ1 +.D̃(x. ω2 . ω2 . ω1 + ω2 + ω3 ) · ζ̃ (x. ω3 ) = 0(3) . ω)e−iωτ dω. ω1 )ζ̃ (x. +χ̃ (3) (x. 2π  n 1 ζ (x. . τn ) = (n) 2π × . ω1 . ω1 . ω) = 1. ω) = ζ̃ (x. ω1 )ζ̃ (x. ω + ω ) · ζ̃ (x.181) (n) 1 n and the tensors on the right-hand sides of (3. respectively.185) .. ω) = [(x. (2) (x. (3. 1 2 1 (3) (x.. ω − ω1 − ω2 ). ω1 . ω1 )D̃(x. ω) = 0 1 + χ̃ (x... ω2 + ω3 ) : ζ̃ (x. . ω2 )ζ̃ (x. ω3 ) . ω − ω1 − ω2 ) dω1 dω2 + ··· . (3. ω1 )D̃(x.. τ1 . ω) · D̃(x. ω1 )ζ̃ (x.181) can be expressed from the equa- tions (x. .. ζ̃ (x. dτn (3. D̃(x. ω ... ζ̃ (x.

ω − ω1 ) ≡ χ̃ (2) (x. thus  N D(x. ω)] · E(x) + B(x. the displacement D is expanded in terms of a series of complex (envelope) functions. Again. ω).. . ω − ω1 ).3.192) . ω). t) = Dν (x)e−iω t . (3. We will treat the time-averaged linear dispersive energy H for a classical monochromatic field at nonzero frequency ω. our notation slightly differs from that in Drummond (1990). (3. ω1 .. A similar extension of notation is conceivable (2) also in tensors ζ̃ (x. t) = Eν (x. t) d3 x. Bleany and Bleany 1985). ω N . For a permittivity (x. each of which has a restricted bandwidth. Landau and Lifshitz 1960. t).188) ν=−N where D−ν = (Dν )∗ (3. ω) in Peřina (1991) comprises sums instead of the integrals.190) Here. ω1 .. this can be written in terms of a complex amplitude E (Bloembergen 1965. t) = 2 Re E(x)e−iωt . ω) ≡ χ̃ (1) (x. (3.191) ν=−N where E−ν = (Eν )∗ (3.186) V ∂ω 2μ where the angular brackets mean the time average and   E(x. (3. .189) and in the monochromatic case ν Dν (x. Let us note that the similar relation for P̃(x. ζ̃ (x. In the more general case.. ω).185) can resemble relation (2. χ̃ (2) (x. ω1 . t). ω − ω1 ).187) It is important to distinguish the monochromatic case from the case of quasi- monochromatic fields.. t) · B(x. Relation (3. (3. .2 Dispersive Nonlinear Dielectric 121 The inverse relation of D̃(x.. the electric-field vector can be expanded as  N E(x... −ω. −ω. t) = Dν (x. ω) comprises χ̃ (x. Such a (n) change does not affect only the meaning of the tensors χ̃ (n) and ζ̃ but also (and above all) the physical unit of their measurement. The relevant nonzero central frequencies are then ω−N ..   ∂ 1 H = E ∗ (x) · [ω(x.4) in Drummond (1990) on the condition that the integrals will be replaced by the sums.

ων ) − ων ν ζ̃ (x. t) = D̃(x. ων )] · Dν (x. (3. t) · [ζ̃ (x. t) = B̃(x. t) · [ων (x.193) should be replaced by ων +δ 1 Dν (x.194) 2π ων −δ Bloembergen (1965) presented the relation (3. If relation (3.197) ν=−N where B−ν = (Bν )∗ . (3. ζ̃ (x. (3.200) 2 . t) = Ẽ(x.  1  ζ̃ (x.199) 2π ων −δ Next. t  ) (t) d3 x. t  ) · B(x. ω) ≈ ζ̃ ν (x) + ωζ̃ ν (x) + ω2 ζ̃ ν (x) ≡ ζ̃ ν (x. (3.186) is exact for monochromatic fields. t). relations (3. t  ) (t) d3 x.195) 2μ By modifying the summation. it must be modified for a quasimonochromatic field as follows:    1  −ν 1  ∂ H (t ) (t) = E (x.196) 2μ To achieve a completeness.122 3 Macroscopic Theories and Their Applications and in the monochromatic case ν Eν (x. ω) can be approximated near ω = ων by a quadratic Taylor polynomial. (3. t) 2 ν=−1 ∂ων  1 + B(x. (3. t) 2 ν=−N ∂ω  1 + B(x.190) and (3. ω). 2π ων −δ ων +δ 1 Eν (x. we supplement relations (3. t  ) · B(x.198) ων +δ 1 Bν (x. t) = E ν (x)e−iω t . t) = Bν (x. ω)e−iωt dω. (3. ω)e−iωt dω. (3.191) with the expansion of the magnetic induction field  N B(x.193) In the case of quasimonochromatic fields. ων )] · Eν (x.188) and (3. we obtain the energy integral in terms of the electric displacement fields    1  −ν N ∂  H (t  ) (t) = D (x. ω)e−iωt dω.186) as sufficiently accurate for such a case.

t) = 0. t) − ζ̃ ν (x) · D̈ν (x. t)eiωt dt. t) + B (x. t) = ζ̃ ν (x) · Dν (x. t). t).203) We introduce also Λ̃(x. the prime stands for the partial derivative ∂ω . t).201) ∂ω 2 ∂ For brevity. t) · ζ̃ ν (x) · Ḋ (x. t) = μΛ̇(x. which is a functional of (components of) the dual vector potential. (3. t) d3 x. t) = Λ̃(x. B(x. t). the Taylor polynomial is not in a standard form.205) 2π ων −δ Each quasimonochromatic field obeys the Maxwell equations ∇ × Eν (x. t) = Ḋν (x. t) = ∇ × Λ(x.3. t) 2 ν=−N  1 −ν  ν 1 −ν ν − Ḋ (x. ∇ · Bν (x. ω) ≈ ζ̃ ν (x) − ω2 ζ̃ ν (x). t). ∇ · Dν (x. A canonical theory of linear dielectric will be obtained using the causal local Lagrangian. Drummond (1990) considers a Lagrangian L(Λ−N . t) · B (x. ∇ × Hν (x.204) ων +δ 1 Λν (x. t). we rewrite relation (3..202) Here we deviate slightly from Drummond (1990). .196). but he does not remove the time dependence from the right-hand side. ω) = Λ(x. 2 ν 1 ν H (x. (3. t) + iζ̃ ν (x) · Ḋν (x.196) in the form N    1  H (t ) (t) = D−ν (x. ω) − ωζ̃ (x. 2 μ (3. Drummond speaks of time aver- ages and he indicates the time average on the left-hand side and partially on the right-hand side in (3. t) · ζ̃ ν (x) · Dν (x. (3.2 Dispersive Nonlinear Dielectric 123 so that ∂ 1  ζ̃ (x.. Further explanations can be found in Drummond (1990) if necessary. (3. D(x. (3. which comprises the brackets (ω − ων ) and (ω − ων )2 . t) = −Ḃν (x. Λ N ). Moreover.207) μ .. t) = 0. (3.206) where  1  Eν (x. t) = B (x. Using the notation Ḋ ≡ ∂t∂ D. ω)e−iωt dω. This is defined as Λ.

t)] · ζ̃ ν (x) · [∇ × Λν (x. t) − ∇ × iζ ν (x) · D−ν (x. t)] · ζ̃ ν (x) · [∇ × Λ̇ν (x. t) − iD−ν (x.206) as Hamilton’s equations. (3. t)Λ̇ν (x. t)] · ζ̃ ν (x) · [∇ × Λν (x.208) while generating the Maxwell equations (3. (3. (3. Since Λν can be specified to be transverse fields. t)] · ζ̃ ν (x) · [∇ × Λ̇ν (x. t)]  1  − [∇ × Λ̇−ν (x. a canonical Lagrangian must be found that corre- sponds to (3. (3. t) + ζ ν x) · Ḋ−ν (x. t) = B−ν (x.209) 2 The canonical momenta are 1   Πν (x. t) · ζ̃ ν (x) · Ḋ (x. Healey 1982).208) In order to quantize the theory. t)] 2  −ν ν + μΛ̇ (x. t) · ζ̃ ν (x) · Ḋν (x. t) 2 ν=−N μ  − D−ν (x. t) . On the basis of (3.124 3 Macroscopic Theories and Their Applications The components of the dual vector potential fulfil linear wave equations. t)] · ζ̃ ν (x) · [∇ × Λ̇ν (x. t) · Bν (x. t) · ζ̃ ν (x) · Dν (x. t)] 2 ν=−N 1  − [∇ × Λ̇−ν (x. Drummond (1994) derived the Lagrangian using the method of indeterminate coefficients in the form N   1 L = L0 = μΛ̇−ν (x. t)] d3 x. The use of restricted variations can be realized using transverse functional derivatives (Power and Zienau 1959. We can rewrite also the Lagrangian of Drummond in the form  N  1 1 −ν L = L0 = B (x. t) 2 ν=−N − [∇ × Λ (x. t) · Λ̇ (x. t) d3 x. t)  1 −ν  ν − Ḋ (x.210) 2 where we introduce for brevity the fields (3.202) we can infer the Hamiltonian function of the form N   1 H = H0 = [∇ × Λ−ν (x. It is next necessary to derive a Lagrangian whose Lagrange’s variational equations correspond to obvious wave equations and whose Hamiltonian corre- sponds to (3.211) 2 . the variations can also be restricted to be transverse. t) d3 x.208). t)] −ν  − i[∇ × Λ−ν (x.203) again.

2 Dispersive Nonlinear Dielectric 125 On the contrary. t) 3 ν =−N ν =−N ν =−N 1 2 3 · ζ̃ (x. t) is a nonlinear energy density 1    ν1 N N N U (x.216) . t)Dν4 (x. t)Dν3 (x.. the Legendre transformation. t)|4 .210) considered as a par- tial differential equation. the field operators Λ̂ν and Π̂ν are introduced. A reason is that each Λ̇ν is to be found from (3. t)] = iδi⊥j (x − x )δμν 1̂. t)Dν3 (x. ων3 . (3. t) d3 x.213) In order to give an example. In the corresponding quantum theory.214) 4 Drummond (1990) has presented the quantization of the nonlinear medium using a treatment of modes defined relative to the new Lagrangian. t) = N D (x. Π̂ j (x . it is also interesting to note that Λ̂−ν = (Λ̂ν )† . t)δ−ων1 . t) (3) × δ−ων1 . (3.3.212) where U N (x.ων2 +ων3 ν2 (2) 1     ν1 N N N N + D (x. was not performed in Drummond (1990). i. The canonical momenta have the form (3. · ζ̃ (x.ων2 +ων3 +ων4 + ··· . a substitution of Λ̇ν in Hamilto- nian (3. (3.208) with an expression in Πν and Λν . ων2 . t). −ων1 − ων2 ) : Dν2 (x. −ων1 − ων2 − ων3 ). t) = ζ̃ |D (x. a one-dimensional case is treated and the nonlinear refractive index as the lowest nonlinearity of most universal interest.215) Since these operators are not Hermitian. (3. t) 4 ν =−N ν =−N ν =−N ν =−N 1 2 3 4 .210) also in the nonlinear case. (3. For N = 1. Π̂−μ = (Π̂μ )† . 1 (3) 1 U N (x. ω . which obey the transverse commutation relations of the form μ [Λ̂iν (x. The objective is the total Lagrangian and Hamiltonian of the form L = L 0 − U N (x. Also this theory simplifies a great deal if the plane wave one-dimensional propagation is considered.e.Dν2 (x. H = H0 + U N (x. t) d3 x. The local Lagrangian method is used as the foundation of a nonlinear canonical Lagrangian and Hamiltonian.

. with (n) 1(n) the (n + 1)th-rank unit tensor for n odd and ζ̃ (x. t)}. âk  have the standard commutators † [âk .195). This feature of the theory is due to the dependence of the Hamiltonian (3. ω) = ζ̃ (ω)1. n ≤ 3. ζ̃ (x. In part of the exposition. Conversely.2. Drummond (1994) has presented the total energy in the length L. L t  1 W (t) = μH (x. (3.202) can be completed with terms which make up the Hamiltonian dependent solely on 2 Re{Λν (x. . . . . Traditionally. Then. Drummond (1994) points out that the solution of this commutator problem is the inclusion of the important physical property of a real dielectric. the dual vector potential has the expansion    ∂ω ∂k Λ̂ (x.219) and ωk are solutions to the equations  ζ̃ (ωk ) ωk = k .2 Propagation in One Dimension and Applications Drummond (1994) discusses in detail a simplified model of a one-dimensional (n) dielectric.221) k .k  1̂. t)).220) μ This enables one to write  † Ĥ0 = ωk âk âk (3. .218) k 2Lk ζ̃ (ω k ) † where âk . t) is the magnetic strength vector. τ ) dτ dx. . . Instead of the time average of energy (3. . where ζ̃ (x. the right-hand side of relation (3. ωn )1(n) . 3. ωn ) = 0(n) for n even. (3. a set of Fourier transform fields is defined and the annihilation operators âkν and b̂kν are introduced. (3. âk  ] = δk. .186). . t) (Λ̇ν (x. t) = μ1 B(x. the description of the nonlinear medium assumes that the dispersion terms are negligi- ble. The Hillery–Mlodinow theory which does not take account of dispersion (Ho and Kumar 1993) has the electric- field commutation relation with the magnetic field modified from its free-field value.208) on both the real and imaginary parts of the components Λν (x. t) = 1 âk eikx . ω1 . (3. The operators âkν correspond to the normal modes while b̂kν generate additional necessarily vacuum modes. (Π̂νj )† commute. . (3. Neglecting the unphysical modes. τ ) Ḋ(x.217) 0 2 t0 Here H (x. t) + 2 E(x. ω1 . ωn ) = ζ̃ (n) (ω1 .126 3 Macroscopic Theories and Their Applications This entails that Λ̂iν . single polarization components are considered only.

t)Ψ̂(x. Φ̂† (x2 . t)] = δ̃(x1 − x2 )1̂. (3. the free particles interact via the Hamiltonian nonlinearity. With respect to practical applications. The photon-density amplitude field reminds us of the so-called detection operator (Mandel 1964. (3.228) . The total polariton number operator is N̂ = Ψ̂† (x. A polariton- density amplitude field is simply defined as  1  i[(k−k 1 )x+ω1 t] Ψ̂(x. (3. that lead to second harmonic and parametric interactions.223) L k where k 1 = k(ω1 ) is the centre wave number for the first envelope field.227) L k where v is the central group velocity at the carrier frequency ω1 .226) A polariton-flux amplitude can also be approximately expressed as  v  i[(k−k 1 )x+ω1 t] Φ̂(x. t) = e âk . (3. it is necessary to define photon-density and photon-flux amplitude fields. t)] = v δ̃(x1 − x2 )1̂. t) = e âk .225) L Δk where the range of Δk is equal to that of k − k 1 . (3. This field has an equal-time commutator of the form [Ψ̂(x1 . This flux has an equal-time commutator of the form [Φ̂(x1 . t) dx.222) k When there is a nonlinear refractive index or ζ̃ (3) term. t). t). such as ζ̃ (2) terms.2 Dispersive Nonlinear Dielectric 127 and (reintroducing D̂ 1 )  † Ĥ = ωk âk âk + U N ( D̂ 1 ) dx.3. It is this coupling that leads to soliton formation. (3. Ψ̂† (x2 . (3. It is also possible to involve other types of nonlinearity.224) where δ̃ is a version (L-periodic) of the usual Dirac delta function 1  iΔk(x1 −x2 ) δ̃(x1 − x2 ) ≡ e . Mandel and Wolf 1995).

â(k) are corresponding annihilation operators defined so that. On simplified assumptions. t)Φ̂(x. (3. An alter- native moving frame transformation is x t − vx ξ= . (3.232) 4 Here ω(k) are the angular frequencies of modes with wave vectors k describing the linear photon or polariton excitations in the fibre including dispersion. the electric displacement field D̂(x) is expressed as  (k)kv(k) D̂(x) = i â(k)u(k..233) In terms of the waveguide. (3. t) .229) n̄ where n̄ is the photon-number scale and t0 is a timescale. r) is included here to show how the simplified one-dimensional quantum . t) by the scaling  vt0 φ̂(x. (3. t)φ̂ (x. t) is appropriate for the system. (3. at equal times. A common choice is to define the dimensionless field φ̂(x. â † (k)] = δ(k − k  )1̂. [â(k  ). r)|2 d2 r = 1. (3.230) t0 x0 Here x0 is a spatial length scale introduced to scale the interaction times. (3. Φ̂† (x. This scaling transforma- tion is accompanied by a change to a comoving coordinate frame. r)eikx dk + H. (3. r) and |u(k.235) Here v(k) is the group velocity and (k) is the dielectric permittivity.c. The mode func- tion u(k.128 3 Macroscopic Theories and Their Applications Operationally. The first choice of an altered space variable gives x v −t vt ξv = . the nonlinear Hamiltonian is (cf. t) = Ψ̂(x.τ = . defined so that the expec- tation value φ̂ † (x.231) x0 t0 The quantization technique developed by Drummond (1990) was applied to the case of a single-mode optical fibre (Drummond 1994). τv = .214)) † 1 (3) Ĥ = ω(k)â (k)â(k) dk + ζ̃ D̂4 (x) d3 x.234) 4π where x = (x. t) is the photon-flux expectation value in units of pho- tons/second.

where v = v(k 1 ) = ∂ω . t) = Ψ̂(xv + vt. . while two.237) ∂x ∂t 2 ∂x2 . the nonlinear term χe describes an interaction potential V (xv . t). t) = − − χe ψ̂ (x v . this ∂k k=k ∂k k=k 1 reduces to the usual quantum nonlinear Schrödinger equation  ∂ ω ∂ 2 † i ψ̂1 (xv .240) This interaction potential is attractive when χe is positive as it is in most Kerr media. which he reduces to a more usual form again.. It is known that this potential has bound states and is one of the simplest exactly soluble known quantum field theories. allowing solitons to form. (3. he obtains an “unusual form” of the quantum nonlinear Schrödinger equation. Upon modifying the time variable. t) = + iχ e Ψ̂ (x. ω = ∂ 2 ω2 . t) ψ̂1 (xv . t) = Φ̂(x. it is preferable to substitute flux amplitude operator  1 Ψ̂(x. The repulsive and attractive cases were inves- tigated by Yang (1967. (3. Drummond (1994) associates an idea of the spatial pro- gression with the flux amplitude operator. 1968). Since the operators there have their standard . . In calculations. t) Ψ̂(x. (3.239) m where m = ω is an effective mass of a particle. t) Ψ̂(x. t). t) ψ̂1 (x v . Similarly.2 Dispersive Nonlinear Dielectric 129 theory relates to vector mode theory. When the interaction Hamiltonian describing the evolution of the polariton field Ψ̂(x. This theory is one-dimensional and tractable and does not need renormalization. t) (3.241) v into equation (3. (3.5 μm.3. (3.and three-dimensional versions do need renormalization. the following Heisenberg equation of motion for the field operator propagating in the +x-direction can be found   2  ∂ ∂ iω ∂ † v + Ψ̂(x. the second derivative ω can be expressed as  ω = .238) ∂t 2 ∂ xv2 1 where ψ̂1 (xv . 1 .236) 4(k 1 )c2 After taking the free evolution into account. t) in the slowly varying envelope and rotating-wave approximations is considered. xv ) = −χe δ(xv − xv ). In the case of anomalous dispersion which occurs at wavelength longer than 1.237). t). the coupling constant χe is introduced 3[χ̃ (3) (ω1 )]2 [v(k 1 )]2 χe ≡ |u(r)|4 d2 r. In a comoving reference frame.

ω. ω . (3. t) dω dz. and fails for pulse duration much shorter than this. ω. ω) is a macroscopic frequency-dependent coupling which can be assumed to be independent of z. t)] = δ(z − z  )δ(ω − ω )1̂. (3. The problem is whether these commutators are well defined. t)Ψ̂(z. t)] dω Ψ̂(z. ω. ω. ω)[ Â(z.244) and r (z. t)] dω dz −∞ 0 ∞ ∞ + ω † (z. they must have equal-time commutators. δx j is the atomic displacement operator. For this reason. In contrast.. t). The Raman effect can be included macroscopically through a continuum Hamiltonian term coupling photons to phonons of the form (Drummond and Hardman 1993) ∞ ∞ ĤR =  Ψ̂† (z. t) 0 (3. t) +  (z. ω. t) ∂z ∂t 2 ∂z 2 ∞  † −i r (ω)[ Â(z. ω.D(x̄ j )D(x̄ j )δx j . t)Ψ̂(z. ω. t)r (z.245) and ∂ Â(z. especially when high enough intensities are present. t). the existence of a corresponding set of phonon operators must be taken into account.242) j Here D(x̄ j ) is the electric displacement at the jth mean atomic location x j . t) Ψ̂(z. t) + † (z. The treatment of the quantum theory can start from the classical theory developed by Gordon (1986). t) − ir (ω)Ψ̂† (z. t)Ψ̂(z. the nonlinear Schrödinger equation requires corrections due to refractive-index fluctuations for pulses longer than about 1 ps. (3. the resulting equa- tion (Drummond 1994) appears as the quantum nonlinear Schrödinger equation with time and space interchanged. ω. and ζ Rj is a Raman coupling tensor. The corresponding coupled set of nonlinear operator equations is   2  ∂ ∂ iω ∂ † v + Ψ̂(z. ω. t) = −i Â(z. In order this interaction to be quantized. † (z  .130 3 Macroscopic Theories and Their Applications meaning. t) = + iχe Ψ̂ (z. An important physical effect in propagation is that from molecular excitations.243) −∞ 0 where [ Â(z.246) ∂t . The Raman interaction energy of a fibre is known to be [Carter and Drummond (1991)]  . Here the Raman excitations are treated as an inhomoge- neously broadened continuum of modes localized at each longitudinal location z. t) Â(z. (3. Such an interpretation means that the operators have equal-space commutators. WR = ζ Rj .

τv ): ∂ i ∂2 φ(ζ. τv ) ≈ [i f φ † (ζ.248) 0 t0 χ χ t0 z χe t2 ω ζ = . (3. an operator equation can be transformed to an equivalent pair of c-number stochastic equations for Ψ(z. τv )]φ(ζ.249) z0 χ |k | v The last terms appearing in Equation (3. the use of an enlarged nonclassical phase space can increase computation times. τv ). the disagreeable terms can be neglected. Thus. It is often more useful to employ phase-space distributions or operator distributions such as the Wigner representation (Wigner 1932) and the Glauber–Sudarshan P- representation (Glauber 1963.   ∞ ν dν |k  |v 2 h(τv ) = 2 r sin(ντv ) . the operator equations are transformed to complex Itô stochastic equations which involve only c-number commuting variables. τv ) + iΓR (ζ. τv ) ± φ(ζ. t) and A(z. f = .245) and (3. the Wigner function defined on a classical phase space is useful.3. Γ† . The Wigner function represents symmetrically ordered operators which . Γ represents the quantum noise of a field introduced by the electronic nonlinearity and ΓR is the thermal noise due to the phonon coupling. t). with Γ. k  = − 3 . but a coloured noise prop- erty. In numerical simulation.247) are stochastic functions. For this reason. τv ) ∂ζ 2 ∂τv2 τv +i h(τv − τv )φ † (ζ. n̄ = 2 . (3. the generalized P-representation is mentioned and the positive P-representation is used.231). Using this method. τv )φ(ζ. ΓR → ΓR . In relation (3. equiv- alent stochastic differential equations are obtainable. Sudarshan 1963). In practical terms. i → −i. Substituting the integrated phonon variables into the equations for the photon field gives the following equation for a new function φ(ζ. the known exact solutions (Yang 1968) of the quantum non- linear Schrödinger equation can be hardly utilized at typical photon numbers of 109 . In sufficiently intense fields. z 0 = 0 . τv ) dτv −∞  if + Γ(ζ. ΓR being independent noise terms.247).247) n̄ There is a corresponding Hermitian conjugate equation for φ † obtained by making † † the substitutions φ → φ † . ω.246) are ready for the positive P-representation. equa- tions (3. In other words. This is because there are third-order derivative terms in the Fokker–Planck equation for the Wigner function that have no stochastic equivalent. τv )φ(ζ. In the review (Drummond 1994). (3. Γ → Γ† . On the transformation (3. ΓR .2 Dispersive Nonlinear Dielectric 131 The phonon operators do not have white-noise behaviour. The Wigner func- tion does not have an exact stochastic equation.

It has been concluded that both equal-time and equal-space commutation relations are valid for the quantum soliton description. they make the new time a “position” and the position a “time” variable and then get a classical nonlinear Schrödinger equation. On the simplifying assumption that the medium is homogeneous and isotropic. † † G (2) 12 (τ ) ≡ Ê 2 (z. They begin with a single three-level cascade atom and with a χ (2) crystal. The two theories have been shown to provide similar outcomes of a homodyne measurement. Drummond and He (1997) have proposed the investigation of a quantum soliton. (3. t) Ê 1 (z. the Doppler effect. Haus and Lai 1990). They calculate cross correlation of the idler Ê 1 (z. which is utilized as an amplifier. Milonni and Maclay (2003) have shown how radiative recoil. Such an object consists of a superposition of a second-harmonic photon with a localized pair of subharmonic photons. In Matsko and Kozlov (2000) the work with physical units could be amended. and spontaneous and stimulated radiation rates are set up when the radiator is embedded in a host medium having a negative index of refraction. They compare this model with the simple one. because a cutoff is not considered or is taken to be infinity. Yariv 1989. t). . They have simplified. which occurs in parametric waveguides.132 3 Macroscopic Theories and Their Applications have a diverging vacuum noise term. Korolkova et al. Shen 1984). implicitly. t) . t + τ ) Ê 1 (z. which is described by coupled parametric amplifier equations (Boyd 2003. t) . t) and the signal Ê 2 (z. t + τ ) Ê 2 (z.251) They complete the observation of antibunching and oscillations in the reverse cor- relation with an interesting physics of the three-level atomic system.250) The neglect of the Langevin noise seems to be admissible especially for large detun- ing of the pump. Further they present the theory and the results of the three-level cascade scheme. They calculate also reverse correlation. t + τ ) Ê 1 (z. In changing to dimensionless variables. the classical propagation equation for the slowly vary- ing electric-field envelope by introducing a new time measurement in dependence on a position. Matsko and Kozlov (2000) have presented an approach which absorbs the results of two previous studies (Drummond and Carter 1987.177)). (2001) have studied a quantum soliton in a Kerr medium. (3. The system is analogous to the quark model of the meson. (3. t) Ê 2 (z. Then he has applied the results to treat basic emission and absorption processes for atoms in dispersive dielectric host media. Using this simple approach to quan- tization. t + τ ) Ê 2 (z. † † G (2) 21 (τ ) ≡ Ê 1 (z. As the χ (3) nonlinearity in silica optical fibres is low. Raymond Ooi and Scully (2007) have studied three-level extended medium. Milonni (1995) has considered the classical expression for the field energy and lifted the restriction to the magnetic susceptibility independent on frequency (cf.

The mutual coupling of these modes emerges naturally in this formalism. Here Ê in(+) (t) is a positive- frequency part of the input field and it fulfils the commutation relation [ Ê in(+) (t). after they introduced and studied the “modes of the universe” (Lang 1973. As a first step one had to relate the field operators inside and outside the cavity. the incident external field mode. Milburn and Walls (1981) have shown that the cavity of a degenerate parametric oscillator admits only a 50% amount of squeezing (in the steady state). (3. and t̃ is a respective transmission coefficient. (1990a) are expounded or formulated as exercises.3. Ujihara 1975. These authors have cleared up the relation of this subtle property of squeezed light and its generation with the concept of light propagation. It is in order to mention a book of Scully and Zubairy (1997).253) . Following Lang (1973). 1977). it had become clear that the use of squeezed states (Walls 1983. the one-sided empty cavity is described also by the relation   2l Ê cav (t) = r̃ Ê cav t − (+) (+) + t̃ Ê in(+) (t). Whereas it was obvious that the field operators inside the cavity remain the usual quantum-mechanical annihilation oper- ators of one or a small number of harmonic oscillators.3 Modes of Universe and Paraxial Quantum Propagation The laser physics and the optical engineering are typical of their calculation meth- ods and the effort for their improvement is apparent. (3. r̃ is a real amplitude √ reflection coefficient. and the output field mode. The laser cavity is coupled to the outer space and the mode coupling can be investigated in detail. The modes of universe are discussed. 3. Not only the interpretation but also the derivation of the Langevin-like “noise” terms was presented by Lang and Scully (1973). Yurke was first to realize that the pessimistic conclu- sions do not hold as the noise reduction in the transmitted field can be quite different from that in the intracavity field (Yurke 1984).1 Quasimode Description of Spectrum of Squeezing Toward the end of the 1980s. Ê in(−) (s)] = K δ(t − s)1̂. and Carmichael (1987). Gardiner and Collett (1985). and are used to define the intracavity quasimode.3 Modes of Universe and Paraxial Quantum Propagation 133 3. which include the interior of the imperfect cavity of interest. Loudon and Knight 1987) in the interferometry can lead to the enhancement of signal-to-noise ratios. the connection of the field operators outside the cavity with the “Langevin-noise operators” was established as late as 1980s by Collett and Gardiner (1984).252) c where l is the cavity length. The parax- ial description of light propagation can be quantized. Detectors of radiation with a spatial resolution motivate the inclusion of the optical imaging in quantum optics.3. t̃ = 1 − r̃ 2 . 1976. where the results of Gea-Banacloche et al.

The factor at the delta function in the equal- time commutator of the field is K c and from this we can calculate the equal-space commutator (t > 0. Ê >(−) (−0. To approve this change we denote the rightward and leftward travelling positive- frequency parts as Ê >(+) (z. s)] = [ Ê >(+) (−ct. t) + E<(+) (z.259) can be rewritten in the form. t). t) mean the slowly varying smooth ampli- tude with the property Ω E>(+) (z. Ê cav (+) (t) ≡ Ê >(+) (+0. Barnett and Radmore 1988).258) We propose to consider the definition  t c â(t) = E (+) (t  ) dt  (3. definition (3. (3.260) = K cδ(−ct + cs)1̂ = K δ(t − s)1̂. (3. (3. t) = Ê >(+) (z. t) = Ê <(+) (z.263) c E<(+) (z. For the full Fox–Li quasimode (Fox and Li 1961.261) Hence. (3. (3.259) K 2l t− 2lc cav instead of (3.256) (+) Ê out (t) = (+) Eout (t)e−iΩt . output.134 3 Macroscopic Theories and Their Applications where Ê in(−) (t) = [ Ê in(+) (t)]† and Ω K = . a single-mode annihilation operator â(t) is defined  2l (+) â(t) = Ê (t)eiΩt . (3. Ê out (+) (t) ≡ Ê <(+) (−0.254) 20 cA with Ω being a quasimode frequency. (3.255) K c cav It is convenient to use the slowly varying amplitudes Ein(+) (t). t) and E<(+) (z.262) K c2l 0 where (cf.258)) E>(+) (z. k0 = . t)e−ik0 z+iΩt . t)eik0 z+iΩt . respectively. t) and Ê <(+) (z. t) so that Ê in(+) (t) ≡ Ê >(+) (−0. which is in a better agreement with the quantum field theory. Ê in(+) (t) = Ein(+) (t)e−iΩt . (+) (3. 0)] (3. Ê >(−) (−cs. Eout (+) (t). t).255). s > 0 without loss of generality) [ Ê in(+) (t). (3.  l 1  (+)  â(t) = E> (z. Ê in(−) (s)] = [ Ê >(+) (−0.264) . t). and Ecav (+) (t) for the input. 0). t).257) (+) Ê cav (t) = Ecav (t)e−iΩt . t) dz. (3. and cavity fields related to the cavity frequency Ω.

we see that the annihilation operator â(t) corre- sponds to the same quasimode in all times. (3.266) K 2l t− 2lc Recalling the space integration. (3.270) K in where we replaced the average over the short-time interval by the value of the function (at the upper limit).269) 2l 2 Noting that   t c 1 c b̂(t) = Ein(+) (t  ) dt  2l K 2l t− 2lc  1 (+) = E (t).3 Modes of Universe and Paraxial Quantum Propagation 135 Performing the time integration on both sides of relation (3.265) c where  t c b̂(t) = Ein(+) (t  ) dt  . . (1990a) define.268) as a quantum Langevin equation  d √ 1 (+) â(t) = −Γâ(t) + 2Γ E (t). (3. (3. we obtain that   2l â(t) = r̃ â t − + t̃ b̂(t). Gea-Banacloche et al. for arbitrary measurement times. whereas b̂(t) is appropriate to many distinct modes. (3.3. (3.268) dt 2l where c 1 2 Γ= t̃ . (3. In the situation when it holds that     2l 1 2 2l d r̃ â t − ≈ 1 − t̃ â(t) − â(t). we rewrite equation (3.252).271) dt K in Further. the spectrum of squeezing of the output field via the quadrature variances. They present a microscopic effective Hamiltonian model of balanced homodyne detection.267) c 2 c dt in the short cavity round-trip time limit we get  d √ c â(t) = −Γâ(t) + 2Γ b̂(t).

136 3 Macroscopic Theories and Their Applications They refer to the fundamental papers (Collett and Gardiner 1984. Caves and Schumaker 1985. Yurke 1985). with a balanced homodyne detector one measures the combinations 1 iθ (+) . the operator is introduced N ˆ (δω) = √ T (+) Ã out Eout (t)eiδωt dt.273) K As shown also by Yurke (1985) and Carmichael (1987). Gardiner and Col- lett 1985.272) T 0 where  1 N = . As an approximation. (3. (3. where this concept of spectral squeezing was originally treated.

274) 2 and from this the natural generalization of the single-mode quadrature concept is . Eoutθ (t) = e Eout (t) + e−iθ Eout (−) (t) . (3.

275) outθ out out 2 We may wonder why a non-Hermitian operator is taken for such a generalization of the Hermitian operator. Almost the exact reverse holds. the connection between single quasimode squeezing and spectral squeezing is explored and the difference in the noise reduction inside and outside the cavity is clarified in a way that lends itself to a simple visualization. In this theory. In particular. Their one-dimensional theory solves the problem of how to describe the quantized radia- tion field in a leaky cavity using Fox–Li modes. ˆ (δω) + e−iθ Ã (3. Finally. increased noise in the conjugate quadrature. if one-photon resonance and initial atomic coherences involving the middle level are present. Dutra and Nienhuis (2000) have unified the concept of normal modes used in quantum optics and that of Fox–Li modes from semiclassical laser physics. The classical Langevin formalism is further replaced by the alternative Fokker–Planck formalism for the calculation of the spectrum of squeezing. These analyses are based on the mean values and the normally ordered variances of quantum operators for which classical Langevin equations may be written down. however. It has been shown that without one-photon resonance and initial atomic coherences involving the middle level. in fact. the effect of finite measurement time on the quadrature variances is briefly analysed. Finally. unlike conventional . This general Fokker–Planck formalism was applied to the two-photon correlated-spontaneous-emission laser. ˆ (δω) = 1 eiθ Ã Ã ˆ † (−δω) . the intracavity field may be perfectly squeezed while the outside field is not only unsqueezed but has. the maximum squeezing of the ultracavity mode is 50% while the detected field can be almost per- fectly squeezed. Gea-Banacloche et al. (1990b) have first analysed measurements of small phase or frequency changes for an ordinary laser and calculated the extra cavity phase noise for a phase-locked laser.

which helps determine the Hamiltonian. Brown and Dalton (2002) have considered three-dimensional unstable optical systems. one is usually interested in the spatial dependence of temporally steady-state fields. For the single wave interacting nonlinearly with matter. t). Aiello (2000) has derived simple relations for an electromagnetic field inside and outside an optical cavity. cf. limiting himself to one.and two-photon states of the field. as a consequence of natural cavity modes having been used. Abram and Cohen (1991). where the temporal evolution by the Hamiltonian is supplemented by the spatial progression with the momentum operator. An alternative proposal is made that the quantum-mechanical equivalent of the classical steady-state condition is the description of the system by a stationary state of a suitable Hamiltonian. t)] = δ(z − z  )1̂. In this theory a non-Hermitian envelope-field operator Ψ̂(z. (3. A number of concepts and properties resulting from the standard canonical quantization procedure have been suited to the non-Hermitian modes by the exact transformation method. He has expressed input– output relations using a nonunitary transformation between intracavity and output operators. They have defined non-Hermitian modes and their adjoints in both the cavity and external regions. t) exp[i(kz − ωt)].2 Steady-State Propagation In Deutsch and Garrison (1991a) it is assumed that in the case of amplifier. Ψ̂† (z  .3. In the application to the optical field. (3. 3. The results are applied to the sponta- neous decay of a two-level atom inside an unstable cavity.3 Modes of Universe and Paraxial Quantum Propagation 137 models. It is no attempt at a reformulation of one-dimensional propagation. with the property [Ψ̂(z.277) An 2 (ω) where A is the beam area and n(ω) is a dispersive index of refraction. t). There is a formal resemblance to a nonrelativistic many-body theory for a complex scalar field (Deutsch and Garrison 1991b). i. system and reservoir operators no longer commute with each other. In contrast to Deutsch and Garrison (1991a). the total Hamiltonian can be written as Ĥ = Ĥenv + Ĥint .3. t) = e Ψ̂(z. the vector potential (or electric field in the lowest order) corresponding to a carrier plane wave of a given polariza- tion e is expressed as follows:  2πω Ê ω(+) (z. (3. we will not consider a carrier wave Hamiltonian. we will make a simplification.278) .276) is introduced.e.

(3.283) ∂t  The introducing of the carrier wave Hamiltonian has revoked the considering of the Schrödinger picture (Deutsch and Garrison 1991b). along with the envelope pic- ture which we have confined ourselves to. t) dz. (3. . Relation (3. Ĥ ].0 . the generality will not be exer- cised and we will treat only the vacuum input and the case of degenerate parametric amplifier. (3. t) i = − [Ψ̂(z.279) ∂z ∂z where Ĥenv is the Hamiltonian governing the free progression of the envelope and Ĥint is a general interaction Hamiltonian. t) =− . the state |Φ evolves by the Schrödinger equation ∂|Φ i = − ( Ĥenv + Ĥint )|Φ .138 3 Macroscopic Theories and Their Applications ic Ĥenv = − 2n(ω)   ∂ Ψ̂(z. the label (ss) will be omitted. In fact.283) will be independent of time. t) c ∂ Ψ̂(z. (3. For application under consideration there will be exact frequency matching between the carrier frequencies of the various waves which interact so that the Hamiltonian in equation (3.283):   Ĥenv + Ĥint |Φ ss = λ|Φ ss . In the standard Heisenberg picture.284) For the stationary solutions. (3. t) ∂ Ψ̂† (z. thus the steady state (ss) solutions are identified with the stationary solutions to equation (3.281) ∂t n(ω) ∂z The solution is   ct Ψ̂(z.277) is the positive-frequency component of the electric-field in the envelope picture similarly as relation (4. (3. t). t) − Ψ̂(z.5b) in Caves and Schumaker (1985) is this component in the interaction picture.280) ∂t  or for a linear medium ∂ Ψ̂(z. The enve- lope picture is essentially the modulation picture in Caves and Schumaker (1985). t) × Ψ̂† (z. t) = Ψ̂ z − .282) n(ω) In the standard Schrödinger picture. the equation of motion for the envelope-field operator reads ∂ Ψ̂(z.

we can consider a coherent state corresponding to a constant one-photon wave function.286) can be rewritten in the normal ordering form   1 D̂[α] = exp − |α(z)| dz 2 2     × exp α(z)Ψ̂† (z) dz exp − α ∗ (z)Ψ̂(z) dz (3.3 Modes of Universe and Paraxial Quantum Propagation 139 (i) In the case Ĥint = 0̂.293) .288)  α 1 â ≡ â ≡ α ∗ (z)Ψ̂(z) dz.290) Since [â.1. (3.291) relation (3. we may consider either the limits −∞. the stationary solutions are the translation- invariant states.292) and the corresponding one-photon state is . ∞ in the integral on the left-hand side in (3. (3. As an example. (3.3. α ≡ 1 α(z)Ψ̂† (z)|0 dz. To have a unitary representation of the translation.  * . . â † ] = 1̂.279) or the spatial periodicity.285) i. (3. ρ ρ On substituting |Φ = |{α} into relation (3. We define a functional displacement operator   D̂[α] ≡ exp [α(z)Ψ̂† (z) − α ∗ (z)Ψ̂(z)] dz (3. (3.286) = exp(ρ â † − ρ â).e. We prefer the latter possibility. we derive that α(z) and λ should make the vacuum . (3. (3.287) where  ρ= |α(z)|2 dz.289) ρ ρ The coherent state is defined as the displaced vacuum |{α} ≡ D̂[α]|0 . in the case of vacuum propagation.284) and applying then the operator D̂ † [α] from left to both sides.

296) dz and hence λ = 0.302) or to the relation with the insertion.301) Condition (3.295) n(ω) dz with λ|0 . On substituting again Φ into the new relation.297) Instead of a translation-invariant wave function.301) is equivalent to (  − λ N̂ )|Φ = 0 (3. (3.298) on the left-hand side of (3.294) ∂z The eigenvalue is λ again. The idea takes into account that the wave function of any number of particles should be the eigenfunction of the operator  exp(i Â)|Φ = exp(iλ N̂ )|Φ (3. we may try one that is an eigen- function of the translation operator. and applying the operator D̂ † [α]. (3. When the boson number operator commutes with the operator  = Ĥenv + Ĥint .140 3 Macroscopic Theories and Their Applications state the eigenstate of the operator    i c  †  ∂ Ψ̂(z) + α(z)1̂ D̂ † [α] Ĥenv D̂[α] = − Ψ̂ (z) + α ∗ (z)1̂ 2 n(ω) ∂z  † ∗   ∂ Ψ̂ (z) + α (z)1̂   − Ψ̂(z) + α(z)1̂ dz.299) behind λ on the right-hand side. we obtain the right-hand side in the form . (3. (3. On equating † c dα(z) D̂ [α] Ĥenv D̂[α]|0 = −i Ψ̂† (z) |0 dz n(ω) dz c dα(z) − i α ∗ (z) dz|0 (3. this problem can be generalized by the insertion of the number operator N̂ . we see that dα(z) =0 (3. (3. N̂ = Ψ̂† (z)Ψ̂(z) dz.284).300) or exp(i(  − λ N̂ ))|Φ = |Φ .

3.307) 2 n(ω) with 1 κ(z) = g0 (z) exp[iφ(z)]. χ (2) (z) is the second-order susceptibility coupling the pump to the degenerate signal and idler fields and Ep (z) is the pump amplitude.306) where ωp is the pump frequency. (3. and β(z) is the remaining phase originating from the product χ (2) (z)Ep∗ (z). (3.309) n(ω)c π φ(z) = − + Δk z + β(z). Expression (3. (3. Substituting for Ê ω(+) (z.310) 2 Here g0 (z) is the standard power gain coupling constant (Yariv 1985). (3. (3. (3.308) 2 4π ω (2) g0 (z) = |χ (z)||Ep (z)|.304) n(ω) dz The solution to equation (3.c.3 Modes of Universe and Paraxial Quantum Propagation 141    λ Ψ̂† (z) + α ∗ (z)1̂ Ψ̂(z) + α(z)1̂ dz. ωp = 2ω. dx dy dz.305) c where λ is any real number.296) is thus generalized. Δk = 2k − kp is the phase mismatch at the degenerate frequency. t) from relation (3. t)]2 2  + H.305) for the complex amplitude is suffi- cient for the fulfilment of the relation Â|Φ = λ N̂ |Φ . c dα(z) − i = λα(z).303) Condition (3. . (3.304) reads  iλn(ω) α(z) = α(0) exp z .277) gives the interaction Hamiltonian in the envelope picture i c Ĥint = [κ ∗ (z)Ψ̂2 (z) − κ(z)Ψ̂†2 (z)] dz. the interaction Hamiltonian can be written (Hillery and Mlodinow 1984) as follows:  1 Ĥint =− χ (2) (z)Ep∗ (z) exp[−i(kp z − ωp t)][ Ê ω(+) (z. (ii) In the case of the degenerate parametric amplifier. (3.

Ψ̂(z) |0 {ξ } = 0 = C̃ˆ |0 . (3.315) where the commutator C̃ˆ is   ˆ ic d  C̃ = − exp[iθ(z)] exp[−iθ (z)] cosh[r (z)] exp[iθ (z)] sinh[r (z)] n(ω) dz d  − cosh[r (z)] sinh[r (z)] dz  + κ ∗ (z) exp[iθ (z)] sinh2 [r (z)] − κ(z) exp[−iθ (z)] cosh2 [r (z)] Ψ̂† (z)  d  + cosh[r (z)] cosh[r (z)] dz d  − exp[iθ (z)] sinh[r (z)] exp[−iθ (z)] sinh[r (z)] dz + κ ∗ (z) exp[iθ (z)] sinh[r (z)] cosh[r (z)]   ∂ − κ(z) exp[−iθ (z)] sinh[r (z)] cosh[r (z)] Ψ̂(z) + Ψ̂(z) . They define a functional squeezing operator   1 Ŝ[ξ ] ≡ exp [ξ (z)Ψ̂†2 (z) − ξ ∗ (z)Ψ̂2 (z)] dz .312) Similarly as in case (i).314) we rewrite the eigenvalue problem in the λ-independent form   Ŝ † [ξ ]( Ĥenv + Ĥint ) Ŝ[ξ ]. (3.313) Applying the operator Ψ̂(z) to both the sides of (3. (3. (3.313) and taking into account that λΨ̂(z)|0 = 0 = Ŝ † [ξ ]( Ĥenv + Ĥint ) Ŝ[ξ ]Ψ̂(z)|0 . (3. ξ (z) and λ should be solutions of the equation Ŝ † [ξ ]( Ĥenv + Ĥint ) Ŝ[ξ ]|0 = λ|0 . (3.311) 2 with z-dependent squeezing parameter ξ (z) = −r (z) exp[iθ (z)].316) ∂z . The squeezed vac- uum is defined as |0 {ξ } ≡ Ŝ[ξ ]|0 .142 3 Macroscopic Theories and Their Applications To solve the time-independent Schrödinger equation. Deutsch and Garrison (1991a) assume that the eigenstate is a squeezed vacuum state corresponding to a two-photon wave function.

318) dz On introducing the complex amplitude ζ (z) = − exp[iθ (z)] tanh[r (z)]. (3.321) When β(z) = 0 and the phase difference θ (z) − φ(0) is small. when moreover g0 (∞) > |Δk|.320) dz which may be useful for guessing the boundary condition r (z) |z=0 = 0. the squeezing param- eter r (z) integrates values of experimental parameter 12 g0 (z  ).319) we can write equations (3.3 Modes of Universe and Paraxial Quantum Propagation 143 The eigenvalue condition requires that the real and imaginary parts of the coefficient at Ψ̂† vanish yielding the desired propagation equations dr 1 = g0 cos(θ − φ). the squeezing  parameter θ. z  ∈ [0.3. (3. θ(z) |z=0 = φ(0). (3.317) dz 2 dθ = −g0 coth(2r ) sin(θ − φ).317) and (3. (3. (3.318) in the compact form d(−ζ ) = κ − κ ∗ζ 2. z].

The presence of the singular operator Ψ̂(z)Ψ̂† (z) indicates that λ has no finite value in general. The problem of quantum propagation of paraxial fields was considered first by Graham and Haken (1968). The revived interest is indicated by Kennedy and Wright (1988). Such an envelope implies that the wave is paraxial and monochromatic.317) and (3. Deutsch and Garrison (1991b) begin with generalizing the results of Lax et al.3 Approximation of Slowly Varying Envelope The macroscopic approach to the quantum propagation aims at a quantum version of the slowly varying envelope approximation. which develop the classical theory of a strictly monochromatic wave in an inhomogeneous nonlinear (perhaps amplifying) medium. we find that λ = O Δz 1 . The generalization is made only to a quasimonochromatic wave and the . 1 c sinh3 r λ− sin(θ − φ) dz. (z) converges to the function of the experimental parameters φ(z) − arcsin g0Δk (∞) .318) again.3. Resorting to the partition of the field into finite elements oflength  Δz in each of which we can define local field operators. A direct solution of (3.313) requires that the real and imaginary parts of the coefficient at Ψ̂† (z) vanish yielding the propagation equations (3. (1974). (3.322) Δz n(ω) cosh r 3.

Ψ(x. where qT is the transverse part of q. by letting the wave function { f λ (q)} depend on a small positive parameter θ. qz ). (3.2 Here eλ (k) are the orthogonal polarization unit vectors and the reduced Planck constant is introduced in view of the possible later quantization. In contrast to Deutsch and Garrison (1991b). We define f λ (q) by the relation f λ (q) = Fλ (q + k0 ). t) exp[i(k0 z − ω0 t)]. t = 0) = 3 eλ (k)Fλ (k)eik. and A0 is a normalization constant.x d3 k. qz .325) (2π) 2 2|k| λ=1.θ . the positive-frequency component of the vector potential satisfies the free-field wave equation 1 ∂ 2 (+) ∇ 2 A(+) (x.327) θ k0 θ k0 where 1 V= . (3. A(+) (x.323) c2 ∂t 2 The approximation of the slowly varying envelope is introduced by expressing A(+) (x. henceforth referred to as an envelope field. In the Coulomb gauge. Fλ (k) are thus momentum-space wave functions. In other words. t) as envelope modulating a carrier plane wave propagating in the z-direction with the wave number k0 and the frequency ω0 = ck0 .333). θ ). a paraxial wave function {Fλ (k)} is concentrated in a small neighbourhood k0 = k0 e3 of the wave vector of the carrier wave. θ ) = V fλ . q = qT + qz e3 .144 3 Macroscopic Theories and Their Applications quantum theory is presented in the simplest system of codirectional propagation considering only the free-field dynamics. (3. (3. t) = A0 Ψ(x. 2 . t) is a vector-valued function. (3. Let us assume that   √ qT qz f λ (qT . The intuitive notion of a paraxial field is that it is composed of rays making small angles with the main propagation axis. The initial positive-frequency component can be expressed as follows:  1 c  A(+) (x.326) where q is the relative wave vector. which we will specify before relation (3. we stress that we express the con- centration in a small neighbourhood of q0 = 0. (3. Let us observe that q = (qT .328) θ 4 k03 .324) Here. t) − A (x. f λ (q) ≡ f λ (q. t) = 0.

336) ∂t . ∂ (+) 1 i A (x. θ ) = θ n f λ (η).329) θ k0 qz ηz = 2 . the integro-differential form of the wave equa- tion for A(+) (x. t = 0. θ ). θ ).325).x d3 k. t. where Fλ (k) ≡ Fλ (k. (3. we note that (0) f λ (η. t.3.329) and (3.333) |q + k0 | λ=1. θ ) = (cΩ − ω0 )Ψ(x. Substituting from (3. where qT ηT = . θ ) is investigated. In Deutsch and Garrison (1991b). Let us note that ∇ = (∇T . θ ) = c(−∇ 2 ) 2 A(+) (x. we find that 1 Ψ(x. ∞  (n) f λ (η.335) (2π) 2 ∂ with F̃(k) being the Fourier transform. t.324) at t = 0. t.334) gives ∂ i Ψ(x. (3.324) into (3. t. Substituting integral (3. (3.3 Modes of Universe and Paraxial Quantum Propagation 145 and we have introduced the notational convention that an overbar indicates a dimen- sionless function of the scaled variables (and perhaps θ).332) lead to the wave function being θ -dependent.330) θ k0 The functions of interest are those that have a convergent power-series expan- sion in θ .331) n=0 In contrast to Deutsch and Garrison (1991b). (3. the parameter θ has been introduced. (3. and choosing c A0 = 2k0 . (3.325) for the envelope field defined by (3.2 Here. with the / momentum-space wave function given by equation (3. θ )eiq. a differ- ence from Deutsch and Garrison (1991b). t) ≡ A(+) (x. This relates to defining a dimensionless “momentum” vector η = (ηT . (3. A(+) (x. θ ) = f λ (η). θ ). t.332) Relations (3. (3.334) ∂t 1 where (−∇ 2 ) 2 is an integral operator defined by 1 1 (−∇ 2 ) 2 F(x) = 3 |k| F̃(k)eik. ηz ). θ ).326). which is not present in integral (3. ∂z ).x d3 q. θ ) = 3 (2π) 2   k0 × eλ (q + k0 ) f λ (q.

θ 2 ω0 t. θ ) = H(θ )Ψ(ξ . ζ = θ 2 k0 z.341) ∂τ where 1 H(θ ) = [ Ω(θ ) − 1 ]. θ 2 k0 ∂ζ∂ Ω(θ ) = .346) ∂τ m=0 .343) k0 This provides the possibility of expanding the differential operator H(θ ). ∞  (n) H(θ ) = θnH . τ ). (3. = k02 − 2ik0 − ∇T2 − 2 . τ.342) θ 2  Ω θ k0 ∇ T . z. . θ ). θ ) has the expansion ∞  (n) Ψ(ξ . t. τ. τ ). θ ).337) ∂z ∂z ∂z The scaled configuration-space variables ξ = (ξ T . θ ) ≡ Ψ(xT . (3. t. 1. τ. . (3. . ζ ) are ξ T = θ k0 xT . (3.146 3 Macroscopic Theories and Their Applications where    1 ∂ ∂ ∂2 2 Ω ≡ Ω ∇T . n = 0. θ ) = √ Ψ(θ k0 xT .339) After expressing the envelope field in the form 1 Ψ(x. 2. θ ) = θ n Ψ (ξ . The dimensionless amplitude Ψ(ξ . (3. θ 2 k0 z. τ ) = H Ψ (ξ . (3.344) n=0 (n) where the differential operators H are just defined by the formal expression. (3.345) n=0 It is evident that the terms satisfy the equations ∂ (n) n (n−m) (m) i Ψ (ξ . τ.338) and the dimensionless time variable τ = θ 2 ω0 t. (3.336) as follows: ∂ i Ψ(ξ . .340) V we can rewrite relation (3. (3. (3.

347) (2π) 2 λ=1. θ ) = θ n Kλ (η).348) |q + k0 | where ẽλ (q) = eλ (q + k0 ). τ = 0. θ ) where / w(η.333) as  1 Ψ(x. θ ) = 1 + θ 2 (2ηz + ηT2 ) + θ 4 ηz2 . We rewrite equa- tion (3. θ ) Kλ (η. (3. θ )eiq·x d3 q.352) Considering the expansion ∞  (n) Kλ (η. (3. θ ) f λ (η)eiη·ξ d3 η. θ ) = √ .354) (2π) 2 λ=1.349) Reexpressing (3. (n) (n) Ψ (ξ ) = 3 (3.347) in terms of the scaled variables gives  1 Ψ(ξ .2 Deutsch and Garrison (1991b) claim that the preceding arguments can be used to identify the subspace of the photon Fock space consisting of the paraxial states .351) w(η. (3.2 where the function Kλ (q) is defined by  k0 Kλ (q) = ẽλ (q).3 Modes of Universe and Paraxial Quantum Propagation 147 In Deutsch and Garrison (1991b).350) (2π) 2 λ=1. θ ) = 3 Kλ (q) f λ (q. the discussion of the classical equation of motion is completed by considering the initial-value problem.353) n=0 we obtain the initial expansion of the envelope fields  1 Kλ (η) f λ (η)eiη·ξ d3 η.3. (3. (3. (3. θ ) = 3 Kλ (η. (3.2 The scaled kernel function is ẽλ (η.

when |η| → ∞. G)1̂ = Fλ∗ (k)Gλ (k) d3 k 1̂.360) Let us observe that  † † â †2 [F ] = Fλ1 λ2 (k1 .361) λ1 λ2 where Fλ1 λ2 (k1 . k2 )âλ1 (k1 )âλ2 (k2 ) d3 k1 d3 k2 . â † [F ] = 1̂. k2 ) = Fλ1 (k1 )Fλ2 (k2 ) (3.148 3 Macroscopic Theories and Their Applications of the field.355) λ=1. (3. âλ (k).356) λ=1.363) 2 λ1 λ2 . † Let us recall the standard plane wave creation and annihilation operators âλ (k).2 If F is normalized.358) where f is a normalized one-photon wave function.362) exemplifies a normalized symmetric function. then   â[F ]. Before we generalize the definition of a state with exactly one photon present. and the wave packet creation and annihilation operators  † † â [F ] = Fλ (k)âλ (k) d3 k (3. faster than any power of |η|−1 . 2 = √ Fλ1 λ2 (k1 . We now define  † † ĉ [ f ] = ĉλ (q) f λ (q) d3 q.357) are the creation operators corresponding to the envelope field. we return to the original oper- ators.2 and its conjugate. They resorted to the space S. | f . â [G] = (F . We are led to the definition of the state with exactly two photons present  1 † † |F . the infinitely differentiable functions that decrease. (3. k2 )âλ1 (k1 )âλ2 (k2 )|0 d3 k1 d3 k2 . Let us proceed with the generalized commutation relations   †  â[F ]. (3.2 where † † ĉλ (q) = âλ (q + k0 ) (3. (3.359) λ=1. 1 = ĉ† [ f ]|0 . (3. (3.

. . . (3. . . . . . ..364) where F is this time a normalized symmetric wave function of m photons and |F. d3 km . m = √ ..3. 0 = F|0 . km ) m! λ1 λm † † × âλ1 (k1 ) . ĉλm (qm )|0 d3 q1 .. m . (3. . This almost completes the definition of the Fock space. .   1 |F.   1 | f .368) where f λ1 . . m ≥ 1.. .366) m=0 where F (m) are (unnormalized) symmetric m-photon wave functions.λm (q1 + k0 . . since any element of this space has the form ∞  |Φ = |F (m) ... . .λm (q1 . m . f λ1 .λm (q1 .367) m=0 Similarly... . (3. d3 qm . and any element of the Fock space has the form ∞  |Φ = | f (m) .. qm ) = Fλ1 . Fλ1 . . . . (3. . The pure state |Φ is normalized if and only if the func- tions F (m) are jointly normalized by ∞  (F (m) . qm ) m! λ1 λm † † × ĉλ1 (q1 ) .. qm + k0 ) (3. . and F (0) is a complex number..370) where f = F. .λm (k1 . âλm (km )|0 d3 k1 .365) with F a complex unit. In general.. (3. (3. . F (m) ) = 1.. m ≥ 1.. 0 = f |0 . . . (3...369) and | f .3 Modes of Universe and Paraxial Quantum Propagation 149 where F is a normalized symmetric two-photon wave function. m = √ . m ≥ 1.371) m=0 .

qm + k0 ) (3. . (3.377) for all times.380) m=0 we may expand equation (3..372) and f (0) = F (0) . 2 . a unitary operator T̂ (θ ) is of interest such that T̂ (θ )| f  . . qz1 . (3..381) m=0 . (3. . θ ) = θ m |Φ(m) (t) .374) In the description of the dynamics using the Schrödinger picture.λm (qT1 . qm ) = Fλ(m) 1 . qTm . (3. . (3.373) where (cf.λm (q1 .λm (q1 + k0 . . . θ ) . qzm . θ ) . m = | f (θ ).327)) m f λ1 .375) in the form ∂  i |Φ (t.. θ ) . . m .. . θ ) = − Ĥ  (θ )|Φ (t.375) ∂t  where |Φ(t.. . (3.. (3.... (3.λm T1 .. . θ ) = V 2 q qz1 qTm qzm  × f λ1 .. θ ) |t=0 = T̂ (θ )|Φ (t = 0) . θ θ 2 θ θ (3.. . we can rewrite (3. . .378) into the coupled equations for the coefficients of the series ∞  |Φ (t. .379) Using the expansion ∞  Ĥ  (θ ) = θ m Ĥ (m) . ∂ i |Φ(t.150 3 Macroscopic Theories and Their Applications where f λ(m) 1 .378) ∂t  where Ĥ  (θ ) = T̂ † (θ ) Ĥ T̂ (θ ). (3. θ ) = T̂ † (θ )|Φ(t. θ ) .376) Defining the state |Φ (t. θ ) = − Ĥ |Φ(t. For the subsequent analysis. the paraxial approximation means mainly the evolution of the initial state |Φ(t.

t = 0.380) to an arbitrary operator M̂(t).383) m=0 We can then rewrite the equation of motion ∂ i M̂(t) = − [ M̂(t). M̂  (t. Ĥ  (t.379) and (3.385) ∂t  We may expand this equation into coupled equations similarly as (3. (3. we introduce the operator Ψ̂(x) by relation (3. Since T̂ (1) = 1̂.333). 1) = M̂(t). definition (3.3. (3. θ ) = − [ M̂  (t. or any operator M̂(t) can be expressed as ∞  M̂(t) = M̂ (m) (t). with Ψ(x.384) ∂t  in the form ∂  i M̂ (t. θ ). θ ) = θ m M̂ (m) (t). Ĥ (t)] (3. θ )]. M̂  (t.383) simplifies for θ = 1. (3. According to Deutsch and Garrison (1991b).3 Modes of Universe and Paraxial Quantum Propagation 151 Describing the dynamics in the Heisenberg picture. (3. (3. θ ) = T̂ † (θ ) M̂(t)T̂ (θ ).378).386) relation (3.382) ∞  M̂ (t.387) m=0 In both the pictures. we should generalize relations (3. θ ) .339) can be used whenever it is advantageous.

θ ) .→ Ψ̂(x) on the left-hand side and f λ (q.

Ψ̂ j (x )] = 3 K̂ λi(n) (q) K̂ λ(m) j (q)e d q. This operator can be expanded as ∞  Ψ̂(x) = Ψ̂(n) (x).389) (2π) 2 λ=1. we can compute the commutators between fields of different orders that are  (n) (m)†  1 iq·(x−x ) 3 [Ψ̂i (x).2 Using this expansion.390) (2π) 2 λ=1.→ ĉλ (q) on the right-hand side. (3.2 .388) n=0 where  1 Ψ̂ (x) = (n) 3 K̂(n) λ (q)ĉλ (q)e iq·x 3 d q. (3. (3.

∂ † [Ψ̂i (x. This formalism is applied to show that Mandel’s local-photon-number operator and Glauber’s photon-counting operator reduce. 3. (3. (3. we first expound some general concepts as spatially multimode squeezing and spatial entanglement.393) 20 c . î(ρ. t). t) and â † (z. In addition. t). (m)† [Ψ̂i(n−m) (x). to the same true number operator.392) ∂t In the zeroth order.3. it is shown that the O(θ 2 )-difference between them vanishes for experiments described by stationary coherent states. ρ. and formally resembles a nonrelativistic many-particle theory. Ψ̂ j (x . the nth-order commutator can be expressed as †  n [Ψ̂i (x). In this section. A nonperturbative quantization of a paraxial electromagnetic field has been achieved by forcing the plane waves involved in the expression for the vector- potential operator to obey paraxial wave equations at the time origin (Aiello and Woerdman 2005). ρ. t) as  ω0 Ê (+) (z. ρ. The observed quantity is the surface photocurrent density operator. t)](n) = 0̂. the theory yields a quantized analogue of the classical parax- ial wave equation. and describe some optical devices that are able to generate light beams with these properties (Kolobov 1999). Kolobov (1999) enriches exposition of the usual quantum optics by new facts. ρ. Ψ̂ j (3.4 Optical Imaging with Nonclassical Light In optical imaging with nonclassical light or quantum imaging it is important to know how quantum entanglement properties of light beams in the spatial domain can be exploited in order to improve the quality of processing of images and of parallel signals (Gatti 2003). The time moments are completed with space points ρ. t). ρ.391) m=0 The equal-time commutation relations are preserved by the dynamics in each order of the approximation scheme. This operator can be written in terms of space. t) = i exp[i(k0 z − ω0 t)]â(z. in the zeroth order. Let Ê (+) (z. where ρ is the position vector in the transverse plane of the wave. an Hermitian operator.and time-dependent photon annihilation and creation operators â(z. It is connected with exis- tence of very small photodetectors or pixels.152 3 Macroscopic Theories and Their Applications As expected. Then we provide some references to interesting approaches in this field. t) mean the positive-frequency operator of the electric field of a quasiplane and a quasimonochromatic wave travelling in the +z- direction. Let the photodetection plane be located at the point with longitudinal coordinate z normal to the z-axis. Ψ̂ j (x )](n) = (x )].

t) 2 If the correlation function g (2) (ξ . ρ. The quantum theory of photodetection provides the following expressions for the mean value of the photocurrent density operator î(ρ. 0) ≥ g (2) (ξ . â(z. t  ) = G (2) (ξ . But â(z. ρ.397) This correlation function is proportional to the probability of detecting a photon at time t  and at the spatial point ρ  under the condition that the previous detection happened at time t and point ρ. ρ. t + τ ) : g (2) (ξ . ρ  . one may speak of the antibunching in space–time. τ ). t)â(z. t  ) = : Î (ρ.395) ) * 1 {δ î(ρ. t. t). One can define the degree of second-order spatio-temporal coherence as : Î (ρ. Analo- (2) gously. (3. t) î(ρ  . τ ). ρ  . t) and â † (z. δ î(ρ  . When the intensity of light is stationary in time and uniform in the transverse area of the light beam. t  )] = 0̂ (3. t. if g (2) (0.399) . τ ). t) are not the standard-modal annihilation and creation operators.398) Î (ρ. (3. ρ. t)â(z. t) Î (ρ + ξ . (3. ρ  . t) determines the mean photon-flux density in photons per cm2 per second at point ρ and time t. this correlation function depends only on the time difference τ = t  − t and the spatial difference ξ = ρ  − ρ between two points. ρ.394) and are normalized so that the mean value â † (z. The antibunching in space–time is a purely quantum-mechanical phenomenon. t  )}+ = î(ρ. (3. and its space–time correlation function 12 {δ î(ρ. t  ) : − î(ρ. Indeed. t  )}+ : î(ρ. τ ) of a classical electromagnetic field stationary in time and uniform in space must satisfy g (2) (0.396) Here Î (ρ. τ ) has its maximum at ξ = 0 and at τ = 0. â † (z.3 Modes of Universe and Paraxial Quantum Propagation 153 Here ω0 is the carrier frequency of the wave and k0 is its wave number. t  )] = δ(ρ − ρ  )δ(t − t  )1̂. t  ) . t) Î (ρ  . t) . t) Î (ρ  . They obey the commutation relations [â(z. t). τ ) (3. δ î(ρ  . 0) < g (2) (ξ . t) . it follows from the Schwarz inequality that the correlation function g (2) (ξ .3. t) δ(ρ − ρ  )δ(t − t  ) 2 + η2 : Î (ρ. it is natural to speak of the bunching in space–time. The second contribution to the correlation function of the photocurrent density operator is proportional to the normal. ρ. t) is the photon-flux density operator. t). [â(z. ρ. ρ  . t). ρ. t  ) : . 0) > g (2) (ξ . τ ) = . g (0. t) = â † (z. G (2) (ρ.and time-ordered space–time intensity cor- relation function G (2) (ρ. t) = η Î (ρ.

{·. the correlation function of the photocurrent density operator 12 {δ î(ρ. (3.404) 2 where î(t) = â † (t)â(t). it cannot be explained within the framework of semiclassical theory. (3. Ω) = î(ρ. δ î(t)}+ exp(iΩt) dt = î(t) . One can show that in semiclassical theory the sum of the second and third contributions is always nonnegative. The photocurrent noise spectrum is defined as a Fourier transform of the pho- tocurrent correlation function. As follows from relation (3. t) + G̃ (2) (q. Ω) as follows A (δ î)2 (q. for a light field stationary in time and uniform in the transverse plane. (3. δ î(ρ  . G̃ (2) (q. Ω) − î(ρ. t)}+ 2 × exp[i(Ωt − q · ρ)] dt d2 ρ.403) This formula is a generalization of the standard quantum limit for a single-mode field described by the photon annihilation and creation operators â(t) and â † (t). t) 2 δ(Ω)δ(q). Ω) = î(ρ. Therefore the semiclassical minimum value of the photocurrent density noise is given by the shot noise in space–time. ) * A 1 (δ î)2 (Ω) = {δ î(0). Since the antibunching in space–time means an exactly oppo- site inequality. t) dt d2 ρ. the second one is the spatio-temporal Fourier transform of the intensity correlation func- tion. A (δ î)2 (q. The noise spectrum of the photocurrent density operator is the spatio-temporal Fourier transform of this correlation func- tion. Ω) = {δ î(0. .e.401) can be negative and compensate partially or even completely for the shot-noise contribution for some frequencies Ω and spatial fre- quencies q. from the temporal domain into the space–time one.402) and the last one from the space–time-independent product of two mean photocurrent densities.396). (3.401) Here the first contribution comes from the shot-noise term in relation (3. we can write the noise spectrum A (δ î)2 (q. ·}+ indicates an anticommutator. In quantum theory the sum of the second and third terms in relation (3. when the light field is treated as a c-number. t) . t  )}+ depends only on the time difference τ and the spatial difference ξ .396). δ î(ρ.154 3 Macroscopic Theories and Their Applications for arbitrary ξ and τ . ) * A 1 (δ î)2 (q. 0). i. Ω) = exp[i(Ωt − q · ρ)]G (2) (ρ. respectively. t).400) Using the photodetection formula (3. (3.396).

z) is the spatial coordinate.405).405) Here â(k) and â † (k) are the photon annihilation and creation operators of a spa- tial mode with the wave vector k.e. At the start of such a study. t) = i ω(k)â(k) 20 (2π)3 × exp[i(k · r − ω(k)t)] d3 k. For a wave travelling in the +z-direction. i. Evolution of the quantized field due. as purely temporal evolution. (3. On comparison of relation (3.e. one sees that  1 ω(k) â(z. But the simpler question of quantized field propagation in free space is considered first. for instance. For a complete quantum-mechanical description. to the interaction with an atomic medium is described in terms of the Heisenberg equations for annihila- tion and creation operators. Equation (3. where r = (x. â(k )] = 0̂. In the Heisenberg representation (3. [â(k). this density matrix remains constant as time evolves. we would like to have a formula that determines the field operator at any point ρ in the transverse plane at coordinate z given the field operator over the plane z = 0. but such particularities appear consistently in the review article. Such a description for transparent nonlinear media when the field interaction with atoms is described in terms of an effective Hamiltonian is much appreciated. i. this operator is written in the form of the modal decomposition  +  1 Ê (+) (r. â † (k )] = (2π)3 δ(k − k )1̂.405) determines the Heisenberg field operator Ê (+) (r. Such a description of field dynamics is not well suited to the problem of field propagation in free space or a medium.393) with relation (3.405). (3. with k = |k|.406) The factor (2π )3 is not usual.407) . t) in all points r and t of the space–time as a solution of the initial-value problem. through the modal operators â(k) and â † (k) given at time t = 0 as Schrödinger operators. t) = â(k) (2π)3 k0 × exp{i[q · ρ + (k z − k0 )z − (ω(k) − ω0 )t]} d2 q dk z . In the continuum limit. it would be more appropriate to have a quantum-mechanical analogue of the classical wave-optical propagation and diffraction theory. (3. we have to specify the density matrix of the field for the continuum set of modes k.3 Modes of Universe and Paraxial Quantum Propagation 155 An opinion of many workers in quantum optics is expressed in Kolobov (1999). Let Ê (+) (r.3. ρ. t). y. The operators â(k) and â † (k) obey the canonical commutation relations [â(k). be the positive- frequency operator of the electric field in a vacuum. The difficulties associated with the quantum-mechanical description of field propa- gation in free space or a nonlinear medium lie in the usual procedure of field quan- tization. the frequency ω(k) is given by the free-space dispersion relation ω(k) = kc.

(3. ρ.409) where    i ∂ 1 2 δ̃(r − r ) ≈ 1 − − 2 ∇⊥ δ(r − r ).412) This differs from relation (3. This constant is u(k)v(k) ξ 2 (k) = . In this approximation.410) k0 ∂z 2k0 with ∇⊥2 being the transverse Laplacian with respect to ρ. The positive-frequency operator of a quantized electric field in a transparent dielectric medium can be written in a form similar to that for a vacuum (Klyshko 1988). . t) = i ξ (k) ω(k)â(k) 20 (2π)3 × exp[i(k · r − ω(k)t)] d3 k. ρ.394) are not proved in this connection. which describes the strength of the field in the medium as compared to that in a vacuum. ρ. that is. [â(z. ρ. (3.156 3 Macroscopic Theories and Their Applications The normalization of these operators is such that the free Hamiltonian of the elec- tromagnetic field can be written as ω0 Ĥ0 = â † (z. t)â(z. it is based on the book of Klyshko (1988).408) c V The commutation relations (3. is reproduced in Kolobov (1999). the equation for the slowly varying operator â(z.405) in the factor ξ (k). which is related to the equation for propaga- tion of a quantized field in a nonlinear parametric medium. but equal-time ones are derived. (3. â † (z  . t). t) reads   ∂ ∂ i 2 â(z. t). ρ.   1 + Ê (+) (r. and ρ(k) is the so-called generalized anisotropy angle. (3. In part. t)] = δ̃(r − r )1̂. ρ. (3. ρ  . Here we have reproduced only the expression derived in the quasimonochromatic and paraxial approxima- tions from the literature. u(k) = ∂ω(k) ∂k is the group velocity.411) ∂t ∂z 2k0 An unpublished result of Sokolov.413) c2 cos ρ(k) Here v(k) = n(k) c is the phase velocity of light in the medium. t) = −c + c ∇⊥ â(z. (3. t) d3 r. the angle between the electric field and the induction.

t) = E p exp[i(k p z − ω p t)]. Using this notation one can introduce the slowly varying operator â(z. One will describe the parametric interaction in the medium in terms of an effec- tive Hamiltonian. t)]2 d3 r + H. ρ. (3. t) = iξ exp[i(k1 z − ω0 t)]â(z. Relation (3. we neglect the quantum fluctuations of the pump wave.3. ρ. ρ.412) neglects the use of and summa- tion over the parameter μ too.3 Modes of Universe and Paraxial Quantum Propagation 157 The fact that the electromagnetic field is a vector field is not emphasized in Kolobov (1999). The dispersion relation ω(k) is not single valued. t) of the quantized field in the medium (cf. (3.  ω0 Ê (+) (z. ρ. The slowly varying operator â(z. The medium is illuminated by a monochromatic plane wave. ρ. (3. t) in the paramet- ric medium is described by the following equation: ∂ i â(z. Relation (3. but has at least two branches. The evolution of the slowly varying amplitude operator â(z. The annihilation opera- tors correspond to these branches. t)]. and g is the strength constant of the parametric interaction proportional to the amplitude E p of the pump wave and the susceptibility constant χ (2) of the medium.407). t) + [ Ĥ0 + Ĥint .414) 20 c Here we have denoted by k1 the wave number of the wave in the medium. the pump. equation (3. These branches should be distinguished by a parameter μ. t) is given by an equation identical to relation (3. It is assumed that a χ (2) nonlinear parametric medium fills a vol- ume V . t). â(z. The pump wave propagates in the +z-direction and has the frequency ωp and wave number kp .  1 ω(k) â(z.415) but here ω(k) means a dispersion relation for the medium. ρ. ρ. (3. t) = iω0 â(z.416) We choose the frequency ωp of the pump wave in the form ωp = 2ω0 and consider the amplitude E p as a c-number. ρ. t) = â(k) (2π )3 k0 × exp{i[q · ρ + (k z − k0 )z − (ω(k) − ω0 )t]} d2 q dk z .e. ρ. but it is mentioned in respect of the book by Klyshko (1988). i. ρ.393)).c. Ê p(+) (z. (3.418) ∂t  .412) yet neglects the use of and summation over the appropriate parame- ter ν.417) c V Here n 0 gives the density of active atoms in the parametric medium. ρ. Under usual assumptions the parametric interaction can be described by the fol- lowing effective Hamiltonian n0 g Ĥint = i exp[i(kp − 2k1 )z][â † (z.

(3. Ω) exp{i[k z (q. Ω) − k1 ]z}. −Ω) exp[iΔ(q.420) where + k z (q. Ω) = B̂ B̂ a(s + k z (q.421) with q = |q| a z-component of the wave vector with frequency ω0 + Ω and spatial (s. Ω)z]. Let fl (ρ) mean these eigenmodes. q. t) and â † (z. ˜ q. When an active nonlinear medium is placed in a resonator. B̂ a(s. Ω) dz. Δ(q. ∂ ˆ (z.426) l .423) One lets u = ∂ω(k ∂k1 1) mean the group velocity of the wave in the crystal. a description may employ discrete transverse modes of the cavity. â(z. −q. On appropri- ate derivations. For B̂ a(s. q. Ω). q. ρ. ˜ q. We express the Fourier transform â(z. −Ω) − kp . Ω). q. q. In terms of â(z. t) exp(iΩt) exp(−isz) exp(−iq · ρ) dt d2 ρ dz (3. ρ.424) ∂z where σ = 2nu0 g is the coupling constant of the parametric interaction. and Lugiato and Marzoli (1995) have adopted this approach. t) it is given by relation (3. Ω) similarly defined as B̂ frequency q.408). (3. Ω) + k z (−q. t). Lugiato and Gatti (1993). Ω) − k1 . (3. Gatti and Lugiato (1995). q. (3. The set of functions fl (ρ) satisfies both the condition of orthonormality fl∗ (ρ) fl  (ρ) d2 ρ = δll  (3. One introduces the Fourier transform of the space–time photon annihilation operator â(z.158 3 Macroscopic Theories and Their Applications Here Ĥ0 is the free-field Hamiltonian in the medium. q. q. ρ.422) One introduces the mismatch function Δ(q. ˜ q. ρ. Ω). Ω) = ˆ (z. (3. Kolobov (1999) presents the equation of propagation for the operator ˆ (z. Ω) = σ ˆ † (z.425) and completeness  fl∗ (ρ) fl (ρ  ) = δ(ρ − ρ  ). q.419) ≡ e−isz â(z. Ω) = k z (q. (3. Ω). Ω) = k 2 (ω0 + Ω) − q 2 . Ω) = â(z. Ω) with the aid of a new operator ˆ (z. Ω) it holds that (s.

Their transverse components are ±q and their z-components are k z (q. q. Ω) = U (q. We can express the space–time correlation function of the photocurrent density operator (3. one arrives at the following transformation â(z. From the commutation relations † (3. Ω) and k z (−q. As a result of the parametric down-conversion. with frequencies ω0 + Ω and ω0 − Ω. (3. The Fourier transforms B̂a l (Ω) are defined as ∞ B̂ a l (Ω) = exp(iΩt)âl (t) dt.  â(ρ. (3. Ω) is introduced in Kolobov (1999). δ î(ρ  . ˜ q. t). respectively. Ω) + V (q. t) over the eigenmodes fl (ρ). (3. Ω) inside the crystal is described by equation (3. −Ω). q. Solving this equation and respecting relation (3.3. −q.424). −Ω).421).429) −∞ Noise spectrum of the photocurrent density for the lth mode as an analogue of the noise spectrum (δ A î)2 (q. It is not assumed that the photocurrent density is uniform in space. ) *  1 {δ î(ρ. by relation (3. ˜ q. â(l. (3.420) between the operators ˜ q.394) together with relation (3.b). âl  (t  )] = δll  δ(t − t  )1̂.427) l where âl (t) are operator-valued expansion coefficients that have the meaning of photon annihilation operators for the lth mode. −Ω).3 Modes of Universe and Paraxial Quantum Propagation 159 One can expand the slowly varying field operator â(ρ.396) in terms of the noise spectra (δAî)2 l (Ω) of the individual eigenmodes fl (ρ) of the cavity. Ω)â(0. The evolution of the slowly varying operator ˆ (z. Ω) and ˆ (z.430) 2π −∞ (i) Generation of multimode squeezed states of light The generation of multimode squeezed states of light by a travelling-wave opti- cal parametric amplifier was described by Kolobov and Sokolov (1989a. a pump photon ωp splits into signal and idler photons. Ω)☠† (0.428) It is noted that the derivation is valid for the field operators outside the cavity. Ω) and k(−q. Ω). t) = fl (ρ)âl (t). (3. but that it is stationary in time and the photocurrent density fluctuations for different eigenmodes are uncorrelated.431) .425) it is easy to see that âl (t) and âl (t) obey the commutation relation † [âl (t). t  )}+ = fl (ρ) fl (ρ  ) 2 l ∞ 1 A × (δ î)2 l (Ω) exp[−iΩ(t − t  )] dΩ. and wave vectors k(q.

π.396). Ω) = 0. Ω)|2 − |V (q.432) Γ where  [Δ(q. Ω) obey the free-field At the input to the crystal. (3.438) (2π)3 . (1985) in the case of co-propagation and by Yurke (1985) for counter-propagation. Ω) = exp i k z (q. q . operators â(0. [â(0. (3. Ω) and V (q.433) 4 The functions U (q. Ω) equal to    Δ(q. (3. β = |β| exp(iϕβ ) = αU (0. Ω) × cosh(Γl) + sinh(Γl) . and (3. Equation (3.395). Ω )] = (2π)3 δ(q − q )δ(Ω − Ω )1̂. q. Ω) and V (q. commutation relation ˜ q. Ω)|2 d2 q dΩ ≡ î l + î s . Ω)]2 Γ= |σ |2 − . q. (3. (3. Ω) U (q.437) Phase modulation predominates for θ (q. Ω) V (q. 2Γ    Δ(q. Ω) = ± π2 and amplitude modulation for θ (q.431) is used. Ω) − ϕβ .431) involves the spatial frequency q. Ω) = exp i k z (q. Ω) have the property |U (q.400). (3.434) ˜ q. (3. It is assumed that. Relation (3. Upon leaving the crystal. which first enters as a wave with complex amplitude α and q = 0. along with the pump wave. (3. a monochromatic plane wave of frequency ω0 is incident normal to the input surface of the crystal. Ω)|2 = 1. Ω) = ψ(l. 0) + α ∗ V (0. 0). Ω) and ☠† (0. The mean of the photocurrent density operator has the forms η î = η|β|2 + |V (q.436) The type of noise modulation of the resultant field in space–time is determined by the angle θ (q. The mean of the photocurrent density operator and its noise spectrum is found from relations (3. ☠† (0.435) The broad-band squeezing in a three-wave interaction was discussed by Caves and Crouch (1987) and in a four-wave interaction by Levenson et al.160 3 Macroscopic Theories and Their Applications with coefficients U (q. Ω) − k1 − l 2 σ × sinh(Γl). this wave will serve as a local oscillator wave with the complex amplitude β. Ω) − k1 − l 2  iΔ(q. Ω).

442) where g(q. ˜ q. Ω) + exp[iψ(z. Ω − Ω )] d2 q dΩ . Ω) = U (q. Ω − Ω ) (2π)3 + g ∗ (q . (3. When the com- plex components are used. ˜ q. Ω) + Re{exp(−2iϕβ )g(q. q.431) simplifies to the form ☠μc (l.440) Tc Sc where 1 δs = 2 δ(q. (3. With any angle ψ(z. Ω)]☠† (z. One introduces the function δ(q. q. −Ω). −q. Ω)} η2 + [δ(q .441) qc Ωc is the degeneracy parameter for spontaneous parametric down-conversion (Mandel and Wolf 1995). q. λ = c. Ω)V (−q. are introduced with the property ☠1c (z. Ω)]â(z.446) . Ω) + i☠1s (z.3 Modes of Universe and Paraxial Quantum Propagation 161 where the subscript l refers to the local oscillator field and the subscript s indicates the spontaneous parametric down-conversion. (3. Ω)]. q. Ω) − exp[iψ(z. Ω) for a while. “complex quadrature components” are first defined. Ω) + i☠μs (l. q.3.444) ☠2c (z. s. Ω)]â(z. transformation (3. Ω) .439) The mean of the photocurrent density operator î s can be written as δs î s = η . Tc = 2π Ωc is its coherence time.443) Under homodyne detection. Ω) = exp[iκ(q. −Ω) modulate the local oscillator wave in space and time. q. Ω) and (−q. Ω). Ω)]☠† (z. q. the down-conversion waves (q. Ω)] exp[±r (q. Ω)   = −i exp[−iψ(z. q. Ω )g(q − q . (3. The noise spectrum of the photocurrent density has the form A   (δ î)2 (q. q. Ω) + i☠2s (z. Ω)][☠μc (0.445) In other words. q. Ω)|2 . (3. Sc = ( 2π qc )2 the coherence area. Ω )δ(q − q . −q. (3. Ω) = exp[−iψ(z. (3. q. and Ωc and qc the widths of the frequency and spatial frequency spectra of spontaneous parametric down-conversion. q. Ω) = î + 2η2 |β|2 δ(q. Ω). Ω) + i☠μs (0. slow quadrature components ☠μλ (z. q. Ω) = |V (q. Ω) d2 q dΩ (3. q.

Δ(q. Ωm ) = ± π2 . Ω). or in a forward geometry. If the phase-matching condition is not perfectly met. Ω = 0. −Ω)|. In Kolobov (1999). In space–time language this can be said as follows. one must pay attention to nonzero carrier frequencies q and Ω. Ω) = î 1 − η + η cos2 [θ (q.447) 2 In relation (3. κ(q. Ω)]. Ω) at the output surface of the crystal are defined with ψ(l. Ωm . −Ω)]. Ω)U −1 (q. Ω) are two other squeezing parameters.162 3 Macroscopic Theories and Their Applications where + corresponds to the component with μ = 1 and − to the component with μ = 2. Δ(0. Ω) is affected to the first order in Δ(q. Ω)] exp[−2r (q. The components ☠μλ (0. when Δ(0. Three kinds of measurement can be distinguished based on the three types of phase matching. Ω) ≈ 0. q. Ω)V (−q.449) Maximum squeezing occurs at frequencies qm . as proposed by Yuen and Shapiro (1979). Ω)]. 2 1 ψ(l. . Ω)] . q. the phase θ (q.446). q. They are said to belong to such a phase-matching surface. A backward four-wave mixer can produce multimode squeezed light with a much larger spatial bandwidth than a forward four-wave mixer and another scheme. which are necessary for reducing fluctuations in the number of photoelectrons below the Poissonian limit. Yuen and Shapiro (1979) were the first to propose a degenerate mixing process as a possible source for squeezed light. Ω) = arg[U (q.448) Leaving out the integral term in the noise spectrum of the photocurrent density (3. Ω) and r (q. and (c) q = 0. Ωm ) = 0. (b) q = 0. In the case of degenerate phase matching and  kΩ < 0 one infers that in the region of frequencies Ω < Ωm and spatial frequencies q < qm the noise of the photocurrent density operator is reduced below the shot- noise level.442). In the case of nondegenerate phase matching in a crystal. Ω)]  + sin2 [θ (q. the case of the frequency and angle-degenerate phase match- ing. according to Kumar and Shapiro (1984). 0) = 0. one chooses a complex amplitude of the local oscillator wave with θ (qm . Ω = 0. (3. (3. Ω) = arg[U (q. A four-wave mixer can be set up either in a backward geometry. 1 κ(q. Ω)] exp[2r (q. Ω)U −1 (q. Ω)lamp and the squeezing parameter is not yet influenced to the first order. one can rewrite it in the form A   (δ î)2 (q. These angles are 1 ψ(0. Ω) at the input surface to the crystal are defined in the coordinate system with ψ(0. Ω) and the components ☠μλ (l. 0) > 0. is considered. q. Ω = 0. In fact. To reduce shot noise to the highest extent. q. Ω) = arg[V (q. The frequencies Ωm and qm determine the minimum time Tm and the minimum area of photodetector Sm . Ω)] = |U (q. q. when (a) q = 0. (3. Ω)| ± |V (−q. 2 exp[±r (q. which fulfil the condition Δ(qm .

−q. q. Ω)z]. ρ. a phase conjugate wave is generated in the medium that propagates in the opposite direction to the probe wave (Fisher 1983). Ω). Ω)b̂in (−q.452) ∂z Here κ is a coupling constant proportional to the product of the two pump wave amplitudes and to the nonlinear susceptibility χ (3) of the medium. q. μ = p.454) † b̂out (q. μ = p. ρ. respectively. The process occurs in a transparent χ (3) nonlinear medium. illuminate the slab at a small angle to the z-axis. q. Ω) = ˜ˆc (l. Ω) + V (q. Two counterpropagating plane monochro- matic pump waves E 1 and E 2 of angular frequency ω0 and wave vectors k1 and k2 . Ω) + kcz (−q. q. −Ω) exp[−iΔ(q. Ω)âin (q. Ω) = ˆμ (z. The solution of equations (3. mean the wave vectors of the probe and conjugate waves. Ω) ∝ U (q. as follows: ˜ˆμ (z. respectively. One describes the probe and conjugate waves by two corresponding slowly varying operators ˆp (z. Let kμ (q. q. t) and ˆc (z. ρ. q.453) with kp. −Ω) exp[−iΔ(q. Ω) may follow the classical one (Fisher 1983).451) and (3. Ω). (3. q. −Ω) − k1z − k2z . A quasiplane and quasimonochromatic probe wave of carrier frequency ω0 enters the medium from the left and propagates in the +z-direction. Ω) and ˜ˆc (z = l. Ω) is a phase-mismatch function given by Δ(q. Δ(q. (3. Ω)z]. Ω)b̂in (q. c. Ω) = −iκ ˜ˆ c (z. (3.cz (q. Ω) ∝ U (q. −Ω). In the nonlinear interaction between the two pump waves and the probe wave. (3. (3. −Ω). ˜ˆμ (z. c. The input–output transformation. (3. One introduces the Fourier trans- forms of these space–time operators.452) with the boundary conditions ˜ˆp (z = 0. . −q.450) These operators evolve in the nonlinear medium according to the equations ∂ ˜ † ˆ p (z. Any of them can be compared with the optical parametric amplifier. q.2z the corresponding projections for the pump waves.451) ∂z ∂ ˜ † ˆc (z. t).3. Ω) = iκ ˜ˆ p (z. t) exp[i(Ωt − q · ρ)] dt d2 ρ.3 Modes of Universe and Paraxial Quantum Propagation 163 The exposition is restricted to the backward four-wave mixing. Ω) = kpz (q. Ω) + V (q. Ω) being the projections of the probe and conjugate wave vectors onto the positive z-direction and k1. † âout (q. It is assumed that the medium has the form of a plane slab the thickness of which (distance between two surfaces parallel to the ρ plane) is equal to l.455) Notably one obtains two independent processes of multimode squeezing. which yet differs from the solution by the inclusion of the incoupling and outcoupling beam splitters. Ω)âin (−q. Ω) = ˜ˆp (0. is presented in Kolobov (1999).

(3. Even though here a cavity is investigated. One is led to the idea of a “realistic” medium and its “quantum nature”. The source is described by the master equation. The functions f pli (r. φ) f p l  i  (r. . 2π ∞ f pli (r. f pli (r. Another source is based on a cavity. the cavity modes still seem to have been determined in the paraxial approximation. (3. The frequency dependence can be used to filter the probe signal. in which situation the paraxial approximation has been used. are the radial and angular indices. φ) satisfy the conditions of orthonormality. such a discrete eigenset is given by the Gauss– Laguerre modes  cos(lφ) for i = 1. (Here Δ ≡ ΔOPA . Ω).458) 0 0 The eigenfrequencies of these modes are given by ω pl = ω00 + (2 p + l)ζ. and it still can generate multimode squeezed states. . .460) pli ∂t i i=1 p. a nonlinearity of the equations is meant. l = 0. r = x 2 + y 2 is the radial and φ is the angular variable. but also a better quantization desired.) This func- tion depends on the spatial frequency in the optical parametric amplification and the forward four-wave mixing and does not depend on it in the backward four- wave/mixing. For a cavity-based geometry a more natural language for the description of mul- timode squeezing is that of discrete eigenmodes of the resonator. In the case of the cavity with spherical mirrors.456) sin(lφ) for i = 2. not a travelling wave. φ)r dr dφ = δ pp δll  δii  . ∂ 1  2 ρ̂ = [ Ĥint . Obviously.457) 2δl. The functions L lp are the Laguerre polynomials. Here we are provided by the literature with another example.164 3 Macroscopic Theories and Their Applications The investigation of the difference is concentrated on comparison between the phase-mismatch functions Δ(q. φ) = f̃ pl (r ) × (3. ρ̂] + ˆ ρ̂. (3. It is a subthreshold optical parametric oscillator. ΔBFWM .459) where ω00 is the lowest eigenfrequency of the resonator and the parameter ζ depends on the curvature of mirrors and the distance between them (Yariv 1989).0 π w 2 ( p + l)! w 2 w2 w where w is the waist of the beam + and p. which just as we saw at the beginning has not yet been employed. respectively. concretely this scheme in a cavity with spherical mirrors (Lugiato and Marzoli 1995). 1.       2 p! 2r 2 l 2r 2 r2 f̃ pl (r ) = √ Lp exp − 2 . . ΔFFWM . Λ̂ (3. From these properties one determines “spatial” squeezing bandwidth q = kl1 and q = ∞ in the paraxial approximation. 2.l .

The method of description seems to be known in quantum optics. and â pli (t) inside the cavity.463) where ω pl − ωs Δ pl = .465) In relation (3.466) Using the Fourier transform ∞ B̂ g pli (Ω) = exp(iΩt)ĝ pli (t) dt.467) −∞ . (3. respectively. ĉ pli (t) of the vacuum fluctuations entering it. (3. This property says that a subthreshold oscillator is being investigated. The † operators ĉ pli (t) and ĉ pli (t) correspond to the operator-valued Langevin forces and describe the vacuum fluctuations entering the cavity through the outcoupling mir- rors. 2π ∞ γ  †2  Ĥint = i Ap  (r. φ) − Â2 (r.464) γ with ω pl and ωs the eigenfrequencies of the eigenmodes of the resonator and the frequencies of signal photons. These operators obey the commutation relations † [ĉ pli (t).460). b. from which we conclude that 0 < Ap < 1. (3. a coupling constant is expressed as the product γ Ap .461) pli pli pli pli pli pli pli describes the damping of the mode pli due to cavity decay through the outcoupling mirror with the rate γ . Every mode is damped and the rate constant for each of the modes is the same. where the term Λ̂ pli   ˆ ρ̂ = γ 2â ρ̂ â † − â † â ρ̂ − ρ̂ â † â Λ̂ (3.3 Modes of Universe and Paraxial Quantum Propagation 165 ˆ ρ̂.462) 2 0 0 where Ap is the coupling constant proportional to the nonlinear susceptibility χ (2) of the medium and the amplitude of the pump wave.462).3. Instead of solving the master equation (3. † + â pli (t) = −γ [(1 + iΔ pl )â pli (t) − Ap â pli (t)] + 2γ ĉ pli (t). (3. g = a. φ) r dr dφ. + b̂ pli (t) = 2γ â pli (t) − ĉ pli (t). (3. we can write a set of indepen- dent Langevin equations (Walls and Milburn 1994) for the annihilation and creation † operators â pli (t) and â pli (t) inside the cavity. ĉ pli (t  )] = δ pp δll  δii  δ(t − t  )1̂. Kolobov (1999) refers to Collet and Gardiner (1984) for the input–output rela- tions for the field operators b̂ pli (t) in the wave outgoing from the cavity. The interaction Hamiltonian Hint is given by Lugiato and Marzoli (1995). c. (3. Such a simplification should still be explained.

166 3 Macroscopic Theories and Their Applications to equations (3. (3. δ î(ρ  .469) [1 + iΔ pl (Ω)][1 − iΔ pl (−Ω)] − A2p with Δ pl (±Ω) = Δ pl ∓ Ω γ . t). which is a discrete equivalent of the multimode squeez- ing transformation (3. A relation holds  A 4Ap (δ î)2 pl (Ω) = î(ρ.463).468) with the coefficients [1 − iΔ pl (−Ω)][1 − iΔ pl (Ω)] + A2p U pl (Ω) = . we arrive at the following squeezing transformation between the Fourier transforms of the incoming and outgoing operators: B̂ b pli (Ω) = U pl (Ω)B̂ c pli (Ω) + V pl (Ω)B̂ † c pli (−Ω). A lens allows one to compensate for this deterioration and even further to improve the resolving power. ˜ + L . Ω) at the exit plane of the crystal and â(l â(l. t  )}+ = f̃ pl (r ) f̃ pl (r  ) cos[l(φ − φ  )] 2 p.471) where ϕβ is the phase of the local oscillator and Ω̃ = Ω γ is the dimensionless fre- quency and η = 1 (Collett and Walls 1985. [1 + iΔ pl (Ω)][1 − iΔ pl (−Ω)] − A2p 2Ap V pl (Ω) = . equation (3.430). The slowly varying operators ˜ q. t) 1 + (1 + Δ2pl − A2p − Ω̃2 )2 + 4Ω̃2  × [2Ap + Re{exp(−2iϕβ )(1 − Δ2pl + A2p + Ω̃2 − 2iΔ pl )}] . Savage and Walls 1987). (3. (3.l ∞ 1 A × (δ î)2 pl (Ω) exp[−iΩ(t − t  )] dΩ.420)) follows: . We will assume that the plane of photodetection lies at a distance L from the exit plane of the nonlinear crystal and is parallel to it. but the fol- lowing relation is presented (Lugiato and Marzoli 1995): ) *  1 {δ î(ρ. Ω) at the photodetection plane are for the free propagation related as (cf. q.470) 2π −∞ which is an analogue of relation (3. (ii) Free propagation and diffraction of multimode squeezed light With respect to the free propagation and diffraction of multimode squeezed light it is shown that propagation in free space in general deteriorates the resolving power of low-noise measurements with squeezed light.431). The calculation of the photocurrent noise spectrum is not included. (3.

˜ + L . t) = exp −i â z. For a lens of focal length f . −q.473) with the coefficients Ũ (q. Along with the free propagation one is interested in the dependence of the field operator at the plane of photodetection on that at the input to the nonlinear crystal. Ω) = θ(q. From the results of the analyses performed it is natural to choose z = l.476) 2f c 2f A From this relation it follows that the noise spectrum (δ î)2 (q. Ω).474) and the like for Ṽ (q. Ω) − k0 ]L}U (q.e. Ω) = Ũ (q. phase) of the squeezing changes more rapidly in dependence on the spatial frequency than on the output from the crystal. −Ω). (3. Ω) − k0 ]L}â(l.e. Relation L (0) θ̃ (q. Even the phase shifts produced during wave propagation inside the nonlinear crystal can be com- pensated for.475) 2k0 where a paraxial and quasimonochromatic approximation is assumed. (3. Ω) − . The increase is related to the diffraction. Ω). i. (3. says that the orientation angle (i. (3.3 Modes of Universe and Paraxial Quantum Propagation 167 ˜ + L . q. The deterioration is reversible. Ω) + k z(0) (−q. the output plane of the crystal. The resolving power of the low- noise observation has decreased. q. t − 4f + c . −Ω) − 2k0 ] 2 z q2 L ≈ θ(q. Ω) has been conserved. where L is negative. New quantities are provided with a tilde. Ω). ρ. â(l ˜ q. Ω) = exp{i[k z(0) (q. provided that the object plane has the position −2 f relative to the lens and the image plane has the position 2 f relative to the lens. It suffices to choose k0 L = −lamp . Ω) = exp{i[k z(0) (q.3.477) 2kl . Concerns with correct quantization call attention to the proposal of imaging some plane inside the crystal onto the detection plane.472) where k z(0) (q. Ω) is the z-component of the wave vector in free space. Ω)☠† (0. Ω)â(0. where l amp is the amplification 1 k0 length. The minimum area lamp Sm of low-noise detection is proportional to k + 2L . (3. It is a simple generalization of the above description. This imaging is understood as a general choice z = l + L. Ω) + [k (q. Ω) + Ṽ (q. (3. −ρ. â(l ˜ q. the optical imaging is represented by the field operators     k0 ρ 2 1 ρ2 â(z + 4 f.

The scheme of homodyne detection is closely related to holographic measurements.478) 2kΩ  (1)  if kΩ has the opposite sign to kΩ . A number of references for interference mixing. the photo- electron number collected by a pixel with the area Sd during the time interval Td is considered as an example. even though a statistical definition of the mode is peculiar. The outgoing light from the two ports of the second beam splitter is detected by two photodetector arrays. the result is independent of Sd and Td . Thus geometrical imaging of the plane inside the crys- tal at the distance L given by (3. (iv) Spatially noiseless optical amplification of images . If it holds that Sd ≥ Sc and Td ≥ Tc . Ω) in the vicinity of the matching surface to become indepen- dent of spatial frequency. when squeezing is significant. For general- ization and analogy in the case of spatial modulation.477) onto the photodetection plane broadens the range of spatial frequencies at which one has a noise reduction below the shot-noise level. One example of nonde- structive modulation in space is an opaque screen with apertures larger than the coherence area of squeezed light Sc . Here we concede the efficiency of models simplified to several modes: The average number of photons necessary for a single low-noise measurement is given by a quantity.b). The amplitude modulation in space is not advantageous for creation of optical images with a regular (sub-Poissonian) photon statistics. In order to assess physical possibilities for low-noise measurements. the coherence area Sc limits the number of modes on an illuminated spot of area S on the input to the nonlinear crystal. which we could obtain on choosing as “average” model simplified to two modes. Kolobov (1999) remarks on what follows. (3. which can be trans- mitted in a time interval T . The sub-shot-noise microscopy utilizes a Mach–Zehnder interferometer. The length of the slab will be  kΩ L (1) = −lamp (1)  . η ≈ 1. (iii) Noiseless control of multimode squeezed light With respect to the noiseless control of multimode squeezed light. the detection of faint phase objects as proposed by Kolobov and Kumar (1993) is described. As a natural generalization of the analysis in Caves (1981). For high quantum efficiency. the minimum detectable spatially varying phase change is defined. which provides such nondestructive modulation in time are presented. Whereas the coherence time Tc limits the number of images. the statistics of photoelectrons is sub-Poissonian. see Sokolov (1991a. An improvement of the frequency behaviour of the noise spectrum can be achieved by inserting into the light beam a slab of a dispersive medium with wave number k (1) (Ω).168 3 Macroscopic Theories and Their Applications for the phase θ̃ (q.

Let b̂in (ξ . which is an analogue of the band of (temporal) frequencies.481) 2L where T is the intensity transmission coefficient of the cavity outcoupling mirror. t) of the cavity mode closest to resonance with input signal is described by the equation ∂ c b̂(ξ . . Kolobov (1999) refers to Kolobov and Lugiato (1995) for such a proposal. Caves (1982)) and it should be extended to the spatial domain. In a paraxial approximation.482) where ωc is the longitudinal cavity frequency closest to the frequency ωs of the signal field. (3. the detuning parameter is defined as Δ = ωc − ωs . The exposi- tion is self-contained. t) ∂t 2k √ + σ b̂† (ξ .479) λf λf where f is the focal length of the lens and λ is the wavelength of the light. (3. Let â(ρ. In (3. spectral bandwidth of images should be within the bandwidth of the cavity employed. t) stand for the photon annihilation and creation operators in the image plane P4 . L is the perimeter of the cavity. t) in the object plane by the following transformation performed by the lens L 1 :  1 2π b̂in (ξ . (3. and c is the light velocity in a vacuum. In general.480) σ is the constant of parametric interaction proportional to the pump amplitude and k is the wave number of the travelling wave inside the cavity.3. t). t) mean the photon annihilation and creation operators in the object plane P1 and let ê(ρ. t) = −(κ + iΔ)b̂(ξ .3 Modes of Universe and Paraxial Quantum Propagation 169 Noiseless amplification has been defined in phase-sensitive amplifiers (see. (3. k = 2π λ . respectively. which per- form the spatial Fourier transformation and broaden a narrow region of transverse vectors q. One considers a ring-cavity degen- erate optical parametric amplifier and monochromatic images. for example.480) Here κ is the cavity decay constant equal to cT κ= . t) is expressed through the operator â(ρ. t) and â † (ρ. t) mean the field operators in the input and the output planes of the optical parametric oscillator. t) = â(ρ. the slowly varying field operator b̂(ξ . Many areas of physics would benefit from the possibility of noiseless amplification of faint optical images. t) exp −i ξ · ρ d2 ρ. t) and ê† (ρ. The operator b̂in (ξ . Astronomy and microscopy come to mind. t) − i ∇⊥2 b̂(ξ . t) and b̂out (ξ . The optical parametric amplifier is combined with input and output lenses. t) + 2κ b̂in (ξ .

170 3 Macroscopic Theories and Their Applications

The output field operator b̂out (ξ , t) is the sum of two waves, one of which is
reflected from and another transmitted through the outcoupling mirror of the cavity,

b̂out (ξ , t) = 2κ b̂(ξ , t) − b̂in (ξ , t). (3.483)

To express the output field operators in terms of the input operators one takes the
spatio-temporal Fourier transform of b̂(ξ , t),

b(q, Ω) = b̂(ξ , t) exp[i(Ωt − q · ξ )] d2 ξ dt. (3.484)

The spatio-temporal Fourier transforms of b̂in (ξ , t) and b̂out (ξ , t) are similar.
The transformation of the field amplitude from the object plane P1 to the input
plane P2 given by relation (3.479) is equivalent to the following relation between
the spatio-temporal Fourier transform B
bin (q, Ω) and the temporal Fourier transform
â(ρ, Ω)
B̂ λf
bin (q, Ω) = λ f â − q, Ω ,
˜ (3.485)

where we have used

â(ρ, Ω) = â(ρ, t) exp(iΩt) dt. (3.486)

Since the lens L 2 has the same focal length as L 1 , we have an identical relationship
between the Fourier transforms B̂bout (q, Ω) and ê(ρ,
˜ Ω) in the output plane P3 and in
the image plane P4 ,
˜ Ω) = λ f B̂ λf
ê(ρ, bout − ρ, Ω , (3.487)

˜ Ω) = ê(ρ, t) exp(iΩt) dt.
ê(ρ, (3.488)

It can be derived that

˜ Ω) = u(ρ, Ω)â(ρ,
ê(ρ, ˜ Ω) + v(ρ, Ω)☠† (−ρ, −Ω), (3.489)

with u(ρ, Ω) and v(ρ, Ω) given by

[1 − iδ(ρ, Ω)][1 − iδ(ρ, −Ω)] + |g|2
u(ρ, Ω) = ,
[1 + iδ(ρ, Ω)][1 − iδ(ρ, −Ω)] − |g|2
v(ρ, Ω) = . (3.490)
[1 + iδ(ρ, Ω)][1 − iδ(ρ, −Ω)] − |g|2

3.3 Modes of Universe and Paraxial Quantum Propagation 171

Here one has introduced the dimensionless coupling strength g of the parametric

g= , (3.491)

and the dimensionless mismatch function δ(ρ, Ω),
Δ Ω ρ
δ(ρ, Ω) = − + , (3.492)
κ κ ρ0

with ρ0 defined as

ρ0 = f . (3.493)
2π L

To ensure a linear amplification regime, the coupling strength g must be |g| < 1.
In Kolobov (1999) it has been shown that multimode squeezed states of light
come about as a natural generalization of single-mode squeezed states. They can be
produced in experiments when just one spatial mode of the field is cut out by means
of a high-Q optical cavity. Travelling-wave configurations are most convenient for
the generation of multimode squeezed states. To observe them, one must employ a
dense array of photodetectors.
Many new physical phenomena are connected with multimode squeezing. Mul-
timode squeezed states offer a few applications including optical imaging with
sub-shot-noise sensitivity, sub-shot-noise microscopy, and noiseless amplification
of optical images. There are some other phenomena related to multimode squeez-
ing, such as the similarity of homodyne detection of multimode squeezed states to
the scheme of optical holography, an application of these states to optical image
recognition with photon-limited images (Morris 1989), a possibility to improve a
quantum limit in optical resolution with the use of nonclassical light (den Dekker
and van den Bos 1997). Multimode squeezed states can be applied in the field of
optical pattern formation, which studies the spatial and spatio-temporal phenomena
that arise in the structure of the electromagnetic field in the plane orthogonal to the
direction of propagation. For instance, the filamentation of a laser beam initiated by
quantum fluctuations of light in its transverse area (Nagasako et al. 1997, Lugiato
et al. 1999).
Björk et al. (2004) have shown that the use of entangled photon pairs in an imag-
ing system can be simulated with a classically correlated source sometimes. They
have considered two schemes with “bucket detection” of one of the photons. In con-
trast, entangled two-photon imaging may exhibit effects that cannot be mimicked
by any classical source when bucket detection is not used (Strekalov et al. 1995).
Caetano and Souto Ribeiro (2004) have investigated theoretically and experimen-
tally the transfer of the angular spectrum of the pump beam to the down-converted
beams. They have demonstrated that the image of a given object placed in the pump

172 3 Macroscopic Theories and Their Applications

can be formed in the twin beams by manipulating the entangled angular spectrum
and performing coincidence detection.
Gatti et al. (2004) have analytically shown that it is possible to perform coher-
ent imaging by using the classical correlation of two beams obtained by splitting
thermal light. They have presented a formal analogy between two such classically
correlated beams and two entangled beams produced by parametric down conver-
sion. The classical beams can qualitatively reproduce all the imaging properties of
the entangled beams. These classical beams are spatially correlated both in the near
field and in the far field even though to an imperfect degree.
Bache et al. (2004) have presented a theoretical study of ghost imaging which
uses balanced homodyne detection to measure signal and idler fields arising from
parametric down conversion. They have used a general model describing the three-
wave quantum interaction with respect to finite size and duration of the pump pulse.
They have shown that the signal–idler correlations contain the full amplitude and
phase information about an object located in the signal arm, both in the near-field
(object image) and the far-field (object diffraction pattern) cases. One may pass
from the far-field result to the near-field result by simply performing inverse Fourier
transformation. The analytical results are confirmed by numerical simulations.
Bennink et al. (2004) have reported two distinct experimental demonstrations
of coincidence imaging. They have shown that uncertainties of distance and mean
direction of two classical fields must obey an inequality. With the use of entangled
photons they formed two images whose resolution had a product that was three
times better than is possible according to classical diffraction theory. For the sake of
comparison, a similar experiment was performed with light in a classical mixture of
states (cf. Gatti et al. 2003). While the resolution of the image was good in the far
field, the uncertainty product obeyed the classical inequality in the near field.
Valencia et al. (2005) presented the first experimental demonstration of two-
photon ghost imaging with a pseudothermal source. They have introduced the con-
cepts of two-photon coherent and two-photon incoherent imaging. Similar to the
case of entangled states, a two-photon Gaussian thin lens equation connects the
object plane and the image plane. Specifically, the thermal source acts as a phase
conjugated mirror.
Altman et al. (2005) have probed the quantum image produced by parametric
down-conversion with a pump beam carrying orbital angular momentum. With one
detector fixed and the other scanning, the usual single-spot coincidence pattern is
predicted (Monken et al. 1998) to split into two spots, which has been demonstrated.
Mosset et al. (2005) have presented the first experimental demonstration of noise-
less amplification of images that yielded spatially integrated intensity (of the pho-
todetection process) for different lateral detector sizes. Achieving two-beam and
single-beam conditions, they have compared phase-insensitive and phase-sensitive
schemes with theory.
Quantum imaging is a branch of quantum optics that investigates the ultimate
performance limits of optical imaging imposed by quantum mechanics. The use of
quantum-optical methods enables one to solve the problems of image formation,
processing and detection with sensitivity and resolution which exceeds the limits

3.4 Optical Nonlinearity and Renormalization 173

of classical imaging. The most important theoretical and experimental results in
quantum imaging can be found in Kolobov (2007).

3.4 Optical Nonlinearity and Renormalization

Abram and Cohen (1994) mainly applied a travelling-wave formulation of the theory
of quantum optics to the description of the self-phase modulation of a short coherent
pulse of light. They seem to have been first to use a renormalization (Kubo 1962,
Zinn-Justin 1989). The renormalized theory successfully describes the nonlinear
chirp that the pulse undergoes in the course of its propagation and permits the cal-
culation of the squeezing characteristics of self-phase modulation. The description
of the propagation of a short coherent pulse of light inside a medium that exhibits
an intensity-dependent refractive index (Kerr effect) has become relevant to opti-
cal fibre communications, all-optical switching, and optical logic gates (Agrawal
1989). Neglect of dispersion and the Raman and Brillouin scattering leads to the
description of self-phase modulation. In classical theory it is derived that, in the
course of its propagation, the pulse becomes chirped (i.e. different parts of the pulse
acquire different central frequencies), which influences also its spectrum. Abram
and Cohen (1994) have pointed out many difficulties in the investigation of the
quantum noise properties of a light pulse undergoing self-phase modulation. The
traditional cavity-based formalism truncates the mutual interaction among the spa-
tial modes to a self-coupling of a single mode (or only a few modes) and cannot
give a reasonable approximation to the frequency spectrum produced by self-phase
modulation. In spite of the difficulties, papers based on a single-mode description of
a field indicated that the slowly varying approximation can produce squeezed light
(Kitagawa and Yamamoto 1986, Shirasaki et al. 1989, Shirasaki and Haus 1990,
Wright 1990, Blow et al. 1991) and others treated squeezing in solitons (Drummond
and Carter 1987, Shelby et al. 1990, Lai and Haus 1989), an effect that was verified
experimentally (Rosenbluh and Shelby 1991). Blow et al. (1991) have shown the
divergence of nonlinear phase shift, which Abram and Cohen (1994) treat through
the process of renormalization.
Let us review the basic features of the quantization of the electromagnetic field
in a Kerr medium and discuss the relevance of the renormalization procedure to the
treatment of divergences of effective medium theories. We consider a transparent,
homogeneous isotropic, and dispersionless dielectric medium that exhibits a nonlin-
ear refractive index. We examine the situation similar to Abram and Cohen (1991).
The Hamiltonian for the electromagnetic field in a Kerr medium is

1 2  3
Ĥ (t) = B̂ (z, t) +  Ê 2 (z, t) + χ Ê 4 (z, t) dz, (3.494)
2 4

where the integration along the direction of propagation z is denoted explicitly, but
integration over the transversed directions x and y will be implicit. We use the
Heaviside–Lorentz units for the electromagnetic field without  = c = 1, χ is the

174 3 Macroscopic Theories and Their Applications

nonlinear (third-order) optical susceptibility. From the perspective of the substitu-
tion of (3.25), the Hamiltonian (3.494) can be written as (Hillery and Mlodinow

1 1 1χ 4
Ĥ (t) = B̂ 2 (z, t) + D̂ 2 (z, t) − D̂ (z, t) dz, (3.495)
2  4 4

where the displacement field

D̂(z, t) =  Ê(z, t) + χ Ê 3 (z, t). (3.496)

The canonical equal-time commutators are (cf. (3.28))

[ Â(z 1 , t), D̂(z 2 , t)] = −icδ(z 1 − z 2 )1̂, (3.497)
[ B̂(z 1 , t), D̂(z 2 , t)] = −icδ  (z 1 − z 2 )1̂. (3.498)

It is convenient to adopt a slowly varying operator picture in which the zeroth-
order dynamics of the field governed by the linear medium Hamiltonian are already
taken into account exactly, while the optical nonlinearity can be treated within the
framework of perturbation theory. In such a picture, a field operator Q̂(z, t) evolving
inside a nonlinear medium is related to the corresponding linear medium operator
Q̂ 0 (z, t) by

Q̂(z, t) = Û −1 (t) Q̂ 0 (z, t)Û (t). (3.499)

The unitary transformation Û (t) is given by

Û (t) = T exp − Ĥ1 (τ ) dτ , (3.500)

where T ≡ T denotes the time ordering and

Ĥ1 (t) = − D̂04 (z, t) dz (3.501)
4 4

is the interaction Hamiltonian. The full nonlinear Hamiltonian (3.495) can be writ-
ten as follows:

Ĥ0 (t)
; <= >
H ( D̂0 , B̂0 ) = H0 ( D̂0 , B̂0 ) + Ĥ1 (t), (3.502)

where Ĥ0 (t) is the linear medium Hamiltonian. Following the traditional modal
approach to relation (3.499), Kitagawa and Yamamoto (1986) developed a single-
mode treatment of the self-phase modulation. Clearly, such an investigation is valid

3.4 Optical Nonlinearity and Renormalization 175

only inside an optical cavity with a sparse mode structure. In this situation, the time
evolution cannot be interpreted as space progression. In developing a travelling-
wave theory for self-phase modulation, Blow et al. (1991) obtained a solution simi-
lar to the single-mode solution. They encountered a nonintegrable singularity upon
normal ordering, a thing what is termed in quantum field theory as “ultraviolet”
divergence. To avoid an infinite nonlinear phase shift due to the Kerr interaction so
simply described, Blow et al. (1991, 1992) introduced a finite response time for the
nonlinear medium as regularization in the Heisenberg picture. Alternatively, Haus
and Kärtner (1992) considered the group velocity dispersion for the propagation of
pulses in the medium as a regularization.
At any rate, the regularization is a subsequent sophistication of the simple model
as known in classical theory. The response time and the group velocity dispersion
are necessary ingredients of a complete description of the propagation of the elec-
tromagnetic excitation in fibres. But they have no influence on the effects associated
with the vacuum fluctuations under study. A systematic way of dealing with the
vacuum fluctuations in quantum field theory is the procedure of renormalization
(Itzykson and Zuber 1980, Zinn-Justin 1989). The renormalization is known also
in classical field theory. In order to obtain finite results, the procedure of renormal-
ization redefines all the quantities that enter the Hamiltonian. The renormalization
point of view is that the new Hamiltonian is the only one we have access to. It
contains the observable consequences of the theory and the parameters are the ones
we obtain from experiments. The bare quantities are only auxiliary parameters that
should be eliminated exactly from the description (Stenholm 2000). The re-defined
(renormalized) quantities are able to incorporate the (infinite) effects of the vacuum
fluctuations. We will provide the definitions of broad-band electromagnetic field
operators and treat the propagation of light in a linear medium. The normal ordering
is considered as the simplest renormalization, e.g. in the case of the effective linear

1 1 2
Ĥ0 (t) = B̂0 (z, t) + D̂0 (z, t) dz.

To this end, the Hamiltonian Ĥ0 (t) is to be written in terms of the creation and
annihilation operators. The normal ordering allows us to subtract the vacuum-field
energy up to the first order from the effective Kerr Hamiltonian (3.495). However,
when this Hamiltonian is used to describe propagation disregarding the richness of
the quantum field theory, the normal ordering gives rise to additional divergences
that can be attributed to the participation of the vacuum fields. Upon renormaliza-
tion, involving also the refractive index, the divergences are removed.
As the Kerr nonlinearity involves the fourth power of the derivative ∂t∂ Â(z, t), it
cannot be in the general case renormalized to all orders with a finite number of cor-
rections. Inspired by nonlinear optics, the slowly varying amplitude approximation
decouples counterpropagating waves and the renormalization to all orders becomes
possible. All types of optical nonlinearity χ (n) give rise to divergences which require
the renormalization. In the treatment of parametric down conversion (Abram and

176 3 Macroscopic Theories and Their Applications

Cohen 1991), the problem of divergences and the need of normalization were not
In Abram and Cohen (1994), the broad-band electromagnetic field operators are
defined and the propagation of light in a nonlinear medium is treated. In the absence
of the optical nonlinearity, χ = 0 and the linear-medium displacement field has the
usual proportionality relationship to the electric field,

D̂0 (z, t) =  Ê 0 (z, t). (3.504)

The magnetic field and the displacement field in the linear medium obey the equal-
time commutation relation

[ B̂0 (z 1 , t), D̂0 (z 2 , t)] = −icδ  (z 1 − z 2 )1̂. (3.505)

The operators V̂0± (cf. (3.80), (3.81) of Abram and Cohen (1991)) reappear as the

ψ̂+ (z, t) = √ V̂0+ (z, t), (3.506)
ψ̂− (z, t) = √ V̂0− (z, t). (3.507)

For ψ̂± (z, t), the equations of motion in the Heisenberg picture may be calculated
by the use of commutator (3.505) as follows:

∂ i   ∂
ψ̂± (z, t) = Ĥ0 , ψ̂± (z, t) = ∓v ψ̂± (z, t), (3.508)
∂t  ∂z

where v = √c is the speed of light inside the dielectric exhibiting the refractive
index . Their solutions are

ψ̂+ (z, t) = ψ̂+ (z − vt, 0), (3.509)
ψ̂− (z, t) = ψ̂− (z + vt, 0). (3.510)

The equal-time commutators of the copropagating field operators can be obtained
from the definition and commutator (3.505) as follows:

[ψ̂+ (z 1 , t), ψ̂+ (z 2 , t)] = −ivδ  (z 1 − z 2 )1̂, (3.511)
[ψ̂− (z 1 , t), ψ̂− (z 2 , t)] = ivδ  (z 1 − z 2 )1̂. (3.512)

For the counter propogating fields, the corresponding operators commute with each

[ψ̂+ (z 1 , t), ψ̂− (z 2 , t)] = 0̂. (3.513)

3.4 Optical Nonlinearity and Renormalization 177

Operators (3.506) and (3.507) permit us to express the linear medium Hamiltonian
(3.503) as

1  2 
Ĥ0 (t) = ψ̂+ (z, t) + ψ̂−2 (z, t) dz, (3.514)
thus separating it into a sum of two mutually commuting partial operators, one for
each direction of propagation.
In the homogeneous medium it is possible to separate the electromagnetic field
operators ψ̂± (z, t) into positive- and negative-frequency parts

ψ̂± (z, t) = φ̂± (z, t) + φ̂± (z, t), (3.515)

defined as

1 i ψ̂± (z  , t) 
φ̂± (z, t) = ψ̂± (z, t) ± V.p. 
dz , (3.516)
2 π −∞ z − z

where V.p. denotes the Cauchy principal value. The operators φ̂± (z, t) and φ̂± (z, t)
can be considered as creation and annihilation operators, respectively, for a right (or
left)-moving electromagnetic excitation which at time t is at point z. The equal-time
commutators of φ̂± (z, t) are somewhat complicated,

† v ∂ 1
[φ̂± (z 1 , t), φ̂± (z 2 , t)] = P ∓ iπ δ(z 1 − z 2 ) 1̂,
2 ∂z 1 z1 − z2

[φ̂+ (z 1 , t), φ̂− (z 2 , t)] = 0̂, (3.517)

where P refers to the familiar generalized function P 1z . Nevertheless, an impor-
tant simplification results when only unidirectional propagation is considered. On
introducing the operators

D̂0 (z, t) = [φ̂+ (z, t) + φ̂− (z, t)], D̂0(−) (z, t) = [ D̂0(+) (z, t)]† , (3.518)
B̂0(+) (z, t) = √ [φ̂+ (z, t) − φ̂− (z, t)], B̂0(−) (z, t) = [ B̂0(+) (z, t)]† (3.519)

and considering the relations

D̂0 (z, t) = [ψ̂+ (z, t) + ψ̂− (z, t)],
B̂0 (z, t) = √ [ψ̂+ (z, t) − ψ̂− (z, t)] (3.520)

and relation (3.515), we verify that

D̂0 (z, t) = D̂0(+) (z, t) + D̂0(−) (z, t),
B̂0 (z, t) = B̂0(+) (z, t) + B̂0(−) (z, t). (3.521)

178 3 Macroscopic Theories and Their Applications

Using the new operators, the equal-time commutation relation (3.505) can suit-
ably be modified as
[ B̂0(+) (z 1 , t), D̂0(−) (z 2 , t)] = − cδ  (z 1 − z 2 )1̂. (3.522)
For a right-moving electromagnetic excitation, we observe that
√ (+)
φ̂+(+) (z 1 , t) = 2 B̂0(+) (z 1 , t),

2 (+)
φ̂+(+) (z 2 , t) = D̂ (z 2 , t), (3.523)

where the subscript (+) refers to k > 0. Using (3.523), we obtain that

[φ̂+ (z 1 , t), φ̂+ (z 2 , t)](+) = −ivδ  (z 1 − z 2 )1̂(+) . (3.524)

Similarly, for a left-moving electromagnetic excitation, we note that
√ (+)
φ̂−(−) (z 1 , t) = − 2 B̂0(−) (z 1 , t),

2 (+)
φ̂−(−) (z 2 , t) = D̂ (z 2 , t), (3.525)

where the subscript (−) refers to k < 0. From this

[φ̂− (z 1 , t), φ̂− (z 2 , t)](−) = ivδ  (z 1 − z 2 )1̂(−) . (3.526)

The electromagnetic creation and annihilation operators allow us to speak of the
normal order, for instance, when we write Hamiltonian (3.514) in the form

. t) dz. (3.528) and † † F̂+ (z̄.527) We can define annihilation and creation wave-packet photon operators F̂+ (z̄. t) + φ̂− (z. t) = F(z − z̄)φ̂+ (z. t) dz. (3. t)φ̂+ (z.529) respectively. (3. † † Ĥ0 (t) = φ̂+ (z. Here F(z) is a complex function: 1 F(z) = √ exp(iK z) F̃(z). t) dz (3.530) vK where v K is the central (carrier) frequency and F̃(z) is the wave-packet envelope function peaked at z = 0 and k = 0. t)φ̂− (z. t) = F ∗ (z − z̄)φ̂+ (z.

3. we obtain that the operators F̂+ and F̂+ follow the boson commutation relation . (3.531) † where F  denotes the spatial derivative.4 Optical Nonlinearity and Renormalization 179 On the usual assumption of narrow bandwidth and iv F  (z)F ∗ (z) dz = 1.

(1984). (3.532) Let us remark that the commutation relation (3. t) = 1̂. † F̂+ (z̄. F̂+ (z̄. t). A coherent state appropriate to ρ F is defined as  . Now a coherent pulse can be considered whose shape is described by ρ F(z) with a scaling factor ρ. where the formalism of the counterdirectional coupling was derived or rather this pitfall underestimated.532) is relation (A5) in Milburn et al.

it obeys the approximate quantum field eigenvalue equation √ φ̂+ (z. When we examine right-moving pulses.535) has kindled the interest in the Glauber factorization conditions and the theory of coherence (see also Ledinegg (1966)). (3.524) can be modified. (3. t)|ρ F =  v K ρ F̃ ∗ (z − z̄) exp[−iK (z − z̄)]|ρ F .528).529) can now be written as F̂(η̄) = F(η − η̄)φ̂(η) dη. (3. at the same time.538) .534) and. t)|ρ F = ρ|ρ F (3.537) dropping the subscript +.536) and simplify relation (3. t) = φ̂(η.509). we can introduce moving-frame coordi- nate η = z − vt (3. (3. φ̂+ (z.535) The approximation made in the derivation of (3. Similarly. 0) ≡ φ̂(η). The right-moving narrow-bandwidth wavepacket operators (3. (3.533) It satisfies the “single-mode” eigenvalue equation F̂+ (z̄. the commutation relation (3. † |ρ F = exp ρ F̂+ − F̂+ |0 . whenever we use the coordinate η explicitly.

In this case.520). For a Kerr medium. An important feature of these operators is that their commutator is a δ function † [âk0 (z 1 ). t) = F̂(η̄. the interaction Hamiltonian Ĥ1 (t) is expressed as χ Ĥ1 (t) = − 4 D̂04 (z. the interaction Hamiltonian can be written as χ  4 Ĥ1 (t) = ψ̂+ (z − vt) + ψ̂− (z + vt) dz. .537).539) In the moving-frame representation F̂(η̄. (3. The slowly varying amplitude approximation according to Abram and Cohen (1991) is used in Abram and Cohen (1994).545) 16 2 are the parts of the Hamiltonian (3. (3. (3.495) may be written up to the first order in χ as Ĥ (t) = Ĥ0 (t) + Ĥ1S+ (t) + Ĥ1S− (t) + O(χ 2 ). 0). no coordinate transformation has been explained. (3.541) Abram and Cohen (1994) have analysed the approximations that enter the quan- tum treatment of propagation in a Kerr medium and outline the corresponding renor- malization procedure. (1991) and Shirasaki and Haus (1990). Drummond (1990).540) Under the same narrow-bandwidth condition.543) that commute with Ĥ0 (t). (3.542) 4 According to (3. the transformation leaves the time coordinate unchanged. It is feasible to find a connection with the approaches leading to narrow-band field operators (â) contained in Shirasaki and Haus (1990). can be approximated by δ  (z 1 − z 2 ) ≈ −ik0 δ(z 1 − z 2 ).544) where χ Ĥ1S± (t) = − ψ̂±4 (z ∓ vt) dz (3. In terms of application of (3. (3. âk0 (z 2 )] = vk0 δ(z 1 − z 2 )1̂. (1990) and used in papers by Blow et al. and Blow et al.180 3 Macroscopic Theories and Their Applications where η̄ = z̄ − vt and F̂ † (η̄) = F ∗ (η − η̄)φ̂ † (η) dη. t) dz. the commutator of the Abram–Cohen operators.543) 16 2 The exact Hamiltonian (3. which is a delta function derivative.

3. t)φ̂(z.543) may be approximated by its first-order stationary component Ĥ1S . Z is a function we give only asymptotically (cf.551) where κ = 3χ 4 2 . Ĥ1S+ will be referred to as the slowly varying amplitude Hamiltonian. Now we turn to the renormalization. φ̂(η. Abram and Cohen 1994) v 2 Z  Λ. Ĥ1S = Ĥ1S+ + Ĥ1S− . the perturbative Hamiltonian (3. In the framework of the rotating-wave approximation. we obtain that χ   Ĥ1S+ = − 6 φ̂ † φ̂ † φ̂ φ̂ S dz. t) .546) ∂t  This first-order approximation to the equation of motion can be solved formally using the corresponding time-evolution operator (cf.552) Λ→∞ 2π . t)φ̂(z. (3. (3. (3. t) dz − φ̂ † (z.4 Optical Nonlinearity and Renormalization 181 In view of approximation (3. t) = Ĥ1S+ (t).544) for the electromag- netic field in a Kerr medium can be written as κ Ĥ (t) = φ̂ † (z. (3.548) Therefore.544).549) 16 2 where   6 φ̂ † φ̂ † φ̂ φ̂ S = φ̂ † φ̂ † φ̂ φ̂ + φ̂ † φ̂ φ̂ † φ̂ + φ̂ † φ̂ φ̂ φ̂ † + φ̂ φ̂ † φ̂ † φ̂ + φ̂ φ̂ † φ̂ φ̂ † + φ̂ φ̂ φ̂ † φ̂ † . the equation of motion of a right-moving field operator can be written in the interaction picture as follows: ∂ i   φ̂(η.550) Upon the normal ordering. (3. (3.499)) t  i Û S+ (t) = T exp − Ĥ1S+ (τ ) dτ . t)φ̂(z. t) dz. t) dz 2 − κ Z φ̂ † (z. t)φ̂(z. (3. (3. t)φ̂ † (z.547)  −∞ The classical slowly varying approximation has its quantum counterpart on a double assumption: (1) the initial state of the field is a narrow-bandwidth state and (2) the nonlinearity is weak enough so that the full nonlinear Hamiltonian (3.

etc.553) The third term in equation (3. At the same time. (3. (3. In particular.551) are familiar. a formal series (in ) of “counterterms” is added to the Hamiltonian in order to remove the divergences that arise upon nor- mally ordering the results of calculation (Itzykson and Zuber 1980).557) with † Ĥ0R (t) = φ̂R (z.560) 2 . For Λ fixed. t)φ̂R (z. t)φ̂R (z. t) dz. (3.R (t) = ( j + 1)Z j φ̂R (z. t)φ̂R (z. a renormalized Kerr Hamiltonian ĤR (t) may be defined by introducing a counterterm of order  as follows: ĤR (t) = Ĥ (t) + 2κ Z φ̂ † (z.558) ∞  ( j) Ĥ1S+. Whereas the first two terms in equation (3.R (t). (3. In the renormalization procedure. t) dz. (3. or at least whose Hermitian parts. The Hamilto- nian itself exemplifies that it is not sufficient for removing divergences. would relate to an experiment.R (t) =  j κ j+1 Ĥ1S+. the third term. t)φ̂R (z. t) (3. t). arises in the normal ordering proce- dure. which is divergent.555) are further quantities which.554) † √ φ̂R (z.182 3 Macroscopic Theories and Their Applications where Λ is a high-frequency cutoff. a renormalized refractive index is defined n nR = .551) exchanges the sign and for Λ → ∞ it is an infinite change in the inverse of the refractive index. (3. t) dz. (3. t)φ̂(z. The renormalized field operators √ φ̂R (z. Such a relationship is no more required from the bare quantities.556) 1 + κ Z The renormalized Kerr Hamiltonian can be written in terms of the renormalized field operators as ĤR (t) = Ĥ0R (t) + Ĥ1S+. but at the same time renormalized parameters and renormalized field operators are introduced.R (t). t) = 1 + κ Z φ̂ † (z. this last term vanishes if  → 0. t) = 1 + κ Z φ̂(z.559) j=0 where ( j) (−1) j+1 † † Ĥ1S+.

This set-up measures simultaneously the expecta- tion values of the operators M̂0 (η) and M̂ π2 (η). we have neglected the Kerr medium here. The detection of a light pulse by a bal- anced homodyne detector can be expressed in a moving frame through the measured quantum operator + M̂θ (η) = exp(iθ ) K LO FLO (η̄ − η)φ̂(η) + H.561) where FLO (η) is the coherent amplitude of the local oscillator (LO) pulse peaked at η = η̄ and θ is the phase difference between the local oscillator and signal pulses.565) where Δ M̂θ (η) = M̂θ (η) − ρ F| M̂θ (η)|ρ F (3. Abram and Cohen (1994) have calculated the quantum noise properties of a coherent pulse undergoing self-phase modulation in the course of its propagation by eliminating the vacuum divergences through the renormalization procedure. K LO = K .R (t) is the “usual” Kerr term and  j κ j+1 Ĥ1S+. The instantaneous intensity of the signal pulse peaked at η = 0 is introduced as I (η) = v Kρ 2 F̃ ∗ (η) F̃(η). The nonlinear phase Θ has exactly the same value as in classical nonlinear optics. in order to obtain the quantum noise spectrum S0 (k) = exp[ik(η1 − η2 )] ρ F|Δ M̂θ (η1 )Δ M̂θ (η2 )|ρ F dη2 dη1 . For simplicity. z) has the expectation value in the coherent state F ∗ (η. z) = κv K z I (η) (3.c.563) where Θ(η. z)]. the central frequency of the local oscillator is the same as that of the incident pulse.. In homodyne detection. The fact that equa- tion (3.3. z) = φ̂(η. (3.564) is the nonlinear phase shift produced by self-phase modulation.4 Optical Nonlinearity and Renormalization 183 (0) ( j) here κ Ĥ1S+. t = vz ) instead of φ̂(η).563) obtained by invoking renormalization corresponds directly to what is observed experimentally in a propagative configuration underlines the validity of this approach. (3.566) . z) = ρ F|φ̂(η.562) It can be obtained that the operator φ̂(η. but it is included when we write φ̂(η. z)|ρ F = v Kρ F ∗ (η) exp[−iΘ(η. Two-point correlation functions were examined. The one-point averages were first determined.R (t) is the jth quantum correction. (3. (3.

(3. v) = α| exp i [u M̂ 2 (η) + v M̂0 (η)] |α . z)] dη.573) 4π 2 According to equation (3. It is appropriate to give a physical interpretation of the above results and to discuss the case of squeezing that can be observed in the propagation of a coher- ent pulse. z)] dη. (3. Using the linked cluster theorem.572) where |α is the continuous-wave coherent state. (3. v) exp[−i( pu + qv)] du dv. the characteristic function (3. For narrow-bandwidth signals. This result is similar to that obtained by linearizing the self-phase modula- † tion exponential operator exp(iγ âk0 âk0 ) around the mean field (Shirasaki and Haus 1990). z)][1 + iΘ(η. the phase properties of quantum noise in equation (3. In exper- iment. z)] exp[2iΘ(η.572) can be written in terms of connected averages. z) dη. π (3. (3. such a noise is detected at frequencies several orders of magnitude below the carrier optical frequency and its spectrum is considered to be flat throughout the typical bandwidth. z)][1 − iΘ(η.184 3 Macroscopic Theories and Their Applications and k is the spatial frequency at which the quantum noise is measured. z)] dη. First.574) . (3.570) S4 = ILO (η̄ − η)[−iΘ(η. the quantum characteristic function is defined as   C(u. p) = C(u. (3.571) where ILO (η̄ − η) is the local oscillator instantaneous intensity of the local oscillator pulse.568) S2 = ILO (η̄ − η)[1 + Θ2 (η. z)] exp[−2iΘ(η.563). the expectation value of the Wigner distribution is √ q 0 + i p0 = I eiθ . (3.569) S3 = ILO (η̄ − η)[iΘ(η. Two lowest order connected averages are feasible and describe the moments of the Wigner distribution 1 W (q.567) with S1 = ILO (η̄ − η)Θ2 (η.567) can be visualized by examining the field fluctuations. The low-frequency noise spectrum Sθ (0) can be decomposed into four terms Sθ (k) ≈ Sθ (0) = S1 + S2 + exp(−2iθ)S3 + exp(2iθ )S4 .

4 Optical Nonlinearity and Renormalization 185 while the principal squeeze variances of the quantum noise are 1 2(BS ∓ |C|) = √ (3. (2003) have studied the propagation of ultrashort pulsed beam beyond the paraxial approximation in free space. (3. both with and without the second space derivative term. Bespalov et al. 3π arg C = 2Θ + arctan(Θ) − .578) 2 In the Abram–Cohen theory. (2002) have investigated the propagation of light in the (1 + 1)- dimensional approximation. to the scalar wave equation is common. the Gaussian noise is a good approximation up to nonlinear phase shifts of the order 103 rad. |C| = Θ. Lu et al. the quadrature will not always be the principal quadrature due to the chirp. A comparison with rigorous nonparaxial results obtained by numerical method is carried out.3. it does not exclude that a quantization of the field and of the equations will be necessary in the near future. When the phase of the local oscillator is constant along the pulse profile. Blow et al. Bergman and Haus 1991). The nonparaxial corrections to an arbitrary paraxial solution are given in a series form. Restriction to the transverse components and. They have paid attention to the two series expansions of the index of refraction of an isotropic optical medium in Born and Wolf (1968). finally.576) where Θ ≡ Θ(η. (3. the use of a matched local oscillator has been proposed such that its phase Θ(η) varies in a way that matches the signal chirp. bandwidth 100 GHz. 1992. Spatial and tempo- ral distributions are considered. Although this analysis is completely classical.575) 1 + Θ2 ± Θ + = 1 + Θ2 ∓ Θ. the Sagnac interferometer can be used for this purpose (Shirasaki and Haus 1990. Alternatively. On these expansions they have based two wave equations. z) and we have used the characteristics of quantum noise (Peřinová et al. and intensity 1 W propagating in a silica fibre. the variance will not be the same everywhere. for the case of a coherent beam of central frequency 2×1015 Hz. a local oscillator pulse that is much shorter than the signal pulse can sample only the central portion of the signal in order to measure the appropriate squeezing. To circumvent this problem. They have presented a method to derive the non- linear wave equations suitable for describing dynamics of extremely short pulses. (3. 1991) 1+ 1 BS = 1 + Θ2 . . particu- larly when the principal quadrature is measured at the peak of the pulse. Besides the Kerr effect.577) 2 2 Moreover.

Tatarinova and Garcia (2007) set the problem in the framework of the classical nonlinear optics and so the renormalization is not needed. Dalton et al. (2004) have considered the nonlinear propagation of randomly distributed intense short photon pulses in a photon gas. They have provided a quantum description of the transition radiation emitted by a charged particle in passing from one dielectric medium to another. and linear dielec- tric media. Fragmentation of incoherent photon pulses in astrophysical contexts and in forthcoming experiments using very intense short laser pulses has been predicted. arbitrary in other respects. 3. the appropriate. Irrespective of these and some other approximations. The quasimode functions approximate the true mode functions of a classical optics device when they are obtained on the assumption of an ideal electric permittivity function and the permittivity function describing the device does not deviate much from the ideal one. The quantum Hamiltonian has been derived in a generalized multipole form. The authors have shown that plane-wave photons can be related to the normal-mode ones within the framework of scatter- ing theory. (1999b) have presented a macroscopic canonical quantization of the electromagnetic field and radiating atom system in dielectric media based on expanding the vector potential in terms of quasimode functions. They have considered excited atoms and changes in the spontaneous emission rates for both electric and magnetic dipole transitions of the atoms within or near dielectric media. Dalton et al. In the . where n 0 is the linear refractive index. Kovalev et al. lossless. The renormalization and the Bogoliubov renormalization group are different con- cepts (Shirkov and Kovalev 2001). Shukla et al. They have used the quantization schemes discussed to determine the fluctuation properties of various field components. n(I ) is such that n(0) = 0. The procedure of analytical solution begins with an approximate transformation of the nonlinear Schrödinger equation onto eikonal equations. easy. dispersionless. (2000) and Tatarinova and Garcia (2007) have expounded the renormalization-group approach to the problem of light- beam self-focusing. The topics treated have included the normal-mode expansion and the plane-wave expansion. is studied.186 3 Macroscopic Theories and Their Applications In this book the “fully” relativistic quantum electrodynamics is not treated. The case of nonlinear self-focusing accompanied by multipho- ton ionization has been explicitly analysed.5 Quasimode Theory Glauber and Lewenstein (1991) have developed quantum optics of inhomogeneous media with linear susceptibilities. (1996) have carried out canonical quantization of the electromag- netic field and radiative atoms in passive. calculation provides results which are in good agreement with numerical simulations. Nev- ertheless. its importance for quantum optics is going to be appreciated in the nearest future. The propagation of a laser beam of intensity I in a nonlinear medium with a refractive index n 0 + n(I ). Plane waves in Glauber and Lewenstein (1991) are such quasimodes.

In a generalization of Helmholtz’s theorem (Dalton and Babiker 1997).5 Quasimode Theory 187 coupled-mode theory (see. t) = (R) . Here we present part of the theory (Dalton et al. t)]2 . t) d3 R.583) 2 ∂t 2μ(R) The conjugate momentum field Π(R. (3. t) can be decomposed uniquely in the generalized transverse and longitudinal components F() () ⊥ (R.586) The true mode functions satisfy the generalized Helmholtz equation 1 ∇× [∇ × Ak (R)] = ωk2 (R)Ak (R). (3. t) 2 1 Lc (R. F (R. t). t)] = 0 (3. t)]2 H  (t) = + d3 R. (3.g.2) the “ideal” waveguide modes are also such quasimodes. t) is given by the relation  1 ∂A(R. t) + F (R. ∇ × F (R. Section 6. It is assumed that a classical linear optics device is described with the spa- tially dependent electric permittivity (R) (and the magnetic permeability μ(R)). t) = 0. The generalized Coulomb gauge condition for the vector potential A(R. a vector field F(R. t) = −Π(R. t). D(R. t) is obtained from the Lagrangian density Lc (R.579) is used. (3. t) as ∂A(R.585) 2(R) 2μ(R) We have the electric displacement field D(R. t).580) with ∇ · [(R)F() () ⊥ (R.584) ∂t The Hamiltonian is   [Π(R. (3. (3. t) = F() () ⊥ (R. t) in the form F(R. e.581) The macroscopic Lagrangian is given by the relation L  (t) = Lc (R. (3.3. t) Π(R. t)] = 0. t) ∇ · [(R)A(R. t) = (R) − [∇ × A(R. t). t)]2 [∇ × A(R.582) where the Lagrangian density Lc (R. 1999b). (3. which will be completed below.587) μ(R) .

t) = pk (t)(R)Ak (R).597) . (3. t)] = 0.596) which are the generalized Helmholtz equation. respectively. Let λα denote the angular frequency of the quasimode. t) d3 R. As the vector potential A(R. the field (R) ˜ (R) A(R.595) ˜ (R)U∗α (R) · Uβ (R) d3 R = δαβ . the gauge conditions.579). t) d3 R. (3. t) fulfils the generalized Coulomb gauge condition ∇ · [˜ (R)F(R. (3. (3. idealized ver- sions of the true mode functions Ak (R).594) μ̃(R) ∇ · [˜ (R)Uα (R)] = 0.591) These variables are complex. (3. It is assumed that these functions produce quasimode functions Uα (R). (3. One has equations 1 ∇× [∇ × Uα (R)] = λ2α ˜ (R)Uα (R).593) k It is assumed that the exact electric permittivity and magnetic permeability functions do not deviate much from artificially chosen functions ˜ (R). (3. and orthonor- mality conditions. t) satisfies the generalized Coulomb gauge condition (3. (3. μ̃(R). Expansions of the vector potential and conjugate momentum field in terms of the true modes Ak (R) are  A(R. (3.589) Generalized coordinates qk (t) and generalized momenta pk (t) can be introduced by the relations qk (t) = (R)A∗k (R) · A(R.588) The true mode functions satisfy the orthogonality and normalization conditions respecting (R) as a weight function. (3.592) k  Π(R.188 3 Macroscopic Theories and Their Applications where ωk are real and positive angular frequencies and satisfy the generalized Coulomb gauge condition ∇ · [(R)Ak (R)] = 0.590) pk (t) = A∗k (R) · Π(R. (R)A∗k (R) · Al (R) d3 R = δkl . t) = qk (t)Ak (R).

(3.β 2 α. (3.607) μ(R) (R) α (R) Making a specific choice for K (Dalton et al. (3.600) ˜ (R) α. and the scalar potential Φ̃(R. must be specified (Glauber and Lewenstein 1991).β (R) Using expansion (3. T (3. Therefore.599) need not be considered. can be chosen in the form ˜ (R) A(R. t) = Q α (t)K αβ Uβ (R).608) . t) = ψ̇(R.5 Quasimode Theory 189 (R) where F(R. (3. The vector potential is given as  ˜ (R) A(R. where the complicated form of the coefficients α Q α (t)K αβ may and may not involve Q α (t) = ˜ (R)U∗α (R) · Ã(R. t) d3 R (3. t).β α where W = (KT )−1 M−1 (K∗ )−1 . t) (3. the expansion (R)  A(R. Another choice is based on the gauge transformation Ã(R. (3.583) as 1 ∗ 1 ∗ L  (t) = Q̇ α (t)(W−1 )αβ Q̇ β (t) − Q (t)Vαβ Q β (t). an arbitrary function of R and t.605) and ˜ (R) ∗ ˜ (R) Mαβ = (R) Uα (R) · Uβ (R) d3 R. (3. t). t) = Q α (t)K αβ Uβ (R) (3. t) = A(R. t).603) 2 α.604) ∗ V = K HK .606) (R) (R)   1 ˜ (R) ∗ ˜ (R) Hαβ = ∇× U (R) · ∇ × Uβ (R) d3 R.602) α. t) − ∇ψ(R. (3. any field. (3.601) and matrix elements K αβ of a suitable matrix K. t).β C exists.3. 1999b) K = (M∗ )−1 .602).598) where ψ(R.582). one can write the Lagrangian (3.

t) (Π̂(R. the conjugate momentum field Π(R. P̂β (t)] = iδαβ 1̂ = [ Q̂ †α (t).609) −1 −1 V = M HM .612) 2 α.616) The nonzero equal-time commutators are † [ Q̂ α (t). t) = Q α (t)(M−1 )βα Uβ (R). t) is still expressed by equation (3.612) now becomes a quantum Hamiltonian Ĥ  (t). Similarly. Pα∗ (t) → P̂α† (t).615) ((3. (3.584) does not depend on the gauge condition (3.611) β The Hamiltonian is given by the relation 1 ∗ 1 ∗ H  (t) = Pα (t)Wαβ Pβ (t) + Q (t)Vαβ Q β (t).β α As the same Lagrangian is used and definition (3.190 3 Macroscopic Theories and Their Applications one has W = M. (3. . t)) is introduced.β 2 α. Q ∗α (t) → Q̂ †α (t).602) ((3.614)) when the quantum operator Â(R. (3. Let us imagine the replacements (3. P̂β (t)]. we obtain that  Π(R.614) α The classical generalized coordinates Q α (t) and generalized momenta Pα (t) are replaced by quantum operators according to the prescriptions Q α (t) → Q̂ α (t). (3. the Hamiltonian given through equation (3. t) and conjugate momentum field Π(R.584) and exchanging α ↔ β.588).602) it follows that  ˜ (R) A(R. (3.617) The vector potential A(R.610) The generalized momentum coordinates Pα (t) for the electromagnetic field are  Pα (t) = (M−1 )αβ Q̇ β (t). (3.584). As from (3. Π̂(R. (3.613) α. t) now become field operators Â(R. t).616)) in the expression (3. t) = Pα (t)˜ (R)Uα (R). (3.615) Pα (t) → P̂α (t). (3.β (R) respecting (3. t).

(3. âk for the true modes (see Dalton et al.623) Wαα † The nonzero equal-time commutators are standard. Â−α (t).594). (1999b) have defined an effective quasimode angular frequency μα as follows: + μα = Wαα Vαα . (3.621) and (3. Defini- † tions (3. It is shown to involve a Bogoliubov transformation (Dalton et al. From Equations (3.621) 2 2ηα   η α 1 †α (t) = Q̂ †α (t) − i P̂ † (t).618) 2 α 1  †  V̂Q−Q (t) = P̂α (t)Wαβ P̂β (t) + Q̂ †α (t)Vαβ Q̂ β (t) . (3. (3. As usual with quantum harmonic oscillators.β α =β Dalton et al. (3.5 Quasimode Theory 191 It has the form Ĥ  (t) = ĤQ (t) + V̂Q−Q (t). (1996)) and the quantities just introduced can be obtained from the expansions of two sets of functions (R)Ak and ˜ (R)Uα (R) in terms of the other.621) and (3. anni- hilation and creation operators for each of the quasimodes are introduced as usual linear combinations   ηα 1 Âα (t) = Q̂ α (t) + i P̂α (t). [ Âα (t). Âβ (t)] = δαβ 1̂. † The relationship between the annihilation. (3. (1999c).622) one expresses the generalized coordinates and momenta Q̂ α (t).3.619) 2 α. where 1  †  ĤQ (t) = P̂α (t)Wαα P̂α (t) + Q̂ †α (t)Vαα Q̂ α (t) .622) can be completed with the equations for Â−α (t). creation operators âk .622) 2 2ηα α where  Vαα ηα = . where −α denotes that the operators are associated with the quasimode function Û∗α (R).620) It may differ from λα in (3. P̂α (t) as follows:   † .

(3. Q̂ α (t) = Âα (t) + Â−α (t) .624) 2ηα .

192 3 Macroscopic Theories and Their Applications  1 ηα † .

0 1   √ V αβ RWA V̂Q−Q (t) = ηα ηβ Mαβ + √ †α (t) Âβ (t). 0 1  √ Vα. (3. the Hamiltonians become  1  ĤQ (t) =  †α (t) Âα (t) + 1̂ μα . t) and Π̂(R.634) α. t) = K αβ Âα (t)Uβ (R) + K αβ †α (t)U∗β (R) .β ηα ηβ α =β Similarly. t) should be expressed in terms of annihilation and creation operators.624) and (3. (3. (3. (1999b) have considered also an approximate form of this Hamiltonian. P̂α (t) = Âα (t) − Â−α (t) .618) and (3. t) = ˜ (R) Âα (t)Uα (R) − †α (t)U∗α (R) . Dalton et al.  RWA V̂Q−Q (t) ≈  vαβ †α (t) Âβ (t).−β + √ †α (t) Âβ (t) + H.628) 2 α.631) α i 2  The quantum Hamiltonian in the rotating wave approximation is ĤRWA (t) = RWA ĤQ (t) + V̂Q−Q (t). One finds that    ˜ (R)   ∗ Â(R.c.625) into the Hamiltonians (3.β α =β . (3.630) α.−β † non−RWA V̂Q−Q (t) = − ηα ηβ Mα. (3.626) α 2 V̂Q−Q (t) = V̂Q−Q RWA (t) + V̂Q−Q non−RWA (t).625) i 2 On substituting (3.619).632) or μα ≈ λα + vαα .β ηα ηβ α =β non−RWA and V̂Q−Q (t) stands for the nonrotating wave correction term.β 2ηα (R)   1 ηα   Π̂(R. A simplification is related to the approximation μα ≈ λ α (3.627) RWA where V̂Q−Q (t) means the rotating-wave contribution. (3. Further. (3.633) where we have added a term. (3. the field operators Â(R.629) 4 α. (3.

(3.636) (R)    1 ˜ (R) ∗ (H1 )αβ = ∇× − 1 Uα (R) μ0 (R)    ˜ (R) · ∇× − 1 Uβ (R) d3 R.638) where the evolution operator Û (t) given by  i Û (t) = exp − Ĥ (0)t (3. We should denote more exactly Ĥ (t. (3. The state vector |ψ(t) evolves as |ψ(t) = Û (t)|ψ(0) . V̂ (t.635) 2 λα λβ 2 ˜ (R) (M1 )αβ = ˜ (R) − 1 U∗α (R) · Uβ (R) d3 R. t  ) are t-independent and the notation Ĥ0 (t  ). In spite of simplifications this grad- uation is present in quantum optics. t  ) is written as a sum of an unperturbed Hamiltonian Ĥ0 (t. but with wave functions replaced by field operators. In this theory the Hamiltonian Ĥ (t. First we will review basics of the Schrödinger picture approach to the scattering theory and then outline the Heisenberg picture approach to this theory (Dalton et al. t  ). (3. and Ĥ (t. t  ) and an interaction term V̂ (t. The methods developed apply also to optical and acoustic scattering in classical physics (Reed and Simon 1979).1 Relation to Quantum Scattering Theory The phenomenon of scattering occurs in various situations in optics. and Ĥ (t  ).639)  is unitary. So the formulation in the Schrödinger picture is appropriate. In quantum field theory. is used. . the scattering theory is a natural continuation and generaliza- tion of the analysis of collisions. 1999a). t) in the Heisenberg picture and Ĥ (t.637) (R) 3. In the classical particle mechanics. The first variable denotes the explicit time dependence of the Hamiltonian. For simplicity it is assumed that Ĥ0 (t. t  = 0. t  ). The single-channel scattering theory may be adequate for many applications in quantum optics. 0) in the Schrödinger picture.5. the scattering theory resembles its simplified form for quantum mechanics. (3. it is formulated in the Heisenberg picture. t  ). t. t. Accord- ingly.3. It must be and has been reconstructed for wave functions in quantum mechanics. V̂ (t  ).5 Quasimode Theory 193 where 0 1 1 + (H1 )αβ vαβ = λα λβ (M1 )αβ + + . t  = 0.

(3. Newton 1966)   Û0 (t)|ψin −∞. Newton 1966) that are sufficient for the asymptotic conditions to hold are that ( · · ·  are the norms of the state vectors) 0 V̂ Û0 (τ )|ψin  dτ < ∞ for a dense set of |ψin .643) 0 If the asymptotic conditions hold.647) but may not satisfy Ω̂Ω̂† = 1̂ (see below). − |ψ(t) → 0 as t → (3.642) −∞ ∞ V̂ Û0 (τ )|ψout  dτ < ∞ for a dense set of |ψout .194 3 Macroscopic Theories and Their Applications When a scattering experiment is described. The conditions (Taylor 1972. So they may not be unitary. (3. (3. .645) t→−∞   and Ω̂− = lim [Û † (t)Û0 (t)] t→+∞    i i = lim exp Ĥ (0)t exp − Ĥ0 (0)t .640)  we have the so-called asymptotic conditions (Taylor 1972. the state vector |ψ(t) should approach freely evolving state vectors as t → ±∞. So with the unitary free evolution operator Û0 (t) given by  i Û0 (t) = exp − Ĥ0 (0)t . which are based on the so-called input states |ψin and output states |ψout . (3.641) Û0 (t)|ψout +∞.646) t→+∞   The Møller wave operators are isometric.644) and are defined through the relation Ω̂+ = lim [Û † (t)Û0 (t)] t→−∞    i i = lim exp Ĥ (0)t exp − Ĥ0 (0)t (3. They satisfy (Ω̂ = Ω̂+ or Ω̂− ) the relation Ω̂† Ω̂ = 1̂. then the Møller wave operators Ω̂± exist which map |ψin and |ψout onto the state vector at t = 0: |ψ(0) = Ω̂+ |ψin = Ω̂− |ψout (3. (3.

(3. The Møller wave operators are not unitary if there exist bound energy eigenstates for the Hamiltonian Ĥ .3. Ŝ † Ŝ = Ŝ Ŝ † = 1̂.656) |ψout . (3. (3. Ŝ Ĥ0 (0) = Ĥ0 (0) Ŝ. (3.652) From this relation and its Hermitian conjugate it follows that Ŝ and Ĥ0 (0) commute. |ψI (∓∞) = (3.650) Mapping (3.5 Quasimode Theory 195 The scattering operator Ŝ maps the input vector |ψin onto the output vector |ψout . |ψin = Ŝ † |ψout . Namely equations (3.641) and (3. the unperturbed Hamiltonian is invariant under the unitary transfor- mation Ŝ or the unperturbed energy is conserved in a scattering process. (3. |ψout = Ŝ|ψin . (3.655) In physical systems the limits |ψI (∓∞) may exist.647) it is obvious that it involves two Møller operators † Ŝ = Ω̂− Ω̂+ .651) The Møller wave operators satisfy the important intertwining relation Ĥ (0)Ω̂± = Ω̂± Ĥ0 (0). Some of the previous results simplify in the interaction picture. (3.648) become  |ψin .653) In other words.648) and from equations (3. On this assumption the simpli- fication occurs.654) Λ̂ is called unitary deficiency and is a sum of all the projectors onto the bound states of Ĥ (0). (3.648) can be inverted.649) It could be easily verified that the scattering operator Ŝ is unitary. It can be derived that Ω̂Ω̂† = 1̂ − Λ̂. In this picture state vector is defined through the equation † |ψI (t) = Û0 (t)|ψ(t) .644) and (3.

662) dt where we have used both the intertwining relation and the equation V̂I (t) = ĤI (t) − Ĥ0I (t). (3. Ω̂I− (+∞) = 1̂.661) In the case of Ω̂I+ (t) equation (3. Ĥ0 (0)| f = ω f | f . In the Schrödinger picture. (3. Newton 1966) for the scattering operator. (3.  i ∞ Ŝ = T exp − V̂I (t1 ) dt1 . The formal solution of the problem consisting in this equation with the boundary conditions (3.660)   where we have used the intertwining relation. Ω̂I+ (−∞) = 1̂. (3. (3. the Møller wave operators are associ- ated with the interaction-picture operators Ω̂I± (t). (3.   † i i Ω̂± (t) = Û0 (t)Ω̂± Û0 (t) = exp I Ĥ0 (0)t exp − Ĥ (0)t Ω̂± .657) Schrödinger picture operators are transformed to interaction-picture ones via the equation † ÂI (t) = Û0 (t) ÂÛ0 (t). Ĥ0I (t)].659) dt † where Ĥ0I (t) = Û0 (t) Ĥ0 (0)Û0 (t).663)  −∞ where T means the time ordering. Ω̂I− (−∞) = Ŝ † .664) . Taking the limits as t → ±∞ one sees that (Taylor 1972. If this operator is time independent it is found that d ÂI (t) i = [ ÂI (t). scattering processes are often spoken of in terms of transitions between initial and final states that are eigenstates of Ĥ0 (0).196 3 Macroscopic Theories and Their Applications and |ψI (+∞) = Ŝ|ψI (−∞) . (3.659) becomes dΩ̂I+ (t) i = V̂I (t)Ω̂I+ (t). A possible time dependence of the operator  has not been designated. Especially.661) provides one with a Dyson expression (Taylor 1972.658) where  is any Schrödinger operator. (3. Newton 1966) Ω̂I+ (+∞) = Ŝ. So an initial state |i and a final state | f have the properties Ĥ0 (0)|i = ωi |i .

. G(t1 . b̂nin (tn )|ψin . (1999a) as follows: G(t1 . tn ) = ψout |b̂1out (t1 )b̂2out (t2 ) . So 2πi f | Ŝ|i = f |i − δ(ω f − ωi ) f |T̂ (ωi + i0)|i . It obeys the Lippmann–Schwinger integral equation T̂ (z) = V̂ (0) + V̂ (0)Ĝ 0 (z)T̂ (z). . where z is a complex energy variable. (3. .669) is the resolvent operator associated with Ĥ0 (0).673) . tn ) = ψ(0)|b̂1 (t1 )b̂2 (t2 ) .3. (3.666) where Ĝ(z) = [z 1̂ − Ĥ (0)]−1 (3. As conservation of the unperturbed energy holds.672) Therefore. we read relations (8) and (10) in Dalton et al.5 Quasimode Theory 197 where ωi and ω f are frequencies. b̂nout (tn )|ψout . † b̂kout (t) = Ω̂− b̂k (t)Ω̂− . .670) −∞  Glauber’s theory of photodetection assumes multitime quantum correlation func- tions of the form (Glauber 1965) G(t1 . . . . . . . 1999a) † b̂kin (t) = Ω̂+ b̂k (t)Ω̂+ . b̂n (tn )|ψ(0) . . (3.667) is the resolvent operator. . (3. −∞  ∞ 0 1 Ĥ0 (0) Ω̂− = 1̂ + Ĝ 0 (ω − i0)T̂ (ω − i0)δ ω1̂ − dω. . . . The Møller wave operators can also be related to the T̂ (z) operator. tn ) = ψin |b̂1in (t1 )b̂2in (t2 ) .665)  The T̂ (z) operator is defined through the relation T̂ (z) = V̂ (0) + V̂ (0)Ĝ(z)V̂ (0). (3. . . (3. (3.668) where Ĝ 0 (z) = [z 1̂ − Ĥ0 (0)]−1 (3. This fact is expressed using the transition operator T̂ (z) (Taylor 1972. the matrix element f | Ŝ|i is zero unless ω f = ωi .671) We will define the input and output operators through the relations (Dalton et al. . Newton 1966). So ∞ 0 1 Ĥ0 (0) Ω̂+ = 1̂ + Ĝ 0 (ω + i0)T̂ (ω + i0)δ ω1̂ − dω.

(1999b) have outlined quasimode theory of the lossless beam splitter and Dalton et al. In this case the quasimode functions are plane waves.678) The input and output operators are related to the interaction-picture operators for long times. † b̂kout (t) − Û0 (t)b̂k (0)Û0 (t) → 0̂ as t → +∞.680) n2 0V . μ̃(R) = μ0 everywhere. † † b̂kout (t) = Ω̂− b̂k (t)Ω̂− = Ω̂− (1̂ − Λ̂)b̂k (t)(1̂ − Λ̂)Ω̂− . (3. (3. The coordinate axes are chosen such that refractive index equals n for |z| > d2 and it equals unity for |z| ≤ d2 . (3. To obtain quasimodes they choose ˜ (R) = n 2 0 . we may write relation (3. b̂kin (t) − b̂kI (t) → 0̂ as t → −∞. (3.676) We may summarize that † b̂kin (t) − Û0 (t)b̂k (0)Û0 (t) → 0̂ as t → −∞.679) Dalton et al. b̂kout (t) − Ŝ b̂kI (t) Ŝ † → 0̂ as t → −∞. b̂kin (t) − Ŝ † b̂kI (t) Ŝ → 0̂ as t → +∞. b̂kout (t) − b̂kI (t) → 0̂ as t → +∞.675) in the form b̂kout (t) = Ŝ b̂kin (t) Ŝ † . They have found references to the true mode and quasimode theories of the beam splitter (see there). For the sake of quantization. They assume √ that the device consists of two trihedral pieces of glass of refractive index n (n > 2) separated by a thin air gap of width d. (1999d) have continued and extended it in relation to scattering theory. (3. they assume that the field is contained in a box of volume V = L 3 and the quasimode functions 1 Uα (R) = + eα exp(ikα · R).198 3 Macroscopic Theories and Their Applications Further from the relations † (1̂ − Λ̂)b̂k (t)(1̂ − Λ̂) = Ω̂+ b̂kin (t)Ω̂+ .675) Using definition (3. (3. (3.677) Let us recall that operators in the interaction picture are introduced as † b̂kI (t) = Û0 (t)b̂k (0)Û0 (t).649).674) it holds on substitution that † † b̂kout (t) = Ω̂− Ω̂+ b̂kin (t)Ω̂+ Ω̂− .

β α =β The wave vectors kα for the quasimodes are 2π kα j = να j .681) n For the treatment to be simple. ναy = νβy ≡ ν y (3. (3. . ναz are integers. On calculating the matrix elements (M1 )αβ . (3. νβz = ∓ν y . ναy . (1999d)) simplifies   1 (n 2 − 1) 2π vαβ = sin νy d 2 2L L  2 c √ (n − 1) (poln(α) = poln(β) =) × 2 (3. ναy > 0. y.684) α.685) L where ναx .3. (3. The interesting directions are represented by ναx = 0. the authors may consider only two directions of propagation: one along √12 (e y − ez ) and the other along √12 (e y + ez ). The angular frequencies are c λα = kα . j = x. The complete expression for the coupling constant (see Dalton et al. ναz = ∓ναy . (H1 )αβ one obtains that vαβ is zero unless ναx = νβx ≡ νx . In general.5 Quasimode Theory 199 where eα are polarization vectors and kα are wave vectors (eα · kα = 0).687) n 1 (poln(α) = poln(β) =⊥) for the appropriate values of α and β. On the simplifying √ assumption it follows that ναz = ∓ν y . (3.683) α  V̂Q−Q (t) =  vαβ †α (t) Âβ (t). ⊥) is the same. and λα = nc ν y 2π L 2 = λβ . the beam splitter is described by the Hamiltonian Ĥ  (t) = ĤQ (t) + V̂Q−Q (t).686) and the polarization type (. (3. z. So only the quasimodes of the same angular frequency are coupled.682) where the unperturbed Hamiltonian ĤQ (t) and the coupling Hamiltonian V̂Q−Q (t) in the rotating-wave approximation are given by the relations  ĤQ (t) =  μα †α (t) Âα (t).

The cavity region I lies between a perfect . with nonaxial wave vectors. The sums over ν y can be converted to integral over k y using the prescription  ∞ L → dk y . with an atom–cavity coupling constant g. Also we are afraid that the appropriate operators do not converge in the rotating- wave approximation when only two directions of the incident light are consid- ered.5. the mode functions of which are nonzero outside the cavity and zero inside. (3. The form of the travelling and trapped mode functions for this cavity is derived in Dalton and Knight (1999a) and the mode–mode coupling constants are calculated in Dalton and Knight (1999b). (3. The weak dependence of the coupling constants on the mode frequency difference demonstrates that the conditions for Markovian damping of the cavity quasimode are satisfied. whose mode function is nonzero inside the cavity and zero outside. and which have the same axial wave vector as the cavity quasimode. Also the atom can decay directly via the Markovian damping with a rate constant Γ0 to certain external quasimodes.200 3 Macroscopic Theories and Their Applications Sums over quasimodes α with the same νx .b) have given a justification of the standard model of cavity quantum electrodynamics in terms of a quasimode theory of macroscopic canonical quantization. We will speak of the atom–field interaction in the following subsec- tion. The quasimodes are treated for the representative case of the three-dimensional Fabry–Perot cavity. 3. A standard model used in cavity quantum electrodynamics and laser physics may be pictured as follows. The standard model may be specified as a typical cavity model. Radiative atoms are located in the optical cavity.2 Mode Functions for Fabry–Perot Cavity Dalton and Knight (1999a.688) νy 2π 0 Let us note that  L 1 (n 2 − 1)   c √ (n 2 − 1) (poln =) vαβ → sin k y d 2 . the three- dimensional planar Fabry–Perot cavity. The atoms are coupled directly to a cavity quasimode. ν y reduce to sums over ν y and the sign of νz .689) 2π 2 4π n 1 (poln =⊥) The application of quantum scattering theory to the beam splitter is justified in the usual situation where integrated one-photon and two-photon detection rates are finite for incident light field states of interest (Dalton et al. An optical cavity is produced by a perfect mirror and a semi-transparent mirror. 1999d). The cavity quasimode decays via Markovian damping with a rate constant Γc to certain external quasimodes. With modification made above we have not yet rederived the results of these authors.

. L 2π k y = ν y  . and both are larger than the dielectric layer thickness d. For the quasi-cavity the electric permittivity function ˜ (R) is given as ⎧ ⎨ 0 for −(L + d̃) ≤ z < −d̃. The external region length L is much greater than the cavity length l. In both cases μ = μ0 everywhere. .3. .693) Here it has been assumed that in (3. One makes the thin. . The electric permittivity function (R) for the true cavity is given as ⎧ ⎨ 0 for −(L + d) ≤ z < −d. ±1. i. as there are no magnetic media involved. The external region III lies between the dielectric layer plane at z = −d and a second perfect mirror in the z = −(L + d) plane. Each wave vector k is written in terms of its axial component k z ez and its trans- verse component kτ . y = ± L2 . where the dielectric constant κ̃ is related to the refractive index ñ through κ̃ = ñ 2 . the general form of the true mode functions and the quasimode functions is the same.690) ⎩ 0 for 0 ≤ z ≤ l. be invariant   to transition from the plane x = − L2 to the plane x = + L2 and from the plane   y = − L2 to the plane y = + L2 . .694) L . The mode functions and the necessary partial derivatives of these functions must have the period L  in x and y. (R) = κ0 for −d ≤ z < 0.5 Quasimode Theory 201 mirror in the z = +l plane and a thin layer of dielectric material with dielectric constant κ = n 2 of thickness d. (3.e.691) ⎩ 0 for 0 ≤ z ≤ l.692) 2π k x = νx . Obviously. (3. .692) It can be derived that the mode functions have the form Ak (R) = exp[ikτ · (xex + ye y )]Zk (z). . (3. ˜ (R) = κ̃0 for −d̃ ≤ z < 0. ±1. located between the z = 0 and z = −d planes (region II). . It is assumed that the three regions constitute a rectangular cuboid with bound-   aries also at x = ± L2 . We may interpret the notation of the true mode functions as general as far as it is useful. kτ = k x ex + k y e y . (3. (3. ±2. νx = 0. ν y = 0. ±2. strong dielectric approximation (Dalton and Knight 1999a). An artificial cavity is described by a modified thickness d̃ and a modified refractive index ñ.

are defined with respect to the  and ⊥ polarizations.701) The coefficients are     αi ai = α0 . (3. (3.697) c τ = ex cos φ + e y sin φ (3.699) The polarization vectors are ei = τ cos(θ1 ) − n sin(θ1 ). βt .693) ⎧ ⎨ αi ei exp(ikiz z) + αr er exp(ikrz z) region III.695) ⎩ γu eu exp(ikuz z) + γv ev exp(ikvz z) region I. αr .696) where ωk kω =. and γv are coefficients. (3. In region II the wave vectors are kt = nkω [τ sin(θ2 ) + n cos(θ2 )]. (3. In the  polarization case we have the following parameters.700) er = −τ cos(θ1 ) − n sin(θ1 ). Dalton and Knight (1999a. (3.703) ar 2i exp[−ikω (L + d) cos(θ1 )] and α0 will be characterized below. The quantities. Zk (z) = βt et exp(iktz z) + βs es exp(iksz z) region II. where αi .702) αr ar where     ai 1 exp[ikω (L + d) cos(θ1 )] = (3.704) ks = nkω [τ sin(θ2 ) − n cos(θ2 )]. In region III (n = ez ) the wave vector of the forward-propagating wave is ki = kω [τ sin(θ1 ) + n cos(θ1 )]. γu . (3. We refer to the original work for the latter.693) and (3. βs .705) .b) investigated not only the travelling modes but also the trapped modes. (3.695). (3. which are contained in relations (3.202 3 Macroscopic Theories and Their Applications In (3.698) and that of the backward-propagating wave is kr = kω [τ sin(θ1 ) − n cos(θ1 )]. (3.

We have concentrated on the quasimodes. We have assumed that the γ0 dependence is useful for γ0  α0 and the α0 dependence is useful for α0  γ0 .707) es = −τ cos(θ2 ) − n sin(θ2 ). (3. (3.710) bs 2 − exp(−iφ0 ) and β0 exp(iξ0 ) (β0 ≥ 0) will be characterized below. (3. (3.713) γv gv where     gu 1 exp[−ikω l cos(θ1 )] = (3. (3.715) . (3. We obtain that α0 ≈ cos[kω (L + d + l) cos(θ1 )] γ0 − Λ cos(θ1 ) cos[kω (L + d) cos(θ1 )] sin[kω l cos(θ1 )].708) The coefficients are     βt bt = β0 exp(iξ0 ).711) The polarization vectors are eu = ei . In region I the wave vectors are ku = ki .709) βs bs where     bt 1 exp(iφ0 ) = (3. (3. (3. We have calculated the dependence of α0 and β0 on γ0 and that of γ0 and β0 on α0 . n sin(θ2 ) = sin(θ1 ). kv = kr .3.714) gv 2i exp[ikω l cos(θ1 )] and γ0 will be characterized in what follows.706) The polarization vectors are et = τ cos(θ2 ) − n sin(θ2 ).712) The coefficients are     γu gu = γ0 .5 Quasimode Theory 203 where Snell’s law holds. ev = er .

725) . In region I the wave vectors are given in relation (3. In region III the wave vectors are given in relations (3.699). We obtain that γ0 ≈ cos[kω (l + L + d) cos(θ1 )] α0 − Λ cos(θ1 ) cos[kω l cos(θ1 )] sin[kω (L + d) cos(θ1 )]. (3.711).720) = ex sin φ − e y cos φ.719) where σ =τ ×n (3.709). The polarization vectors are et = es = σ . In region II the wave vectors are given in relations (3.702).204 3 Macroscopic Theories and Their Applications where Λ = n 2 kω d  1.723) The coefficients are given in relation (3.704) and (3. In the ⊥ polarization case we have the following parameters. (3.721) The coefficients are given in relation (3. The polarization vec- tors are eu = ev = ei . (3. (3.718) α0 cos(θ2 ) We should also assume that cos[kω l cos(θ1 )] ≈ 0 or an external field far from reso- nance. (3. β0 exp(iξ0 ) cos(θ1 ) ≈− sin[kω l cos(θ1 )]. where     ai 1 exp[ikω (L + d) cos(θ1 )] = (3.717) β0 exp(iξ0 ) cos(θ1 ) ≈ sin[kω (L + d) cos(θ1 )]. The polarization vectors are ei = er = σ . where     bt 1 exp(iφ0 ) = (3.696) and (3.705).724) bs 2 exp(−iφ0 ) and β0 exp(iξ0 ) (β0 ≥ 0) will be characterized below. (3.716) γ0 cos(θ2 ) We should also assume that sin[kω l cos(θ1 )] ≈ 0 or a resonant field.722) ar 2i − exp[−ikω (L + d) cos(θ1 )] and α0 will be characterized below. (3.

730) α0 We should also assume that cos[kω l cos(θ1 )] ≈ 0 or an external field far from reso- nance. for both polarizations.734) gv γu − gu γv = 0 .727) cos(θ1 ) β0 exp(iξ0 ) ≈ − sin[kω l cos(θ1 )].731) αr γv with   1 − 12 iΛ cos(θ1 ) 1 iΛ cos(θ1 ) T≈ 2 (3.728) γ0 We should also assume that sin[kω l cos(θ1 )] ≈ 0 or a resonant field. (3.3. where     gu 1 exp[−ikω l cos(θ1 )] = (3. (3.732) − 2 iΛ cos(θ1 ) 1 + 12 iΛ cos(θ1 ) 1 for  polarization and 0 2 2 1 1 − 12 iΛ [cos(θ 2 )] − 12 iΛ [cos(θ 2 )] T≈ cos(θ1 ) 2 cos(θ1 ) 2 (3. for instance. (3.726) gv 2i − exp[ikω l cos(θ1 )] and γ0 will be characterized in what follows. We obtain that α0 ≈ cos[kω (L + d + l) cos(θ1 )] γ0 [cos(θ2 )]2 −Λ cos[kω (L + d) cos(θ1 )] sin[kω l cos(θ1 )]. (3.733) 1 2 iΛ [cos(θ 2 )] cos(θ1 ) 1 + 12 iΛ [cos(θ 2 )] cos(θ1 ) for ⊥ polarization. (3. the coefficients in region III are connected to those in region I by the relation     αi γ =T u .713).5 Quasimode Theory 205 The coefficients are introduced in relation (3. We obtain that γ0 ≈ cos[kω (l + L + d) cos(θ1 )] α0 [cos(θ2 )]2 −Λ cos[kω l cos(θ1 )] sin[kω (L + d) cos(θ1 )]. The field must fulfil the boundary conditions  ar αi − ai αr = 0.729) cos(θ1 ) β0 exp(iξ0 ) ≈ sin[kω (L + d) cos(θ1 )]. (3. These values have been obtained using wave optics. in which.

Analyses of cavities have been mentioned in the beginning of Section 3.735) gv where relation (3. The notation Uα (R) should be used for the quasimode functions which Aα (R) still denotes.206 3 Macroscopic Theories and Their Applications The eigenfrequencies ωk are given as solutions of the transcendental equation     gu −ar ai T = 0.736) n kω kω2 may be mentioned.738) (L  )2 L 0 independent of the polarization. cos(θ1 ) = 1− (3.637).737) (L  )2l0 independent of the polarization.1.3. . Coupling constants between a single axial cavity quasimode and axial external quasimodes depend on the external quasimode frequency slowly. They have shown that the coupling problem for modes in a three- dimensional Fabry–Perot cavity is equivalent to a similar problem in a one-dimensional Fabry–Perot cavity.697) and the relations   |kτ |2 |kτ |2 cos(θ2 ) = 1 − 2 2 . The selection rules allow coupling between axial cavity quasimodes and axial external quasimodes.636) and (3.739) They have also found that Mαβ and Hαβ are zero for quasimodes of different polarizations. When the external field is off a resonance. They have concluded that the approximation is good if the cavity is of sufficiently high quality and if the precise spatial dependence of the field does not weigh. we put  2 |α0 | = (3. we put  2 |γ0 | = (3. (3. Bar- nett and Radmore (1988) have shown that even the mode-strength function which characterizes the true modes may be approximated using quasimodes and a phe- nomenological coupling. The coupling constants between different quasimodes are calculated according to relations (3. Dalton and Knight (1999b) have found that Mαβ = Hαβ = 0 if ναx = νβx or ναy = νβy . To achieve an approximate normalization of the near-resonant mode. and |kτ |2 is a parameter. (3. One may conclude that the conditions for irreversible Markovian damping of the cavity quasimode are satisfied.

lists differ- ent particles within atom ξ . t) satisfies a generalized Poisson equation ∇ · [(R)∇φ(R. 2. . . 2003).3 Atom–Field Interaction Within Cavity First we complete what is needed for the description of a system of radiative atoms interacting with the electromagnetic field in the presence of a neutral dielectric medium on the assumptions made before.α    jL (R. t) = qξ α u[rξ α (t) − Rξ (t)] × [ṙξ (t) − Ṙξ (t)] ξ. (3. t) = qξ α [rξ α (t) − Rξ (t)] ξ. qξ α . Mξ α are the charge and mass for the ξ α particle. respectively. . t) = qξ α δ R − rξ α (t) ṙξ α (t). t). t). atom-true field mode couplings are used as a basis for the pseudomode model.3. One defines  1 PL (R.α 0   × δ R − Rξ (t) − u[rξ α (t) − Rξ (t)] du  1 + qξ α [rξ α (t) − Rξ (t)] × Ṙξ (t) ξ. 3.5. In the generalized Coulomb gauge condition.741)  1 ML (R. t). .5 Quasimode Theory 207 In the work of Garraway (1997a.742) where Rξ (t) is the position of the centre of mass of the atom ξ .    ρL (R. (3.b). In certain situations quasimodes can be associated with pseudomodes (Dalton et al. t). lists different radiative atoms and α = 1. ξ. t) given in terms of the positions rξ α (t) and velocities ṙξ α (t) of the charged particles forming the radiative atoms. t)] = −ρL (R. and magnetization density ML (R.740) ξ.α 0   × δ R − Rξ (t) − u[rξ α (t) − Rξ (t)] du. (3. . the scalar potential φ(R. . t) = qξ α δ R − rξ α (t) .743) . polarization density PL (R. 2. (3. whose mass is Mξ . current density jL (R. We add that the radiative atoms are sta- tionary and electrically neutral. The radiative atoms possess charge density ρL (R.α Here ξ = 1.α 0   × δ R − Rξ (t) − u[rξ α (t) − Rξ (t)] du.

t) · PL (R. t) as ∂A(R. t) is the reduced polarization density: PL (R. t) [∇ × A(R. (3.208 3 Macroscopic Theories and Their Applications The macroscopic Lagrangian is given by the relation 1 L  (t) = Mξ α ṙ2ξ α (t) − Vcoul (t) + Lc (R.745) ∂t In the Lagrangian Vcoul (t) is the Coulomb energy given by the relation [(R)∇φ(R. t) is obtained from the Lagrangian density Lc (R. t) · ∇ × A(R. t) 3 H  (t) = + Vcoul (t) + dR ξ.748) ∂t and the particle momenta are obtained from L  (t) as pξ α (t) = Mξ α ṙξ α (t) 1   + qξ α uB Rξ (t) + u[rξ α (t) − Rξ (t)]. (3. t). t) = PL (R.747) The conjugate momentum field Π(R. t) Π(R. t) 2 1 Lc (R. t) 3 + d R − [∇ × A(R. t)] · ML (R. t) − (R)∇φ(R. t) = (R) − PL (R. t)]2 3 Vcoul (t) = d R (3.α 2Mξ α 2(R)  2  Π (R. t)]2 + + d3 R 2(R) 2μ(R) Π(R.744) 2 ξ. t)]2 2 ∂t 2μ(R) ∂A(R. (3. t du × [rξ α (t) − Rξ (t)]. t) − PL (R. t du ξ. (3. t) d3 R (R)  qξ2α  1     + u∇ × A Rξ (t) + u rξ α (t) − Rξ (t) . t) (3. t) = (R) − [∇ × A(R.749) The multipolar Hamiltonian is  p2ξ α (t) PL 2 (R.750) . t) is given by the relation  1 ∂A(R. t).α 2Mξ α 0   2 × rξ α (t) − Rξ (t) .α where the Lagrangian density Lc (R. t) d3 R.746) 2(R) and PL (R. t) · + ML (R. 0 (3.

586).β α  − Q̇ ∗α (t)Nα (t).β 2 α. t) = (R)Ak (R) PL (R . (3. (3.3. t) is given as  1 pξ α (t) ML (R. t) = qξ α u[rξ α (t) − Rξ (t)] × ξ. t) · A∗k (R ) d3 R . (3.745) as follows: 1 L  (t) = Mαξ ṙ2αξ (t) − Vcoul (t) 2 α.α 0 Mξ α   × δ R − Rξ (t) − u[rξ α (t) − Rξ (t)] du. The reduced polarization density is given in terms of true mode functions as  PL (R.756) β . with the Lagrangian density (3. ˜ (R) ∗ L α (t) = U (R) · PL (R.754) (R) α Making the choice (3. 1999b).752) k Using expansion (3. (3. one has N(t) = M−1 L(t).602).ξ    1     + qαξ ṙαξ (t) · uB Rξ (t) + u rξ α (t) − Rξ (t) .744). t) d3 R. (3.751) We have property (3. (3.ξ 0 1 ∗ 1 ∗ + Q̇ α (t)(W−1 )αβ Q̇ β (t) − Q (t)Vαβ Q β (t) 2 α.755) The generalized momentum coordinates Pα (t) for the electromagnetic field are  Pα (t) = (M−1 )αβ Q̇ β (t) − Nα (t).608) for K (Dalton et al. one can write Lagrangian (3.5 Quasimode Theory 209 where the reduced magnetization density ML (R. t du × rξ α (t) − Rξ (t) α.753) α where N(t) = K∗ L(t).

β α where X(t) = M−1 D(t)M−1 .ξ Mμξ 0 0    × δ R − Rξ (t) − u  rμξ (t) − Rξ (t) d u  du     2 ˜ (R) ∗ ˜ (R) × rμξ (t) − Rξ (t) ∇× U (R) · ∇ × Uβ (R) (R) α (R)    ˜ (R) ∗   − rμξ (t) − Rξ (t) · ∇ × Uα (R) rμξ (t) − Rξ (t) (R)   ˜ (R) · ∇× Uβ (R) d3 R d3 R .ξ α 1 ∗ 1 ∗ + Pα (t)Mαβ Pβ (t) + Q (t)Vαβ Q β (t) 2 α.β 2 α.ξ 2Mαξ 2 α.757) 2 α. (3.758) with  ˜ (R) ∗ Rβ (t) = − ML (R.760) (R) . (3. (3.210 3 Macroscopic Theories and Their Applications The Hamiltonian is given by the relation  p2αξ (t) 1 ∗ H  (t) = + Vcoul (t) + N (t)Mαξ Nξ (t) α.β α  1 ∗ + Pα∗ (t)Mαβ Nβ (t) + Q (t)(M−1 )αβ Rβ (t) α. t) · ∇ × Uβ (R) d3 R.β 2 α. (3.β α 1 ∗ + Q (t)X αβ (t)Q β (t).759) (R) 2 1 1   qμξ   Dαβ (t) = u u  δ R − Rξ (t) − u rμξ (t) − Rξ (t) μ.

761) α . radiative energy (two terms). ˜ (R) where ∇ × (R) Uβ (R) means that the vector R and the derivatives with respect to its components are replaced by the vector R and the derivatives with respect to the components of R . polarization energy. electric interaction energy. Coulomb energy. and diamagnetic energy. (3. t) = (M−1 L(t))α ˜ (R)Uα (R). magnetic interaction energy. The reduced polarization density can be expanded in terms of the quasimode functions ˜ (R)Uα (R) as  PL (R. The terms in the Hamiltonian are the particle kinetic energy. respectively.

t) = PLξ (R.741) with the same modification. t) 3 V̂coul (t) + d R = V̂coul IA (t) + V̂pol IA (t) + V̂cont (t). t) by relation (3.5 Quasimode Theory 211 The quantization for the radiative atom charged particles is the familiar prescriptions involving Hermitian operators rαξ (t) → r̂αξ (t).743) and (3.740) with the sum over ξ omitted and PLξ (R.747) as   ∇ · (R)∇φξ (R. t) = −ρLξ (R. t) are given by equations (3. t) d3 R (R)  qξ2α  1     + u∇ × Â Rξ (t)1̂ + u r̂ξ α (t) − Rξ (t)1̂ . (3. t) by (3.763) where P̂L (R.−β † + − ηα ηβ Mα.β ηα ηβ α =β 0 1  √ Vα. t) (3. M̂L (R. t) · M̂L (R. t) .765) and   PLξ (R.c. t) · P̂L 2 (R. One modifies (3.α 2Mξ α 0   2 × r̂ξ α (t) − Rξ (t)1̂ . pαξ (t) → p̂αξ (t). (3.766) . t) 3   + d R− ∇ × Â(R. The full quantum multipolar Hamiltonian is  p̂2ξ α (t) P̂L 2 (R. t). One has (Dalton and Babiker 1997) P̂L 2 (R. 4 α.751) in the operator form.761) and (3. (3.−β + √ †α (t) Âβ (t) + H. t du ξ. t) − (R)∇φξ (R.762) with the usual commutation rules applying.3. The polarization energy term and the Coulomb energy term can be combined to be equal to the sum of intra-atomic Coulomb and polarization energy terms plus intra-atomic contact energy terms.764) 2(R) One defines ρLξ (R.β ηα ηβ α =β Π̂(R. (3. t) 3 Ĥ  (t) = + V̂coul (t) + dR ξα 2Mξ α 2(R)  1  † + μα Âα (t) Âα (t) + 1̂ α 2 0 1  √ Vαβ + ηα ηβ Mαβ + √ †α (t) Âβ (t) 2 α.

t) V̂cont (t) = d3 R.767) ξ 2(R)  P̂Lξ (R.771) ξ. The reduced polarization density becomes    ˜ Rξ (t)   P̂L (R.η 2(R) ξ =η To obtain the electric dipole approximation one neglects the magnetic and dia- magnetic interaction energy terms. In (3. t) · P̂Lη (R.β  Rξ (t) The atom–electromagnetic field interaction energy then assumes the forms     1 ηα ˜ Rξ (t) E1 V̂A−F (t) =   ξ.α. t) IA V̂pol (t) = d3 R. t) 3 IA V̂coul (t) = d R. t) = (M−1 )αβ   μ̂ξ (t) · U∗β Rξ (t) ˜ (R)Uα (R). (3. (3. t) · (R)∇ φ̂ξ (R.770) ξ where μ̂ξ (t) is the dipolar moment for the ξ atom.768) ξ 2(R)  P̂Lξ (R. t) = μ̂ξ (t)δ R − Rξ (t) . (3.212 3 Macroscopic Theories and Their Applications respectively. (3. (3.769) ξ.764)      (R)∇ φ̂ξ (R.α i 2  R (t) ξ    . The polarization density is given in its dipolar approximation    P̂L (R. t) · P̂Lξ (R.

(3. (3. × Âα (t)μ̂ξ (t) · Uα Rξ (t) − †α (t)μ̂ξ (t) · U∗α Rξ (t) .772)    μ̂ξ (t) · Π̂ Rξ (t). (3.775) ξ.α 2Mξ α .774) with  p̂2ξ α (t) ĤA (t) = + V̂coul IA (t) + V̂pol IA (t).773) ξ  Rξ (t) The quantum Hamiltonian in the electric dipole approximation and rotating wave approximation is  ĤE1. t =   .RWA (t) = ĤA (t) + ĤQ (t) + V̂A−F E1 (t) + V̂Q−Q RWA (t). (3.

respectively. and the coupling of two optical cavities.4 Several Sets of Quasimodes The quasimode theory of macroscopic quantization (Dalton et al. relations (3.777) ˜ (l) (R)Uα(l)∗ (R) · Uβ(l) (R) d3 R = δαβ .β . On the contrary.776) μ̃(l) (R) ∇ · ˜ (l) (R)Uα(l) (R) = 0. and (3. (3. It follows that Markovian spontaneous emission damping occurs for the radiative atoms. In other situations. This model has been justified in terms of the quasimode theory of macroscopic canonical quantization (Dalton and Knight 1999a. This suggests problems such as reflection and refrac- tion at a dielectric boundary.579).5. (3. 1991) being an example.750). With λα(l) the angu- lar frequency of the (l.b).747). .745).594). (3.b).c). the linear cou- pler (Lai et al. (1999d). 1 ∇× [∇ × Uα(l) (R)] = (λα(l) )2 ˜ (l) (R)Uα(l) (R). . μ̃(l) (R).744). In the analysis. each with its own set of associated mode functions.746). N ). The generalization allows for the case where two or more quasipermittivities are introduced. it suffices to consider just a single quasipermittivity function in order to obtain suitable quasi- mode functions (Dalton et al.5 Quasimode Theory 213 The coupling constants describing energy exchange processes between a radia- tive atom placed in the cavity and nonaxial external quasimodes vary slowly with the external quasimode frequency. (3. 1999b. t) = Q α(l) (t)K αβ Uβ (R).c) has been generalized (Brown and Dalton 2001a. the linear coupler. (3. along with their associated sets of quasimode functions. it is appropriate to construct quasimode functions via the introduction of two distinct quasipermittivities. which are defined as the solutions of N separate Helmholtz equations involving the quasiper- mittivities and quasipermeabilities ε̃ (l) (R). the standard model in cavity quantum electrodynamics has been considered. . (3. (3. 3. A full quasimode treatment of the beam splitter has been given in Dalton et al. In the model the basic processes are described by a cavity damping rate.778) Expansion of the vector potential A(R) directly in terms of the quasimode func- tions Uα(l) (R) is not possible. Instead we can write (R)  (l. a radiative atom spontaneous decay rate. their coupling with the (axial) cavity quasimodes consists in reversible photon exchanges as characterized through one-photon Rabi frequencies. (3. and an atom–cavity mode coupling con- stant. (3. .m α. (3. (3.595). 1999b.748).3.586). and (3.m) (m) A(R. In some situations.596) are generalized. We assume N sets of quasimode functions Uα(l) (R) (l = 1. α) quasimode.779) ˜ (R) l. such as a single laser cavity or a beam splitter. The theory comprises the above relations (3.

.ξ 0 × [rξ α (t) − Rξ (t)] 1   (l)∗ 1  (l)∗ + Q̇ α (t)(W−1 )αβ (l. . . but the notation is not changed.m) (m) Q (t)Vαβ Q β (t) 2 l. space integrals are cured with excluding infinitesimal volumes containing these discontinuities from their domain. The quasimode functions Uα(l) (R) for N > 1 are not all linearly independent. .781) ⎝ .214 3 Macroscopic Theories and Their Applications which involves a double sum over all quasimodes.2) . Y(2.607). 1 L  (t) = Mαξ ṙ2αξ (t) − Vcoul (t) 2 α. . . ⎠ Y(N . Y(1.m) ˜ (l) (R) (l)∗ ˜ (m) (R) (m) Mαβ = (R) Uα (R) · U (R) d3 R.. (3. (3. When applying the theory to situations where the true or quasi permittivities and permeabilities contain discontinuities.m) (m) Q̇ β (t) − (l. Here only quasimodes with the frequencies are retained that are important for the quantum optics system. composed of K(l.. .606).1) Y(2. .780) α Some of the matrices become block matrices of the form (given for an arbitrary case Y) ⎛ (1. . In that theory only the lower energy atomic orbitals are included.β 2 α. t du α. .N ) A matrix Y becomes the block matrix Y and a matrix C becomes the column block matrix C. .N ) ⎟ ⎜ ⎟ C = ⎜ . ⎠ C(N . Y(N .. Relations (3. .m α.753) is generalized.1) Y(N .N ) ⎟ ⎜ ⎟ Y=⎜ . (3. ⎟. (3.N ) and column block matrices of the form (given for an arbitrary case C) ⎛ (1. ⎟ (3.m) . The square matrix K has become a block matrix K.754) are generalized (l. The set of quasimodes arising from N different quasipermittivities is overcomplete. Relation (3.ξ  1   + qαξ ṙαξ (t) · uB Rξ (t) + u[rξ α (t) − Rξ (t)]. . and (3.β α  − Q̇ α(l)∗ (t)Nα(l) (t)..1) (1.N ) ⎞ C ⎜ C(2.782) ⎝ .783) (R) (R) β .2) .N ) ⎜ Y(2. It is solved by following an analogy with the theory of linear combinations of atomic orbitals (Coulson 1952).2) ⎞ Y Y .

788) (R) β  ˜ (l) (R) (l)∗ Rα(l) (t) = − ML (R.m) 1 ˜ (l) (R) (l)∗ Hαβ = ∇× U (R) μ(R) (R) α  ˜ (m)(R) (m) · ∇× U (R) d3 R.m α.787) 2 l.m α.m) (m) N (t)Mαξ Nβ (t) α.758).m) (m) + Pα(l)∗ (t)Mαβ Nβ (t) l.784) (R) β (l) ˜ (R) (l)∗ L α(l) (t) = U (R) · PL (R.786) j β Relation (3.m α.m) + Q (t)X αβ (t)Q (m) β (t). (3.β α 1   (l)∗ (l.m) Dαβ (t) = u u  δ R − Rξ (t) − u[rμξ (t) − Rξ (t)] μ.ξ 2Mαξ 2 l.m) (m) + Pα (t)Mαβ Pβ (t) + Q (t)Vαβ Q β (t) 2 l.m) (m) 1   l∗ (l. (3.3.785) (R) α Relation (3.756) is generalized.m)∗ (m) Rβ (t) 2 l.5 Quasimode Theory 215  (l.m α. (3. Relations (3.β α The block matrix X(t) is given by (3. (3. t) d3 R.757) is generalized:  p2αξ (t) 1   (l)∗ H  (t) = + Vcoul (t) + (l.m α.  Pα(l) (t) = (M−1 )αβ (l.ξ Mμ.β α  (l.β 2 l.ξ 0 0   × δ R − Rξ (t) − u  [rμξ (t) − Rξ (t)] du  du    ˜ (l) (R) (l)∗ ˜ (m) (R) (m) × [rμξ (t) − Rξ (t)] ∇ × 2 U (R) · ∇ × U (R) (R) α (R) β  ˜ (l) (R) (l)∗ − [rμξ (t) − Rξ (t)] · ∇ × Uα (R) [rμξ (t) − Rξ (t)] (R)   ˜ (R) (m) (m) · ∇× U (R) d3 R d3 R .m α. (3. where the notation has the actual mean- ing. (3.789) (R) . t) · ∇ × Uα (R) d3 R.ξ α 1   (l)∗ (l.  qμξ 2 1 1   (l.β 1   (l)∗ + Q (t)(M−1 )αβ (l.760) and (3.759) are generalized.m) (m) Q̇ β (t) − Nα(l) (t).

216 3 Macroscopic Theories and Their Applications

Relation (3.761) is generalized:

PL (R, t) = (M−1 L(t))α(l) ˜ (l) (R)Uα(l) (R). (3.790)
l α

Replacements (3.615), (3.616) are adapted,

Q α(l) (t) → Q̂ α(l) (t), Q α(l)∗ (t) → Q̂ α(l)† (t), (3.791)
Pα(l) (t) → P̂α(l) (t), Pα(l)∗ (t) → P̂α(l)† (t). (3.792)

The nonzero equal-time commutators are (cf. (3.617))

[ Q̂ α(l) (t), P̂β (t)] = iδlm δαβ 1̂ = [ Q̂ α(l)† (t), P̂β(m) (t)]. (3.793)

The electromagnetic field is real, which somewhat complicates the quantization
(Brown and Dalton 2001a).
Relations (3.618) and (3.619) are generalized,

1    (l)† 
ĤQ (t) = P̂α (t) + Q̂ α(l)† (t)Vαα
(l,l) (l)
P̂α (t)Wαα Q̂ α (t) ,
(l,l) (l)
2 l α
V̂Q−Q (t) =
 (l,m) (m) (l,m) (m)

× P̂α(l)† (t)Wαβ P̂β (t) + Q̂ α(l)† (t)Vαβ Q̂ β (t) . (3.795)
l,m α,β
(l,α) =(m,β)

Relations (3.621), (3.622), and (3.623) are generalized,
ηα(l) (l) 1
Âα(l) (t) Q̂ α (t) + i P̂α(l) (t), (3.796)
2 2ηα(l)
ηα(l) (l)† 1
Âα (t) =
Q̂ (t) − i P̂α(l)† (t), (3.797)
2 α 2ηα(l)


ηα(l) = (l,l)
. (3.798)

It follows simply from (3.793) that these annihilation and creation operators obey
the following equal-time nonzero commutation relations:

[ Âα(l) (t), Âβ (t)] = δlm δαβ 1̂. (3.799)

3.5 Quasimode Theory 217

Relations (3.624) and (3.625) are generalized,


Q̂ α(l) (t) = Âα(l) (t) + Â−α (t) , (3.800)

ηα(l) (l) (l)†

P̂α(l) (t) = −i Âα (t) − Â−α (t) . (3.801)

Relation (3.626) is generalized,

ĤQ (t) =  Âα (t) Âα (t) + 1̂ μα(l) ,
(l)† (l)
l α

Relation (3.620) is generalized,
(l,l) (l,l)
μα(l) = Wαα Vαα . (3.803)
Let us recall that
V̂Q−Q (t) = V̂Q−Q
(t) + V̂Q−Q
(t), (3.804)
where V̂Q−Q (t) is generalized,

V̂Q−Q (t) =
2 ⎛ ⎞
 / V (l,m)
⎝ ηα(l) ηβ(m) Mαβ
(l,m) αβ ⎠ Âα(l)† (t) Â(m)
× +/ β (t), (3.805)
(l) (m)
l,m α,β ηα ηβ
(l,α) =(m,β)

and V̂Q−Q (t) is also generalized,
V̂Q−Q (t) =
4 l,m α,β
(l,α) =(m,β)
⎡⎛ ⎞ ⎤
/ V (l,m)
α,−β ⎠ (l)†
× ⎣⎝− ηα(l) ηβ(m) Mα,−β Âα (t) Âβ (t) + H.c.⎦ .
(l,m) (m)†
(l) (m)
ηα ηβ

Relations (3.630) and (3.631) are generalized,

  ˜ (m) (R)
Â(R, t) =
l,m α,β 2ηα(l) (R)

(l,m) (l)
× K αβ Âα (t)U(m) (l,m)∗ (l)†
β (R) + K αβ Âα (t)U(m)∗
β (R) , (3.807)

218 3 Macroscopic Theories and Their Applications

 ηα(l) (l)
Π̂(R, t) = −i ˜ (R)
× Âα(l) (t)Uα(l) (R) − Âα(l)† (t)Uα(l)∗ (R) . (3.808)

The theory comprises the above relations (3.764), (3.767), (3.768), and (3.769).
Relation (3.772), which comprises the annihilation and creation operators, is gener-

ηα(l) ˜ (l) (R)
V̂A−F (t) = −i
ξ l,α
2 (R) Rξ (t)

× Âα(l) (t)μ̂ξ (t) · Uα(l) Rξ (t) − Âα(l)† (t)μ̂ξ (t) · Uα(l)∗ Rξ (t) . (3.809)

Let us recall that in the rotating wave and electric dipole approximations we can
write the quantum Hamiltonian as

ĤE1,RWA (t) = ĤA (t) + ĤQ (t) + V̂A−F
(t) + V̂Q−Q
(t). (3.810)

(l,m) (l,m)
The coupling constants Mαβ , Vαβ can be calculated from the matrices M and
H using relations (3.609) and (3.610) (Brown and Dalton 2001a). For the usual
case where μ̃(l) (R)=μ(R) and the overlap between the set l and the set m of mode
functions is small, to good accuracy we have

(l,m) 1, i = j, α = β,
Mαβ ≈ (l,m) (l,m) (3.811)
(M1 )αβ + (M2 )αβ , otherwise,

(l,m) λα(l)2 , i = j, α = β,
Vαβ ≈ (l,m) (l,m) (3.812)
(H1 )αβ + (H2 )αβ , otherwise.

Relations (3.636), (3.637) are generalized and also modified,

(l,m) ˜ (l) (R) ˜ (m) (R)
(M1 )αβ = ˜ (R) −1 − 1 Uα(l)∗ (R)
(R) (R)
· U(m) 3
β (R) d R, (3.813)

(l,m) ˜ (R) ˜ (m) (R)
(M2 )αβ = ˜ (R) + − 1 Uα(l)∗ (R)
(R) (R)
· U(m) 3
β (R) d R, (3.814)

3.5 Quasimode Theory 219

(l,m) 1 ˜ (R)
(H1 )αβ = ∇× − 1 Uα (R)
μ(R) (R)
˜ (R) (m)
· ∇× − 1 Uβ (R) d3 R, (3.815)

(l,m) 1 (m)2 ˜ (l) (R) (m)2 
˜ (m) (R)
(H2 )αβ =− ˜ (R) λα + λβ
2 (R) (R)
× Uα(l)∗ (R) · U(m) 3
β (R) d R. (3.816)

Relation (3.632) is generalized,

μα(l) ≈ λα(l) (3.817)

and relation (3.633) is also generalized,

μα(l) ≈ λα(l) + vαα
. (3.818)

Relation (3.634) is generalized,
 (l,m) (l)†
V̂Q−Q (t) ≈  vαβ Âα (t) Â(m)
β (t). (3.819)
l,m α,β
(l,α) =(m,β)

Relation (3.635) is generalized,

e. Such an accord is described also by the quantum theory based on a microscopic model of the dielectric media (Hynne and Bullough 1990).m) (l. (z). The true mode approach has continued the previous literature. homogeneous dielectric material of refractive index n 1 . We shall assume that space has been divided into two regions. but with refractive index n 2 . Region 2.820) λα(l) λ(m) β The foregoing theory has been applied to reflection and refraction at a dielectric interface (Brown and Dalton 2001b).m)  (H1 )αβ − (H2 )αβ + / . is then . (3. is assumed to contain material obeying the same restrictions. is assumed to be filled with linear. Here we will expound the quasimode approach in part. The permittivity function for the system.m) 1 αβ + (M ) (l. The analysis has been very thorough including the quantum scattering theory in the Heisenberg picture. The behaviour of the intensity for a localized one-photon wave packet has been examined. Allen and Stenholm (1992) or Carniglia and Mandel (1971). which has exhibited agreement with the classical laws of reflection and refraction.m) 2 αβ 2 (l. Region 1.m) 1 vαβ = λα(l) λ(m) β (M ) (l. which is formed by the points with z ≥ 0. (l. which consists of the points with z < 0.g.

z < 0. but are evanescent in the other region. From the generalized Helmholtz equation (3.822) (ñ 2 )2 0 . With the two values. ˜ (2) (z) = (3. (z) = (3.823) n 22 0 . two sets of plane waves are associated. respectively. each set effectively being restricted to just one of the regions. For the reflection and refraction problems we choose the two quasipermittivity functions (Brown and Dalton 2001b). the functions are not confined in one region. ñ 2  1. one being n 21 0 in all space and the other being n 22 0 in all space. z < 0. We could try to use two quasipermittivity functions.220 3 Macroscopic Theories and Their Applications  n 21 0 . z ≥ 0. The vanishingly small refractive index in one region means that all incident waves in the other region except some with angles of incidence smaller than the critical angle produce only an evanescent wave in the region with negligible refractive index. we can determine the form of the quasimode functions U(1) (2) α (R) and Uα (R). Instead we choose two sets of quasimodes. A completeness of the union of these sets is also available.821) n 22 0 . ˜ (1) (z) = (3. The quasimode functions are to an excellent approximation given by the formulae α (R) ≈ Nα U(1) (1)  / . We will treat the case of out- of-plane polarization.  (ñ 1 )2 0 . The union of these sets does not enjoy the mutual orthogonality of all functions. z ≥ 0. where the quasirefractive indices ñ 1 and ñ 2 are positive constants which fulfil ñ 1 .  2 n 1 0 . which are associated with the quasipermittivities ˜ (1) (z) and ˜ (2) (z). z < 0. The spatially confined nature of these types of mode functions is also used when applying a quantum scattering theory approach to energy transfer from one region to another. At a closer look. z ≥ 0. An effective mutual orthogonality is present.776).

√ × r̃1∗ exp(ik(1) αi · R) + r̃1 exp(ikαr · R) Θ(z) (1) /  + t̃1 r̃1∗ exp(ik(1) αt · R)[1 − Θ(z)] σ. (3.824) α (R) ≈ Nα U(2) (2)  / .

(3. √ × r̃2∗ exp(ik(2) αi · R) + r̃2 exp(ikαr · R) [1 − Θ(z)] (2) /  + t̃2 r̃2∗ exp(ik(2) αt · R)Θ(z) σ.825) .

. (3. .826) 1− i tan(θ̃αi ) tanh(θ̃αt ) with (l) (l) |(kαi )τ | tan(θ̃αi )= (l) . It is obvious that the subscript α should be replaced by a variable ki(l) . the subscript α has been preserved. Approximate expressions are obtained also by consid- ering r̃l = −1 and t̃l = 0. the scalar product of the functions simplifies to a Dirac delta function when we choose   32 1 1 Nα(l) = √ . √ The usual formulae are obtained by multiplying the right-hand sides by r̃l . (3. |k(2) αt | = |k |.X = Nα.3. In the case of the continuous sets. which also simplifies them. 2 t̃l = (l) (l)  . . (l) (l)  .829) L (l) where X = x. The continuous sets of the mode functions in the case where the evanescent waves have been neglected are discretized easily. The use of quantization box of volume L 3 is (l) (l) immediate.828) 2π n l 0 As this value does not depend on ki(l) .X . The corresponding z-component should be negative for l = 1 and it should be positive for l = 2. We assume that the propagation vectors kαi and kαr obey (l) (l) (l) 2π kαi. − for l = 2. We have concluded that the discrete sets are not always obtained so easily as desired.5 Quasimode Theory 221 where + (l) 1 + i tan(θ̃αi (l)  ) tanh(θ̃αt ) r̃l = .. (3. |(kαi )τ | ñ 2 (1) ñ 1 (2) |k(1) αt | = |k |.X are integers. which may justify an alternative phase factor.X = kαr. y and Nα. + for l = 1. On the modification the two sets of the functions are continuous contrary to the notation. (3. We should complete the case X = z. We note that r̃l are complex the case where units and t̃l r̃l∗ = |t̃l |.1 − i tan(θ̃αi ) tanh(θ̃αt ). (kαi )z / (l) (l) 2 (l)  |(kαi )τ |2 − |kαt | tanh(θ̃αt )=± (l) .827) n 1 αi n 2 αi Here Θ(z) is the step function and Nα(l) are normalization constants appropriate to + the evanescent wave has been neglected. .

z (l) . (3.222 3 Macroscopic Theories and Their Applications For X = z the quantization box should not suggest the periodic boundary con- dition. Then of course (l) π kαi. We assume that the mode function vanish for z = 0 and z = L. The modes with in-plane polarization are not considered for simplicity (Brown and Dalton 2001b).z = −kαr.z (l) = Nα.830) L (l) The integer Nα.z should be negative for l = 1 and it should be positive for l = 2. .

however. Peřinová. It seems that the separation of the equations for the medium polariza- tion is not a sufficient ground for the theory to be considered microscopic. an expression of the noise which is quantal in essence fits in the framework of the electromagnetic- field quantization. LLC 2009 . V. It is important that the motion of the medium polarization may be damped and losses may be included. The equivalence between the Green-function approach and the method of continua for media suitable for both approaches has been demonstrated. it differs formally from the method of continua of harmonic oscil- lators. continua of harmonic oscillators have been considered. The Green-function approach has been elaborated on for various media. 4. on which we shall concentrate ourselves in what follows. one can see the presence and correlation of fluctuations of the electric-field strength in the vacuum state of the field and the ground state of the matter. when the polarization of the medium is described by separate equations. but we adopt this nomenclature. A quantum noise is considered for the field commutators not to depend on the time. only the inclusion of a nonlinearity of the medium was under development in the course of writing this book. Many papers have been devoted to the Green-function approach to the quantiza- tion of the electromagnetic field in a medium. in the quantum theory of solids. 223 DOI 10. Quantum Aspects of Light Propagation. In the framework of this model. The magnetic properties are usually neglected.  C Springer Science+Business Media. Lukš.Chapter 4 Microscopic Theories A divergence from the macroscopic theories emerges. In the framework of this approach the electric permittivity of the medium can be derived. A standard microscopic approach is expected from the quantum theory of solids. whereas we refer to a microscopic theory. Even though the Casimir effect is not regularly connected with the propagation. As this theory rather begins with a quantum noise.1007/b101766 4. Still. A. The description of the fields can be quantized. but they must be included in the phenomenological quantum description of negative-index materials.1 Method of Continua of Harmonic Oscillators Many scientists would call the following exposition a macroscopic theory for a lossy medium.

The appropriate relation contains the complex relative permittivity of the medium (ω) as a linear transform of the coupling constant g(ω) between the light and the dressed matter-field B̂(k. one can introduce. with fields in the reciprocal space. and the reservoir in the Cartesian product of the direct and a reciprocal space are con- sidered. ω) for polari- tons. 1991) using a simple version of this model (Kittel 1987). as usual. Huttner and Barnett (1992b)). Diagonalization by the Fano technique is performed (Fano 1961. It is proven that also the coupling constant dependent on at least the frequency of the reservoir “elementary” mode fulfils the assumptions for further diagonalization. cf. t) and B̂(x. Their analysis is restricted to a one-dimensional model and to transverse electromagnetic fields. After introduc- ing the Lagrangian densities. and thus a quantization scheme to describe the losses must introduce the medium explicitly.. 2000). ω). but not losses. the radiation is coupled to the mat- ter (it is a field again) and the matter is coupled to the reservoir (it is a field of the dimension increased by unity). Huttner and Barnett (1992a) use the model of Hopfield (1958) and Fano (1956). The complex relative permittivity (ω) fulfils the Kramers–Kronig relations. The total transition to a direct space for the reservoir is not usual. whose operator exhibits the same dependence on the wave vector and the frequency as the operator of the reser- voir field. ω). but is conceivable.1. ω). Only at the very beginning the radiation and the matter in the direct space. (Rosenau da Costa et al. t) (see. but couplings rather form a chain. The vector potential depends on the spatial coordinate and the time as usual and it has the form of the integral of the vector potential for a unit density of polaritons with the wave vector k and the frequency ω multiplied by the polariton operator Ĉ(k. Taking into account the frequency decompositions of the fields Ê(x. The operator of this field shares the dependence on the wave vector and the frequency with the operator of the reservoir field.1 Dispersive Lossy Homogeneous Linear Dielectric Huttner and Barnett (1992a) have started from the observation that the macroscopic approach to the theory of the electromagnetic field in a medium is a quantization scheme that does accept dispersion. having first treated the quantization of light in a purely dispersive dielectric (Huttner et al. This diagonal- ization gives rise to the (dressed) matter-field B̂(k. Barnett and Radmore 1988). So it does not deal with a fun- damental property of the susceptibility. The description of the matter and reservoir is first diagonalized. In contrast with the vacuum theory (the theory of the electromagnetic field in a vacuum) a macroscopic field emerges this way whose operator depends also on the frequency. Losses in quantum mechanics are treated by coupling to a reservoir. This diagonalization gives origin to the field Ĉ(k. the Kramers–Kronig relations. The matter is not quite identical with the reservoir. in an “almost” . A rigorous treatment is contained in the book of Klyshko (1988). Huttner and Barnett (1992a) work. the effect of choice of the type of coupling between light and matter on the definition of the conjugate variables for the components of the vector potential is discussed.224 4 Microscopic Theories 4.

2) where −E 0 .4) 2 . The one-dimensional model has been expanded to three dimensions. In Huttner and Barnett (1992b) the derivation of such equal-space commutation relations is provided in the case of a linear dielectric. The Hamiltonian is first derived. That is to say that the fields ĉ+ (x. the positive and negative propagating components. ω) are introduced. s) = √ ĉ± (x. ω) and ĉ− (x.79)). then diagonalized. (4. (4. Lai and Haus 1989. ω) exp(−iωs) dω. <= > 2 Lem = (Ȧ + ∇U ) − (∇ × A) (4. two Maxwell equations are transformed onto two spatial Langevin equations. ∞ 1 ĉ±dir (x.1) 2π 0 and that may be why Huttner and Barnett (1992a) name the papers devoted to the phenomenological approach to quantization (Levenson et al. the field is expressed in terms of space-dependent amplitudes. Using these defini- tions. From this. and the expansions of the field operators are transformed. Let us note that Huttner and Barnett (1992b) in the introduction mention also the popular approach (Huttner et al. Potasek and Yurke 1987. ω) are used and the spatial Langevin force f̂ (x. or (3.1 Method of Continua of Harmonic Oscillators 225 conventional manner.507). (3. simple equal-space commutation relations between the operators in the “application” times s and s  follow. <= > 2 B 1 . ω) is introduced. 1985. In contrast with the macroscopic theories this technique is not derived from a Lagrangian and has not been justified in terms of a canonical scheme. Caves and Crouch 1987.4.3) 2 2μ0 is the electromagnetic part which is expressed in terms of the vector potential A and the scalar potential U . Respecting the frequency decomposition of the field D̂(x. 1990). The propagation of light in the dielectric is anal- ysed. These dif- fer almost negligibly due to the imaginary part of the refractive index (see (3. The equal-space commutation relations between the operators at the frequencies ω and ω can also be derived. Huttner et al. Huttner and Barnett (1992b) have started the canonical quantization from a Lagrangian density L = Lem + Lmat + Lres + Lint . ρ 2 Lmat = (Ẋ − ω02 X2 ) (4. the fields ĉ± (x. and their spatial equations of evolution are obtained. t). The canonical scheme and losses cannot be easily unified. 1990). but this has been solved in Huttner and Barnett (1992b). in which spatial progression equations are derived and quantization of the field is performed imposing the equal-space commutation relations.506).

8) (2π) 2 We shall underline the newly introduced quantities in order to differentiate between the quantities in real and reciprocal spaces. comprising a field Yω of the continua of harmonic oscillators of frequencies ω. 0 ∗ Lint = −α[A∗ · Ẋ + A · Ẋ + ik · (U ∗ X − U X∗ )] ∞ ∗ − v(ω)(X∗ · Ẏω + X · Ẏω ) dω. but it can be written in terms of the proper dynamical variable X. t) = 0 E(r. The total Lagrangian can be written in the form  L= (Lem + Lmat + Lres + Lint ) d3 k.226 4 Microscopic Theories is the polarization part. (4. Let us recall that E∗ (k.10) 0 . (4. t) = E(−k.6) 0 is the interaction part with coupling constants α and v(ω).9) where the prime means that the integration is restricted to half the reciprocal space and the Lagrangian densities become Lem = 0 (|E|2 − c2 |B|2 ). (4. The former has an integral expression and that is why we go to the reciprocal space. and ∞ Lint = −α(A · Ẋ + U ∇ · X) − X · v(ω)Ẏω dω (4. For example the electric field is written as 1 E(r. see below. α could be a tensor. ∞ Lres = ρ (|Ẏω |2 − ω2 |Yω |2 ) dω. t) = 3 E(k. The displacement field is defined by D(r. (4.7) As U̇ does not appear in the Lagrangian. The interaction between the light and the polarization field has the coupling constant α and the interaction between the polarization field and other oscillator fields used to model the losses has the coupling constant v(ω). U is not a proper dynamical variable. ρ ∞ 2 Lres = (Ẏω − ω2 Y2ω ) dω (4. Lmat = ρ(|Ẋ|2 − ω02 |X|2 ).5) 2 0 is the reservoir part. used to model the losses (reservoirs). In general. t). It comprises both the annihilation and the creation operators. t). t)eik·r d3 k. modelled by a harmonic-oscillator field X of frequency ω0 (the polarization field). t) − αX(r.

2. The polarization field X and other oscil- lator fields Yω (the matter fields) are decomposed into transverse and longitudinal parts. t) U (k. (4. It can be derived that D is a purely transverse field.13) where L⊥ em = 0 (|Ȧ| − c k |A| ). The transverse part contains only transverse fields and is  ⊥ L = (L⊥ ⊥ ⊥ ⊥ em + Lmat + Lres + Lint ) d k. are introduced. t) = Aλ (k. and the transverse fields are decomposed along them to get  A(k. c. t) = X⊥ (k. t) + X  (k. is also given in Huttner and Barnett (1992b). containing only longitudinal fields. (4. L can now be used to obtain the com- ponents of the conjugate variables for the fields ∂L λ −0 E λ ≡ λ∗ = 0 Ȧ . (4. t)eλ (k) (4. 2 0 ∞  ⊥∗ ⊥ L⊥int = − αA · Ẋ + v(ω)X ⊥∗ · Ẏ ω dω + c. The total Lagrangian can then be written as the sum of two independent parts. For con- venience.1 Method of Continua of Harmonic Oscillators 227 As usual in quantum optics. we choose the Coulomb gauge. The scalar potential in the reciprocal space   α κ · X(k.12) and Yω can be expressed similarly.16) ∂ Ȧ . 2 2 2 2 ⊥ L⊥ ⊥ 2 mat = ρ(|Ẋ | − ω0 |X | ). one can restrict oneself to transverse components of other fields and omit the superscript ⊥ . For example X can be written as X(k. k · A(k. Unit polarization vectors eλ (k). so that the vector potential A is a purely transverse field. 2 2 ∞ ⊥ L⊥ res = ρ (|Ẏω |2 − ω2 |Y⊥ ω | ) dω. which are orthogonal to k and to one another. t) = i . t) = 0.4. λ = 1.15) λ=1. t)κ (4.2 and similar expressions for the other fields.11) 0 k where κ is a unit vector in the direction of k. 3 (4.14) 0 The longitudinal part.

(4. Ê (k . As usual. ω v(ω) ∗ + + ρω |Yω | + 2 2 (X · Qω + c. including the interaction between the polar- D∞ 2 ization and the reservoirs and ω̃02 ≡ ω02 + 0 [v(ω)] ρ2 dω is the renormalized frequency of the polarization field.17) ∂ Ẋ δL λ Q λω ≡ λ∗ = ρ Ẏ ω − v(ω)X λ . t)] = iδλλ δ(k − k )1̂. 1989) by pos- tulating equal-time commutation relations between the variables and their conju- gates λ λ ∗ i [ Â (k. (4.228 4 Microscopic Theories ∂L λ Pλ ≡ λ∗ = ρ Ẋ − α Aλ . namely αρ |A|2 .25) where all quantized operators are denoted by a caret. Q̂ ω (k . .21) 0 ρ ρ is the energy density of the matter fields. Fields are quantized in a standard fashion (Cohen-Tannoudji et al. (4. (4. t). |P|2 Hmat = + ρ ω̃02 |X|2 ρ ∞ . c.) dω (4. the annihilation operators are introduced. k̃ being defined by k̃ = is the / k 2 + kc2 with kc ≡ ωc α2 c = ρc2 0 . P̂ (k . c.24)  λ λ∗ [Ŷ ω (k.20).18) δ Ẏ ω The famous ambiguity is worth mentioning: The conjugate of A can be −0 E (with the coupling αρ A · P). The Hamiltonian for the transverse fields is  H= (Hem + Hmat + Hint ) d3 k.|Q |2 . t)] = − δλλ δ(k − k )1̂. (4.19) where   Hem = 0 |E|2 + c2 k̃ 2 |A|2 (4.20) + electromagnetic energy density. has already been classified 2 into (4. Thus any gauge determines a type of coupling. as well as −D (with the coupling E · X). (4. t)] = iδλλ δ(k − k )δ(ω − ω )1̂. t).22) ρ is the interaction energy between the electromagnetic field and the polarization. α Hint = (A∗ · P + c.23) 0 λ λ ∗ [ X̂ (k. t). Part of the interaction energy with the matter.) (4.

1 Method of Continua of Harmonic Oscillators 229  0 λ λ .4.

It is important that the matter can be formally decoupled from the reservoir by the Fano technique and a dressed matter field obtained. (4.29) are obtained. â(λ. t) . (4. t) + ωb̂ω† (λ. t) + P̂ (k. k. and the k integration has been restored to the full reciprocal space.27) 2ω̃0 ρ   ρ λ 1 λ b̂ω (λ. the equal-time commutation relations for the creation and annihilation operators [â(λ. â † (λ .31) λ=1. t) − i Ê (k. (4. t) d3 k. (4. k. −k. t). t)b̂ω† (λ. [b̂(λ. b̂† (λ . t) + H. k.2 0 ∞   + V (ω) b̂(λ. k.33) / / v(ω) ω ω̃0 ckc2 where V (ω) = ρ ω̃0 . (4.28) 2ω̃ ρ From the equal-time commutation relations for the fields (4. k. k. t) = k̃c  (k. k.2  ∞ Ĥmat = ω̃0 b̂† (λ. t) 2 λ=1. d3 k. t). t). t) = −iωŶ ω (k. k .25). k. k.24). k. k. t)b̂† (λ. k. t) + Q̂ ω (k. t) + H. k. Λ(k) ≡ k̃ .30) where  Ĥem = k̃câ † (λ. It is worth mentioning that the Maxwell–Lorentz equations can be derived from the Hamiltonian. t)] = δλλ δ(ω − ω )δ(k − k )1̂ (4. t) dω λ=1.23). t)â(λ. t) . t) 2 0  †  + b̂ (λ. t)b̂† (λ. k . c.26) 2k̃c   ρ λ i λ b̂(λ. (4.32)    Ĥint =i Λ(k) â(λ. Following Fano (1961). t)] = δλλ δ(k − k )1̂. dω d3 k. (4. k. The dressed . t)b̂(λ. t) = ω̃0 X̂ (k. b̂ω (λ . The normally ordered Hamiltonian for the transverse fields is Ĥ = Ĥem + Ĥmat + Ĥint . (4. t)b̂ω† (λ. k. t)b̂ω (λ. t)] = δλλ δ(k − k )1̂. † [b̂ω (λ. the polarization and reservoir parts of the Hamiltonian can be diagonalized. t) (4. k. c. k. k. −k. k .2  + â † (λ.

−k. t). (4. k.34) B̂(λ. ω. ω )b̂ω (λ. k. ω. t) 0 . ω. t) ∞ + α1 (ω. t) + β0 (ω)b̂† (λ. t) and B̂(λ. k. respectively. t)] = δλλ δ(k − k )δ(ω − ω )1̂. t) = α0 (ω)b̂(λ. k. k. t) are introduced. B̂ † (λ . ω . k . ω.230 4 Microscopic Theories matter field creation and annihilation operators B̂ † (λ. [ B̂(λ. which satisfy the usual equal-time commutation relations. k.

36) The coefficients of the Bogoliubov transformation are calculated. t. we are also interested in the inverse transforma- tion. (4. (4. t).37) 2 ω − ω̃02 z(ω) 2 where z(ω) is defined by ∞  1 V(ω )  z(ω) = 1 − lim dω .38) 2ω̃0 ε→+0 −∞ ω − ω + iε   ω − ω̃0 V (ω) β0 (ω) = . ω ) are defined as follows. t) = α0 (ω) B̂(λ. ω) dω (4. [ B̂(λ. −k. ω. † + β1 (ω. k. α1 (ω. ω ) = δ(ω − ω ) + .35) The coefficients α0 (ω). t). ω. β1 (ω. ω )b̂ω (λ. (4.42) 0 . that is to say the formulae   ω + ω̃0 V (ω) α0 (ω) = . (4. A useful definition of an “eigenoperator” is presented Barnett and Radmore (1988).39) 2 ω − ω̃02 z(ω) 2    ω̃0 V ∗ (ω ) V (ω) α1 (ω. ω ) = (4. From relation (4. As usual with the substitutions. k. k.35) it can be seen that the diagonalization is performed independently for every pair of the counterpropagating modes of the polarization field (“the modes” here are only formally similar to those of the electromagnetic field) and that it is performed using a Bogoliubov transformation.41) 2 ω + ω ω2 − ω̃02 z(ω) are derived. ω. Ĥmat ] = ω B̂(λ. t) − β0 (ω) B̂ † (λ. In the study no constant V(ω) ≡ |V (ω)|2 occurs. k. −k. (4. It is interesting that the diagonalization is performed once for the polarization and reser- voir parts of the Hamiltonian and once for the total Hamiltonian.40) 2 ω − ω − i0 ω2 − ω̃02 z(ω) and     ω̃0 V (ω ) V (ω) β1 (ω. The Hamiltonian expressed in the modal annihilation operators is considered. t) dω . It is given by the relations ∞  ∗  b̂(λ. ω ). β0 (ω). (4.

ω)â † (λ. The diagonalization of the total Hamiltonian is formally very similar to √ the diago- nalization of its matter part. t) 0 . ω) B̂ † (λ.45) 0 for the coefficients of the Bogoliubov transformation seem to be familiar. ω)β1∗ (ν. ω . −k. ω) B̂(λ. Ĉ(λ. t) ∞ + α̃1 (k. (4. ω.43) The conditions ∞   I ≡ |α0 (ω)|2 − |β0 (ω)|2 dω = 1. A dimensionless coupling constant ζ (ω) = i ω̃0 [α0 (ω)+ β0 (ω)] is defined and the annihilation operators are introduced (by a Fano type of technique). ω ) − β1 (ν. It has been shown that the diagonalization cannot be performed on the common assumption of white noise (the Markov-type coupling). t) + β̃0 (k. k. ω . nothing more. ω)â(λ. t) − β1 (ω . Without it. t) = α̃0 (k. 0 (4. t) dω . k. ω)α1 (ν. −k. k. It is commented on free charges and a conducting medium being beyond the scope of Huttner and Barnett (1992a). ω ) ≡ α1∗ (ν.44) 0 ∞   I (ω.4. t) = α1∗ (ω . k.1 Method of Continua of Harmonic Oscillators 231 and ∞   b̂ω (λ. we are farther from the original Lorentzian formulation. We need not grieve for the assumption of the white noise. ω . ω ) dν = δ(ω − ω ) (4. ω. ω ) B̂(λ. k.

ω. ω). t) dω . + β̃1 (k. (4. (4. ω) = . ω) = 1 − lim dω . β̃0 (k. (4.49) k̃c 2  ∗ (ω)ω2 − k 2 c2 where the complex relative permittivity (ω) is introduced 1  2 2  (ω) = 1 + k c − (k 2 c2 + ωc2 )z̃ ∗ (k. ω) 2 where ∞  ωc2 |ζ (ω )|2 z̃(k. ω ) B̂ † (λ. ω) = . ω. independent of k.47) k̃c 2 ω − k̃ 2 c2 z̃(k. ω. ω ).46) where the coefficients α̃0 (k. ω) . ω ) are rather complicated and are derived in the form    ωc2 ω + k̃c ζ (ω) α̃0 (k. (4.48) 2(k̃c)2 ε→+0 −∞ ω − ω + iε or alternatively    ωc2 ω + k̃c ζ (ω) α̃0 (k. ω .50) ω2 . (4. ω). α̃1 (k. −k. and β̃1 (k.

t). ω ) = δ(ω − ω ) + c  ∗ .232 4 Microscopic Theories    ωc2 ω − k̃c ζ (ω) β̃0 (k. t)..56) The vector potential is now given by  1 ωc2  Â(r.. (4. t). t) + ↠(−k. ω) = . . dω d3 k.52) 2 ω − ω − i0  (ω)ω2 − k 2 c2 and   ωc2 ζ ∗ (ω ) ζ (ω) β̃1 (k.58) 20 .8) being modified for the operators expresses the operators Â(r. 2k̃c0  k̃c Ê(k. 0)e−iωt .. (4. t) = 3 eλ (k) (2π) 2 20 λ=1.57) 0 ω2 (ω) − k 2 c2 Relation (4. t).. t) = i [â(k.. k. k. 0)e + H. t). . . ω.55) they have a harmonic time dependence Ĉ(λ.2 ∞  ζ ∗ (ω) −i(ωt−k·r) × Ĉ(λ.53) 2 ω − ω − i0  ∗ (ω)ω2 − k 2 c2 The operators Ĉ(λ.. t) + ↠(−k. (4. ω. k. t) = Ĉ(λ. t). t). by the rela- tions   Â(k.54) and being operators for eigenmodes. t)] = δλλ δ(k − k )δ(ω − ω )1̂ (4. t).. ω. Let us note that on the substitution into (4. k . (4. t). ω. Ê(k. X̂(k. Ŷω (k. Ê(k. Ê(r.. (4. P̂(k. in terms of the operators Â(k. t). Ĉ † (λ . c.. t)]. [Ĉ(λ. ω ) = . k.. k. ω. k. t)].8) for Â(k.. t). ω.. (4. (4. ω. ω . ω. t) also satisfy the usual commutation relations. t) = [â(k. Q̂ω (k. Ĥ ] = ωĈ(λ. t). t) and Ĉ † (λ.. t). [Ĉ(λ. k. ω.51) 2 k̃c  ∗ (ω)ω2 − k 2 c2   ω2 ζ ∗ (ω ) ζ (ω) α̃1 (k. k. ω.

Ê(r.4. Ê(k. are introduced. ω . 0) and b̂ω (k. t).35). ω. dω. t). b̂ω (k. t)e−ik·r d3 r.63) 0 Scω|n(ω)|2 . t). t) are introduced. . Ŷω (k. t) and b̂ω (k.43).28). t). ω. ω. ω.56). 0).46).59) 0 and for the operators B̂(k. t) by the relation (the slightly modified relation (4. ω. 0). ω)Ĉ† (−k.. 0). Q̂ω (k. 0)... 0 (4. (4. ω.27). t) by the relation ∞   B̂(k. ω. On the substitution into the intermediate result for these operators by the relations 1 Â(k.. 4π 0 (4. b̂†ω (k. 0) by relation (4. (4. (4. ω. These substitutions solve the Cauchy or initial problem. 0) are introduced.61) (2π) 2 . On the substitution into the intermediate result for these operators by the formulae (4. t) = α̃1∗ (k. the operators â(k. which have the time dependence (4. t) = 3 Ê(r. 0) are introduced. t) and ↠(k.. 0) by relation (4. t)e−ik·r d3 r. the operators b̂(k. ω.. ω.. ω. t) are introduced.. respectively. 0)e−iωt + H.. t) = 3 Â(r.1 Method of Continua of Harmonic Oscillators 233 the annihilation and creation operators â(k. ω)Ĉ(k.42) and (4. the operators Â(r. t) − β̃0 (k. 0). t) dω (4. ω)Ĉ† (−k. t) dω . On the substitution into the intermediate result for the operators b̂(k. On the substitution into the intermediate result for Ĉ(k. 0). On the substitution into the intermediate result for the operators â(k.26). ω.. Huttner and Barnett (1992b) restrict themselves first to a one-dimensional case when describing the propagation in the dielectric.2) from Huttner and Barnett (1992b)) ∞   â(k. b̂† (k. t). the operators B̂(k.62) where  η(ω) A(ω) = . c. 0) and B̂(k.60) the operators Ĉ(k. ω. and (4. 0). X̂(k. the operators Â(k. P̂(k. (2π) 2 1 Ê(k. b̂(k. t) − β̃1 (k. ω. are introduced. t) by relations (4. The vector potential is considered in the simpler form ∞ 1   Â(x. . 0)e−iωt + ĉ− (x. ω )Ĉ(k. 0) are introduced and on the substitution into the intermediate result for the operators B̂(k. t) = √ A(ω) ĉ+ (x. t) = α̃0∗ (k.

ω. Jeffers and Barnett (1994) modelled the propagation of squeezed light through an absorbing dispersive dielectric medium.66) |ζ (ω)| n(ω) and Im{K (ω)} > 0. (4. ω. (4. t) = e dk. and the operators ĉ± (x. ω . ω. n(ω) is the complex refractive index defined as the square root of the relative permittivity (ω) with a positive real part η(ω). (4. t). ω.65) c ζ ∗ (ω) |n(ω)| eiφ(ω) = . t) = − √ eiφ(ω) Ĉ(k. ω. t) commutes with all the Langevin operators f̂ (x  . ω. (4. ω. that ĉ+ (x. f̂ † (x  . Since the magnetic field can be expressed similarly as the vector potential. t). ω . dielectrics with a thin absorption line. i. finally. t)eikx dk (4.e. t) are introduced  Im{K (ω)} iφ(ω) ∞ Ĉ(k. t) and f̂ † (x  . t) for all x  > x.70) and. t). t)eikx ĉ± (x. t) + = ±iK (ω)ĉ± (x. ĉ∓ (x.69) further of the equal-space commutation relations † [ĉ± (x. ω . ω. t)] = δ(ω − ω )1̂. He formulated a canoni- cal quantization of the electromagnetic field in a closed Fabry–Pérot resonator with a dispersive slab. while ĉ− (x. t)] = 0̂ (4. Huttner and Barnett (1992b) remind of the simple commutation relations [ f̂ (x. . t) and f̂ † (x  . t)] = δ(x − x  )δ(ω − ω )1̂. ĉ± (x. Hradil (1996) considered “lossless” disper- sive dielectrics. ω.68) 2π −∞ and it also enters a rather similar expression for the electric displacement operator. t) commutes with all the Langevin operators f̂ (x  .64) π −∞ K (ω) ∓ k where the complex wave number K (ω) and the phase factor eiφ(ω) are expressed as n(ω)ω K (ω) = . ω . t) for all x  < x. (4. ω.67) ∂x where the Langevin-noise operator is ∞ i f̂ (x. t). † [ĉ± (x.67) have been obtained from the Maxwell equations for the monochro- matic fields. ω . the spatial quantum Langevin equations of progression can be obtained as ∂ ĉ± (x. ω . ω . t) ± 2Im{K (ω)} f̂ (x.234 4 Microscopic Theories S is a cross-sectional area. ω. Equation (4. ω. ω.

Gruner and Welsch (1995) have contributed to the stream of papers aiming at a description of quantum properties of the dispersive and lossy dielectrics includ- ing the vacuum fluctuations.1 Method of Continua of Harmonic Oscillators 235 Wubs and Suttorp (2001) have solved the initial-value problem for the damped- polariton model formulated by Huttner and Barnett (1992a. Expectation value of the Poynting vector operator is computed. for example.1. Quantization of the electromagnetic field in inhomogeneous. two techniques suitable for nonequilibrium processes are utilized: the Heisenberg equation of motion and the diagrammatic Keldysh proce- dure. The calculations fit into the development of the theory of thermal scanning microscopy. (1987) and later by Glauber and Lewenstein (1991). t − τ ) dτ . Input–output relations are unitary and no additional quantum-noise terms are required. and lossy dielectrics is performed with the help of a procedure which is still attributed to Huttner and Barnett (1992b). A canonical quantization scheme for radiation fields in linear dielectrics with a space-dependent refractive index has been developed by Knöll et al. (2003) analyse radiative heat transfer between two dielectric bod- ies. For application. i. It is remarked that in nonlinear models the Keldysh formalism provides a framework for the perturbation expansion.4. 4. 1991). Equations specialized to the case of a dielectric layer with a uniform density of oscillators are usual expressions.e.71) 0 . First.177) and a constitutive relation comprising an integral term. Knöll et al. (4. They study it in terms of a symmetrized cor- relation function. They try to expound and supplement the paper by Huttner and Barnett (1992b) from the point of view of the quantization of the phenomenological Maxwell theory.b) and have found that for long times all field operators can be expressed in terms of the initial reser- voir operators. Hillery and Drummond (2001) have studied the scattering of the quantized elec- tromagnetic field from a linear dispersive dielectric in the limit of “thin” absorption lines. see. the quantization of radiation in a dispersive and lossy dielectric is per- formed. t) + χ (τ )E(r. (1986. This begins from the classical Maxwell equations (3. Janowicz et al. 1990. Knöll and Welsch (1992) and a related work (Knöll and Leonhardt 1992). To this end.176) with (3. t) = 0 E(r. The field is represented by means of the dual vector potential. They have investigated the transient dynamics of the spontaneous- emission rate of a guest atom in an absorbing medium. ∞  D(r. dispersive. fluctuations of radiation field in the ground state of the coupled light–matter system.2 Correlation of Ground-State Fluctuations The quantization of the radiation imbedded in a dielectric with a space-dependent refractive index has been expounded in the book by Vogel and Welsch (1994).

2 0 which comprises a sum over λ which is absent from relation (3. Let us remember relations (4. (4. ω) have been introduced.72) and the Helmholtz equation is presented. but even the most general ones are presented. â j (r . / In these relations. Continuing the use of the operators â(r. ω) = 0 (ω)E(r.76) .75) −∞ is applied (the putting of an identity matrix to be multiplied by a function and sub- sequent transverse projection of the columns). The effect of the medium is entirely determined by the complex permittivity (ω).62) here) has been written. Ĉ(k. ∞ Δi j F(r) = δi⊥j (r − r )F(r ) d3 r . k. ω)Ĉ(λ. ω). The Huttner–Barnett quantization scheme is introduced with a diagonalized Hamiltonian  ∞ Ĥ = ωĈ † (λ. ω) (4.74) (2π)3 −∞ k which can be interpreted also as a transverse projection of the columns of 3 × 3 iden- tity matrix multiplied by the δ function. ω) and f̂(r. (4.6). The definition of the transverse δ function is presented as ∞   1 ki k j δi⊥j (r) = δi j − 2 eik·r d3 k. ω) − 0 F(ω)f̂(r.73) λ=1. etc. the commutators [âi (r. (4.236 4 Microscopic Theories is transformed into the Fourier space to yield D(r. (4. ω) from Huttner and Barnett (1992b). It still has no tensorial character. The vector-field operators â(r. for example.3). ω) dω d3 k. ω) being a generalization of the component ĉ(x. The operator constitutive equation (in the Fourier space) D̂(r. ω) and the identity 2 ζ (ω) = ω Im{(ω)} uti- lized. (4.14) of Huttner and Barnett (1992b). Then. The frequency-dependent field operators are introduced in the three-dimensional case. but very complicated expressions are obtained. ana- logues of their frequency decompositions of the vector-potential operators. k. ω) and f̂(r. ω). (4.21) of Huttner and Barnett (1992b) (cf. The transverse projection of the columns of an identity matrix multiplied by other functions has proved to be useful. ω )] can be expressed in terms of such projections. ω) = 0 (ω)Ê(r. have been presented. the vector â(r. k.7) of Huttner and Barnett (1992b). Not only the equal-time commutation relations.5). ω) ωc2 ∗ √ should be replaced by Ĉ(λ. ω). and (4. an analogue of relation (5. electric- field strength operators. Here. (4. the operator Δi j .

78) 0 i j with the abbreviation Δr = r − r .4.81) is considered. k. t). Cf. (4. (4. there exists no decomposition into first-order equations. The theory is applied to the determination of the correlation of the ground-state fluctuations of the electric-field strength. t + τ ) Ê n (r + Δr. The operators â(r. ω) are not independent operators. t) 2  + Ê n (r + Δr.80) and enter the expansion for Â(r. k). the partial differential equation for the operator â(r. τ ) = 0| Ê m (r. ω). t + τ ) |0 (4. Let us recall the usual annihilation operators â(λ. The symmetric correlation function of the electric-field strength 1  K mn (Δr. In the three-dimensional case. there exists no generalization of relation (4. We remark that  K mn (Δr. In the three-dimensional case. The canonical commutation relations are i ⊥ [ Âi (r. â † (λ . k). Ê j (r . Huttner and Barnett (1992b) who in the one-dimensional case intro- duce forward and backward-propagating fields and show that such a definition ensures the causal (one-sided) independence of the respective operators of the oper- ator f̂(r.77) π differs from the classical equation (4.79) A test of the consistency of the theory in the limit (ω) → 1 has been accomplished.72) by an additional term. ω) derived from Ĉ(λ. t)] = − δ (Δr)1̂. which satisfy the commu- tation relations [â(λ. ω) and f̂(r. τ ) = n R (ω) 4π 2 c2  0   ω .1 Method of Continua of Harmonic Oscillators 237 where   0 F(ω) = 0 Im{(ω)} (4..64) and no equation for such quantities. k )] = δλλ δ(k − k )1̂ (4. 0). On substituting into the phenomenological Maxwell equations. t) Ê m (r. ω) is obtained which is a Helmholtz equation with a right-hand side.

∞ exp − ωc n I (ω) × cos(ωτ )Δi j sin n R (ω)Δr dω. (4.82) 0 n R (ω)Δr c .

90) ω0 n I (ω) ≈ n I (ω0 ) ≡ n I0 . (4. is now given by the group velocity in the medium provided that the spatial distance is not too large. (4. Their solution based on a Green-function . they assume a transparent medium.89) Further. The absorption causes a spatial decay of the correlation of the field fluctuations.85) has been studied. such as a fibre. for which it may be justified to put approximately ω n R (ω) ≈ n R0 + n R1 . also the dispersion of the group velocity needs a consideration.88) (ωτ )−1  1. With increasing distance.84) Restricting attention to optical frequencies within an interval of the width 2Δω. ω β= n R (ω). (4. and group velocities and group velocity dis- persion on the dynamics of the field fluctuations within a frequency interval (4. The light cone of strong correlation. the classical Maxwell equations can be con- sidered as quantum operator equations. phase. 4.87) c and times of order of ω−1 .91) The influence of absorption.83) + n R (ω) = Re{ (ω)} = Re{n(ω)}.85) Δω  1. Gruner and Welsch (1995) let (βΔr )−1  1.238 4 Microscopic Theories where Δr = |Δr| and + n I (ω) = Im{ (ω)} = Im{n(ω)}. (4. (4. (4. (4. (4. (4. which in empty space is determined by the speed of light in vacuum. ω0 − Δω < ω < ω0 + Δω.2 Green-Function Approach On allowing for a frequency-dependent complex permittivity that is consistent with the Kramers–Kronig relations and introducing a random operator noise source asso- ciated with the absorption of radiation.86) ω0 where ω0 is an appropriately chosen centre frequency and assuming that dispersion and absorption are small on lengths of order of β −1 .

ω) + K2 (ω)Ã(r. .1 Dispersive Lossy Linear Inhomogeneous Dielectric Gruner and Welsch (1996a) have expounded a quantization scheme which starts with phenomenological Maxwell equations instead of Lagrangian densities and is consistent with the Kramers–Kronig relations and the familiar (equal-time) canoni- cal commutation relations for the vector potential and electric field. ω) is the “Fourier transform” of the (known) operator vector-potential Â(r. (4. the equations comprise (ω). especially.93) π −∞ ω − ω ∞ 1 Re{(ω )} − 1  Im{(ω)} = − V.2. ω.. ω). ˆ (r. This function has the analytical continuation in the upper complex half-plane. (Ω). which satisfies the relation (−Ω∗ ) =  ∗ (Ω). ω) is the “Fourier transform” of the operator-noise current. In fact. multilayered dielectrics. a hypothetical addition of a nontrivial solution of the homogeneous Helmholtz equation would violate the boundary condition at infinity. (4.p. 0) dω + H.92) The real and imaginary parts of the relative permittivity satisfy the well-known Kramers–Kronig relations ∞ 1 Im{(ω )}  Re{(ω)} − 1 = V. t) and ˆj̃n (r. The quantization scheme is based on the Helmholtz equation with the source term Δà ˆ ω) = ˆj̃ (r.4.p.96) 0 where quantum mechanically also the frequency-dependent operators can be time dependent. the frequency-dependent complex relative permittivity introduced phenomenologi- cally. 0) = ˆ (r. When Im{(ω)} > 0.2 Green-Function Approach 239 expansion of the vector-potential operator seems to be a natural generalization of the mode expansion applicable to source-free radiation in nearly lossless dielectrics. (4. (4. is the principal value of the integral. t = 0.95) n where à ˆ (r.p. 4.94) π −∞ ω − ω where V. à (4. from the exposition it can be seen that the vector-potential operator is introduced by the relation ∞ Â(r. dω . In the phenomenological classical Maxwell theory.  dω . This is realized for homogeneous and inhomogeneous. c.

(4. r . (4. ω). This operator can be chosen in the form (cf. (4. ω) dω d3 r. ω. ω). ω. 0) d3 r dω + H. ω) + K 2 (ω)G(r.240 4 Microscopic Theories ˆ (r. f̂ j (r . The Hamiltonian Ĥ is diagonal in the operators f̂(r.. (4. 0). ω)ˆj̃n (r . 0) = −Â(r.101) 0 where the Green function G(r. Ê(x  . ω)≡ ˆj̃n (r. t) is uniquely determined by a linear transfor- mation of the source operator ˆj̃n (r. ω) = F(ω) ω f̂(r.104) are replaced by the relation i [ Â(x. ω) · f̂(r. ω) = δ(r − r ) (4.77). 0) = G(r. ω. Ê j (r . f̂ j (r .105) A0 . r . 0). Another required property is ˙ 0) Ê(r. Gruner and Welsch (1995)) ˆj̃ (r. 0)] = − δ (r − r )1̂. r .. ∞ Ĥ = ωf̂† (r. Relation (4.103) and the canonical field commutation relations i ⊥ [ Âi (r.98) 0 and these operators have the usual properties † [ f̂ i (r.102) and the boundary condition that it vanishes at infinity. (4.99)   † † [ f̂ i (r. ω )] = 0̂. r . ω )] = [ f̂ i (r. ω) satisfies the equation ΔG(r. ω). ω )] = δi⊥j (r − r )δ(ω − ω )1̂. Gruner and Welsch (1996a) illustrate this procedure in linearly polarized radiation propagating in the x direction. t). (4. (4.97) n c2 with F(ω) given in (4. ω).104) 0 i j Relation (4. f̂ j (r . c. ω)≡ à Hence. ω).104) must be verified by straightforward calculation. 0)] = − δ(x − x  )1̂. (4. For the sake of clarity. the operator à ˆ (r.100) From the foregoing considerations it follows that (when all appropriate condi- tions are fulfilled) the operator of the vector potential can be defined by the relation ∞ Â(r.

k = ±Re{n(ω)} ωc )). ω) (it would be possible to introduce the opera- tors â(x.4. The operators f̂ (x.2 Green-Function Approach 241 where A is the normalization area perpendicular to the x direction. ω) are replaced by the operators â± (x. It is shown that when losses in the dielectric may be disregarded. the concept of quantization through the mode expansion can be recognized. Im{(ω)} → 0. which satisfy the commutation relations ω .

106) c . ω )] = exp −Im{n(ω)} |x − x  | δ(ω − ω )1̂. â± (x  . ω). (4. † [â± (x.

ω )] = 2Im{n(ω)} exp ∓iβ(ω) (x + x  ) c c ω . ω). † ω ω [â± (x. â∓ (x  .

the operators â± (x. (4. Ê Δω (x. These operators become independent of x in the limit Im{n(ω)} ωc |x − x  | → 0.105) is in an obvious contradiction with a macroscopic approach. sin Re{n(ω)} ω |x − x  |  × exp −Im{n(ω)} |x − x | c c Re{n(ω)} ωc × θ [±(x − x  )]δ(ω − ω )1̂.107) where θ (x) is the Heaviside function. As the commutation relation (4.and backward-propagating fields.  1 ω ω . ω). ÂΔω (x. but it holds that the operator-valued Langevin noise is space dependent. 0). (4. 0). as the forward.108) AR (ωc )0 where ωc is the centre frequency for suitably defined operators. it is important that Gruner and Welsch (1996a) have derived the relation i [ ÂΔω (x. 0). are governed by quantum Langevin equations. Ê Δω (x  . 0)] = − δ(x − x  )1̂. As could be expected. The theory further reveals that the weak absorption gives rise to space-dependent mode operators that spatially progress according to quantum Langevin equations in the direct space.

the operators â± (x. There exists a straightfor- ward generalization of the quantization method based on a mode expansion (Khos- ravi and Loudon 1991. Agarwal 1975).105) is performed by straightforward calculation. A general proof of this rela- tion is not present. As an example of inhomogeneous structure. which is more complicated. causality reasons are only pointed out. (4. two bulk dielectrics with a common interface are considered. 1992. The verification of the commutation relation (4. ω) = ± 2Im{n(ω)} exp ∓iRe{n(ω)} x f̂ (x. The problem of determining a classical Green function reappears. F̂± (x. 1996). .109) i c c In other words. The behaviour of short light pulses propagating in a dispersive absorbing linear dielectric with a special attention to squeezed pulses has been studied (Schmidt et al. ω) progress in space. ω).

(4.31). t)][b̂ω† (. (1998).115) and the components b̂(. t) dω d3 k 0 ∞  + V (ω)[b̂ (. relation (4. −k. The result of the substitution from relations (4.112) where the nonlinear interaction term Ĥnl is given by Ĥnl = − f [X(r)] d3 r (4. t). k. ω)).2 Dispersive Lossy Nonlinear Inhomogeneous Dielectric Emphasizing the important differences from the linear model.114) where  ∞   Ĥl = ω0 b̂† (. Schmidt et al. The total Hamiltonian can be written as Ĥ = Ĥl + Ĥnl . (1998) have derived evolution equations for the field operators and shown that additional noise sources appear in the nonlinear terms.2. t) must be appropriately defined (see.32). t) + ωb̂ω† (. Linear relation- ships between quantum (operator-valued) fields are introduced following Huttner . k. we denote here as Ĥl⊥ . (4.113) and the Hamiltonian Ĥl that governs the linear dynamics can be written as  Ĥl = Ĥl + Ĥl⊥ . t)b̂(.111) While in the linear case it is sufficient to quantize only the transverse fields. k. (4. (4. the Lagrangian and Hamiltonian for the nonlinear dielectric are introduced by Schmidt et al.12). k. Ĥnl couples the transverse and longitudinal fields. t)b̂ω (.2) has been denoted by Ll (r) and this relation with L replaced by Ll (r) has been utilized in the Lagrangian density in the relation L(r) = Ll (r) + Lnl (r). k. t)] dω d3 k † 2 0 (4. cf. k. In general. k. −k.. in the nonlinear case such a procedure would result in a loss of generality. (4. The Lagrangian density (4. t) + b̂(. (4.242 4 Microscopic Theories 4. Schmidt et al. b̂ω (. (1998)) for b̂ (k). b̂ (k.33) into relation (4. t)b̂ω (.110) where moreover Lnl (r) = f [X(r)].30) which we have denoted as Ĥ . −k.

ω) = [f̂(r. (1998) do not attempt at diagonalization of the nonquadratical Hamiltonian Ĥ . We now approach the following representations of the matter fields.122) ∂t To the number of the relations. Ĥ ]. They avoid the difficulty with the generalization of the def- initions (4. ω) as ∞ ⊥ X̂ (r) = ˆ ⊥ (r. ˆ ⊥ 0 ˆ  X̃ (r. we may be afraid that this correctness will not endure the change to the nonlinear case.118) α π 0 ˆ (r. This change is reflected in the equations of motion for the basic fields and the vector-potential field in the Heisenberg picture. b̂(k). ω) dω + H. (4.95) with à and the explicit relation (4.121) ∂t ∂ ˆ ˆ ω) + [Ã(r. (4. c.112). (4.4. the notation of which is still the same as of the Hamiltonian in (4. (4. ω) being connected with the field f̂(r. c. ω) = −iω[(ω) − 1]Ã(r. ω) as the solution of equation (4. ω).117) 0 where . Schmidt et al. à (4. ˆ ω). ω). relation (4. ω). which nevertheless hold in linear and nonlinear cases.123) c2 π 0 . ω) = [à nl (4.35) to the nonlinear functions of the operators â(k). and (4. ω). and the vector-potential field ∞ Â(r) = ˆ (r. the nonhomogeneous Helmholtz equation belongs  Δà ˆ ω) = ω ˆ (r. (4. ω) + Im{(ω)} f̂(r.46) and (4. ω) + [f̂ (r. ω). The relations hold for all times and both in linear and in nonlinear cases.116). ˆ (r. ω).56).116) 2ρ ω̃0 0 the transverse matter field X̂⊥ (r) can be expressed in terms of the field f̂(r. ∂  i f̂ (r.119) 0 If the validity of the expressions (4. Ĥnl ]. The longitu- dinal matter field X̂ (r) can be expressed in terms of the field f̂ (r.  . ω) .2 Green-Function Approach 243 and Barnett (1992b) as well as Gruner and Welsch (1995). ω). Ĥnl ]. ω) = [f̂ (r. B̂(k.119).30). c. ω) as  ∞   X̂ (r) = [α0∗ (ω) − β0∗ (ω)]f̂ (r.120) ∂t ∂ i f̂(r. Ĥ ] = ωf̂(r. (4. ω) dω + H. ω) dω + H. ω) + [f̂(r. Ĥ] = ωÃ(r. b̂ω (k).97).117) is related to the time evolution of the kind of (4. Ĥ ] = ωf̂ (r. i Ã(r. X̃ (4. ω) + K2 (ω)Ã(r...  Im{(ω)} f̂(r.

125) ∂t  where we. ∂t  nl c π 0 2 (4.127) over ω.129) ∂x (ω) . The scheme is related to the familiar model of classical susceptibilities and applied to the problem of prop- agation of quantized radiation in a dispersive and lossy Kerr medium. t) = K2 (ω1̂) ˆ (r. ˆ (x. (4. t). ω) + à (4. ω) obey the nonlinear new noise sources. Then the total Hamiltonian (4. ω) into components à field à ˆ (x. t). The basic equations are applied to the one-dimensional case and propagation equations for the slowly varying field amplitudes of pulse-like radiation are derived. ω). ω) ≡ à Here it is noticed that f̂(r.126) Relation (4. ω. for the sake of clarity. ω) and È (x. propagating in + − the positive and negative x-directions. à dynamics.112) can be reduced to a one-dimensional single-polarization form. ω.124) where 1̂ˆ is the identity superoperator and relation (4.118) indicates ˆ (r. t) + K2 1̂ˆ + Ĥ× Ã(r. t). ω). ω) = ±iK (ω) à ± f̂ (x. ω) ∓ iN Im{(ω)} ñ (x.65)). ω) ≡ f̂(r. ω). All of the fields f̂ (r. ω) = à à ˆ (x.. ω. in a single-mode optical fibre. The wealth of operator-valued fields serves the expres- sion of the dispersion and absorption in the nonlinear medium. Ô]. (4.125) can be written in the form    ∂ 1 ˆ (r. Respecting the nota- tion K 2 (ω) = c−2 ω2 (ω) (cf.127) where the elimination of the field X̂(r) using relations (4. t).116) and (4. In the linear theory it is possible to separate the two transverse polarization directions from each other and from the longitudinal direction. Let us consider the propagation in the x direction of plane waves polarized in the y direction. this is not possible for nonlinear media. (4. As has already been stated. we can see that K 2 (ω)à ˆ à ˆ (r. respectively. ˆ ω. write ∂t∂ to the right from the notation 1̂ˆ and the action of Ĥnl× on an operator Ô is defined by Ĥnl× Ô ≡ [ Ĥnl . (4. ˆ (x. By integration of (4. (4. In practice. ω. an equation adequate to the linear and nonlinear cases is obtained. t) = ω  Δà Im{(ω)} f̂(r.128) + − ˆ (x. only one transverse polarization direction will be excited.122) implies that ∂ 1 ω1̂ˆ = 1̂ˆ + Ĥnl× .244 4 Microscopic Theories ˆ (r. ω. ω) are the solutions of spatial equations of progression where à ±  ∂ ˆ ˆ (x. The one dimensionality of the problem permits one to decompose the ˆ (x. ω). ω). f̂(r. à ˆ (r. ω.

ω. It is shown that the classical Maxwell equations together with the constitutive relations except relation (4.134) .132) ˆj̃(r. which are related to the operator-valued noise polariza- tion P̃ˆ . ω).131) 0 Integrating (4.130) over ω. (1998) have developed three-dimensional quantization presented in part in Gruner and Welsch (1996a) concerning dispersive and absorbing inhomogeneous dielectric medium.2 Green-Function Approach 245 / with the normalization factor N = 4π0 Ac2 . (4. ω) dω.132) and (4. We remark that à ˆ (x.127). (4.4.133) It follows from relations (4. t). The approach directly starts with the Maxwell equations in the frequency domain for the macroscopic electromagnetic field.2. t). ± ± Similarly as from relations (4. t) = ±iK 1̂ˆ + Ĥnl× Ã ˆ (x. ω). ω) ≡ à ˆ (x. ω. the operators Â(+) ± (x) can be introduced.3 Elaboration of Linear Theory Dung et al. t) ± ∂x ∂t   Im{(ω)} ∓ iN f̂ (x. ω. Adequately to the derived equations which we consider to be mere approxima- tions in the nonlinear case.71) can be transferred to quantum theory. On considering the charge and current densities. ω) = iωρ̃(r.133) that ρ̃ˆ and ˆj̃ fulfil the equation of con- tinuity: ∇ · ˆj̃(r.119). ± (x) à ± (4.130) (ω) In analogy to (4. an equation appropriate to the linear and nonlinear cases is obtained. one can arrive from relation (4. (4. The operator-valued noise-charge density ρ̃ˆ and the operator-valued noise-current density ˆj̃ are introduced. ∞ Â(+) = ˆ (x. ω) = −∇ · P̃ˆ (r. ω) = −iωP̃ˆ (r. ˆ ω). (4. 4. Schmidt et al.129) at relation   ∂ ˆ ∂ 1 ñ (x. ρ̃ˆ (r. one concentrates oneself on the noise-charge and noise-current densities. ω. (1998) study the narrow-bandwidth field components and narrow-bandwidth pulses. The theory has been applied to narrow- bandwidth pulses propagating in a dielectric with a Kerr-like nonlinearity.123) and (4.

r . (4. r .137) where G(r.135) is an ordinary δ function instead of a transverse δ function. s. ω) d3 s. (4. ω) · e j . (1998) start from the partial differential equation ˆ (r. The commutation relation (4.139) π 0 ∂ xm −∞ c2 where εkm j is the Levi-Cività tensor and G i j (r.99) must be modified to the form † [ f̂ i (r. ω) dω. ˆ (4. ω). ω) = ei · G(r. (4. where Ẽ = iωà and Ẽ = −∇ ϕ̃ˆ .141) In the sense of the Helmholtz theorem there exists a unique decomposition of the ˆ into a transverse part Ẽ electric field Ẽ ˆ  . f̂ j (r . Dung et al. the Coulomb ˆ ⊥ and a longitudinal part. Dung et al. .138) c together with appropriate boundary conditions. ω) is the tensor-valued Green function of the classical problem.100) remains valid and relation (4.140) and [ Ê i (r). B̂k (r )] = εkm j  G i j (r. ω)Ẽ(r. i.246 4 Microscopic Theories The source term ˆj̃ is related to a bosonic vector field f̂ by the relation like (4. Relation (4. (1998) have derived commutation relations ∞  ∂ ω [ Ê i (r).96) is an integral representation of the vector-potential operator. B̂k (r )]. ω) = δ(r − s)1 (4. In the Coulomb gauge. ω) = iωμ0 G(r. ω) − ω (r. because the whole electromagnetic field is considered. ω).97). ω )] = δi j δ(r − r )δ(ω − ω )1̂.135) It is pointed out that the current density ˆj̃ is not transverse. It satisfies the equation   ω2 ∇r ∇r − 1 Δr + 2 (r. the vector field f̂ assumed here is not transverse as well and the spatial δ function in relation (4. s.e. ω) · G(r.136) 0 c2 whose solution can be represented as (here and in part of what follows we use a different notation) Ẽ(r. Ẽ ˆ ⊥ ˆ ˆ  gauge can be introduced. 2 ∇ × ∇ × Ẽ ˆ ω) = iωμ ˆj̃(r. ω) · ˆj̃(s. r . s. Ê k (r )] = 0̂ = [ B̂i (r). (4. (4. Hence.

145) ˙ It is recalled that Â(r) and 0 Â(r) are canonically conjugated field variables.151) j 0 i j . Tomaš 1995)   G(r. ϕ̂(r )] = 0̂ = [ϕ̂(r). (r. respectively. (4.139) suggests that the “canonical” commutators are not so simple as we would expect by the definition.147) i Then.138) that satisfies the boundary condition at infinity is (cf.146) i j  ˙ [ϕ̂(r). r .4. ω) d3 s. the complexity of the commutation relation (4. ω) d3 s. (4. (4. (4.143) ∂ xi  where δi⊥j and δi j are the components of the transverse and longitudinal tensor- valued δ functions δ ⊥ (r) = δ(r)1 + ∇∇(4π|r|)−1 . The commutation relation between the vector potential and the scalar potential is as complicated. (4. [ Âi (r).149) 4πr Relation (4.144) δ  (r) = −∇∇(4π|r|)−1 . B̂k (r )] = − εikm δ(r − r )1̂. (4. In this case..142) i iω ∂  ϕ̃ˆ (r. ω) = .  (r )]. (4.  (4. On the contrary. ω) = ∇r ∇r + K 2 (ω)1 K −2 (ω)g(|r − r |.  j (r )] = 0̂ = [  ˙ (r )]. when one and only one of these quantities is differentiated with respect to the time or comprises such a derivative. ω) = (ω) for all r.  [ Âi (r). ω). (4.148) where exp[iK (ω)r ] g(r. (4.139) can be simplified as i ∂ [ Ê i (r). the theory is applied to the bulk dielectric such that the dielectric function can be assumed to be independent of space. ω) = 1 à δi⊥j (r − s) Ẽˆ j (s. the solution of equation (4.150) 0 ∂ xm and the “canonical” commutator corresponds to the definition ˙ (r )] = i δ ⊥ (r − r )1̂.2 Green-Function Approach 247 the vector and scalar potentials à ˆ and ϕ̃ˆ . The simple commutation relations are ˙ (r). ω) = − δi j (r − s) Ẽˆ j (s. are related to the electric field as ˆ (r.

155) ⊥() ⊥()  [ f̂ i (r. (r. 0) = d s. ω ))† ] = δi j (r − r )δ(ω − ω )1̂. Â j (r )] = 0̂. ω)ˆj̃⊥ (r . The determination of the tensor-valued Green function for three-dimensional configura- tion of dielectric bodies is a very involved problem. 0) = (iω)−1 ∇ · ˆj̃ (r.. 1998) is not absolute. ω )] =[ f̂ i⊥ (r.151). It is referred to Tomaš (1995) for the classical treatment of multilayer struc- tures. Scheel et al. Another application is the quantization of the electromagnetic field in an inhomo- geneous medium that consists of two bulk dielectrics with a common interface. f̂ j (r . (1998) return to the simple configuration which was mentioned in Gruner and Welsch (1996a). (1998) define the vectors f⊥ (r.135) and (4. ω. Dung et al. 0).100) imply that ⊥() ⊥() ⊥() [ f̂ i (r. ( f̂ j (r . in general. Dung et al. ω. if ω → ∞. (4. ω) d3 s. ω. ω. [ϕ̂(r). ω).248 4 Microscopic Theories Moreover. ω) → 1. To make contact with the earlier work. ω) d3 s.150).154) The commutation relations (4. ω).157) It can be derived that the scalar potential operator ˆ 1 ρ̃ (s.152) hold. (1998) have proven that the fundamental equal-time commutation relations of quantum electrodynamics are preserved for an arbitrarily space-dependent Kramers–Kronig dielectric function.152) The commutation relations presented are equal-time Heisenberg picture ones and therefore it is emphasized that they are conserved. The necessity of a new calculation of the quantum electrodynamical commu- tation relations for a new three-dimensional configuration (cf. ω. the commutation relations (4. (4. ( f̂ j (r . ˆ (4. (4. (4. (4. (4. and (4. 0) = μ0 g(|r − r |. 0) d3 r . Dung et al. ω. ω).158) 4π 0 (ω) |r − s| where ρ̃ˆ (r.159) . ω) = δ ⊥ (r − s) · f(s. ω ))† ] = 0̂.156) The representation of transverse vector potential simplifies to Ã(r. 0) 3 ϕ̃ˆ (r. ω) = δ  (r − s) · f(s. It is shown that for the configuration under study. (4. Let us recall that the complex-valued dielectric function (r.153) f (r. ω) depends on fre- quency and space. (4.

Im ω > 0.160) ∂ω∗ Scheel et al. The partial differ- ential equation and the boundary condition at infinity determine the Green function uniquely. so that ∞ 1 Di j (r. with this upper limit. ω) dω × ∇ r = −iπ 1δ(r − r ) × ∇ r . (4. The left-hand side of relation (4.165) ∂τ . s. (4. (1998) have derived the relation ∞ iμ0 ωG i j (r. By comparison of relation (4.162) ∂ω∗ with ωG k j (r. s.161) −∞ c ← Here the left arrow means that the operators ∂ x∂  will first be written as ∂ ∂x  in the m m expansion of the Hamilton operator. s). ω) dω 2π −∞ ∂ = −μ0 G i jdir (r. t − τ ). s. we could derive that iμ0 ωG i j (r. because of causality.e. τ ) are components of the tensor-valued response function that causally relates the electric field E(r. s. s.2 Green-Function Approach 249 It is assumed that the real part (responsible for dispersion) and the imaginary part (responsible for absorption) are related to each other according to the Kramers– Kronig relations. ω) = 0.163) Second derivation of the Cauchy–Riemann equation (4. they arrive at the identity to be proven ∞ ω ← ← − 2 G(r. s.139). ω) is a holomorfic function in the upper complex half-plane of frequency ∂ (r. τ ) = e−iωτ D̃i j (r.138).164) 0 where Di j (r. Based on the partial differential equation (4. s. s.4. (4. (1998) study relation (4.162) consists in the appli- cation of ∂ω∂ ∗ to relation (4. τ ) dτ. (4. Knöll and Leonhardt (1992) calculate the time dependent. By comparison of the right-hand sides of this relation and relation (4.150). Im ω > 0. s. Scheel et al. ω) = 0.137) with a constitutive relation. ∂   ωG k j (r. ∇. This also implies that (r. ω) = D̃i j (r. an integral equation will be presented in what follows. τ ). ω) = eiωτ Di j (r.138) for the tensor-valued Green function. This could be useful in the combination with a time-dependent (direct-space) noise fdir (r. let us say a direct- space Green function. i. t) to an external current jext (s. r . ω) → 0 if |ω| → ∞. s.162) is then the unique solution of the homogeneous problem. ω) are holomorphic functions of ω in the upper complex half-plane. (4. (4.

ω) = Γ(r.170) c2 ω2 K 02 (ω) = 2 0 (ω). s. the integral equation for the tensor-valued Green function can be written as G(r. s. 0) ≡ g0 (r. From the theory of partial differential equations it is known (see. ω)]g(|r − s|. 0) is given by (4. v. s.167) where G(0) (r. (4.ω) d3 r of an integral equation. we first decompose the tensor-valued Green function into two parts.175) ← G2 (r. ω). ω) satisfies the integral equation G1 = G(0)1 + K · G1 d3 v. ω) d3 v. τ ) ≡ e−iωτ G i j (r. ω) → 0 if |ω| → ∞. s. ω) = G1 (r. ω)][∇v ln K 2 (v. ω).176) .250 4 Microscopic Theories where ∞ 1 G i jdir (r. ω2 K 2 (r. ω) = [1 − ∇r ∇s K −2 (s. (4. ω)]1. G(r. e. ω) · G(v. ω) ] + [K 2 (v. ω). ω) + K(r. ω) − K 02 (ω)]g(|r − v|. with K ik (r. s. s. s.168)   K(r. s. where K (ω) ≡ K 0 (ω). v. (4.173) where G1 (r.166) 2π −∞ is the direct-space Green function. s. (4. On introducing 0 (ω) ≡ D d3 r . Garabe- dian (1964)) that there exists an equivalent formulation D of the problem in terms (r. (4. ω) dω (4.169) Here g(r.150). ω) = [∇r g(|r − v|. v. ω) = g(|r − s|.172) To prove the fundamental commutation relation (4. v. an appropriately space- averaged reference relative permittivity. ω) = G (r. (4. ω)∇ s . ω) + G2 (r.g. s. ω)1. s. s. s. ω) are holomorphic func- tions of ω in the upper complex half-plane. (4.171) c It can be seen that the components of the kernel K ik (r. (0) (4. (4.174) with G(0) 1 (r.149). ω) = (r. (4.

it is shown that the scheme also applies to media with both absorption and amplification (in a bounded region of space). r .179) −∞ c2 −∞ c2 Thus. Artoni and Loudon 1998). s.176) Γ is the solution of the integral equation Γ = Γ(0) + K · Γ d3 v. (4. ω)]. (4.177) with Γ(0) (r.139).161). (4. only the noise-current response function iμ0 ωG1 contributes to commutator (4. for which the permittivity is a symmetric complex tensor-valued function of ω.181) Extensions of previous work on the propagation in absorbing dielectrics took linear amplification into account (Jeffers et al. quantization of the Hamiltonian formalism of the electromagnetic field using a method close to the microscopic approach was performed by Tip (1998).180) −∞ c2 ← The outer product of this equation and the operator (−∇ r ) can be taken and together with relation (4. Knöll et al. ω) dω × ∇ r = − G1 (r. respectively. (1998) derive that iμ0 ωG1 and μ0 ω2 Γ are the Fourier transforms of the response functions to the noise-current density and the noise-charge density.179) implies relation (4.176) and recalling that ∇ r × ∇ r = 0. r . ω). ω) = −∇r [K −2 (s. and an S-matrix formalism for scattering from . r .105) from the holomorphic properties of the tensors K and ωG1 that ∞ ω G1 (r. Multiplying the integral equation (4. A proper definition of band gaps in the periodic case and a new continuity equa- tion for energy flow were obtained.173) and (4. we see that the left-hand side of relation (4. 1996. we obtain as the derivation of relation (4.2 Green-Function Approach 251 In relation (4. ω) dω × ∇ r .161) can be rewritten as ∞ ∞ ω ← ω ← − G(r.174) by the function cω2 and integrat- ing over ω. Matloob et al.178) Scheel et al. i j (r. (1999) investigated quantum-state transformation by dispersive and absorbing four-port devices. Under the usual assumptions on the dielectric per- mittivity. (4.4. (4. In addition. ← ← Combining relations (4. ω) =  ji (r. ω)g(|r − s|. 1997. An extension of the quantization scheme to linear media with bounded regions of amplification is given and the problem of anisotropic media is briefly addressed. ω) dω = iπ1δ(r − r ).

252 4 Microscopic Theories absorbing objects was worked out. Using the Minkowski relations the Maxwell–Minkowski equations are derived. (2001) have proven the equivalence of two methods for quantization of the electromagnetic field in general dispersing and absorbing linear dielectrics: the Langevin-noise-current method and the auxiliary field method. He has used an appropriate form of the macroscopic Langevin equation. Clem- mow 1996). A macroscopic electromagnetic field has been quantized in a homogeneous linear isotropic dielectric by the association of a damped quantum-mechanical harmonic oscillator with each mode of the radia- tion field. Suttorb and Wubs (2004) have provided a microscopic justification of the phe- nomenological quantization scheme for the electromagnetic field in inhomogeneous dielectric due to Gruner and Welsch (1995) (cf. references in subsection 4.2. In classical optics the field of a beam can be represented in terms of its plane-wave spectrum (Smith 1997. A time-independent wave equation with the noise polarization and noise magnetization for the vector potential is derived. He has used a macroscopic Langevin equation for it. Results concerning quantum electrodynamics in dispersing and absorbing dielectric media have been reviewed by Knöll et al. The field vectors E and B are expressed in terms of the vector potential in the Weyl gauge. Tip et al. (2001). The time-average power passing through a plane is a sum of powers contributed by the propagating plane waves and by “counter-evanescent” pairs of plane waves with the same trans- verse components of their wave vectors. Matloob (2004b) has introduced a particular damped harmonic oscilla- tor. A canonical quantization scheme for the Langevin equation has been provided. One works with positive frequency parts of the fields. In Matloob (2005) a homogeneous medium is assumed that is isotropic in its rest frame. Petersson and Smith (2003) have illustrated the role of evanescent waves in power calculations for counterpropagating beams. with the electric induction vector independent of the magnetic strength vector and the magnetic-induction vector independent of the elec- tric strength vector. A path-integral formulation of quan- tum electrodynamics in a dispersive and absorbing dielectric medium has been presented by Bechler (1999) and has been applied on the microscopic level to the quantum theory of electromagnetic fields in dielectric media. Matloob (2004a) has paid attention to a damped harmonic oscillator. A macroscopic electromagnetic field has been quantized in a linear isotropic permeable dielectric medium by asso- ciating a damped quantum-mechanical oscillator with each mode of the radiation field. on considering the anisotropy of the material in the laboratory . When a line current is placed over a dielectric slab. The canonical quantization scheme has been followed.3). it is appropriate to insert a plane between the line current and the slab. Counterpropagating and “counter-evanescent” plane waves are defined relative to a selected plane.. The Minkowski relations are presented which are generalized constitutive relations for uniformly moving media. It is shown that the constitutive relations may be convenient. In this way the generation of Čerenkov and transition radiation have been investigated. The electric induction vector depends also on the magnetic strength vector and the magnetic-induction vector depends also on the electric strength vec- tor.

and considering both the trans- verse electric and the transverse magnetic polarized fields. The fields are quantized by expressing the noise-current density in terms of two infinite sets of appropriately chosen bosonic field operators. By expressing the classical Green function in terms of the classical light modes. Brun and Barnett (1998) con- sidered an experimental set-up using a two-photon interferometer. by a nonlocal susceptibility. (1987)) to absorbing media. They have derived that the complete form of the electric-field operator includes a part that corresponds to the free fields incident from the vacuum towards the medium and a particular solution which can be expressed by using the classical Green-function integral representation of the electromagnetic field.4. where insertion of a dielectric into one or both arms of the interferometer is essential. (1998). (1999) extended the field quantization to these material systems whose interaction with light is described. Di Stefano et al. a semi-infinite dielectric. (2001b) have based an electromagnetic-field quantization scheme on a microscopic linear two-band model. (1995) provided expressions for the electromagnetic-field operators for three geometries: an infinite homogeneous dielectric. and a dielectric slab. A simple quantum theory of the beam splitter. Suggestive is a comparative study of fermion and boson beam splitters (Loudon 1998).4 Optical Field at Dielectric Devices Matloob et al. Matloob. The vacuum field fluctuation is expressed. Their approach has provided a deeper under- standing of antibunching (Artoni and Loudon 1999). They have derived for the first time a noise-current operator for general anisotropic and/or spatially nonlocal media. which can be described only in terms of an appropriate frequency- dependent susceptibility.g. (2001a) have presented a one-dimensional scheme for the electromagnetic field in arbitrary planar dispersing and absorbing dielectrics.2.2 Green-Function Approach 253 frame. 4. (2000) have developed a quantization scheme for the electromagnetic field in dispersive and lossy dielectrics with planar interface. they have obtained a generalization of the method of modal expansion (e. Knöll et al. Di Stefano et al. Di Stefano et al. and Loudon (1995). The Green tensor is studied in reciprocal and spatial coordinate space. Di Stefano et al. near a medium boundary. including propagation in all the spatial directions. Artoni and Loudon (1997) applied the Huttner–Barnett scheme for quantization of the electromagnetic field in dispersive and absorbing dielectrics for the calcu- lations of the effects of perpendicular propagation in a dielectric slab and to the properties of the incident light pulse. was introduced by Barnett et al. . A microscopic derivation has shown that a canonical quantum theory of light at the dielectric–vacuum interface is possible Barnett. which can be applied to a Fabry– Pérot resonator. taking into account their finite extent. Fermions can be studied in analogy with bosons (Cahill and Glauber 1999). (1996) and developed by Barnett et al.

Ê(r.187) . it has been applied successfully to the study of input–output relations (Gruner and Welsch 1996b). . . ω). the region on the left of the plate ( j = 0). k. If Θ(z) is the unit-step function. λ j (z) = (4.184) Θ(z − z  ) for j = j  . . ω). Especially. Khanbekyan et al. [Θ(z − z  )]( j j ) = (4. r .186) (2π)2 where k = (k x . if z ∈ jth region. and f̂( j) (r. z  . ω) = λ j (z) j (ω). ω) = (4. The condi- tions have been stated. z  . σ ≡ ρ − ρ  .    G( j j ) (r. it can be represented as a two-dimensional Fourier integral 1 eik·(ρ −ρ ) G( j j ) (z. Khanbekyan et al.182) j=0 where ρ = (x. independent of ρ. k. These relations have been discussed for the case of the plate being surrounded by vacuum. Input–output relations for frequency components of this operator in the direct space have been given as well.   G( j j ) (z. r . then   Θ( j − j  ) for j = j  . ω) = (4. [Θ(z  − z)]( j j ) = (4. The  result of the shift of respective fields will be denoted by G( j j ) (r. under which the input–output relations can be expressed in terms of bosonic operators. multilayer plate).254 4 Microscopic Theories The Green-tensor formalism is well suited to studying the behaviour of the quantized electromagnetic field in the presence of dispersing and absorbing bodies. y) and  1. k y ) is the wave vector parallel to the interfaces and e−ik·σ G( j j ) (z. we shift the fields G(r. ω) and all the regions along the z-axis such that the jth region has the left boundary plane going through the origin ( j > 0) or the right boundary plane going through the origin ( j = 0). otherwise. z  . r . Ê( j) (r. The permittivity is  n (z.185) Θ(z  − z) for j = j  . For simplicity. (2003) studied the quantized field in the presence of a dispersing and absorbing multilayered planar structure (shortly. and f̂(r. . Since the Green tensor depends only on the difference ρ−ρ  . ω) d2 k. ω). The theory applies to effectively free fields and those created by active atomic sources inside and/or outside the plate. j = 1. ω). As a continuation of this work. ω). ω) d2 σ . and the region on the right of the plate ( j = n). (2003) consider n − 1 layers with thicknesses d j .   Θ( j  − j) for j = j  . ρ. (4. n − 1.183) 0. Three-dimensional input– output relations have been derived for frequency components of the electric-field operator in the transverse reciprocal space.

ω) = eq+ (k)eiβ j (z−d j ) + r j/n eq− (k)e−iβ j (z−d j ) . k.195) and q iβ d q iβ  d   1 t0/j e j j t0/j  e j j Ξqj j = q .191)  The Green tensor G( j j ) (z. k. ω) = −ez ez δ(z − z  ) + g( j j ) (z. k. ρ.197) . ω) = ( j) e−ik·ρ Ê( j) (z. ω) = eik·ρ Ê( j) (z. σs = −1).190) (2π)2 where f̂ (z.189) and the bosonic field operator may be written in the integral form as 1 f̂( j) (r. k. k. z  . (4. ω)[Θ(z  − z)]( j j ) . z  . z  . −k. (4. k. ω) d2 k. ω) may be written as a twofold Fourier transform: 1 Ê( j) (r. ρ. −k. ω) = σq Eq (z. ω) d2 k. (4. ω)Ξqj j Eq (z  . (4. k. (4. ω) = ( j) e−ik·ρ f̂( j) (z. ω) d2 ρ.193) (σp = 1. k. In equation (4. ω). ω)[Θ(z − z  )]( j j ) 2 q=p. ω) by the paper (Tomaš 1995) may be written as  δ j j  G( j j ) (z. ( j) q ( j) (4. k.4.192) k 2j where  i   j>  j<  g( j j ) (z. k.2 Green-Function Approach 255 The electric field operator Ê( j) (r.196) βn t0/n Dq j Dq j  with q q Dq j = 1 − r j/0r j/n e2iβ j d j .s j<  j>   + Eq (z. z  . (4. ω)Ξqj j Eq (z  . (4.193) j> Eq (z. ω) d2 ρ. (4. ω) = eq− (k)e−iβ j z + r j/0 eq+ (k)eiβ j z ( j) q ( j) (4.188) (2π)2 where Ê (z. k.194) j< Eq (z. k. ω) = eik·ρ f̂( j) (z.

k. k.202)   (n) (n) Ê(n) (z. k.203) Here. ω) = eq− (k) Ê qin (z. ω) . the transmission and reflection coefficients between the regions j  and j.200)  k  ( j) 1 k ep± (k) = ∓β j + kez . To obtain generally valid input–output relations.199) c q βj q q and t j/j  = t β j  j  /j and r j/j  . we restrict our attention to the electric-field operator.194) and ( j) (4. ω) = − e e−iβ0 z ĵ(0) (z  . ω) · eq− (n) (k) dz  (4. k. β j ≥ 0) (4. The unit vectors eq± (k) in equations (4. (4. k.205) 2βn z . This operator in front of the structure (superscript 0) and behind the structure (superscript n) is decomposed in terms of input and output amplitude operators   (0) (0) Ê(0) (z. the operators μ0 ω iβ0 z z  (0) Ê qin (z. (4. k. (4. (4. ω) . k. ω) · eq+ (0) (k) dz  (4. ω) = − e eiβn z ĵ(n) (z  .201) kj k If both the incoming fields incident on the two boundary planes of the plate are known.s (n) (n)  + eq+ (k) Ê qout (z. one can calculate the fields outgoing from the two boundary planes by means of input–output relations. ω) q=p. as well as the fields generated inside the plate. where + ω kj = ε j (ω) = k j + ik j (k j . ω) = eq+ (k) Ê qin (z. (4.204) 2β0 −∞ and ∞ μ0 ω −iβn z  (n) Ê qin (z. k. ω) q=p.195) are the polarization unit vectors for transverse electric (q = s) and transverse magnetic (q = p) waves ( j) k es± (k) = × ez .256 4 Microscopic Theories (d0 = dn = 0) and / βj = k 2j − k 2 = β j + iβ j (β j . These relations are valid also for evanescent-field components.s (0) (0)  + eq− (k) Ê qout (z. respectively.198) (k = |k|). k j ≥ 0). k. k.

k. Ê q+ (k.212) z=d j .2 Green-Function Approach 257 are input amplitude operators. k. ω) · eq± (k) dz  . ω) 0 10 1  n−1 ( j) φq0+ (k. ω) Ê qin (k. ω). + . (4. ω). ω) = Ê qin. (4.208) j=1 φqn+ (k. ω) = q q (n) Ê qout (k. ω) · eq+ (k) dz  (4. (4. (0) (0) .213) z=0 . Ê q− (k. ω). − . ω) · eq− (0) (k) dz  (4. ω) They relate the output amplitude operators at the boundary planes of the plate to the input amplitude operators at these planes. ω) rn/0 (k. ω) φqn− (k. ω) μ0 ω iβn z z −iβn z  (n)  + e e (n) ĵ (z . ω) r0/n (k. ω) = Ê qin. ω) = e−iβ0 z Ê qout (0) (k. ω) (n) Ê qin (k. ω) + ( j) ( j) ( j) . ω). . k.206) 2β0 z and (n) (n) Ê qout (z.out (z. (4.4.out (k. Ê qin. ω) μ0 ω −iβ0 z 0  + e eiβ0 z ĵ(0) (z  . . . k. k. ( j) ( j) Ê q± (k. ω) tn/0 (k. . ω) φq0− (k. ω) = − (4. k. ω) t0/n (k. ω) = Ê q+ (z. k.out (z. which may have been better denoted say by F̂q± (k. ( j) ( j) . The operators (0) Ê qout (z. ( j) ( j) .207) 2βn 0 are output amplitude operators. 2. ω) ( j) ( j) Ê q+ (k.210) z=0+ and to the amplitude operators associated with the layers μ0 ω dj  e∓iβ j z ĵ( j) (z  . ω). where the Green function is of a very simple form. k. . (4. ω) Ê q− (k. . Ê qin. since .211) 2β j 0 ( j) j = 1. ω) = eiβn z Ê qout (k.out (k. where the Green function has the complicated form and is “hidden” in the input–output relations 0 1  0 1 q q (0) (0) Ê qout (k.209) z=0− . ω) = Ê q− (z. k. n − 1. . (n) (n) .

1994). Dung et al.211) read q q ( j) t j/0 e2iβ j d j q ( j) t j/0 φq0+ = r j/n . 4. The calculations have been performed on the assumption of a dielectric with a single resonance. φqn− = r j/n . (2001) have studied nonclassical decay of an excited atom near a dispersing and absorbing microsphere of given complex permittivity that satisfies the Kramers–Krönig relations laying emphasis on a Drude–Lorentz permittivity. The theory has been applied to the spontaneous decay of a two-level atom placed at the centre of a three-layer spherical microcavity.214) Dq j Dq j q q ( j) t j/n eiβ j d j ( j) t j/0 eiβ j d j q φqn+ = . Dung et al.5 Modification of Spontaneous Emission by Dielectric Media Scheel et al.258 4 Microscopic Theories In equation (4. The tensor-valued Green function of the configuration has been known (Li et al. the coefficients at the operators (4. (2000) have developed a formalism for studying spontaneous decay of an excited two-level atom in the presence of arbitrary dispersing and absorbing dielectric bodies. φq0− = . Weak and strong couplings are studied and in the study of the strong couplings both the normal-dispersion range and the anomalous-dispersion range associated with the band gap are considered. (1999a) have found quantum local-field corrections appropriate to the spontaneous emission by an excited atom.208).2. spherically. The theory applies to effectively free fields and those created by active atomic sources inside and/or outside the plate. The formalism is based on a source-quantity representation of the electromagnetic field in terms of the tensor-valued Green function of the classical problem and appropriately chosen bosonic quantum fields.215) Dq j Dq j It is worth noting that any two planes z = z (0) ≤ 0− and z = z (n) ≥ 0+ for j = 0 and j = n. Among others. All relevant information about the bodies such as form and dispersion and absorption properties is contained in the tensor-valued Green function. (4. (4. the cavity input–output coupling dominates the strength of the atom–field interaction. also can be used in principle for a formulation of the input–output relations.b) have given a rigorous quantum-mechanical deriva- tion of the rate of intermolecular energy transfer in the presence of dispersing and . it has been assumed that the atom is positioned at the centre of the cavity. (2002a. they have found a condition on which the decay becomes purely non- radiative. the significant effect within the band gap is the photon absorption by the wall material. and cylindrically multilayered media (Chew 1995). Whereas in the range of normal dispersion. This function has been available for various configurations such as planarly. Dung et al. the wall being modelled by a Lorentz dielectric. For simplicity. They have shown how the minimal-coupling Hamiltonian simplifies to a Hamiltonian in the dipole approximation. respectively.

In case an interface is considered the propagation direction of the pulse is assumed to be perpendicular to the surface. He has expressed the displacement D(r. 2003) and arrive at the paper (Padgett et al.4. In two steps. but it remains in one term. It has also been shown for excited atoms in a photon crystal with transition frequency in a band gap that their states do not decay radiatively. t  ≤ t. Inoue and Hori (2001) have developed a formalism of quantization of electro- magnetic fields including evanescent waves based on the detector-mode functional defined in terms of those for the widely used triplet modes. He considers the transfer of angular momentum to a dielectric. which produces only a longitudinal force. New forces are produced by a pulse of Laguerre–Gaussian light in comparison with a plane-wave pulse. multislab planar structures. The pulse is assumed to contain a single photon. where they also have made comparison with experiments. then it is assumed that the dielectric is – weakly – attenuating to ensure that the model need . Assuming the system to be in a superposition state of all of the lower levels initially they have determined the conditions of complete opacity or transparency of the medium. The coupling of pulses is most interesting in the limit of parametric generation. the representations cease to manifest the generalized Coulomb gauge used. without adding a noise polarization term. If the pulse is propagated into a dielectric.2 Green-Function Approach 259 absorbing media with spatially varying permittivity. Tip (2004) has used his auxiliary field method to obtain various equivalent Hamiltonians for charged particles interacting with absorptive dielectrics. A simplification is achieved by assuming that the modal func- tion has zero radial index. Matloob and Pooseh (2000) have discussed a fully quantum-mechanical theory of the scattering of coherent light by a dissipative dispersive slab. t) merely in terms of the electric-field E(r. Instead of a planar pulse he consid- ers Laguerre–Gaussian light beams. Specifi- cally. These are radial and azimuthal forces. concentrated in a wave operator. theoretical studies can take a traditional approach. Matloob and Falinejad (2001) have calculated the Casimir force between two dielectric slabs by using the notion of the radiation pressure associated with the quantum electromagnetic vacuum. They have shown that the minimal-coupling scheme and the multipolar-coupling scheme yield exactly the same form of the rate formula. For a transparent dielectric. In the framework of a semiclasical approach. A simplified approach to the quantization is sufficient for the theory of the radiation pressure on dielectric surfaces (Loudon 2002). They have evaluated atomic and molecular radiation near a dielectric boundary surface. Matloob (2001) has postulated an electromagnetic-field Lagrangian density at each point of space–time to be of an unfamiliar form comprising the noise-current density. 2003). they have used the fact that only the field correlation functions are needed for the evaluation of vacuum radiation pressure on an interface. and to microspheres. t  ). He may issue from the book (Allen et al. They have applied the theory to bulk material. Loudon (2003) continues (Loudon 2002) with two changes or extensions. Paspalakis and Kis (2002) have studied the propagation dynamics of N laser pulses interacting with an (N +1)-level quantum system (one upper state and N lower states).

2π ηL ρL w0 c (4.219) The/radial force compresses the dielectric towards the cylinder of radius ρ0 = w0 |l|2 . . he+lets : f̂ ρ (r. or finiteness of the dielectric in the direction of propagation. η ≡ η(ω0 ) is a refractive index. The author introduces the normal-order Poynting operator :Ŝ(r. t):. As usual. t): denote the radial component of the operator :f̂(r. The time integral of equation (4. the author can write the expectation value   ω0 c 2 2c2  ηz 2 1|: Ŝz (r. t):. The single-photon pulse is represented by the state vector |1 . and l is the orbital angular-momentum quantum number. t):|1 = √ − 2 exp − 2 t − |u|2 . the author constructs the normally ordered angular-momentum density operator. He determines that  2ω0 c(η2 − 1) 2  ηz  1|: f̂ z (r.260 4 Microscopic Theories not complicate by including any exit or reflection. t):|1 = exp − 2 t − |u|2 . He introduces the Lorentz force-density operator :f̂(r. He does not speak of the associated Laguerre polynomials. t):. Loudon (2003) gives Laguerre–Gaussian light in terms of the Lorentz-gauge vec- tor potential. where w0 is the beam waist. and u ≡ u k0 . ρ = x 2 + y 2 . If the dispersion is ignored.216) L π L c where L is a conventional length of the pulse. when he restricts himself to the radial index p = 0. k0 = η(ωc0 )ω0 is an angular wave number. (4. t): denote the longitudinal component of this operator. The spectrum of the photon wave packet is a narrow-band Gaussian function. with z-component denoted by : Ŝz (r. He determines that    2ω0 (η2 − 1) 2c2  |l| 2ρ ηz 2 1|: f̂ ρ (r. (4. ω0 is a central frequency of the wave packet.216) is ∞ 1|: Ŝz (r.217) −∞ Further.218) L c Similarly. The theory is quantized. (4. t):|1 dt = ω0 |u|2 . t): and lets : f̂ z (r.l (r) is the modal function. but it is obvious that the degree of a polynomial is zero. t):|1 = − t − L3 π c 2    2c ηz 2 × exp − 2 t − |u|2 .

He determines that  4c2 (η2 − 1) 2  ηz  1|: f̂ φ (r. t) dr. (4. and κ are evaluated at frequency ω0 . The modification of the result (4. (4. <⎪ ⎭ surface bulk 2ω0 η + 1 + κ 2 2 = . is ⎧ ⎫ ∞ ⎪ ⎪ ⎪ ⎪ 2ω0 ⎨ η2 + κ 2 − 1 2 ⎬ 1|: F̂z (t):|1 dt = + c ⎪⎪ (η + 1)2 + κ 2 (η + 1)2 + κ 2 ⎪ −∞ ⎩= >. t): denote the azimuthal component of the operator :f̂(r.4. As usual. or the total linear momentum transfer to the dielectric. The author gives particular attention to the transfer of longitudinal and angular momentum to the dielectric from light incident from free space. He lets the total force on dielectric in {z > 0} and at time t be represented by the force operator F̂(t) = f̂(r. The author extends the theory to the case.221) at z > 0.2 Green-Function Approach 261 Similarly. t):|1 dt = ω0 exp − |u|2 .224) c (η + 1)2 + κ 2 = >.225) z>0 . Next he lets the total torque on the dielectric in {z > 0} and at time t be represented by the torque operator Ĝ z (t) = ĝz (r.223) z>0 and lets F̂z (t) denote the longitudinal component of this operator. η.220) L c ρ ρ w0 where σ is the spin angular-momentum quantum number of the beam. φ = arg(x + iy).217) at z > 0 is ∞   2ω0 κz 4η0 η 1|: Ŝz (r. where space is divided into two regions with a dielectric of real refractive index η0 (ω) at z < 0 and a dielectric of a complex refractive index n(ω) = η(ω) + iκ(ω) (4. t) dr (4. t):. The time-integrated force.222) −∞ c (η0 + η)2 + κ 2 where η0 . (4. (4. t):|1 = − t − ηL 3 π c 2     2c ηz 2 l σ |l| 2σρ × exp − 2 t − − + 2 |u|2 . he lets : f̂ φ (r. < total In pursuit of the torque the author first introduces the operator that represents the density of the z-component of the torque on the dielectric :ĝz (r. t):. < = >.

The extremely low group velocity is caused by the electromagnetically induced transparency of an atomic transition. Very long pulses that are well tuned to a region of anomalous dispersion do not have superluminal peak velocity of a real physical significance. collective vibrational degree of freedom of a massive mirror has been proposed (Zhang et al. (4. <⎪⎭ surface bulk 4(l + σ )η = 2 . The momentum transfer from light to a dielectric material has been calculated by evaluation of the relevant Lorentz force. Following a brief discussion of one of known atomic stopping-light schemes. Similarly (but for κ = 0 only) the angular rotation of the dielectric slab. It has been found that the Hamiltonian used by . Leonhardt and Piwnicki (2001) have analysed the propagation of slow light in moving media in the case where the light is monochromatic in the laboratory frame. < = >. (2005) have paid attention to the photon drag effect. an all-optical scheme has been analysed in detail. The photon drag effect is named after the generation of currents or electric fields in semiconductors. one that obeys a Poisson equation with zero-frequency limit of the permittivity. The atomic interaction potential changes from the Coulomb one to a static potential. 2003). surface transmitted.262 4 Microscopic Theories The time-integrated torque. Tip (2007) has studied the properties of atoms close to an absorptive dielectric using his quantized form of the phenomenological Maxwell equations. Ough- stun and Cartwright (2005) have compared the group velocity with the instantaneous centroid velocity of the pulse Poynting vector for an ultrashort Gaussian pulse. The proposal may realize an Einstein–Podolsky– Rosen state in position and momentum for a pair of massive mirrors at distinct locations by exploiting a nondegenerate optical parametric amplifier. due to such a wave packet. Lombardi (2002) has re-examined the physical significance of different velocities which can be introduced for a wave train. Yanik and Fan (2005) have formulated basic principles that underlie stopping and storing light coherently in many different physical systems. is ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ∞ 4(l + σ )η ⎨ η2 + κ 2 − 1 1 ⎬ 1|:Ĝ z (t):|1 dt = + 2 (η + 1)2 + κ 2 ⎪ ⎪ η2 + κ 2 η + κ2 ⎪ −∞ ⎩= >. A scheme for transferring quantum states from the propagating light fields to a macroscopic. or total angular-momentum transfer on dielectric.226) η + 1 + κ2 = >. due to a normally incident single-photon wave packet. i. and bulk transmitted contributions pro- vided a photon has passed through. Loudon et al. The author has treated the coupling of atoms with longitudinal modes in detail. whose moment of inertia around the z-axis is denoted I may be calculated. The longitudinal interactions of atoms with absorptive dielectrics are responsible for nonradiative decay of the atoms. From this the shift of a slab of mass M may be calculated. < total It is shown that it is meaningful to divide the total transfer of linear momentum into surface reflected.e.

(4. Veselago has shown that. E = E0 ei(k·r−ωt) . First we summarize the basic electromagnetic properties of left-handed materials. although such materials are not available in the nature.228) c2 Propagation is possible if k2 > 0. where μ = 1 and  is real.229) Using them in the identities k · (E × H) = H · (k × E) = E · (H × k). the permittivity of which is negative. Veselago (1967.227) c where k is the wave vector and ω is the frequency. k × H = −ω0 E. (4. 1968) was the first to address the question of propaga- tion of the electromagnetic waves in the medium with both the permittivity 0  and the permeability μ0 μ negative. (4.231) or k · S < 0. is described by the wave equation  ω2 k2 − 2 μ E(r. 4. (2002b) is unitarily equivalent to a special case of the Hamiltonian used by Tip (2007).6 Left-Handed Materials The interest in “left-handed” materials is reflected both in the theory and in the experiment. ω2 k2 = μ. (4. their existence is not excluded by the Maxwell equations Markoš 2005).2. This is always true in dielectrics.233) . The Maxwell equations simplify to the form k × E = ωμ0 μH.4. (4. (4.2 Green-Function Approach 263 Dung et al.230) we can derive that k · (E × H) = ωμ0 μH2 = ω0 E2 . while no propagation is possible in metals. The propagation of the electromagnetic wave.232) where S=E×H (4. t) = 0. Materials with both  and μ negative allow the propagation of electro- magnetic waves.

235) ∂ω ∂ω The permittivity and permeability determine the index of refraction. The energy density (4. the minus sign is used in (4. Z = . since (Landau et al. As the wave vector ω k= ns. and the relative impedance. Veselago also has pointed out that the left-handed material must be dispersive. (4. and k are a left-handed set of vectors. 1984) ∂(ω) ∂(ωμ) > 0. with a centre or carrier frequency ω0 > 0. The three vectors E.236) is used when the square root has a positive imaginary part and the minus sign in (4.237) s·k where s is the direction of the Poynting vector S.   1 ∂ [ω0 (ω0 )] 2 ∂ [ω0 μ(ω0 )] 2 U = 0 E + μ0 H . contrary to conven- tional materials. where they are a right-handed set.238) c .  or μ must depend on the frequency of a monochromatic field. From the Kramers–Kronig relations it follows that  and μ must be complex. since the time- averaged energy density of the electromagnetic field. or nonnegative imaginary part if the real part vanishes. (4. that Im  and Im μ may be neglected. > 0.e. (4. Z ≡ Z (ω). i. (4. Due to continuity. This inspired Veselago to name the materials with both  and μ negative left-handed materials.234) is positive. (4. 2π ω0 .234) 2 ∂ω0 ∂ω0 would be negative.264 4 Microscopic Theories is the Poynting vector.236) is used when the square root has a negative imaginary part. we may assume Im  > 0 or Im μ > 0 small and apply the rule. Many results have been derived on the assumption of a transparent medium.236)  It is assumed that the complex square root has a positive real part. Even though the assumption that Im  = Im μ = 0 is frequently used and it invalidates the rule of the sign. Here the electromagnetic field is assumed to be quasimonochro- matic. The original Gaussian units of measurement can be respected by the replacements 0 → 4π 1 . Strictly speaking. μ0 → 4π 1 . It means that the wave vector k and the Poynting vector S have opposite directions. n ≡ n(ω). by the relations  √ μ n = ± μ. We introduce a phase velocity vector ω vp ≡ s. The time averaging is done over the period of the carrier. The plus sign in (4. H. we should speak of the medium with both Re  and Re μ negative.236) for  < 0 and μ < 0.

Now we introduce a group-velocity vector 1 vg = ∂k s. Re n < 0. Still it holds that − + + + + ω + + + + + k1x = k1x .4. We introduce n 1 = 1 and n 2 = n. but also for the left-handed materials. not the coordinate. the negative refraction occurs when the refractive index n is negative. the negative refraction occurs.240) ∂ω 2c ∂ω ∂ω ∂ω where Z still means the relative impedance. we + − + suppose that s1z > 0. We consider a planar interface in the plane z = 0 between the half-spaces z < 0 and z > 0. k2x = k1x . k2x = ωc n 2 s2x = ωc ns2x . Artificial structures were first proposed. + + From this. b > 0 is the distance from the rear plane to the image.2 Green-Function Approach 265 we have let vp denote nc s. Its imaging is described by the equation a + b = l. Therefore. Quite reasonably. The half-space z < 0 is free and the half-space z > 0 is filled with a left-handed material. (4. For n < 0 or. and s2z > 0 both for the right-handed media and the left-handed materials. We assume an + i(k+ incident monochromatic wave with E+ 1 = E10 e 1 ·r−ωt) . and k+ + − 2 are the respective wave vectors. The image is real and direct. s· > 0. Let us note that k1x = c n 1 s1x = ωc s1x . It is worth noting that the left-handed material can enhance incident evanescent waves (Pendry 2000). and a transmitted one with E2 = E20 e 2 i(k ·r−ωt) . more generally. the slab of a material with  = μ = −1 does not reflect light. As we will show below. and l is the thickness of the slab. A periodic array of very thin metallic wires has the negative permittivity. These remarkable properties have led to the term a perfect lens for the planar slab. and s2 mean their directions. Let s1 . S1 . (4. s1z < 0. s1 . A planar slab of a material with  = −1 and μ = −1 can be compared with a lens. Then k1y = 0 and k2y = 0 by the isotropy. and S2 . Here k1 . On writing the wave vector k in the forms ω√ k=± μ s. ns2x = s1x . which have negative permittivity and permeability in the microwave region of frequencies. k1 . where a > 0 is the distance from the object to the front plane. s · vg > 0. though not amplified. The respective Poynting vectors may be denoted by S+ − + + − + 1 .241) s · ∂ω It has the same direction as the Poynting vector not only in the right-handed media.239) c we obtain that  ∂k 1 ∂(ω) ∂(ωμ) ∂k = Z + Z −1 s. (4. We let a mean the spatial period of . a reflected wave with E − 1 = − i(k− + + + E10 e 1 ·r−ωt) . + − + We consider k1y = 0. Let us illustrate that the derivation of the Snell law is valid also for a negative refractive index.

of the electromagnetic wave through a homogeneous slab read 1 i = cos(nkl) − (Z + Z −1 ) sin(nkl). it was argued that the negative refraction is ruled out by the causality principle (Valanju et al. (Shelby et al. the index of refraction and the impedance have been derived from the numerical data (Smith et al. . 2002). Similarly. The reports on first experiments.b). t. Numerical simulations of the transmission of the electromagnetic waves through the left-handed medium offered an independent possibility to verify the theoretical predictions (Ziolkowski and Heyman 2001.246). 2002).245) and (4. (4. Absorption was suggested as an alternative explanation of the experiment with negative refraction (Sanz et al. 2003). Markoš and Soukoulis 2002a.242) ω(ω + iγe ) where √ 2πc ωp = + a (4. and the reflection amplitude. For- mulae for the transmission amplitude. (1996) have expressed the response of this medium to the external electric field parallel to the wires (of a two-dimensional lattice) using the effective permittivity (a three-dimensional lattice has been considered first) ωp2 eff ≡ eff (ω) = 1 − . The origin of absorption has been traced up (Markoš et al. r = 0 for  = μ = −1.245) t 2 r i = − (Z − Z −1 ) sin(nkl).244) ω2 − ω02 + iΓω Here ω0 is the resonant frequency and Γ is the absorption parameter. The parameter F is the filling factor for the split ring.g. 2002). On the assumption that the wavelength of the electromagnetic wave is much larger than the spatial period of the left-handed structure and on using relations (4. (1999) that a periodic array of split- ring resonators behaves as a medium with negative magnetic permeability.266 4 Microscopic Theories the lattice and r mean the radius of wires.243) a ln( r ) is the plasma angular frequency and γe is the absorption parameter. Combination of both structures gives rise to the material with both negative permittivity and permeability—the left-handed material. (4. e. it was predicted in Pendry et al. Especially. 2001) were followed by some criticism. Pendry et al. (4. r. (4.246) t 2 where k = ωc conventionally. For instance. The response of the regular lattice to the external magnetic field perpendicular to the plane of the ring is given by the effective permeability Fω2 μeff ≡ μeff (ω) = 1 − . t = exp(inkl).

then m 2eff ≥ 0 for the electromagnetic parameters characteristic of (Ruppin 2000). Their numerical calculations have described a two-dimensional photonic crystal. a wave hitting the photonic crystal interface for that frequency will undergo negative refraction similar to a wave hitting the interface of a homogeneous medium with negative index n. Ruppin (2002) has obtained a modification of formula (4.4. a progress has been reported on Dolling et al. (4. Shen (2004) has defined a frequency-independent effective rest mass of a photon. a finite extent line source was placed outside a photonic crys- tal at an angle of 30◦ .234) for the time- averaged energy density. but it has not been asked whether the effective permittivity and the effective perme- ability may be defined.247) 2 ω=0 If the left-handed medium can be modelled as two-time derivative Lorentz material (Ziolkowski 2001).2 Green-Function Approach 267 For applications it would be very interesting to pass from the microwave to the optical frequencies. They concentrated themselves on clarification of some of the controversial issues. we may find it from the relation m 2eff c4 = − res ωn 2 (ω). For the photonic crystal system a frequency range exists for which the effective refractive index is negative. an effect similar to the negative refraction may occur in a right-handed medium. the wave exibits the negative refraction. So. which respects the absorption in the medium. 2003). Otherwise. Foteinopoulou and Soukoulis 2003) and used in experiments (Cubukcu et al. Γe > 0. In their simulations. (2003) have used finite-difference time-domain simulations to study the time evolution of an electromagnetic wave as it reaches the interface. 2003. Foteinopoulou et al. Optical frequencies have been assumed. after a long time. Pendry (2003) has provided an approval of the contemporary reports on experimental proofs of some properties of the materials with negative refractive index. We cite only the time-averaged energy flux S = U vg and the time-averaged momentum density p = Uω k. (2007). The wave is trapped temporarily at the interface. Photonic crystals have been analysed in theory and numerical calculations (Notomi 2000. The source starts emitting at t = 0 an almost monochromatic TE wave. and. reorganizes. or from the GHz to THz region. In 2007. (2003) present in fact all the interesting results of Veselago (1968) in the lossless case. Letting m eff denote this mass. but is restricted to a relative permittivity (ω) and a relative permeability μ(ω) of the following forms: ωp2  ≡ (ω) = 1 + . The source is adjusted to generate a Gaussian beam. ωr2 − ω2 − iΓe ω . Foteinopoulou et al. It is worth noting that the effective medium is two-dimensionally isotropic. Foteinopoulou et al.

For simplicity. Veselago (2002) has presented a miniature review of the progress in negative- index materials. Pendry (2003) has provided an approval of the contemporary reports on experi- mental proofs of some properties of the materials with negative refractive index. (4. a form of Maxwell’s equations suggests a different transformation  π  π Efd = E cos α − Z 0 Z H sin α . If a variation is restricted to the negative-index medium. The dif- ferential of the optical length is the differential of the Euclidean length multiplied by the refractive index. it is a generalization of the result obtained by Loudon (1970) in the case of no magnetic dispersion. the optical length of the sought path is a local maximum. μ = Re μ. Naqvi and Abbas (2003) have noted that principle of duality has a somewhat dif- ferent form in negative-index materials.250) 2 2 For the negative refractive index. the optical length of the path taken by a ray of light in travelling between two points is a local minimum. The formulation stating that the optical length is stationary is correct.248) ω2 − ω02 + iΓh ω At the same time. Engheta (1998) has expressed a continuous transition from the original solution (α = 0) to the dual solution (α = 1) in the case of the positive-index medium  π  π Efd = E cos α + Z 0 Z H sin α . conventionally.268 4 Microscopic Theories Fω02 μ ≡ μ(ω) = 1 − . (4. The subject of that paper which comprises such a review is the formulation of Fermat’s principle. in the version for the negative-index material. μ = Im μ. Γh > 0.249) 2 Γe Γh where   = Re . the replacement α ↔ −α is made. After a similarity to the left-handed medium had been achieved. 2 (4.  π2   π2  Z 0 Z Hfd = E sin α + Z 0 Z H cos α . we may consider differentials only for a Euclidean space. (4. the waveguide transmitted waves at about 6 GHz. (2002) have measured transmission in a hollow metallic waveg- uide. Marqués et al. A standard analysis of a metallic waveguide . The time-averaged energy density is      1  2ω   2ωμ U = 0  + |E| + μ0 μ + 2 |H| . They have found that it behaves similarly as the periodic array of thin metallic wires with its negative effective permittivity both when unloaded and when loaded.   = Im . If a variation of a path is restricted to the positive-index medium.  2π  2 π  Z 0 Z Hfd = −E sin α + Z 0 Z H cos α .251) 2 2 The two transformations are identified when.

D0 = D (0). They have investigated the effective refractive index spectra of a granular composite. They investigate a three-phase composite. Huang and Gao (2003) have been motivated by the paper (Chui and Hu 2002). while they had the longest waveguide with l = 36 mm. It holds that ω0NL (H) = ω0 X. The transmission may have been radiationless. A2 =  3 ω2 h 2 16 D0c20 . (4. (2003) have noticed that so far properties of left-handed materials in the nonlinear regime of wave propagation have not been studied. Quite remarkably. As an application they consider a linear dependence which corresponds to the Kerr nonlinearity. respectively. D ≡ D (|E|2 ).244) by the dependence of the eigenfrequencies of oscillations on the magnetic field.253) where X is one of the stable roots of the equation (1 − X 2 )[(X 2 − Ω2 )2 + Ω2 γ 2 ] |H |2 = α A2 E c2 . the left-handed composite medium can be realized.4. The assertion that the trans- mission of electromagnetic waves occurs due to the split-ring-resonator medium anisotropy has been dismissed by experiment (Marqués et al. a composite medium which is left handed in a certain frequency region can be prepared. They cal- culate the effective permittivity and the effective permeability as well based on the Clausius–Mossotti relation (Grimes and Grimes 1991).2 Green-Function Approach 269 which still utilizes the effective magnetic permeability does not admit a radiation mode and contradicts the experiment. (4.242) and a third-order nonlinear term. The effective nonlinear dielectric permittivity eff (|E|2 ) is found to be a sum of the earlier result (4.252) ω0NL 2 (H) − ω2 + iΓω differs from the earlier result (4. and γ = ωΓ0 . Marqués et al. Zharov et al. the hollow waveguide behavesas a one-dimensional plasma ω2 with effective permittivity eff ≡ eff (ω) = 0 1 − ωc2 . α = ±1 stands for a focusing or defocusing nonlinearity. Especially. h is the width of the ring. They assume that a metallic structure is embedded into a nonlinear dielectric with a permittivity which depends on the strength of the electric field in a general way. D (|E|2 ). (2002) worked with the wavelength of 5 cm. The effective magnetic permeability of the composite structure (for F  1) Fω2 μeff (H) = 1 + (4. Another calculation has associated a mode with the split-ring-resonator medium anisotropy (Kondrat’ev and Smirnov 2003). Numerical results show that. .254) X6 where H is an appropriate component of the field H. by controlling the volume fraction of dispersive spherical particles in nondispersive host medium. in which metallic magnetic inclusions are embedded into the host medium. E c is a characteristic electric field. 2003). by embedding dielectric and magnetic granules into the host medium and controlling the volume fractions of the two sorts of the granules.

But the magnetic term of the energy density is expressed using the frequency-dependent magnetic permeability. the results are applicable to a wide range of interesting and practical situations. and γ̃e = ωγpee and γ̃m = ωγmpe are the respective electric and magnetic loss terms normalized with respect to the electric plasma frequency. although restricted in their range of validity.e. t) = Re −i Cω α(t)∇ × Fω (r) . It has been demonstrated that a negative-index material allows an ultrashort pulse to propagate with minimal dispersion (D’Aguanno et al. where absorption is negligible. It is assumed that the mode function is nor- malized. The negative-index material has been described with a lossy Drude model  2 1 ωpm 1 (ω̃) = 1 − . He derives the electromagnetic energy density in a form which resembles the time-averaged linear dispersive energy (3. i. 2005). ωpe and ωpm are the respective electric and magnetic plasma frequencies. Milonni (1995) selects any frequency region.255) ω̃(ω̃ + iγ̃e ) ωpe ω̃(ω̃ + iγ̃m ) where ω̃ = ωωpe is the normalized frequency. at the same level of generality as the electric term. |Fω (r)|2 d3 r = 1. i. 2004).270 4 Microscopic Theories A wave packet description allows an easy grasp of negative refraction (Huang and Schaich 2004). For some frequency ω a mode function Fω (r) can be considered which satisfies the transversality condition and the Helmholtz equation ∇ · Fω (r) = 0. He assumes a uniform (or homogeneous) dielectric medium. (4. Attention has been paid to the group-velocity dispersion parameter d2 k β2 = dω 2 [Agrawal (1995)].260) ω . (4. For a macroscopic quantization. It has been noted that these points are related to ω < ωpe and that no zero group-velocity dispersion point is present when ωpm ωpe = 1.256) ω2 ∇ 2 Fω (r) + (ω)μ(ω)Fω (r) = 0 (4.186). (4. The frequencies for which β2 = 0 are plotted as zero group-velocity dispersion points in figures. The formulation simplifies and. (4. away from absorption resonances. (4.258) The monochromatic components at frequency ω of the fields are E(r.259)  c  B(r.257) c2 and appropriate boundary conditions. t) = Re{Cω α(t)Fω (r)}.e. The interesting properties of the left-handed materials have been illustrated on the realistic assumption of losses and incidence of a Gaussian beam (Cui et al. μ(ω̃) = 1 − .

The fields are expressed with these relations as in the classical optics. t) = −i 1 μωk e â e − H.266) k.261)   i c H(r. Ignoring that also the number of such narrow-band fields may be greater than one. or comment on.207) for Eν .263) 8π μ(ω) dω The choice  4π μ(ω) Cω = (4. (4. t) = Re − Cω α(t)∇ × Fω (r) . t) = Re[(ω)Cω α(t)Fω (r)].262). The energy is determined as the integrated energy density in the form n(ω) d H = |Cω |2 |α(t)|2 [n(ω)ω]. with an amplitude α(t). Let us remark that this approach leads also to twice as many annihilation and creation operators as expected. (4. .264) n(ω) d[n(ω)ω] dω of the normalization constant gives the energy in the form 1 H = |α(t)|2 . but also with a normalization constant Cω for the electric strength field.2 Green-Function Approach 271 D(r.4. in (4. (4.267) k. As different systems of units have been utilized we will compare some formulae of Milonni (1995) with those of Drummond (1990).262) μ(ω) ω where α(t) = α(0) exp(−iωt).265) 2 which corresponds to a harmonic √ oscillator. . Restricting himself to a single carrier frequency (ν = 1). (4.206) on substituting the first of equations (3. this ratio reminds us of the microscopic model assuming an oscillator medium with a single resonance. Drummond (1990) has presented the following expan- sions: F G G H  ∂k ∂ωk1   1 ik·x−iωk1 t D̂ (x.λ 2V k ζ̃1 (ωk ) 1 F G  G  ∂ωk1   1H ∂k 1 1 ik·x−iωk1 t B̂ (x. Milonni (1995) does mention (Drummond 1990). (4. the first and the second time derivative of the electric displacement which we obtain in (3. c. The approach of Drummond (1990) enables one to respect the inhomogeneity of the medium as noted in (Milonni 1995).259)–(4. For the quantization of this oscillator we do the replacement α(t) → 2ω â(t) where â(t) is an annihilation operator.λ 2V k ζ̃1 (ωk1 ) kλ kλ . (4. but he does not expound. t) = i 1 k × e1kλ âkλ e − H. c.

.273) 2 For a multimode field. we obtain a normalization constant in the form  1 √ n(ω)ω Cω 2ω = ..272 4 Microscopic Theories where V is the quantization volume and (cf.268) μ Regarding wide bandwidths or a simplification merely. (4.220))  ζ̃1 (ωk1 ) ωk1 = k .269) and dividing by . (4. abandoning the Taylor expansion as in relation (4.266).273) is replaced by F G G 2πωβ μ(ωβ ) † . the electric-field operator (4. c.269) k. t) = âkλ ekλ eik·x−iωk t + H.272) 2 n(ω) d[n(ω)ω] dω in conformity with Milonni (1995). we have also  1 √ μ(ω)2π ω Cω 2ω = (4. (4.λ 2V k ζ̃ (ωk ) where −1 ∂ωk ωk ωk ζ̃  (ωk ) = 1− . t) = Cω 2ω âω Fω (r) + âω† F∗ω (r) . Drummond (1990) has com- pleted the following expansion:    ∂ω k   ∂k Λ̂(x. the replacement 2π ↔ 210 is performed.270) ∂k k 2ζ̃ (ωk ) Using relation (4. (4. (3. (4.271) 2 2 d[n(ω)ω] dω Since μnr = nr and for different units in the relations under consideration. The normalization constant is obvious from the expression 1 √   Ê(r.

∗ Ê(r. t) = H d[n(ω )ω ] â β (t)Fβ (r) + â β (t)Fβ (r) .275) c2 .274) β n(ωβ ) dωββ β where Fβ (r) is a mode function for mode β obeying the transversality condition and the Helmholtz equation ωβ2 ∇ 2 Fβ (r) + (ωβ )μ(ωβ )Fβ (r) = 0. (4. (4.

282) cV vp (ωk )V is used.279)   2πωk n(ωk )(ωk ) D̂(r.280) the factor   2πωk n(ωk )(ωk )vg (ωk ) 2π ωk (ωk )vg (ωk ) = (4. t) = i k. D. while the microscopic quantization. On introducing the notation d[n(ωk )ωk ] c γk = = (4. t) = i k.λ γk V † × [âkλ (t) exp(ik · r) − âkλ (t) exp(−ik · r)]ekλ .281) A comparison with the paper (Huttner et al.277) dωk vg (ωk ) and assuming ekλ to be real for simplicity. (4. t) = i k.280)   2πc2 Ĥ(r. n 2± is involved instead of (ωk ) (for μ(ωk ) = 1). 2) a unit polarization vector orthogo- nal to k. and H fields in the forms   2πωk μ(ωk ) Ê(r. we can write the operators for E.2 Green-Function Approach 273 For an effectively unbounded medium plane-wave modes can be used.λ ωk n(ωk )γk V † × [âkλ (t) exp(ik · r) − âkλ (t) exp(−ik · r)]k × ekλ . In the sum (4. (4. because sums are to be compared with integral expressions. Again the replacement 2π ↔ 20 must be done. . (4.278)   2πμ(ωk )c2 B̂(r.λ ωk n(ωk )μ(ωk )γk V † × [âkλ (t) exp(ik · r) − âkλ (t) exp(−ik · r)]k × ekλ . 1991) can be made. It is not so easy as the previous excercise.4. B. i Fβ (r) → √ ekλ exp(ik · r). (4. (4.λ n(ωk )γk V † × [âkλ (t) exp(ik · r) − âkλ (t) exp(−ik · r)]ekλ . t) = i k. In the integral expression. The macroscopic quantization (Milonni 1995) leads only to optical polaritons.276) V with V a quantization volume and ekλ (λ = 1.

etc. The quantization has been performed by generalizing the theory expounded also in this book.2. respectively. the Doppler effect. Moreover. ωTe (r). For nota- tional convenience. .284) √ where the refractive index n = μ. be the operators of the polarization and the magnetization in frequency space. introduces also acoustical polaritons. This should be verified for the material with both μ and n negative. It is assumed that the magnetodielectric medium is causal and linear. let P̃ˆ (r. . Section 4. and also to the spectral density of thermal radiation. Let Ẽˆ (r. ω). ω) and a relative permeability μ(r.287) N . ωPm (r) are the coupling strengths. For instance.284) should be generalized to include the magnetodielectric absorption. ω). γm (r) are the absorption parameters. It is char- acterized by a relative permittivity (r. . ω) = 0 [(r. be the operators of the electric strength. ωTm (r) are the transverse resonance frequencies. The operator-valued Maxwell equations are very similar to the classical ones. The magnetodielectric media include left-handed material. in frequency space.274 4 Microscopic Theories although it is frequently named also a “macroscopic one”. (4. ω) = 1 + . and spontaneous and stimulated radiation rates. (4. and γe (r). Now the magnetodielectric medium has the decay rate Γ = μnΓ0 . We present the electric constitutive relation ˆ ω) + P̃ˆ (r. ω) and M̃ ˆ (r. It follows also from Fermi’s golden rule.283) √ where n =  is the refractive index and Γ0 is the decay rate in free space.286) ωTm (r) − ω − iωγm (r) 2 2 where ωPe (r). purely electric medium has the decay rate Γ = nΓ0 . (4. (2003).285) ωTe 2 (r) − ω2 − iωγe (r) ωPm 2 (r) μ(r. (4. (4. Especially. the spatial argument has been omitted. ω) = 1 + . Milonni and Maclay (2003) have applied the theory of Milonni (1995) to the effects of a negative-index medium on an excited guest atom. ω)−1]Ẽ(r. The atom embedded in a homogeneous.1. P̃ˆ (r. ωPe 2 (r) (r. A quantization scheme for the electromagnetic field interacting with atomic systems in the presence of dispersing and absorbing magnetodielectric media is contained in Dung et al. The spontaneous decay of an excited two-level atom is influenced by the envi- ronment. ω). radiative recoil. the expression (4. ω).

From the viewpoint of the theory the noise terms guarantee that the field commutators do not depend on Re{(r. −ω∗ ) = G∗ (r.. r . ω).136). and M̃N (r. ω) = δ(r − r )1. ω)∇ × Ẽˆ (r. r . In Section 4. ω) = i 0 Im{(r. ω) = κ [1 − κ(r. 0 N (4. in which the following property of the Green tensor is used.290) N N N is the noise current. ω) is the noise magnetization associated with magnetic losses.2. ω) (4. f̂ λ j (r . ω) = iωμ Ẽ 0 G(r. ω) (4. ω) (4. ω) is the classical Green tensor obeying the equation ω2 ∇ × κ(r. the right-hand side of which ˆ includes the noise polarization P̃N (r. ω). ω) and the noise magnetization M̃ˆ (r. ω)∇ × G(r. (4. Dung et al. ω) = μ (r. ω)} and Re{κ(r. ω) − 2 (r. ω). ω) = −iωP̃ˆ (r. ˆ ω) + M̃ (4. λ = e. ω)ˆj̃N (r . N ω2 ˆ ω) = iωμ ˆj̃ (r. (4.292) c2 Cf. ω) − (r.294) π and from the commutation relations for the fundamental bosonic vector fields (λ. r .288) 0 N where κ0 = μ−1 −1 ˆ 0 . We present also the wave equation for Ẽ ˆ (r. ω )] = δλλ δi j δ(r − r )δ(ω − ω )1̂. ω)} f̂ (r. r . (4. ω) being the noise polarization associated with the electric losses due to material absorption.289) c where ˆj̃ (r. G(r. ω). ω)} f̂m (r. ω)]B̃(r.293) N e π and  ˆ (r. r .296) . r . ω). M̃ ˆ (r. and (4. ω)G(r. ω) d3 r . and the magnetic constitutive relation ˆ (r. f̂ λ j (r .138) above.7 will be referred to calcula- tions. ω )] = 0̂. ω). ω). m) † [ f̂ λi (r. All of the commutation relations follow from the choice  ˆP̃ (r. ∇ × κ(r.137).291) where G(r. (4. (4. ω)Ẽ(r. ω) = κ0 M̃N − Im{κ(r.4.2 Green-Function Approach 275 with P̃ˆ N (r. (2003) consider the solution ˆ (r.295)   [ f̂ λi (r. ω)}. ω) + ∇ × M̃ ˆ (r. ω). κ(r. (4.

ω.299) On dividing by iωμ0 .142). ω. (4. but the complete notation should be as in (4.276 4 Microscopic Theories So far the time dependence has not been considered.297) iω where Ẽˆ ⊥ (r. B̃ˆ (r. 0) = 0̂. ω. ω. ω. 0) + iωD̃(r. (4. ω. ω. 0) = 1 Ẽ à ˆ ⊥ (r. 0). ω. 0). The definition of the operator of the magnetic induction may be rewritten as an operator-valued Maxwell equation ∇ × Ẽ ˆ ω. 0) = Ẽ (4. 0).158) above.298) where Ẽˆ  (r. ω.143) ˆ  (r. 0) = ∇ × Ã(r. 0). (4. ω. ω. 0) = iωB̃(r. we obtain that ω2 .301) On the scalar multiplication of equation (4. 0). 0). 0) means the longitudinal part of Ẽˆ (r. (4. Having defined à ˆ (r. We consider also the scalar potential with the property as quantized in (4.157) and (4. cf. equation (4. (4. ˆ (r.289) with the ∇ operator from the left. We may note that the vector potential ˆ (r.289) yields ∇ × H̃ ˆ ω. (4. − ∇ ϕ̃ˆ (r. cf. ˆ (r. 0).300) where we have used also the two constitutive relations. 0) means the transverse part of Ẽˆ (r.143). ω. we may introduce also ˆ ω.

ω)Ẽ ˆ (r. ω). (4.302) 0 N c or . ω) = iωμ (−iω)∇ · P̃ˆ (r. − 2 ∇ · (r.

∇ · B̃ (4. B̂ j (r . ω)Ẽ(r. ω). t)]. ˆ (r. t)] = −iεi jk ∂k δ(r − r )1̂ (4. Ê j (r . (4. ˆ (r. B̂ j (r . ω) + P̃ˆ (r. t). t). ∇ · D̃ (4.303) N or ˆ (r. ˆ (r. (4. ω) = 0̂.306) It can be shown that the equal-time commutation relations [ Ê i (r. ∇ · 0 (r. ω) =  (r.308) .305) 0 N By definition. ω) = 0̂.304) where D̃ ˆ ω) + P̃ˆ (r. t). t)] = 0̂ = [ B̂i (r. ω)Ẽ (4. ω) = 0̂.307)   [0 Ê i (r.

Then we introduce the field operators in the presence of charge particles Ê(r. t) d3 r = qα ϕ̂ r̂α (t). (4. t). t). t) dω d3 r λ=e. t) = Ê0 (r.314) D̂(r. In the minimal-coupling scheme and for nonrelativistic particles. t) α 2m α 1 + ρ̂A (r.310) 4π 0 |r − r | are introduced. the total Hamil- tonian reads  ∞ † Ĥ (t) =  ωf̂λ (r. ω) and operators r̂α and p̂α act.2 Green-Function Approach 277 are preserved. ω. D̂0 (r. t). t) ϕ̂A (r.m 0  1  2 + p̂α (t) − qα Â(r̂α (t). t) = d3 r (4. (4. t) = qα δ r1̂ − r̂α (t) (4. B̂0 (r. t) − 0 ∇ ϕ̂A (r. t) = B̂0 (r. t). B̂(r. t). The interaction of charged particles with the medium-assisted electromagnetic field is studied using Hilbert space on which the components of the bosonic fields f̂ λi (r. t). Ĥ0 (r. t)ϕ̂(r. ω. (4.313) α In this new situation the old operators should be denoted as Ê0 (r. the position and the canonical momentum operator of the αth particle of mass m α and charge qα . (4. t .311) 2 The third and last terms can be written in the forms 1 1  qα qα  ρ̂A (r. t)ϕ̂(r. t). t) = Ĥ0 (r. (4. t) · f̂λ (r. t) − ∇ ϕ̂A (r. respectively. except the bosonic vector fields f̂λ (r.312) 2 2 α α 4π 0 |r̂α (t) − r̂α (t)| α =α     ρ̂A (r. The charge density    ρ̂A (r. t) d3 r = . t)ϕ̂A (r. t).4.315) . t) d3 r. t) = D̂0 (r. t) d3 r + ρ̂A (r. t)ϕ̂A (r.309) α and the scalar potential of the particles ρ̂A (r . Ĥ(r. t). where ∂k means the partial derivative with respect to the kth compo- nent of the vector r. Here r̂α and p̂α are.

Forms of ΓΓ0 valid for an atom at an arbitrary position inside a spherical free-space cavity surrounded by an arbitrary spherical multilayer material environment may be applied (Dung et al. The case of nonabsorbing bulk material is treated first. ω. Cu (t). then the refractive index is purely imaginary and Γ = 0. 1994. ˆ (r. Here |l is the lower state whose energy is set equal to zero and |u is the upper state of energy ωA . |1λ (r. t). ω) |l . ω. 0) = 0.316) 2 α is the operator of the current density of the particles. The case of atom in a spherical cavity is treated second. In the integro-differential equation the vector dA and the tensor-valued Green function still occur in a relatively simple fashion. These coefficients satisfy linear differential equations and the initial conditions Cu (0) = 1 and Cλl (r.318) 3π 0 c3 is the free-space decay rate. both the tensor-valued Green function and the vector dA . These operators obey the time-independent and time-dependent Maxwell equa- tions. for nonabsorbing left-handed material. The authors treat the spontaneous decay of an excited two-level atom. ω) ≡ f̂λ (r. (4.278 4 Microscopic Theories the new operators. whose alternative are the quantum states |1λ (r.317) where ω̃A3 dA2 Γ0 = (4. On this assumption Γ = Re[μ(ω̃A )n(ω̃A )]Γ0 . |ψ(0) = |{0} |u . In the differential equations. Tai 1994). ω)|{0} . The cavity may be con- ceived as a way of removing the singularity of the tensor-valued Green function. the solution reminds us of the Weisskopf–Wigner theory. but taken at the shifted transition frequency. In the literature it has been shown that the Hamiltonian (4. t) = qα r̂α (t) δ r1̂ − r̂α (t) + δ r1̂ − r̂α (t) r̂α (t) (4. where 1   ·     ·  ĵA (r.311) generates the time-dependent Maxwell equations and the Newton equations of motion for the charged particles. 2003. ω. the transition dipole moment. ωA † is a transition frequency. As expected. The decay rate Γ can be expressed in terms of them and the shifted transition frequency ω̃A is utilized. Cel (r. When (ω̃A ) and μ(ω̃A ) have opposite signs. Li et al. In a formal expression of the state |ψ(t) . because the old operators have not been renamed and the new operators are E etc. the rate Γ is given by relation (4. t) and Cml (r. . They solve the problem of the time development of the state |ψ(t) . occur. In contrast. t) are the respective coefficients. The exposition in the literature may have been more explicit.317) without the notation Re.

(4. (b) magnetic matter. those formulae are specialized for the atom situated at the centre of the cavity and the otherwise homogeneous material environment. . The number of clear-cut cavity resonances decreases as the radius of the cavity decreases. and (c) magnetodielectric matter. The decay rate in the cases (a) and (b) inside a band gap is low and. They have used a renormalization group analysis. Also the role of the resonances is similar to the case (a) or (b). millimetre wave. microwave. Every time a comparison is made between (a) dielectric matter. They can explain the result by Pokrovsky and Efros (2002) that.319) 1 + 2(ω̃A ) For strong dielectric absorption and R small we have a different approximation  3 9 Im{(ω̃A )} c Γ≈ Γ0 . Then smaller ones are characterized. The description of an atom should not be derived using the macroscopic field. even if the description. In that article results obtained from experiments at the microwave frequencies have been reported. if the band gap may belong either to the permittivity or to the permeability. But simplifying assumptions are too weak to work for the magnetodielectric matter. Ozbay et al. (4. continue with R −1 . (2007) have provided 17 references to reports on metamaterials appropriate to a wide range of operating frequencies such as radio. by embedding wires in a medium with negative μ. 2003). Results of Scheel et al. In the case (c) this rate is low. Large cavities are considered first. it also increases much due to each cavity resonance. in the case (a). the authors may address also the local-field corrections.317). If the material absorption may be disregarded. The ratio of the decay rates has been calcu- lated in the literature as a function of the (shifted) atomic transition frequency ω̃A . and even vis- ible wavelengths. 2000) have been amended. They have checked the theoretical results numerically. far-infrared. the terms proportional to R −3 and R −1 may be left out. does not involve the field. one does not get a left-handed medium. and an already constant term and the O(R) term. The theory of the authors directly applies to the real-cavity model. (1999a. The complexity relies on the dependence on z = R ω̃cA .320) |1 + 2(ω̃A )|2 ω̃A R The decay may be regarded as being purely radiationless (Dung et al. Using their results. mid-infrared. or when  and μ may be taken for real numbers. But in the overlap of the electric and magnetic band gaps.2 Green-Function Approach 279 In fact. which takes into account the coupling between each resonator. near-infrared frequencies.4. A cav- ity is considered whose radius is much smaller than the transition wavelength. Hence 2 3(ω̃A ) Γ≈ Re[μ(ω̃A )n(ω̃A )]Γ0 . the decay rate increases. Felbacq and Bouchitté (2005) have found a unified theoretical approach to the left-handed materials. The expansion in powers of z must begin with a term proportional to R −3 . such as relation (4.

(2007) have studied the pulsed second-harmonic generation in positive. Let us note that only a d-independent factor is determined here. the sta- bility of hydrophobic suspensions of particles in dilute aqueous electrolytes (Spar- naay 1989. Each particle is surrounded by ions of opposite charge. Milonni 1994). The same expression . where L D is a Debye length. Casimir’s work had its origin in a problem of colloidal chemistry. We expect that a repulsive force between the particles separated by a distance d increases more rapidly than an attractive force as d → L D + 0. perfectly conducting parallel plates.and negative-index media. Finally it emerges that the result does not depend on the function f (k). (4. In the positive-index media and in the presence of phase mismatch two forward-propagating components of the second-harmonic are generated (Bloembergen and Pershan 1962). In calculating this energy. It is the calculation of the difference between the zero-point field energies for finite and infinite plate separations. Milonni (1994) has reviewed a standard calculation of the Casimir force. 4.280 4 Microscopic Theories Roppo et al. the Euler–Maclaurin summation formula is used. The particles are charged. namely. The attractive force between the plates is then obtained as the derivative of the potential energy with respect to the distance d on changing the sign. Such suspensions are said to be stable if the particles do not coagulate. In the pulsed generation. featuring a formal dependence on a function f (k). In calculating it. the sec- ond harmonic signal comprises a pulse which walks off (and is recognized in much work) and a second pulse which is “captured” and propagates under the pump pulse. In 1948 he gave expression for the attractive force per unit area cπ 2 FC (d) = . the Euler–Maclaurin summation formula has been applied.321) 240d 4 where d is a distance between two uncharged.2. The same result has been obtained by considering the radiation pressure (Milonni 1988). The attractive force should be obtained by integrating the pairwise forces between atoms. This difference is interpreted as the potential energy of the system. Now we mention the original idea of Casimir. assum- ing an interatomic force given by the London–van der Waals interaction (London 1930). We realize also that the repulsive force between these particles decreases more rapidly than the attractive force as d → ∞.7 Application to Casimir Effect In this subsection we will pay attention to papers which apply the quantization of the electromagnetic field in dispersive and absorbing media to the Casimir effect. if f (k) satisfies some conditions. The initial difference between a divergent sum over modes of the confined field and a divergent integral is modified to a difference between a convergent sum and a convergent integral. Here k is the wavenumber of a mode.

322) Let us recall that the (Kronecker and Dirac) delta functions in the commutator between noise polarizations are multiplied by π 0 Im{(ω)}. (4. ω. t)K j (r . K(r. t)K j (r . (4. where a0 is the Bohr radius. (4.324) 2 μ0 In Kupiszewska and Mostowski (1990) and Kupiszewska (1992). / t) into the 0 International System of units similarly as D(r. t) and we miss the functional factor δ(ω − ω ) in the Lifshitz theory. t). t) = 4π P(r. t) in the SI).2 Green-Function Approach 281 can be obtained also by a modification of the standard calculation. in Chapter 7. From relation (7. K(r. it would be appropriate to convert K(r. 233 in Milonni (1994). this field satisfies the fluctuation–dissipation rela- tion (the Gaussian system) K i (r. As the function f (k) depends on km ≈ a10 . t)E(r. t) in the Gaus- sian system of units.4. These calculations are presented in Chapters 2 and 3 in Milonni (1994). t) μ0  1 1 2 − 0 E2 (r. t) = Im{(ω)}δi j δ(r − r ) (SI). In the early 1950s. (1978). Milonni (1994) mentions the Lifshitz macroscopic theory (Lifshitz 1956). t) 1. t). corresponding to some real or complex noise polarization. The force between two semi-infinite dielectric slabs separated by a differ- ent dielectric medium or vacuum is derived. t) + B (r. Lifshitz (1956) then calculates the force in terms of the Maxwellian stress tensor. t) = 2 Im{(ω)}δi j δ(r − r ) . He was first to use a random field. t) = 0 E(r. we see that a different function f (x) depends on km d ≈ ad0 . t) = P(r. The physical basis of Lifshitz’s calculations is not so difficult to understand. The case of vacuum between slabs has been treated by Lifshitz. predictions of microscopic theories did not agree with experimental results. so the derivative with respect to the distance and with the changed sign can be taken independent of finding the constant. Much later. The general case of a dielectric medium between slabs has been treated by Schwinger et al. he mentions forces between dielectric slabs. t)B(r. which we present here in the form 1 T(r. He indicates that some results follow the Casimir approach. At zero temperature. (SI).323) 2π We recognize the right-hand side as a half of the appropriate commutator (K(r. t) ≡ K(r. Therefore. t) + B(r.69). using the factor 4π . Exactly. a number of the Bohr radii spanning the distance between the plates. This means that . it can be seen that K(r. Then the fluctuation–dissipation relation becomes 0 K i (r. the Casimir effect is studied on a restriction to the one-dimensional version. He does not expound this theory in fact. p.

Electromagnetic field quantization in an absorbing medium has been readdressed. and the Casimir effect both for two lossy dispersive dielectric slabs and between two conducting plates was analysed by Matloob (1999a. In a certain step of calculations the infinities cancel. The theoretical work on the Casimir force outweighs experimental work.282 4 Microscopic Theories only wave vectors normal to the surface are taken into account in the calculations. So. but he does not consider just the Lorentz force. An introductory guide to the literature on the Casimir force has been published (Lamoreaux 1999). but for a small reflection coefficient.b) and by Matloob et al. The Maxwellian stress tensor has been used to evaluate the vacuum radiation pressure of the electromagnetic field on each slab in terms of vacuum expectation values. the infinities of the stress tensor and the regular expression which diverges to them have been obvious. Attention has been paid to various limits of the general expression and to the Lorentz model of the . Weigert (1996) has considered several modes between the perfectly conducting metallic plates to be in a squeezed vacuum state. It is also assumed that the temperature is zero. The dielectric function of the slabs has been assumed to be an arbitrary complex function of frequency satisfy- ing the Kramers–Kronig relations. two contributions to the Casimir effect have been distinguished. Weigert (1996) admits that the state of the system is not stationary. An infinite expression has been regularized by means of an exponential cutoff function. As a generalization of the previous result. He calls it energy stress tensor. A result has been provided for any slab thickness. the Casimir effect for the case of two non absorbing dielectric slabs has been studied in detail. The atoms and the electromagnetic field have been described with equations of motion and the long-time solution has been found. but he calculates only with a stress tensor. The Casimir force has been calculated in the limit of semi-infinite slabs. In Kupiszewska and Mostowski (1990). At a given time instant the smallest possible expectation value of the energy in a neighbourhood of one of the mirrors is obtained through a calculation. only the stress tensor. It has been noted that the value of the appropriate component of the stress tensor is equal to the energy density for the one-dimensional calculation. Matloob and Falinejad (2001) have investigated the Casimir effect between two dispersive absorbing slabs in three dimensions. an accuracy of better than 10% has been reported (Lamoreaux 1997). The use of the Maxwellian stress tensor has been considered. The medium has been modelled as a continuous field of quantum harmonic oscillators interacting with a heat bath. In Kupiszewska (1992). It has been proposed to generate squeezed modes inside such a cavity and to measure an increase of the Casimir force. These averages have been expressed using the fluctuation–dissipation theorem and Kubo’s formula (Landau and Lifshitz 1980). T = 0. The electromagnetic field has been quantized in the presence of a dielectric medium. the absorbing dielectric slabs have been considered. A simple relation to the imaginary part of a tensor-valued Green function has been recognized. On regarding dielectric measurement. No explicit electromagnetic field quantization has been made. (1999).

He has calculated the Casimir force in a lossless dispersive layer of an otherwise absorbing multilayer by employing the quantized field operators as emerge from the scheme expounded in this chapter. The model has no direct connection with experiment. x cannot be used for the coordinate. The generalized Bogoliubov transformation has been applied also to the description of the field fulfilling the Dirichlet boundary condi- tions (at a conducting plate) and the Neumann boundary conditions at a permeable plate (the Casimir–Boyer model). But the one- dimensionality assumption reaches so far that all the virtual photons. The effect of finite temperature on the Casimir force between two dielectric slabs has also been considered. the Dirichlet or Neumann boundary conditions. let us say for intervals (0. where physical ideas are transparent and the calculations allow easy numerical evaluation.4. which is denoted by x  instead. One-dimensional analogues of three-dimensional concepts and their properties are studied. Boyer (2003) has presented a model. In fact. or just the opposite direction of the wave vector. namely. He has presented the expression obtained and has compared it with the result of Zhou and Spruch (1995) who had applied the surface mode summation method to purely dispersive media. The boundary condition at x  = x is enforced by the partition. The Casimir effect has been calculated for zero and nonzero temperature. should have the same. The boundary conditions at x  = 0 and x  = L are enforced by the walls. He assumes a one-dimensional box of length L at zero temperature T = 0. if considered. The scope of the paper has been the application of the Selberg trace for- mula to such a Riemann surface similar to methods of mathematical physics and quantum chaos. Tomaš (2002) has considered the Casimir effect in a dispersive and absorbing multilayered system using the Minkowski stress tensor method. In the case of an electromagnetic field. L). the energy-momentum tensor has been subjected to the generalized Bogoliubov transformation. He considers boundary conditions. A simplified thermodynamics is evoked. The difference between Casimir energies in two distinct layers has been established and the differ- ence between Casimir forces in two such layers has been presented provided that their refractive indices are equal. (2002) have generalized the so-called thermofield dynamics via an analytic continuation of the Bogoliubov transformations. . It has been achieved that a field in arbitrary confined regions of space and time is described. He introduces the Casimir energy ΔUzp (x. Kurokawa and Wakayama (2002) have introduced a Casimir energy for a com- pact Riemann surface of genus at least 2 and have related it to the Selberg zeta function.2 Green-Function Approach 283 dielectric function. x) and (x. da Silva et al. The Dirichlet condition corresponds to a perfectly conducting boundary condition describing a perfectly conducting material in three spatial dimensions and the Neumann condi- tion is simplified from an infinitely permeable boundary condition describing an infinitely permeable medium for electromagnetic waves. L) for the case. where a partition is present in the box at a position x. As an illustration he has calculated the Casimir force on a dielectric slab in a planar cavity with realistic mirrors.

T ) = 0. The force has been found by the non-perturbative method. This corresponds to the Rayleigh–Jeans spectrum of radiation. The theories do not comprise rigorous. L . attractive forces act. we may say that. Others are imperfect reflection. (4. 2003). Chan et al. and α = 1 for unlike boundary conditions. The temperature effect has been neglected. However.b. 2002).284 4 Microscopic Theories Boyer (2003) lets α be 0 or 1. Bressi et al.g. The zero-point-energy limit is contrasted by the high-temperature energy- equipartition limit. L) = −πc − + − . the interaction cannot be obtained from a pairwise summation. This geometry cannot be treated by perturbation theory due to the rectangular edges. where α = 0 for like boundary conditions for partition and walls.326) He discusses also the Casimir forces at finite temperature. The simplest and commonly used approximation is the proximity force theorem. nonzero temperature. (4. we may state that. A different approximation is the pairwise summation of renormalized retarded van der Walls forces. Theories for realistic geometries have been developed in response to high-precision measurements (Mohideen and Roy 1998. The force between a conducting plate and a permeable plate was given. in Boyer (1974). It has been only one of the real conditions which differ from the ideal situation and assumptions of the the- ory. off the centre of the box. while roughness corrections are more necessary at the smallest distances typical of the experiments. where this method can be used. He mentions a repelling force between the partitions and the walls. Emig (2003) has considered the force between a rectangular corrugated plate and a flat one. Lifshitz’s theory for dielec- tric bodies demonstrates that. For corrugated metal plates. the proximity force approximation is only valid for the roughness spectrum con- taining small enough wave numbers. While mean number of waves spanning the interplate distance (multiplied by 2π) may be informative of the accuracy of the . It was respected that in the most precise experiments the Casimir force between rough metallic plates was measured (Genet et al.325) 24 16 x L−x L For α = 0. In the case of metallic plates. The approximation leads to correct results when the radius of the sphere is much larger than the distance of closest approach. The proximity force approximation has been tested on the case where the force is measured between a plane and a sphere. forces act which repel the partition from the nearest wall. He speaks of an attractive force between the partition and the walls. and a geometry different from the parallel plates. 2001a. Then ΔURJ (x. from the centre of the box. He finds that    1 α 1 1 4 ΔUzp (x. because it is significant at large distances. e. non- perturbative methods for calculating the force. Emig (2003) has developed a novel approach for calculating the Casimir forces between periodically deformed objects. The results do not agree with those from the zeta-function method (Barton 2001) in a situation. in general. it fails at a small corrugation length. For α = 1.

For its application they assume two parallel plates of area A. The system of conducting shells depends on another system of parameters Λ.327). They mention the well-known decomposition of the spectral density into a smooth term and the oscillating contribution. a. w = 0 and is 2 otherwise. b. Many problems are formulated when the perfectly conducting plates of Casimir are replaced by other perfectly conducting surfaces. (4. They show that. a3 ) depends on the three parameters. (4. (4. f vw is 1 for v = 2. It can be utilized that the Casimir problem is not modified or generalized to dielectric media at the same time. The rectangular cavity of sides (a1 . w) ≡   2 . /       α − cos π wv α cos π wv − 1 N (α. Genet et al. are radii of the cylinders. the rectangular cavity has been considered by Lukosz (1971) and Maclay (2000). To this end they have used approximate semiclassical methods and the exact mode-by-mode summation method.330) 360 (b − a)3 Mazzitelli et al. Mazzitelli et al. a2 . 1986). ṽ(w) is the least positive integer v such that cos π wv > ab . They characterize a method according to Schaden and Spruch (1998. where condition w ≥ 0 is replaced by the condition w = 0 (w ≥ 1). for ab ∼ 1.329) sem where E w=0 sem (E w≥1 ) are obtained from relation (4. v. (2003) mention the proximity theorem (Derjaguin and Abrikosova 1957. .. (4. They derive an energy approximation using the periodic orbit theory. v. (2003) have proposed a specific roughness sensitivity and have considered its expectation value. The generalization to a system of conducting shells has also been realized. a2 .328) 1 + α 2 − 2α cos π wv They further write E sem = E w=0 sem + E w≥1 sem .327) 4πa 2 a w≥0 v≥ṽ(w) v a where  is the “quantization” length. (2003) have computed the Casimir interaction energy between two concentric cylinders. cf. where √ cπ 3 ab sem E w=0 = − . Derjaguin 1960).    c b  1 b E sem = − f vw 4 N . In this method the zero- point radiation is described with trajectories of a particle. and so as a real radiation.4. w . Periodic orbit theory relates oscillations in the quantum level density of a given Hamiltonian to the periodic orbits in the corre- sponding classical system. For example.2 Green-Function Approach 285 approximation.  a<  b. E sem ∼ E w=0 sem . Λ ≡ (a1 . 2000). a3 ). (Plunien et al.

1. In contrast. the semiclassical energy for an isolated cylinder vanishes and the exact energy for a cylinder of radius a is c E C = −0. Further they compute the exact Casimir energy for the coaxial cylinders using the mode-by-mode summation method (Nesterenko and Pirozhenko 1997). Ahmedov and Duru (2003) have calculated the Casimir energies with respect to the previous work such as Mazzitelli et al. This difference does not suggest decision whether the larger or the smaller area should be chosen.332) a2 b2 where c E 12 = − 2 2 ∞a 2π ∞ / d × k z2 − y 2 ln [Fn12 (iy. Here A = 2π a. The final result has the form   1 1 E ex = E 12 − 0. But only the numerical calculation shows that the approximation is relatively good for 1 < ab < 4.286 4 Microscopic Theories not of different areas. the theory of periodic trajectories suggests the choice A = 2π  ab. (4. respectively. Obviously. (2001). Again on the condition α ∼ 1 the relations simplify  cπ 3 1 1 E ex ∼ E 12 ∼ − +O . Still the relation cπ 2 A EP = − (4. Then the Casimir energy per unit height is . (4. We assume that the cylinders have the radii r0 < r1 .01356 .336) a2 They present also numerical results. Let us consider the region between two close coaxial cylinders.331) 720 (b − a)3 ? is applied to the plates “wound” into a cylinder. α) = 1 − 1 − n . (4. α)] dy dk z . (4. (2003) and Høye et al. In (z) and K n (z) are the modified Bessel functions and the MacDonald func- tions.335) 360a 2 (α − 1)3 (α − 1)2 The semiclassical approximation is valid.334) In (αy)K n (y) In (αy)K n (y) α = ab .333) 0 n kz dy with   In (y)K n (αy) I  (y)K n (αy) Fn12 (iy. 1.01356 + c. (4.√2π b.

(4. Then the Casimir energy is   π3 RL 15 Δ2 E tor = − 1 + . They have also mentioned some difficulty with calculations of the temperature effect on the Casimir force between real metals of finite conductivity. Let us analyse the ring with a rectangular cross-section.339) 720Δ 16Δ2 where L is the height of the cylinders and ζ (z) is the Riemann ζ -function. Then the Casimir energy per unit volume is Θπ 3 E ≈− . They distinguish five different approaches.343) 1440r 4 Δ2 Δ2 π2 where Δ is the angle between the half-planes of the boundary. Dividing the right-hand side by 2πΘΔ to “correct” the energy density (Ahmedov and Duru 2003).2 Green-Function Approach 287   π3 R 15 Δ2 E cyl =− 1+ .341) 720r 4 Δ3 √ where Δ ≡ θ1 − θ0 . We assume that the cones have the apex angles θ0 < θ1 ≤ π2 . .  2  1 π Δ2 E =− − .4.340) 360Δ3 4π 2 R 2 Let us evaluate two close coaxial cones. We assume that the spheres have the radii r0 < r1 . Let us consider two close concentric spheres. (2003) have begun with the state of the research of the Casimir effect. R = r0r1 . (4. Then the Casimir energy is   π 3 R2 5Δ2 E sph = − 1+ . (4. The Casimir energy is π3RL Rζ (3) E box ≈ − 3 + . Θ ≡ sin θ0 sin θ1 .338) 720Δ3 2π 2 R 2 where L is the length of the flat space measured parallel to the z axis. Let us imagine a flat space which is periodic in the z coordinate unlike the Euclidean space. Let us consider the region between two cylinders or tori in this space provided that the axis of these cylinders is the z axis. According to the fifth approach. and r is the distance from the com- mon vertex.337) 720Δ3 2π 2 R 2 √ where Δ ≡ r1 − r0 . we obtain that π2 E =− + O(Δ−3 ).342) 1440r 4 Δ4 This is similar to the solution of the wedge problem (Deutsch and Candelas 1979). (4. Geyer et al. (4. (4. the description of the thermal Casimir force can be obtained by the Leontovich surface impedance boundary con- dition. (4.

A “transition” impedance function does not exist evidently. the theoretical basis is as follows. which was employed by H. (2003) is an analogue of relation (8. For Au the transition frequency Ω = 6. Geyer et al. They determine the characteristic frequency ωc at each separation distance. but the field correlation functions are written down in conformity with statistical thermodynamics. Raabe et al. The space dispersion is also essential. Casimir himself. cannot be generalized to the case of absorbing bodies. graphs of both impedance functions are plotted. 2003) is calculated for 0 < a ≤ 5 μm and only by the use of the impedance of infrared optics and of anomalous skin effect. (2003) remind of the fact that at the temperature T = 0 only the anomalous skin effect and infrared optics occur. Explicit field quantization is not performed. Relation (43) in Geyer et al. In the region of the anomalous skin effect and in the relaxation domain metal cannot be described by any dielectric permittivity depending only on the fre- quency. For an ideal metal we have Z ≡ 0 and for real nonmagnetic metals |Z |  1 holds (Landau et al. has been derived for the case of a sphere above a plate made of a real metal. or relaxation domain for higher frequencies. but only at larger separations. Boundary conditions are introduced Et = Z (ω)Bt × n. even though it is different in each of the three domains. (4. It is respected that the main contribution to the Casimir free energy and force is given by the frequency region centred around the so-called characteristic frequency ωc = 2ac . An approximate expression (45). (2003) have underscored that one-dimensional quantization schemes are not rigorous enough when the Casimir force between absorbing multilayer dielectrics is calculated. where a is the space separation between two metal plates.36 × 1013 rad/s is obtained. because in such bodies there are no modes. At the beginning they warn that the “mode summation” method. They further present numerical results for T = 70 K and T = 300 K. In a figure. which is not reproduced here. and n is the unit normal vector to the surface (pointing inside the metal). and the region of the infrared optics for yet higher frequencies.344) where Z (ω) is the surface impedance of the conductor. 1984). The relative thermal cor- rection (Geyer et al. Then they characterize three procedures: (1) The electromagnetic field and the material bodies are treated macroscopically. the normal skin effect occurs already. and then they fix the proper impedance function. Et and Bt are the tangential components of the (Fourier transformed) electric and magnetic fields. The correction factors EE(a) (0) (a) to the Casimir energy agree quite well in the region of the infrared optics and in the transition region. which is not reproduced here.62) in the book (Milonni 1994). At these temperatures. the results due to the right and wrong choice differ significantly. .288 4 Microscopic Theories Three domains of frequencies are distinguished: The region of the normal skin effect for low frequencies. Otherwise. the region of the anomalous skin effect. The surface impedance is determined over the whole frequency axis. In the region of the anomalous skin effect.

Raabe et al. ω). . . Let us refer only to Kupiszewska and Mostowski (1990). n − 1. . r .347) 2 . . t)} (4. The third method is used in Raabe et al. n. (3) The electromagnetic field and the material bodies are described macroscopi- cally as in the first procedure. and so. . 1 < j < n − 1. 1968) and including retardation (see references in the cited paper). . In this context they have mentioned (Schwinger et al. zl  +1 ). n. . . n − 1. Before the Casimir force is calculated from the stress tensor. is the jth layer. . (4. Kupiszewska (1992). l = 1. The walls are composed of j − 1 layers l = 1. which have the properties zl+1 = zl + dl . 2004) have reproduced the essential traits of their quantization scheme. . Raabe et al. r . the numbering may differ a little. These layers have the boundaries zl . But the medium-assisted electromagnetic field is quantized by using an infinite set of appropriately chosen bosonic basic fields. ω) for r = r. l = 0. The permittivity is (r. we let Gl (r. l  = 0. r . (2003) have reserved the first method for Lifshitz (1955. . The mentions about the second method com- prise the note that the calculations were carried out only for one-dimensional sys- tems. From the expression of the Green tensor it follows that it conserves its form on the intervals (zl . . If both spatial arguments are in the same layer. r . . is the solution for the case that the medium of the lth layer fills up the whole space. The other is the scattering approach (Jaekel and Reynaud 1991). the bulk part. 1978) and have characterized the paper (Matloob and Falinejad 2001). (2003). r . . . (4. It is assumed that a “cavity”. l = 1. we are referred to the paper (Tomaš 1995). . The bodies are described by appropriate model systems. r . t) + Tm (r.346) where Glbulk (r. which separates walls. . inclusive of the substrate and the superstrate. . a more general tensor T(r. j − 1 and of n − 1 − j layers l = j + 1. t) + Tm (r. there are n + 1 layers. The values of the scattering part of the Green tensor for r = r are obtained in the coincidence limit of the position vectors. r . . r . r .4.2 Green-Function Approach 289 (2) The electromagnetic field and the material bodies are quantized at a micro- scopic level. ω). . ω) denote the form G(r. r . If some of the walls are semi-infinite. . Then they describe the multilayer structure. We introduce the scattering part Glscat (r. n − 1. They consider n − 1 layers of thicknesses dl > 0. ω) = Gl (r. . 1956). . t) = Te (r. ω) = l (ω) for zl < z < zl+1 . . but the authors also refer to Tomaš (2002). Then they add two further methods. n. t) 1 − 1 Tr{Te (r. Simplify- ing assumptions are made.345) For the tensor-valued Green function. They introduce z 0 = −∞. . l = l  . r . . One is the surface-mode approach in the nonretarded limit (van Kampen et al. . z n+1 = +∞. . zl+1 ) × (zl  . ω) − Glbulk (r. l = 1. (2003.

350) into the modified relation (4. r ).352) and (4. t) = D̂(r. t) = B̂(r.348). r .353) by the way of writing the superscripts scat. and (4.348)    Tm (r.349). the time argument t has been dropped. For finite temperatures T . (4.350) into the modified relation (4.350) kB T where 9 0 1: Ĥ Z = Tr exp − .290 4 Microscopic Theories is defined.353) π 0 2kB T On the left-hand side of relations (4. we obtain    ∞ ω ←− Tm (r. (4.349). (4. t) . ω)G(r. t)Ê(r . We may modify relations (4. r ). Then we introduce the stress tensor Tscat (r) = lim  Tscat (r. since the right-hand sides do not depend on t.351) kB T and kB is the Boltzmann constant. where 1 is the second-rank unit tensor and Te (r. r .347). we get   2  ∞ ω ω Te (r. they employ the statistical operator 0 1 −1 Ĥ ρ̂ = Z exp − .348) and (4. r . The Casimir force (per unit area) is given by the zz-component of the stress tensor (4. (4. ω)} dω (4. Although the generalization to a “cavity” containing dielectric medium has been known. On substituting relation (4. t)Ĥ(r .349) The expectation values are calculated in thermal equilibrium. t) (r = r ).354) r →r With respect to a layer. (4. (i) Basic equation. r ) = − coth ∇ × Im{G(r. r .324). (4. a stationary state of the field. (4. (2003. (4.353). In relations (4. 2004) have restricted themselves to the free space between two stacks.355) r →r . ω)} × ∇  dω. it is suitable to introduce the tensor  j (r) = lim Tscat  Tscat j (r.352). Raabe et al.352) π 0 2kB T c2 and on substituting relation (4. ρ̂ may be written explicitly. r ) = coth Im{(r.

4.359) on replacements l . Propagation con- stants  ω2 l (ω) βl = βl (q. σ = s. (4.356) c2 and reflection coefficients for σ -polarized waves at the top (+) and bottom (−) of the jth layer are defined that are σ rn+ = 0. ω) = − q2 (4. p (4.358) βl βl+1 + 1 + ββl+1l − 1 exp (2iβl+1 dl+1 ) r(l+1)+ s and     βl l βl l p βl+1 − l+1 + βl+1 + l+1 exp (2iβl+1 dl+1 ) r(l+1)+ p rl+ =    .360) and the recurrencies for the others are analogous.357) and are calculated from the recurrence relations     βl βl βl+1 − 1 + βl+1 + 1 s exp (2iβl+1 dl+1 ) r(l+1)+ rl+ =  s    .358) and (4. (4.359) βl l βl l p βl+1 + l+1 + βl+1 − l+1 exp (2iβl+1 dl+1 ) r(l+1)+ σ The coefficients rl− are σ r0− =0 (4.2 Green-Function Approach 291 Now a number of concepts and pieces of notation are introduced. which are formally obtained from relations (4.

→ l. l + 1 .

→ l − 1 (4. (4.363) 0 σ scat As Tzz. Also denom- inators of the fractions for multiple reflections σ σ Dσ l = Dσ l (q. j = − coth 2π 2 0 2kB T 9 : ∞    −1 σ σ × Re qβ j exp 2iβ j d j Dσ j r j−r j+ dq dω . j does not depend on the space point in the jth layer.361) and on the change of the subscript + of the reflection coefficient to −. the argument r has been dropped. ω) = 1 − rl+ rl− exp (2iβl dl ) (4.362) are introduced. ∞    ω scat Tzz. . Finally.

(1956). Of course. (4. ∞    −1 σ σ × qκ j exp −2κ j d j Dσ j r j−r j+ dq . To obtain Tzz. (iii) One-dimensional systems. (4.366). π m=0 2 - .363). we C may sim- ply repeat the derivation from relation (4.292 4 Microscopic Theories (ii) Imaginary frequencies. scat j in the zero-temperature limit. 1 1  d2 q . A comparison with the three-dimensional case is made only for T = 0. since in the one-dimensional system normal incidence occurs and the description can be restricted to a single polarization. (4. we intro- duce  ξ 2  j (iξ ) κj = + q 2. replacement ∞ m=0 →  2π kB T dξ can be realized in relation (4. We introduce kB T ξm = 2mπ .364)  Since the permittivity is positive on the positive imaginary frequency axis.363).365) c2 Exploiting the analytical properties of the ω integrand in relation (4. we arrive at an expression of the integral with respect to ω by a residue series. Further one of the integrals is replaced by a multiplication with a constant. For m = 0. Finally ∞   kB T  1 j = 1 − δm0 scat Tzz. m integer. Contrary to the three-dimensional description. the term with ω = 0 is peculiar and it should be replaced by the limit ω → 0+.366) 0 σ ω=iξm which may be regarded as a generalization of the famous Lifshitz formula Lifshitz (1955). the sum with respect to σ is omitted.

it is shown that the Casimir force between two single-slab walls behaves asymptotically as d −6 instead of d −4 in the . For example. (4.→ . Also analytical expressions for specific distance laws in the zero- temperature limit are derived.367) 4π 2 A q where A is the normalization area.

T1 . For an ideal metal |ΔFpp | does not depend on a. for Au. The chosen approach exhibits an increase |ΔFps | with the temperature T2 .4. T ). at distances relevant to Casimir force measurements and to nanomachinery. T2 ) = Fpp (a. for Au. In the case of two parallel plates. We have chosen in this book that their theory is named microscopic just as the theory due to Hopfield (1958) and . (2003) compare the chosen approach with that. Action of the Casimir force on magnetodielectric bodies embedded in media has been analysed in (Raabe and Welsch 2005). The alternative approach provides a negative difference for an ideal metal. They have referred also to (Bruno 2002). T2 ) − Fpp (a. Chen et al. |ΔFpp | decreases with an increase of a. T1 = 300 K and T2 = 350 K. 2003). 2002). T2 ) = Fps (a. T1 = 300 K and T2 = 350 K. a paper devoted to an attractive Casimir magnetic force.g. In the case of a sphere above a plate. |ΔFps | decrease with an increase of a. They concentrate on the difference ΔFpp ≡ ΔFpp (a. They believed that. If the temperatures are fixed. (4. Then Chen et al. For example. (2003) have replied to the comment (Iannuzzi and Cappasso 2003). Results for single-slab walls for periodic multi- layer wall structure are illustrated in figures. They study the difference ΔFps ≡ ΔFps (a. The difference is negative both for a real and for an ideal metal. but a positive difference (more than six times larger at T2 = 350) for a real metal. The thermal Casimir force is denoted by Fps (a. T2 ) − Fps (a. where a is the separation distance and T is a temperature. (4. (2000) to express the thermal Casimir force denoted by Fpp (a. T ). It could be startling that here Raabe and Welsch (2005) declare the macroscopic quantum electrodynamics themselves.368) where T1 and T2 are temperatures. They fix a = 0. They have declared the consensus that exploring the possible existence or design of materials with nontrivial magnetic properties for obtaining a repulsive Casimir force is important. T1 ). Iannuzzi and Capasso (2003) have published a comment on the paper (Kenneth et al. For an ideal metal |ΔFps | is a linear function of a. 2002) (further references see Chen et al. The configura- tions of two parallel plates and a sphere above a plate are considered. T1 ). they utilize a perturbation result from the paper Bordag et al. the Casimir force between two slabs in vacuum was always attractive. e. after the paper (Brevik et al.369) For example. (2003) study the difference of the thermal Casimir forces at different temperatures between real metals. The consistency of expressions derived in the framework of the macroscopic theory with microscopic harmonic-oscillator models is shown. T1 . the difference of the Casimir forces increases with a decrease of the separation distance. they use a perturbation result after Klimchit- skaya and Mostepanenko (2001). Kenneth et al. while T1 ≤ T2 ≤ 350 K.5 μm and T1 = 300 K.2 Green-Function Approach 293 large-distance asymptotic regime.

that the level of representation is rather a mesoscopic one (Raabe and Welsch 2005). (4. It may be controversial that they do not use micro. which is not reproduced here.372) μ0 2 μ0 The integral of the Lorentz force density f(r) over some space region (volume) V gives the total electromagnetic force F acting on the matter inside V F= f(r) d3 r.b). which can be found below relation (68). but without the constitutive relations.e. The exposition begins with the classical Maxwell equations with charges and currents. which . (4. (4. It is worthwhile to mention that the Lorentz force density f(r) = ρ(r)E(r) + j(r) × B(r) (4. namely.371) over V we obtain that d F= T(r) · da(r) − 0 E(r) × B(r) d3 r. a constant. then the total force reduces to the surface integral F= dF(r).or mesoscopic for the model with the two auxiliary fields fe (r.370) is written as ∂ f(r) = ∇ · T(r) − 0 [E(r) × B(r)].371) ∂t where T(r) is the stress tensor. even though with a reservation. ω).294 4 Microscopic Theories Huttner and Barnett (1992a.  1 1 1 2 T(r) = 0 E(r)E(r) + B(r)B(r) − 0 E (r) + 2 B (r) 1.376) The tensor T(r) may be decreased by a constant term. fm (r. Raabe and Welsch (2005) have commented on the role of Minkowski’s stress tensor. (4. If the volume integral on the right- hand side of this equation does not depend on time. space independent tensor (it suffices to consider the position on the surface ∂ V ). ω).373) V On integrating both sides of equation (4. i. (4. It is consensual.374) ∂V dt V where da(r) is an infinitesimal surface element. (4.375) ∂V where dF(r) = da(r) · T(r) = T(r) · da(r).

the operator of the Lorentz force has not been presented in such an explicit form as its expectation value. r ) .380) μ0 2 Raabe and Welsch (2005) expound the quantum theory of the electromagnetic field as described also in this book in Section 4. r ) = 0 Ê(r)Ê(r ) + B̂(r)B̂(r ) μ0  1  1  − 1 0 Ê(r) · Ê(r ) + B̂(r) · B̂(r ) . r ) + f(r . r ) is a generalized stress tensor   1   T(r. (4.381) 2 μ0 In other words. we can quantize relation (4. which is not reproduced here. the notation is generalized to the form T(r. r) = ∇ r+r · {T(r.382) . In our opinion. They have presented also com- mutation relations between the charge density operator.6.378)  ∇ r+r = ∇ + ∇ ≡ ∇r + ∇r . r ) and to write relation (4.2 Green-Function Approach 295 is considered in much work devoted to the related topic. r ) means the density of a generalized Lorentz force f(r. (4. be it under this name or only as a “stress tensor”. (4. 1 T(r.2. (4. They have calculated correlation functions of some operators in thermal states of the field. r ) = T̂(r. As the expectation value of the stress tensor operator is infinite before a quantum correction. follows. r)} 2 ∂   − 0 E(r) × B(r ) + E(r ) × B(r) . (4. the current density operator. r ) + T(r .4. Calculation of the expectation value of the Lorentz force. r ) = 0 E(r)E(r ) − 1E(r) · E(r ) 2  1 1 + B(r)B(r ) − 1B(r) · B(r ) . (4. and the electromagnetic-field operators.381) in the form T(r. The latter is denoted by the same notation as the corresponding classical stress tensor. r ) = ρ(r)E(r ) + j(r) × B(r ).379) 2 and T(r.380) to introduce a generalized stress tensor T̂(r. r ) (just the same notation as in the classical theory).377) ∂t where f(r. Relation (4.371) can be generalized to characterize the density of a generalized Lorentz force f(r.

It is important for the theory of the van der Waals–Casimir forces inside a dielectric fluid. Let us also assume that μ = 1. 2004). Dzyaloshinskii et al. 3) or semi-infinite (0.385) 240  4 d3 d1 which is just one of the formulae underlying the critique. in his opinion. the approach of Raabe and Welsch (2005) is incorrect. Pitaevskii (2006) discusses the reason why. (4. Raabe and Welsch (2006) . t) + ĵ(r. Their relations (75) and (76) for this element. 4). (4. They include also a criticism of basing the calculations on Minkowski’s stress tensor (Tomaš (2002). Then the counterpart of the previous formula based on the Minkowski stress tensor is   cπ 2 1 1 1 F (M) = √ − 4 . which are not reproduced here. t) d3 r. Pitaevskii (2006) defends the validity of the paper (Dzyaloshinskii et al. The generalization of Casimir’s well-known formula is     cπ 2 μ 2 1 1 1 F= + − . The tensor of the van der Waals–Casimir forces was obtained by summation of an appropriate set of Feynman diagrams for the free energy and its variation with respect to the density (Dzyaloshinskii and Pitaevskii 1959). It can be believed that the warning is helpful. They apply the theory to a planar magnetodi- electric structure.384) 240  3 3μ d34 d14 where dk are thicknesses of regions k = 1.383) t→∞ V where again the charge density operator ρ̂(r. (4. (1978). 1960). which has been disqualified or underestimated by the criticism in Raabe and Welsch (2005).296 4 Microscopic Theories To make contact with microscopic approaches. Let us assume that the two walls and the plate are almost perfectly reflecting. On the condition of mechanical equilibrium this tensor differs from a Minkowski-like one by a constant tensor. which leads to a rela- tively simple relation (81). (2003. are rather complicated. Raabe and Welsch (2005) give the relevant stress tensor element Tzz (r) in the interspace 0 < z < d j . 3. where the subscript j has been used in conformity with Raabe et al. which is not repeated here too. separated by a dielectric fluid. (1960) obtained the same force between solid bodies. t) are appropriately expressed. Raabe and Welsch (2005) con- sider a harmonic-oscillator medium and derive that the (steady-state) Lorentz force acting on such a medium in some space region V is F = lim ρ̂(r. as Barash and Ginzburg (1975) and Schwinger et al. t)Ê(r. since one prefers simpler formulae to more complicated ones. finite (1. t) and the current density operator ĵ(r. provided that the simpler ones are not wrong. Its definition is specific in that homogeneity of the dielectric in an interspace (“cavity”) 0 < z < d j is required. To calculate the Casimir force on a plate in a nonempty cavity. t) × B̂(r. Raabe and Welsch (2005) choose five regions. 2.

vacuum-induced torque is present (Rodrigues et al. the measured force is attractive and is approximately 80% smaller than the force predicted for ideal metals in vacuum.e. The repulsive Casimir force of a left-handed material may balance the weight of one of the mirrors. This notion seems to belong to classical optics essentially and to be formed after the general relativity theory. They assume that corrugations are imprinted on both . 2005. The Casimir force is a superposition of the usual normal component and a lateral one in this situation. (2007a) have studied the lateral Casimir force arising between two corrugated metallic plates. (2006) have measured the Casimir force between a gold-coated sphere and two Si samples of higher and lower resistivity. The results were found to be consistent with Lifshitz’s theory. The result of Casimir and Polder (1948) has been rederived by integration of the force density. Rodrigues et al. They have taken two very different permittivities for media outside the plates. The simple formula for the Casimir force has been compared with the result of the more sofisticated Lifshitz theory. This integration is not an argument against the well-known nonadditivity of the Casimir–Polder forces (Milonni 1994 and references therein). There exists a geometry well suited to the aim of an accurate theory–experiment comparison. i. Leonhardt and Philbin (2007a) calculate the Casimir force for a dispersive medium in their set-up inspired by Casimir’s original idea. They have analysed the contributions of system eigenmodes with great attention to surface plasmon polariton modes. 2001. The lowering of resistivity cor- responded to enhancement of carrier density by several orders of magnitude. Chen et al.4. perfectly conducting plate placed in front of a neutral atom. Lenac and Tomaš (2007) have considered the Casimir effect between metallic plates assuming them to be dispersive and lossless and separated by a medium with the (Gaussian) unit permittivity. 2006a). Lam- oreaux 2005) and found to be consistent with this theory. Munday and Capasso (2007) have performed precision measurements of the Casimir–Lifshitz force between two metal surfaces (gold) separated by a fluid (ethanol). Messina and Passante (2007a) have calculated the Casimir–Polder force density on an uncharged. To this aim first-order perturbation theory and the quantum operator associated to the clas- sical electromagnetic stress tensor have been used. and it has been discussed appropriately. A concise quantum theory of light in spatial transformation media has been developed in (Leonhardt and Philbin 2007b). the surface plasmon polariton modes influence the Casimir effect dominantly except the case of thin layers that are supported by a highly reflective medium. The paper (Leonhardt and Philbin 2007a) is interesting for its use of the notion of a transformation medium. namely. When the separation between the metallic plates is small. In general. They consider two per- fect conductors with a metamaterial sandwiched in between. letting it levitate on zero-point fluctuations. 2006. For this situation. Each measurement was compared with theoretical results using the Lifshitz theory with different dielectric permittivities (Bordag et al. 2 = 1 or 2 = ∞ (perfect conductor). that with parallel and periodic corrugations of the metallic surfaces.2 Green-Function Approach 297 maintain their position that the Casimir force should be calculated on the basis of the Lorentz force. Chen et al.

. and λP obey some specific orderings. Fulling (2007) have presented results on the Casimir force in one-dimensional piston models. the pistons are treated as univalent vertices. . . . the force is repulsive. For instance.298 4 Microscopic Theories plates with the same period and along the same direction. (1961). but with a spatial mis- match. and λP (plasma wavelength). (4. B) and some vertices. Limiting cases such as the proximity-force approximation limit and the perfect reflection limit are recovered when the length scales L. The Casimir force is obtained in terms of the stress tensor integrated over space and imaginary frequency. then the force is ( = 1 = c) (B − 3)π F= . respectively. If at each piston the field obeys the Neumann boundary condition. the result is related to an ordinary Neumann interval of length a or 2a. They have based their approach on the familiar result due to Lifshitz and Pitaevskii (1980). which is automatically regularized on application of the finite-difference method to solve for the Green function. λC . Rodriguez et al. They have also numerically examined the rate of convergence of the periodic-orbit expansion. The pistons will tend to move outward. Kuch- ment 2004). Their analysis indicates that the Casimir–Polder force is the dominant loss agent. In the development of ever smaller atomic magnetic traps carbon nanotubes have been considered to become the elementary building blocks. (2007a. and Pitaevskii (2006). λC (corrugation wavelength). . (2007) have calculated atomic spin-flip lifetimes and have estimated tunneling lifetime corresponding to the sum of the Casimir–Polder potential and the magnetic trapping potential.386) 48a 2 When B = 1 or 2. Either end of each bond ends at one of these vertices. At the univalent vertices either a Dirichlet or a Neumann boundary condition is imposed. Dzyaloshinskii et al. for arbitrary dielectric and metallic materials. It is well known that an atom held in a magnetic trap near an absorbing dielectric surface will undergo ther- mally induced spin–flip transitions. When B > 3. In fact.b) have developed a numerical method to compute the Casimir forces in arbitrary geometries. The geometries that have been considered have the property that the bodies have not a contact and they are in the free space. These models are applications of quantum graphs (Roth 1985. The vacuum expectation value of the stress tensor is calculated in terms of the Green function. (2007) have discussed a periodic-orbit approach to calculations of the Casimir forces. The result is valid provided that these amplitudes are smaller than L (mean separation distance). Fermani et al. and the valence of a vertex is defined as the number of bonds meeting there. They have characterized the quantum star graphs mainly. Some of these transitions lead to trapping losses. At the central vertex the field has the Kirchhoff (generalized Neumann) behaviour. the space may consist of B one-dimensional rays of large length L attached to a central vertex. A finite quan- tum graph consists of B one-dimensional undirected bonds or edges of length L j ( j = 1. In each ray a piston is located a distance a from the vertex. They have used the scattering theory in a perturbative expansion in powers of the corrugation amplitudes. Fulling et al.

(4. For the TE polarization all modes are propagating. (2007) remind that the Casimir effect. A plasmonic contribution to the Casimir Aω energy is denoted by E p . This situation has been investigated already by Barton (1987). k is a wavenumber. 1998). R) dk. R = r B − r A . Intravaia et al. energy again. This is balanced in the total Casimir energy by the contribution of photonic modes (cavity and bulk modes). They calculate dispersion relations for the relevant modes numerically. First they outline the method used by reproducing the Casimir–Polder potential energy between two atoms in a thermal field   c ∞ 3 ck W AB (R) = k α A (k)α B (k) coth π 0 2kB T × V(k. α A (k) (α B (k)) is the dynamical polarizability of . They assume the wall located at z = 0 and let r A and r B denote the positions of atoms A and B.2 Green-Function Approach