Publi shed by

World Scientific Publishin g Co. Pte. Ltd.

5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402 , Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Cataloging-in-Publication Data

Name s: Ou, Zhe-Yu Jeff, author.
Title: Quantum optics for experimentalists / Zheyu Jeff Ou
(Indiana University - Purdue University Indianapolis, USA).
For my family, Jenny , Antho ny, and Nathan
Description: Singapore ; Hackensack, NJ : World Scientific, [2017] I
Includes bibliographical references and index.
Identifiers : LCCN 20170182251 ISBN 9789813220195 (hardcover; alk. paper) I
ISBN 9813220198 (hardcover ; alk. paper) I ISBN 9789813220 20 I {pbk. ; alk. paper) I
ISBN 9813220201 (pbk. ; alk. paper)
Subjects: LCSH: Quantum optics.
Classification: LCC QC446.2 .093 20171 DDC 535/.15 --dc23
LC record available at https://lccn .loc.gov/20 170 18225

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Copyright © 2017 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduc ed in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without wrillen permission from the pu blisher.

For photocopying of material in this volume , please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 0192 3, USA. In this case permission to photocopy
is not required from the publisher.

Printed in Singapore by B & Jo Enterprise Pie Ltd

 Preface

Optics is an old field of physics. The basis for classical optics is the elec-
tromagnetic theory of Maxwell. While quantum optics was developed from
classical coherence theory and quantum electrodynamics theory in the 50's
of last century, research activities in this field have exploded only since
the beginning of the new millennium due to its applications in quantum
information and fundamental tests of quantum theory. The technological
advances in this field made it possible to implement many ideas that can
otherwise only be achieved in theoretical models.
The field of quantum optics has been served quite well by a number of
outstanding textbooks, starting with the classic book by Loudon [Loudon
(2000)], followed by the comprehensive books by Mandel and Wolf [Mandel
and Wolf (1997)], Walls and Milburn [Walls and Milburn (2008)], Scully
and Zubairy [Scully and Zubairy (1995)], and most recent one by Agarwal
[Agarwal (2013)]. Most of the currently popular textbooks were written by
theoreticians, who emphasized on the theoretical aspects of quantum opti-
cal fields but left out most of the experimental part. The book by Bachor
[Bachor and Ralph (2004)] concentrates mostly on the experiments in quan-
tum optics. Even though the second edition involv es Ralph, a theoretician,
to includ e many theoretical aspects of quantum optics, the discussions on
exper iments and theory are basically separated. So, there is not a textbook
that deals with theoretical aspects of quantum optical experiments . When
I visited labs around the world, I was frequently asked by students in the
lab about how to describe a photon in a real experimental environment,
for example, how does a photon pass through a Fabry-Perot filter? Those
tudents who are very skillful and hard working in the lab don't have much
time to think deep on this but they cannot find the answer in most of the
textbooks either. The reason behind this is quite simple: theory is built on

vii
 viii Quantum Optics For Experimentalists
simple and easy-to-describe environment but experiments are usually much
more complicated for the theory to cover all. To make things even harder ,
a complete experimental description involves the concept of modes of the
fields, which is not discussed at all in the currently popular textbooks. The
mode concept is based on the classical wave theory of electromagnetic fields
for which most of the textbooks on quantum optics hardly discuss.
First recognized and promoted by W. Lamb (famous for the Lamb shift
of Hydrogen atom) in a paper [Lamb (1995)] with a quite revealing title
of "Anti-photon", the concept of modes is the best way of approach in
understanding quantum optical phenomena. It is closely related to the
experimental reality and leads to the correct physical pictures in the com-
plicated situation of experiments. Yet, most of the textbooks only deal
with single-mode or few-mode approaches and are not enough to give a
complete description of the real world scenarios, in which field excitations
of multiple modes usually occur. As I will show in this book, the approach
by the concept of modes also helps in understanding some fundamental as-
pects in quantum mechanics such as the unification of wave and particle
pictures and the relationship between indistinguishability and the visib ility
of interference. With the concept of modes, the aforementioned questions
that were asked by students working in labs are straightforward to answer
(see Section 6.3.3).
In a broader sense, after the foundation of quantum optics was laid out
around the later part of last century, we have come to an era of engineering
of the quantum technologies to make useful quantum devices for practical
applications. In this process, a transition from classical to quantum physics
is a necessary step because most people working in optics, espec ially those
working in the lab, are quite familiar with the concepts in classical wave
theory. As one will see from this book, the presentation of quantum optics
based on the mode concept of the classical wave theory makes th is transit ion
relatively easy : there is a direct correspondence between classical wave
theory and the quantum theory of light.
The current book stems from the lecture notes for a number of lecture
series on Quantum Optics that I gave in Tianjin University and East China
Normal University . The original notes are in Chinese and here I trans late
them into English and organize them in the form of a book. There are two
parts in this book. In the first part, I try to construct the foundation of
Quantum Optics from the perspective of an exper imenta list and relate t he
theory with experiments . I start with a histor ica l introduction abo ut t he
early development of quantum optics from classical coherent theory and
Preface ix
semi-classical theory of light by discussing the Hanbury Brown and Twiss
experime nt and its significance. This naturally leads to anti-bunching effect
of photons and the need for a quantum theory of light. Before the discus-
sion of quantization of light fields, I first introduce the mode theory of
classical electromagnetic fields and present a number of examples of optical
modes in a variety of situations, from simple to complicated, with which we
constantly deal in laboratories of optics. This approach is different from tra-
ditional textbooks on quantum optics. From here on, it is straightforward
to discuss a variety of quantum states in terms of the multi-mode language,
which is the correct way to describe experiments. Then I expand to dis-
cuss in depth the theory of quantum optics, including Glauber's quantum
coherence and photo-detection theory, the classical-to-quantum correspon-
dence, and the generation and transformation of quantum states. With a
solid theoretical framework closely related to experiment, I can discuss in
the second part the two most commonly used experimental techniques in
quantum optics, that is, photon counting and homodyne detection. These
two techniques correspond to the measurement of discrete and continuous
variables in quantum information. After introduction of each of the two
techniques, I follow with some applications of the experimental technique.
Some of more advanced topics in theory are placed at the end of the chap-
ters and marked as further reading sections. Beginning students or those
who do not want to go deep in theory can skip these sections without loss
of the knowledge for understanding the experimental parts.
The academic significance of the current book lies in the combination
between the theory and experiments in quantum optics: the theory has
the support of the experiments and in return, the experimental description
has a solid theoretical foundations. For theoreticians, this book provides
a correct physical picture for photons in real world. For experimentalists,
on the other hand, this book will guide them in their daily work in the lab
and lead to further discoveries .
I would like to thank Professors Weip ing Zhang of East China Normal
University and Xiaoying Li of T ianjin Un iversity for their hospitality, col-
laboration, and support during my stay in their universit ies and valuab le
discussions and comments on this book.
Ind ianapolis, May, 2017 Zhe - Yu Jeff Ou
 Contents

Preface Vll

Theoretical Foundations of Quantum Optics 1

1. Historical Development of Quantum Optics and A Brief

Introduction 3
1.1 Historical Background . . . . . . . . . . . . . . . 3
1.2 Hanbury Brown-Twiss(HBT)Experiment . . . . . 6
1.3 Fluctuations of Light in Phases and Amplitudes . 7
1.4 Random Variables and Processes ... 9
1.4.1 Discrete Random Variables . . . . . . . . 10
1.4.2 Continuous Random Variables ..... . 11
1.4.3 Joint Probability of Multiple Random Variables 12
1.4.4 Random Processes ...... . 15
1.5 Classical Coherence Theory of Light . . . . . . . . . . . 16
1.6 Classical Interpretation of HBT Effect . . . . . . . . . . 19
1.7 Anti-Bunching Effect of Light and the Need for A Quantum
Theory of Light . . . . . . . . . 20
1.8 The Topics of Quantum Optics 21
1.9 Problems 22

2. Mode Theory of Optical Fields and Their Quantization 25

2 .1 The Classical Theory of Light . . . . . . . . . . . 26
2.1.1 Maxwell Equations . . . . . . . . . . . . 26
2.1.2 The Eigen -Solutions of Maxwell Equation . 28
2.1.3 Special Eigen -Solutions: Box Model and Plane Waves 29

xi
xii Quantum Opt ic s For E xperim ental ists Cont ent s x iii

2.1.4 Special Eigen-Solutions: Continuous k-Space and 3.2.4 Quadrature-Phase Amplitudes and Quantum Noise 67
Plane Wave Solutions for Three-Dimensional Free 3.2.5 Non-Orthogonality and Over-Completeness of
Space . . . . . . . . . . . . . . . . . . . . . . . . . 31 Coherent States . . . . . . . . . . . . . . . . . . 69
2.1.5 The Concept of Mod es and the Decomposition in 3.3 Further Reading: Schrodinger Cat States . . . . . . . . . 70
Terms of Modes . . . . . . . . . . . . . . . . . . . 32 3.4 Squeezed Vacuum Stat es and Squeezed Coh erent States 72
2.2 More l\Iodes of Optical Fields . . . . . . . . . . . . . . . . 36 3.4.1 Squeezing of Quantum Noise . . . . . . . . . . . 72
2.2.1 Paraxial Rays and Gaussian Beams - the Modes of 3.4.2 Squeezing Operators . . . . . . . . . . . . . . . . 73
Optical Resonators 36 3.4.3 Coherent Squeezed States and Squeezed Coherent
2.2.2 Temporal Modes and Generalized Definition of
States . . . . . . . . . . . . . . . . . . . . . . . . . 75
Single-Mode Fields . . . . . . . . . . . . . . . 41
3.4.4 Further Reading: Photon Statistics of a Squeezed
2.3 Quantization of Optical Fields . . . . . . . . . . . . . . 42
State and the Oscillation Effects . . . . . . . . . . 79
2.3.1 Description of Modes by Simple Harmonic Oscillators 42
3.5 Mixed Quantum States . . . . . . . . . . . . . . . . . . . . 81
2.3.2 Quantization of Simple Harmonic Oscillators . . 44
3.5.1 Density Operator for the Description of a Mixed
2.3.3 Field Operators for an Optical Field . . . . . . . . 45
Quantum State . . . . . . . . . . . . . . . . . . . . 82
2.3.4 The Quasi-Monochromatic Field Approximation . 47
3.5.2 Density Operator for Lasers with Random Phases 86
2.3.5 One-Dimensional Approximation of Optical Fields 48
3.5.3 Density Operator for a Thermal State . . . . . . . 87
2.4 Further Reading: Casimir Effect, A Quantum Effect of
3.6 Further Reading: The Operator Algebra of a and at . . . 90
Vacuum due to Mode Change . . . . . . . . . . . . . . 50
2.5 Some Remarks about the Unification of Particles and 3.7 Glauber-Sudarshan P-Distribution and Wigner Distribution 93
Waves in Quantum Theory of Light . 54 3. 7.1 Glauber-Sudarshan P-Representation of Density
2.6 Problems . . . . . . . . . . . . . 56 Operator . . . . . . . . . . . . . . . . . . . . . 94
3.7.2 Wigner W-Representation of Density Operator 97
3. Quantum States of Single-Mode Fields 59 3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.1 Energy Eigen-States and Number States . . . . . . 59
4. Quantum States of Multi-Mode Fields 107
3.1. l Energy Eigen -States of a Simple Harmonic
Oscillator and the Concept of Photon . . . 59 4.1 Multi-Mode Coherent States of Independent Modes 107
3.1.2 Photon Creation and Annihilation Operators and 4.2 Classic al Description of Multi-Mode Optical Fields 109
Photon Number States . . . . . . . . . . . . . . . 60 4.2.1 Continuous Waves (CW) and Stationary Processes 109
3.1.3 q-Space Representation of Photon Number States: 4.2.2 Pulsed Waves and Non-Stationary Processes . . . 110
Wavefunction of a Single-Photon State . . . . . . 61 4.2.3 Coherence of Optical Fields - Phase Correlations . 111
3.1.4 Photon Number States as the Bas e States of the 4.2.4 Hanbury Brown and Twiss Effect - Intensity
State Space . . . . . . . . . . . . . . . . . . . . . . 62 Correlations . . . . . . . . . . . . . . . . . . . . . 112
3.2 Coherent States la) . . . . . . . . . . . . . . . . . . . . . 64 4.2.5 Transform-Limited Pulses - Mode-Locked Optical
3.2.1 Definition of the Coherent State and Its Number Fields . . . . . . . . . . . . . . . . . . . . . . . . . 114
State Representation 64 4.3 Multi-Mode Single-Photon States - Entangled States of
3.2.2 Photon Statistics Distribution and Photon Number Single Photon . . . . . . . . . . . . . . . . . . . 115
Fluctuations of a Coherent State . . . . . . . . . 65 4.3.1 Two-Mode Single-Photon States . . . . 116
3.2.3 Classical Trajectory and Quantum Uncertainty of 4.3.2 Multi -Frequency Single-Photon States -
Simple Harmonic Oscillator . . . . . . . . . . . . . 66 Single -P hoton Wave Packets . . . . . . . . . . . . 116
 X \"
xiv Quanbtm Opti cs For Experim entalists Cont Pnt s

4.4 Mu lti- Mode Two-Photon States - Two-Photon Entangled 5.5 Quantum Noise and Its Reduction by Squee zed States and
State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 Twin Beams . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.4.1 Two-Photon Polarization States . . . . . . . . . 119 5.6 Remarks about Normal Ordering and Its Relation with
4 .4.2 Two-Parti Two-Photon States of Polarization - Classical and Nonclassical Ph enomena 147
Bell States . . . . . . . . . . . . . . . . . . . . . . 119 5.7 Problems . . . . . . . . . . . . . . . . 148
4.4.3 Multi-Frequency Two-Photon States -
Frequency-Entangled States and Time-Entangled 6. Generation and Transforrnation of Quantum States 151
States . . . . . . . . . . . . . . . . . . . . . . . . . 120 6.1 Generation of Quantum States: Nonlinear Interactions
4.5 N-Photon Entangled States . . . . . . . . . . . . . . . . . 121 between Light Fields . . . . . . . . . . . . . . . . . . 151
4.5.1 Two-Mode N-Photon Entangled States - NOON 6.1.1 A Brief Introduction to Nonlinear Optics:
States . . . . . . . . . . . . . . . . . . . . . . . 121 Three-Wave Mixing and Four -Wave Mixing . 152
4.5.2 N-Parti Polarization Entangled States - GHZ 6.1.2 Two-Photon Processes: Parametric Processes . 153
States and W-States . . . . . . . . . . . . . . . 121 6.1.3 One-Photon Process: Frequency Conversion 155
4.6 Two-Mode Squeezed States - Photon Entangled States of 6.1.4 A Photon Number Doubler . . . 156
Continuous Variables . . . . . . . . . . . . . . . . . 122 6.1.5 Multi-mode Parametric Process 157
161
4.6.1 Twin Beams . . . . . . . . . . . . . . . . . 122 6.2 Linear Transformation: Beam Splitters .
4.6.2 Two-Mode Entangled States of Continuous 6. 2 .1 General Formalism . . . . . . . . 161
Variables with Einstein-Podolsky-Rosen 6.2.2 State Transformation of Number States through a
Correlation . . . . . . . . . . . . . . . . 124 Beam Splitter ............... . 162
4.6.3 Squeezed State in Multi-Frequency Mode - 6.2.3 State Transformation of an Arbitrary State .... 165
Spectrum of Squeezing 125 6.2.4 State Transformation of Squeezed States . 167
4.7 Problems . . . . . . . . . . . . 127 6.3 Optical Resonators: Input-Output Theory of an Open
Quantum System and Model of Decoherence . . . . 169
5. Theory of Photo-detection and Quantum Theory of Coherence 129 6.3.1 Classical Wave Model ........... . 170
5.1 Classical Theory of Coherence and Semi-Classical Theory 6.3.2 Further Reading: Intra-cavity Second Harmonic
of Photo-Detection . . . . . . . . . . . . . . . . . . 129 Generation . . . . . . . . . . . . . . . . . . . . . . 171
5.1.1 Classical Coherence Theory . . . . . . . . . 129 6.3.3 Quantum Treatment of Cavity Loss: Decoherence
5.1.2 Semi-Classical Theory of Photo-Detection . 131 of an Open Quantum System . . . . . . . . . . . 174
5.1.3 More on Classical Explanation of Hanbury 6.3.4 Optical Resonators with Nonlinear Interactions . 178
Brown-Twiss Effect . . . . . . . . . . . . . 133 Problems ................. • • • • • • • 184
6.4
5.2 Glauber's Photo-detection Theory and Quantum Theory
of Coherence . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.2.1 Photo-Electric Measurement and Normal Ord ering 136 Experimental Techniques in Quantum Optics and
5.2.2 Glauber's Quantum Theory of Coherence . . 137 Their Applications 191
5.2.3 Connection between Quantum and Classical
7. Experimental Techniques of Quantum Optics I: Photon
Theory and Optical Equivalence Theorem. 138 193
Counting Techniqu e
5.2.4 Classical and Non-Classical States of Light 139
7 .1 The Process of Photo-D etect ion . . . . . . 194
5.3 Anti-bunching Effect . . . . . . . . . . . . 141
7.2 Detection Probabilities of Photo-El ectrons 195
5.4 Photon Statistics and Photon Correlations . . . . . 143
 Contents xvii
xv i Quantum Optics For Experimentalists

9.2 Spectral Analysis of Photocurrents ... ... .. . 271

7.3 Photon Coun t ing . . . . . . . . . . . . . . . . . . . . . . 199
9.3 Homodyne and Heterodyne Detection Techniques . 276
7.3.1 Coinc idence l\Ieasurement with an "AND" Gate 201
9.4 Vacuum Noise and Beam Splitter Mode l of Losses . 279
7.3.2 T ime-Reso lved Coincidence Measurement 205
9.5 Spectral Analysis of Homodyne Detection ..... 280
7.4 Theoret ical Description of Experiments . 207
9.6 Mode Match and Local Oscillator Noise in Homodyne
7.5 Problems . . . . . . . . . . . . . . . . . 213
Detection . .. ..... . . . . • • • • • • · · · · · · 283
9. 7 Balanced Homodyne Detection . . . . . . • • • • • • 285
8. Applications of Photon Counting Techniques:
9.8 Intensity Fluctuations and Self-Homodyne Detection 287
Mu lti-Pho ton Interference an d Entanglement 217
9.9 Photo-detection for Ultra-Fast Pulses ..... 289
8.1 Multi-Photon Int erferen ce in General .. 217 9. 9 .1 General Consideration . . . . . . . . . 289
8. 1.1 Single-Photon Int erference and Two-Photon 9.9.2 Homodyne Detection of Pulsed Fields 290
Int erfer enc e . . . . . . . . . . . . 217 292
9.9.3 Temporal Mode Match
8.1.2 Two-Photon Wave Function . . 220 293
9.10 Problems
8.2 Various Two-Photon Int erference Effects 221
8.2.1 Hong-Ou-M andel Interfer ence 222 10. Applications of Homodyne Det ect ion Technique:
8.2.2 Time-bin Entanglement and Franson Interferom ete r 227 Quantum Measurement of Continuous Variables 295
8.2.3 Distinguishability in Two-Photon Interferen ce and 296
10.1 Squeezing and Correlation of Quantum Noise . . • • •
Quantum Erasers . . . . . . . . . . . . . . . . . . 230 10.1.1 Quantum Noise in Int ensity. . . . . . . . • • • 296
8.2.4 Cavity Enhancement of SPDC by Constructive 10.1.2 Quantum Noise of Quadrature-phase Amplitud es
Multi-pass Two-Photon Interference 236 and its Reduction: Squ eezed Stat es . . . . 297
8.2.5 Mode-locked Two-Photon States 238 10.1.3 Quantum Correlation of Quadrature-phas e
8.3 Multi-Photon Interferenc e Effects ..... . 239 Amplitudes: EPR Entangled Stat es . . . . 300
8.3.1 Multi-Photon Bunching Effects .. . 240 10.1. 4 Quantum Correlation between Int ensiti es :
8.3.2 Generalized Hong-Ou-Mandel Effects and Twin Beam s ... . .... .... • 305
Destructive Multi-Photon Interference .. 246 10.2 Quantum Noise and Its Reduction in Lin ear
8.4 Quantum Interference and Photon Indistinguishability .. 248 Interferometers ........... • • • • • • · · · · · 308
8.4.1 N-Photon State from N Single-Photon States and 10.2.1 Quantum Noise Analysis for a Mach-Zehnd er
Photon Indistinguishability .. .. ........ . 249 Interferometer . . . . . . . . . . . . . • • • • 309
8.4.2 Pair Distinguishability and its Characterization 252 10.2.2 Sub-shot Noise Interferometry with Squeezed
8.4.3 Characterization of Photon Indistingui shability by States ................ . 311
Multi-Photon Bunching Effects ......... . 254 10.2.3 Quantum Noise Analysis for LIGO . 314
8.4.4 Optical Coherence as a Consequence of Photon 320
10.3 Quantum Noise in Amplifiers ...... . .
Indistinguishabili ty 261 10.3.1 Quantum Amplifiers in General .. 320
8.5 Problems 263 10.3.2 Quantum Noise Reduction in Amplifiers . 322
10.3.3 Quantum Correlation for Quantum Noise
9. Experimental Techniques of Quantum Optics II: Reduction in Amplifiers . . . . . . . . . • • • · · 324
Detection of Continuous Photo-Currents 269 10.3.4 Phase-sensitive Amplifiers ....... • • · · · · 328

9.1 Photocurrent and Its Relation to Quantum Measurement 10.4 Complete Measurement of Quantum States: Quantum
State Tomography . . . . . . . . . . . • • • • · 330
Theory .... .... . . .................. . 269
xviii
Qua7lh .mi Opti cs For E xperim entalists

10.5 Comp lete Quant um State Teleportat ion 332

10.6 Problems 336

11. Quantum Noise in Phase 1\Ieasurement 341

11.1 Phase Measurement in General . . . . . . . . . . . . 342
11.1.1 Ult im ate Quantum Limit in Precision Phase
Measurement . . . . . . . . . . . . . . . . . . 344
11.1.2 A Necessary Cond ition for t he Heisenberg Limit . 354
11.1.3 General Cons iderat ion in the Search for Schemes
Reachin g the Fundamental Limit . . . . . . . . . 355
11.2 Phase Measurement Schemes Reaching Heisenberg Limit 359
11.2.1 Schemes Reaching Heisenb erg Limit with PART 1
Conventional In terferometers . . . . . . . 359
11.2.2 Schemes Reaching Heisen b erg Limit with
Theoretical Foundations of Quantum
Unconventional Int erferometers . 364 Optics
11.3 Non-conventional Int erferom eters .. . ... . . . 370
11.3.1 Genera l Considerat ion .. ... ..... . 370
11.3.2 Cohe rent State Boosted SU(l ,l) Int erfero meters 371
11.3 .3 Loss Analysis for the SU(l ,l) Interferometer 377
11.4 Joint Measurement of Conjugate Obs ervables ..... . 380
11.4.1 Classical Measurement Schemes . . . . . . . . . 380
11.4.2 Joint Measurement with EPR Correlated St at es 382
11.4.3 Joint Measurement with SU(l ,l) Int erferom eter 384
11.5 Probl em s . . . . . . . . . . . . . . . . . . . . . . . 387

Appendix A Derivation of the Explicit Expression for U of

a Lossless Beam Splitter 391

Appendix B Evaluation of the Two Sums in Eq. (8.100) 393

Bibliography 397

Ind ex 411
 Chapter 1

Historical Development of Quantum

Optics and A Brief Introduction

1.1 Historical Background

Light is the most commonly seen matter in the universe, it is also the
simplest system in the universe. That is why understanding the behavior of
light has played a pivotal role in the development of physics . Many ground
breaking concepts in physics originated from optics. For example, Fermat's
least action principle was first developed in optics: light propagates between
two points in a medium in the shortest time. Then it was extended to other
physical systems. The first conceptual revolution in quantum mechanics
stemmed from the study of blackbody radiation. Quantum information is
the current hot topic in research, yet many of the protocols of quantum
information were first realized in optical systems. The reason for all this
is straightforward: the simplicity of optical systems makes them easier
than others to implement otherwise complicated models in physics. So,
understanding of light will help us study other physical systems . Quantum
optics is the most complete theory of light. It can explain all the optical
phenomena observed so far.
The development of classical optics has a history of a few hundred years.
Its highest form is Maxwell's electromagnetic theory of light. Although
quantum optics was only developed in recent decades, its origin can be
traced back to the start of quantum mechanics. Planck's quantum theory
of blackbody radiation [Planck (1900)] is about energy quantization in the
atomic emission and absorption of light . It was assumed that the energy
of atomic emission or absorption of light can only be integer multip les
of some small quantity E. Then Einstein in 1905 introduced the concept
of light quanta for the explanation of the photo -electric effect [Einstein
(1905)] . This concept is about quantization of the optical fields and is

3

4 Quantum Optics For Exp e1·imentalists
totally independent of atoms. Therefore, it was generally believed that the
blackbody radiation theory of Planck in 1900 was the start of the quantum
theory whereas the concept of photon was born in 1905 for the start of
quantum optics.
It is worth noting that after the introduction of photon, Einstein in
1909 studied the energy fluctuations in blackbody radiation and proposed
for the first time the duality of wave and particle for light [Einstein (1909)].
This result is earlier than the duality theory of matter waves by de Brog-
lie. Starting from Planck's blackbody radiation energy spectrum formula,
Einstein applied the general argument of thermodynamics to arrive at the
following formula for the energy fluctuations:
--- - -2
(6.E) 2 = hvE +E /Z, (1.1)
where A is the average value of quantity A. hv is the energy of a pho-
ton and Z is some function related to frequency v and thermal energy
kT. In discussing the physical meaning of Eq. (1.1), Einstein first assu-
med that the field of blackbody radiation consists of independent particles
as photons and obtained from the Poisson statistics of random particles
the fluctuation of particle number N as: (6.N) 2 = N. Then using the
famous formula for energy quanta from his paper on photoelectric effect:
E = hv and total energy formula E = NE, he obtained (6.E) 2 P = hvE,
the first term in Eq. (1. 1). For the second term, Einstein assumed that
the field of blackbody radiation consists of independent plan e waves and
. --- -2
obtamed (6.E) 2 w = E /Z. From these, Einstein rewrote Eq. (1.1) as
(6.E) 2 = (6.E) 2 P + (6.E) 2 w· With this, Einstein concluded that the energy
fluctuations of blackbody radiation exhibit the behaviors of both particles
and waves, i.e. , the particle and wave duality. In the concluding remark of
the 1909 paper, Einstein gave the following assertion:
"... the next stage of the development of theoretical physics will bring us a
theory of light which can be regarded as a kind of fusion of the wave theory
and the emission theory . . . a profound change in our views of the nature
and constitution of light is indispensable."
Einstein was calling for a brand new theory of light in which both par-
ticle and wave are unified. Such a theory was constructed by Dirac in 1927
after the completion of the theoretical framework of quantum mechanics
[Dirac (1927)]. Later, after Schwinger, Tomonaga, and Feynman solved
the difficulties of infinity with renormalization method, this theory became
quantum electrodynamics (QED), which is a part of the Standard Model of
physics, the most complete and thoroughly tested theory so far. The most
5
Histori cal D ev elopm ent of Quantum Opti cs and A Bri ef Introduct ion
fcunous prediction of QED is the Lamb shift betv:een 2S and 2P energy
levels of hydrogen atom, which was confirmed experimentally [Lamb and
Retherford (1947)].
However, early studies with QED concentrated mostly on the behavi-
ors of individual photons and electrons. For the case of multiple photons.
it was simply assumed that photons are independent of each other. But
an experiment performed by Hanbury Brown and Twiss in 1956 sho~ed
that it is not the case (Hanbury Brown and Twiss (1956a)]. The collective
behavior of multiple photons is quite different from that of multiple inde-
pendent photons but related to the correlation of optical fields at_ different
locations and times. This is the problem to be solved by the optical cohe-
rence theory which was developed by Mandel and Wolf in 1950s [Mandel
and Wolf (1997)]. Optical coherence theory stems from the classical wave
theory of light. In the early days, it mostly dealt with the phenomen~ of
optical interference and studied the phase correlation between the optical
fields at different locations and times. The experiment by Hanbury Brown
and Twiss, on the other hand, demonstrated for the first time the intensity
correlat ion of the optical fields.
The classical coherence theory of light soon evolved into the semi-
classical theory of light for the interaction between light and media, in
which only the media (atoms or molecules) are described quantum mecha-
nically but the light is still treated as waves with the classical Maxwell
electromagnetic wave theory. It simply introduced randomness and fluctu-
ations for the optical fields and became statistical optics. It can explain
many optical phenomena including the intensity correlation obser~ed _by
Hanbury Brown and Twiss. By 1970s, this theory, with the quant1zat10n
of media only, was successful in explaining many QED effects including
photo-electric effect and Lamb shift . A question was thus naturally raised:
is it still necessary to quantize the optical fields?
Soon after the establishment of the classical coherence theory of light,
Glauber developed the quantum theory of optical coherence in 1963 [Glau-
ber (1963a,b)] . In this theory, Glauber first worked out the the~ry of ph~to-
detection based on QED and then defined multi -order correlat10n funct10ns
similar to the classical coherence theory. This is a completely quantum
mechanical theory of light and is the foundation for quantum optics . Mo-
reover, Glauber's photo-detection theory is also the foundation for the ex-
perimental part of the current book. As illustrated in Fig. 1.1, the box
on the left contains the optical fields that are described by the theory of
quantum optics. The box on the right includes what we observe in the lab
 --- 7
. l t f Quantum Optics and A Bri ef Introdu ction
6 Quantum , Optics For Experimcntal'ists Histori cal D eve opm en o

"TERFEREtKE HALF-SILVERED
LIQUID l " MIRROR
FILTER /mm~ I PHOTOMULTIPLIER
rusE
011tical ltcld :
"'
3 :
Spectrum
Analyzer -\-nIn ---------'Z_~
ffi_ -0 -~--
1,0,
Q,,
-
D
. d)
a1 .:LJ I Oscilloscope I \
MERCURYLENS RECTANGULAR ' SLIDE
-- ~~ ARC APER1URE c.1
Theory
Photo-detection Experimental observation

Fig . 1.1 The relationship between the theory of quantum optics and the experiments.
Glauber's photo-detection theory connects the theory with the experiments. INTEGRATING
MOTOR

• t Reproduced from [Hanbury

with all kinds of modern instruments. Glauber's photo-detection theory on . 1 2 Sketch of the Hanbury Brown-Twiss expenmen .
Fig ..
Brown and Twiss (1956a)].
photo-detection processes bridges between the theory in the left box and
the experiment in the right box. This will be the main task of the current
orrelation for two points outside the region. d T . de-
book. C t H bury Brown an w1ss
After the discovery of the HBT eff ec ' an .
1 d further this newly found spatial correlation techruque for the mea-
1.2 Hanbury Brown-Twiss(HBT)Experiment ve ope t f the size of stars with int ensity stellar interferometry [Hanbury
suremen o · · ·fi f the HB T
. (1956b)} However the real s1gm cance o
Brown an d T w1ss . ' f . 1
Hanbury Brown- Twiss (HBT) experiment [Hanbury Brown and Twiss . t . that they abandoned the traditional method o optica mea-
(1956a)] is one of the earliest experiments in quantum optics. Although expenmen w::h
a sin le detector but instead employed two detect.ors fo~ ~he
it demonstrated a classical phenomenon of light , the technique used in the
experiment is one of the two major experimental techniques in quantum
~~;::e:servation o1
the correlation between the fluctuation~ao~::e~:~::;
This is a breakthrough in experimenta l methodolog! and p y p
optics and has been widely applied in experimental measurements. Be- role in the subsequent development of quantum ~pt1cs. . . -
cause of this, we can claim that the HBT experiment laid the foundation of . 'fi O
f the HBT experiment is the first discovery of m
Another s1gm cance · h hase
experimental quantum optics. Furthermore, this experiment demonstrated t ·t or amplitude fluctuations of optical fields. Bef~re t~1s, t e p
for the first time the fluctuations of intensities or the amplitudes of optical ens1 y . f tical fields have been observed in the mterference pheno-
fields. This correlation is different from optical coherence which is about fluctuations .o op S both the amplitude and phase of an optical
mena of optical waves. o now' . ·11 1
the phase correlations of optical fields. fi ld t fixed but fluctuate ·with time. In the followmg, V:e w1 exp ore
The experimental setup of the HBT experiment is shown in Fig. 1.2. e are no d 1 these fluctuat10ns.
further these phenomena to see how to mo e
The light source of the experiment is a mercury vapor lamp. After exiting
the source, light first passes through an optical filter to select a single
Fluctuations of Light in Phases and Amplitudes
spectral line and filter out stray light. It then is split into two by a beam 1.3
splitter and sent separately to two photo-detectors. A transverse spatial In the wave description of light , an ideal optical wave can be expressed as
translation was introduced on one of the detectors to move it in and out
of the region of spatial coherence. (A temporal delay may also be used to a plane wave:
(1.2)
one of the detectors so that it can be pushed in and out of the region of E(r , t) = Eo cos(k · r - wt+ <po),
temporal coherence). Finally, the output photo -electric currents from the
two detectors are multiplied for correlation measurement. The result of the or as a spherical wave:
observation is that there exists some correlation between the fluctuations (1.3)
E(r , t) = ~o cos(kr - wt+ <po).
of optical intensities at two points within spatial coherence region but no

8 Quantum Optics For Experimentalists
Herc cpo is the initial phase independent of time. For the conven ience of mat-
hematical expression, we usually use comp lex number to represent rosine-
function:
E (r, t) = Eo eicpoei( k-r -w t ) = Ao ei( k-r -wt) .
(1.4)
Hence, the phase and amp litud e of the wave function can be expressed in
a unified way by a complex amplitude Ao = E 0 eicpo. Here, the intensity
of the wave I= E *(r, t) · E(r , t) = Eo · Eo = 1Ao l2 is only related to the
abso lute value of the comp lex amplitude.
In this description, the wave function' s dependence on time is an unin-
terrupted continuous cosine-function. It is used to describe a monochroma-
tic wave of angular frequency w. The wave train is infinit ely long , as shown
in Fig. 1.3.
Fig. 1.3 A continuous and infinitely long monochromatic wave train.
On the other hand, we know that light is emitted when an atom jumps
from an excited state to a lower energy level and the excited state of the
atom has a finite lifetime 6-T. So, the light wave emitted by the atom is a
wave packet of a length traveled by light in the atomic lifetime , as shown in
Fig. l.4(a). Therefore, in order to describe a continuous wave train with a
finite wave packet emitted by the atom, we connect every finite wave packet:
the wave train is still infinit ely long for continuous wave description but
there exist interruptions and discontinuity, as shown in Fig. l.4(b). This
is our first model of light that is close to an actual optical field, where the
amplitude of the wave is constant but its phase cp0 is not fixed: cp0 = cp(t).
It jumps after every period of time t::.T to another random value.
I In classical Newtonian mechanics, if the number of particles is extreme ly
(a) (b)
Fig . 1.4 (a) The finite wave train emitted by atoms. Its size is a length of c!::,.T trave led
by light in the lifet im e !::,.T of the atom's excited state. (b) The infinitely long wave train
formed by connecting finite wave packets emitted by atoms.
The change of the initial phase cp0 with time will affect the observation of
optica l interference fringes. First, the phase difference cp10 - cp20 of different
H istor ica l D eve lopm en t of Qnantum Opti.cs and A Brief Introdnction 9
optical fields is random. This will average out the interference fringe s.
Secondly, even if thf' two interfering fields are from the splitting of one
field, as in a 1\Iichelson interferometer, interference fringes sti ll disappear
when the delay between the two fields is larger than the length ct::.T of th e
finit e wave packet emitted from atoms. This is because when T > 6-T,
t + T) - c.p(t)is random. This leads to the concept of optical coherence
c.p(
(more in Section 1.5), where a second -ord er correlation function:
(E*(t + T)E(t)) (1.5)
"f(T) = (E*(t)E(t))
can be used to describe the correlation of the phases. Its absolute value
gives the visibility of the interference fringes. The range where 'Y(T) is non-
zero corresponds to the coherence time Tc rv 6-T of the field. Here the
average () is taken over the random variable cpo.
In the model above, the amplitude of the field is unchanged and so
is the intensity of the field. But this model cannot explain the intensity
fluctuations observed in HBT experiment. For this, we need to introduce
amp litude fluctuations. In reality , this is because the time of each atomic
emission is random, as shown in Fig. 1.5. So, in our second model of the
optical field, the whole comp lex amplitude A o = Eo eicpo (both magnitude
and phase) is a complex random variable. To better understand this, we
next briefly describe random variables and random processes.
\ I
Fig. 1.5 The random emission of atoms.
1.4 Random Variables and Processes
large, as in an ensemb le of molecules in a regular vapor cell, we are unable
to solve the equ at ions of motion for all the particles. In this case, we can
only us e the method of statistics to take average over these particles so as to
obta in macroscopically measurable quantities. This is classical statistical
mechanics. In classical optics, when the number becomes large for the
light em ittin g atoms or atoms in the medium of light propagation , we are
unable to solve the Maxwell equations to obtain the solut ion s for the optical
 Hisforical Development of Quantum Optics and A Brief lntroduct-ion 11
10 Quantum , Opt'ics For Experimentalists

waves em itted or scattered by each atom. In this case, we must resort to 1,4.2 Continuous Random Variables
statistical methods to treat this type of problems. This is classical statistical For a cont inuous variable X = [x1 , x2], we first discretize it: we divide the
opt ics, which is the basis of classica l coherence theory. The mathematical range [x 1 ,x 2] into Ji.I small regions, i.e., X = [x1,X1 + 6] + [x1 + 6,x1 +
foundation of statistica l mechanics and statistical optics is the theory of 26] + ...+ [x1+ (M -1)6, x 2], where 6 = (x2 -xi)/AJ. Then, similar to the
probability. In this section, we will systematically but briefly present some discrete case, we measure N times the quantity X and record the numb er
major resu lt s and app lications of the probability theory and statistics. of times of the outcomes falling in each region: N1 times in [x1, xi + 6],
N 2 times in [x1 + 6 , x 1 + 26], etc . So, similar to the discrete case, we
1.4.1 Discrete Random Variables can obtain probabilities p 1 ,p 2 , ·· ·,PM for each of the M small regions as
Consider a ph ysical quantity A. When we measure it, there are m different N --+ oo. Obviously, when M is large, Pi ex: 6. Then, we can obtain a
outcomes: A= {A 1, A 2, ... , Am}- For examp le, tossing a coin will give on ly density of probability for region i:
two outcomes: A = {head, tail}; but throwing a dice leads to six different p (x ) = .
11111 Pi
A or uJ\p = Pi = p ( x ) u.J\ (1.10)
values: A= {l , 2, ... , 6}. We make the measurement N times , in which we M-+oo u
hav e N 1 times the value A 1, N2 tim es the value A2, ... , and Nm times the When 6 --+ 0, 6P--+ dP = p(x)dx is the probability of finding X in the
value Am. Obviou sly, N1 + N2 + ... +Nm= N. region [x, x + dx]. For a finite range [a, bl, the probability of finding X in
From statistics of math ematics , we know that when N --+ oo, Ni/N th is region is
will approach a limit P1 = limN-+oo Nif N. Similarly , P2 = limN -+ooN2/N,
... , Pm = lim N-+ooNm/N. Here the limiting values P1,P2, ···,Pm are the Pab
= 1bp(x )dx. (1.11)
probabilities of the physical quantity A taking the value of A 1, A 2 , ... , Am,
respectively. Obviously, 0 ::SPi ::Sl(i = 1, 2, ... , m) and th ey satisfy norma- Obviously, we have the normalization relation
lization condition : P1 + P2 + ...+ Pm = Li Pi = 1.
1
X2

Px 1 x 2 = p(x)dx = 1. (1.12)
The average of the physical quantity A is defin ed as
X1

(A) = A1N1 + A2N2 + ...+ AmNm = ~A· _ (1.6) Similar to the dis crete probabilities P1,P2, ... , the probability density p(x)
- N ~z-
determines the properties of the continuous random variable X, whose
Its variance is average and variance are given by
= ((A - (A) ) 2) = (t? A) = (A 2) - (A) 2
1
X2
(1. 7)
1
X2
var(A)
(X) = x p(x)d x, (6 2 X) = (x - (X) )2p(x)dx. (1.13)
and the standard deviation is a-A = var (A) = (
6 2A) . For example, for Xl Xl

dice throwing, if the dice is a regular cube, we have p 1 = p 2 = ...p 6 = 1/ 6. For the moment of any order , we can find it from the mom ent generating
Then (A) = 3.5, var(A) = 2.9, and a-A = 1. 7. and characteristic functions. They are related to the probability density
Higher order moments are defined as p(x) by
(Ar) =Al Ni + k2N~+ ... + A~Nm = L A[pi =Vr. (1.8)
Mx(O = J e~xp(x)dx, (1.14)

They can be obtained from the moment generating function MA(() = The normalization relation in Eq. (1.12) gives Mx(O) = 1 = Cx(O). Notice
(e~A) = Lr(AT)e /r!: that p(x) and Cx(() are a pair of Fourier transformation. So, we can obtain
the probability density p(x) from the character istic function Cx(fr
(1.9)
(1.15)
Another important function is the character isti c function: CA (() = (e1~A).
 Histori cal D ev elopm ent of Quantum Optics and A Brief Introduction 13
12 Q'uant'llm Optics For Experimenta lists

For cont inuous var iab les, consider a 2-di mens ional vector R = (X, Y).
Pro bab ilit ies {pi} or prob abilit y density p(x) are also kn own as pro b a-
bili ty dist ribu t ions. Some well-kn own prob abili ty distr ibu t ions are
If var iables X, Y are bot h ra n dom vari ables, vecto r R = (X, Y) is a 2-
dim ra ndom vecto r. We can likewise write t he joint prob ab ility of findi ng
(i) Ga ussian norm al distribution: X = [x,x + dx], Y = [y, y + dy] as dP(x. y) = p(x, y)dxdy wit h p(x, y) as
the jo int proba bili ty density. T he norm alizat ion condi t ion is
p (x) = __ l _e-(x-
J2m;- 2
(ii) Poi sson distribution
µ)2/ 2a2 ( (X) = µ , var(X) = a-2),

(discret e vari abl e) N = {O, 1, 2, ... , n, ... }:

(1.16)
J p(x, y)dxdy = l. (1.19)

From p(x, y), we m ay obta in t he prob abili ty densit ies px(x),py(y) for X
>,_
k
P ( k) = k ! e ->-, ((n) = >..
, var (n) = >..
). (1.17) and Y, resp ect ively :

(iii) Binomi al distribution (dis cret e vari able) n :s;N ,p :s;l: px(x ) = J p (x, y)dy , py (y ) = j p(x, y )dx. (1.20)

N!p n (l _ p) N-n
p(n) = n!(N _ n)! , ( (n) = pN, var(n) = N p(l - p)) . (1.18) 1.4.3.2 Condit ional Probability
With th e joint probabilit y, we m ay ask furth er: what is the prob ability of
1.4.3 Joint Probability of Multiple Random Variables finding X = x 2 given Y = y3? This is th e conditional probabilit y with
Now we deal with multiple random variables. Consider first the cas e of th e condition Y = Y3· We writ e it as px( x 2\Y = y3) or abbreviat e it as
tossing two coins. There ar e four possibilities: A = { (HlH2) , (Hl T2) , px (x 2\y3). Since we ask about th e probability ofrandom vari able X and it
(TlH2) , (Tl T2)}. For the case of throwing two dices, th ere are totall y t akes only two values x 1, x 2 in th e example above, th e probabilities should
6 x 6 = 36 possibilities: A= {ll , 12, 13, ... , 21, 22, 23, . .,, 66}. We may ask b e normalized: px( x 1\y3)+px(x2\y 3) = 1. In other words , with Y = y3, we
what are the probabilities for each possibility. have either X = x 1 or X = x 2 and the total probability for th e two possibi-
liti es must b e 1. Similarly , we have px( x 1\Yj) +px(x2\Yj) = l(j = 1, 2, 3) .
To summarize , wh en talking about conditional probability , we are only con-
1.4.3.1 Joint Probability
cern ed about the subset that satisfies the condition , e.g. , y = y3. In th e
For the convenience of discussion , we consider two discrete random varia- example above , the subs et of conc ern is (x 1, y3) , (x 2, y3) and th e conditi-
bles: X = {x1, x2}, Y = {Yi, Y2, y3}. We denote the probability of having onal probability must be normalized in this subset. Notice th e difference
simultaneously X = Xi, Y = Yj as Pij, where i ,j are the indices for the first betw een conditional probabiliti es px(x1\y3) ,Px( x 2\y3) and the joint pro-
and second random variables, respectively. Here, the simultaneous proba- babilities p(x 1 , y3) = p 13,p( x 2, y3) = p23: the former is normalized whe-
bility Pij is called joint probability. In the current example, it has 2 x 3 = 6 reas the latter is not. They are certainly related. Obviously , we have
values. Table 1.1 lists all 6 possibilities and their joint probabilities. px(x1\y3) ex p(x1,Y3) = p13, px(x2\y3) ex p(x2 , Y3) = P23, that is,
Table 1.1 List of joint probability for X Y
px( x i\y 3) = Cpi 3. Normalization condition gives C = l/(p1 3 + p23). The-
)

(X , Y) Yl Y2 Y3 PX refore , we arrive at
x1 Pll Pl2 Pl 3 PX (x1) 3 (1.21)
X2 P21 P22 P23 PX (x2) Px( x 1\y3)= P~ ,
P13 P23
py py(y1) PY (y2) py(y3)
From what we had earli er: py(y3) = p 13 + p23, we obtain Bayes ' theorem
p(x1 , y3) = Px (x1 \y3)py (y3) or the more general form
Of course , we can still ask about the probability Px 1 = Px (x 1) of X = x 1 p(xi , Yj) = Px (xi \yj )py (yj) = py (yj \xi)Px (x i). (1.22)
or the probability Py3 = py(y3) of Y = y3 as before. From Table 1.1, we find
If X , Y are independent of each other , the outcomes of measuring X has
Px(x1) = P11 +P12 +p13 and py(y3) = P13 +p23. Probability normalization
nothing to do with Y. Hence px( x i\Yj) = px(xi) - Similarly, Px(Yj\xi) =
condition requires LPij = l = px(x 1)+px(x2) = py(y 1)+py(y 2)+py(y 3).
 14 Quantum Opt ·ics For Experimentalists Historical Development of Quantum Optics and A Brief Introduction 15

py(yj), Then , from Bayes ' theorem, we have p(xi,Yj) = Px(xi)PY(Y.i) for Treating x, y as the coordinates of a point , we rotate X - Y coordinates
independent random variables X , Y. and obtain th e general form of Gaussian distribution for two variables:
The discussion above also applies to cont inuou s variables. We only need
= __ l_ eA(x-µx )2+ B (x- w-: )(y-µ), )+C(y-µy )
2
(1.26)
p (.r, y )
,
to drop the subscripts i. j in the corresponding equations. 21rJJJ

where B 2 - 4AC < 0, A, C < 0 and N = l /l B 2 - 4ACI is a normalization

1.4.3.3 Coefficient of Correlation
factor. The correlation coefficient between X and Y is PXY = B /~ -
In genera l, p(xi, Yj) =/-p(x i)P(Yj), that is, p(xilYj) depends on the outcomes The general form of Gaussian distribution fork variables X 1 , ... , Xk is
of Y. Conditional probability Px(xi/Yj) then describes the relationship
between X and Y. For examp le, if Px(x 1/y3) = 1, that means whenever
p(x1, ... , xk) = J( 2:JklE exp [- (x - µf E
1
(x - µ)], (1.27)

Y = Y3, we must have X = X1. The reverse is not always true though: when
X = X1, we may not have Y = y3. But if p(x1/y3) = 1 = p(y 3/xi), x 1 and
y3 are comp letely corre lated. On the other hand, if p(xi,Yj) = p(xi)p(yj),
X and Y are completely uncorrelated.
For the case of partial correlation, we use the covariance quantity
where xT = (x 1 , ... , xk ) is a k-dimensional real vector variable, µ is a k-
dirn real constant vector, I b I is the determinant of matrix b , and b - l
is a symmetr ic positive definite k x k matrix. The correlation coefficient
to
between Xi and Xj is Pij = 0 / b 1

describe, which is defined as For a multiple of variables with Gaussian distribution, we have the
Isserlis Theorem about the higher order moments:
(~X ~Y ) = (XY) - (X) (Y). (1.23) (x1X2...X2n) = L IJ\XiXj) (1.28)
By Cauchy-Schwarz inequality, we have j (~X ~Y) 2 :s;(~ 2 X) (~ 2 Y). After
1 (1.29) .
normalization to the right side of the inequality, we obtain the dimensionless
correlation coefficient: where I:: is the sum over all possible ways of pairing X1X2...X2n and IT is
the product over all pairs from x 1x2 ...X2n• (xixj) = b 01 can be obtained
l(~X~Y )j directly from the symmetric matrix b- 1 .
Pxy = x=)=(
~==c:=2 Y=)
~==;=2 (1.24)

Obviously, if X and Y are completely uncorrlated , i.e. , p(xi, yj) =
p(xi)P(Y.i), we have PxY = 0. On the other hand, if pxy = l, it can
be proved that X and Y has a linear dependence: xi = Cyi + D , that is
p(xi) = P(Yi) = Pi or p(xi, Yj) = Pi6ij · Then from Bayes' theorem, we
have Px(xilY.i) = 6ij = py(yj,Xi)- This means X and Y are completely
correlated . }or the general case, we have O :S p xy :S 1 by Cauchy-Schwarz
inequality. So, correlat ion coefficient PXY quantitatively describes the de-
gree of correlation between X and Y.
1.4.3.4 Gaussian Distributions
For Gaussian distribution of two variables, we start with the independent
case where the joint probability density is a product of two Gaussian dis-
tributions:
1.4.3.5 Central Limit Theorem of Infinite Number of
Random Variables
Consider the sum of M random variables X1, ... , XM: X = (X1 + X2 +
...+ XM ), which is also a random variable. If all M random variables are
indep en dent and identical with the same probability distribution p( x), the
Central Limit Theorem of probability theory states that when M is very
large, the probability distribution of X is a Gaussian with its average equal
to M(Xi) = M J xp(x)dx and its variance as M(~ 2Xi)- This conclusion
is independent of the original distribution p( x). This is the reason why
we encounter Gaussian statistics so often and Gaussian distribution is also
called normal distribution.
1.4.4 Random Processes

When the random variable changes with time, we need to investigate the
(1.25) dynamic behavior of the random variable. This is the random process .
 16 Quantum Optics Fol' E.rperi111cntalists Historical Development of Quantum Optics and A Brief Intrndnction 17

Consider a t ime-dependcnt function of the random variable X: YY (t) = random variables and time: E(r,t) = f({a}.t). In later chapters (Chap-
f(X, t). AL a specific time t = t', Y\."(t') = f(X, t') is another random
ters 4 and 5), we will specify these random variables {CY} explicitly. The
variable related to X: 1~\."(t') = {y = f(x , t') ,J: = [x1,:r2]} . But for a quantities we measure in the experiment are the average over these random
specific value x = .r' of X, yx(t) = f(x', t) is a function of time. J.., is a variables. For example, optical intensity (J(r, t) ){Cl}= (E*( r, t)E(r, t)){a}·
possible outcome when we measure X, or a samp le. Then Yx(t) = J(x', t)
is one possible observed process of the random process Yx(t) = f(X, t), or a
realization. A random process is realized with an ensemble of many sample
processes . For example, Y,\."(t) = X cos wt describes a harmonic process
with a random amplitude. Figure l.6(a) shows two observed samples.
s Ep(r, t)

Fig. 1. 7 Sketch of Young's double slit experiment .

As an examp le, let us consider interference experiment of light. The

(a) first one is Young's double slit experiment (Fig. 1.7). At observation point
(b)
P, the optical field is written as
Fig. 1.6 ( a) Two observed samples of random process Yx (t) =X cos wt. (b) One sample
process for the random process Yx(t) = Li Xi cos wit- _ E(r1, t - Ri/c) i[ki-(r -ri)-wt] + E(r2, t - R2/c) i[k2 -(r-r 2 )-wt]
Ep (r ,t ) - ------e R e ·
R1 2

The two sample processes in Fig. 1.6(a) do not look random at all. (1.33)
That is because it only dep ends on one random variable. If it is a function Here the two slits act as two point sources with amplitudes determined
of multiple random variables: Yx(t) = L i Xi cos wit, one sample of the by the field E(r, t) right before the slits at positions r1 , r2. Time delays
observed realization will look more random, as shown in Fig. 1.6(b). originate from different propagation times. If the distances from P to the
Similar to random variables, we can take an average of a random process: two slits are equal: R 1 = R 2 = R, the intensity of light at P is
(Yx(t)) = j J(x, t)px(x)dx. (1.30) (Ip(r, t')) = (IE(r1, t)l2) + (IE(r2, t)l2)
We can also obtain the correlation at two different times: +[(E*(r1, t)E(r2, t)) ei(k1-k2)·r+i<po+ c.c.]
(Yx(t1)Yx(t2)) = j f(x, t1)f(x, t2)Px(x)dx. (1.31) ex 1 + Vbd cos[(k1 - k2) · r + 6<p], (1.34)

Because it belongs to the same process, the definition above is also called where t' = t + R/ c. The cosine-function in the expression above gives the
auto-correlation function. When a continuous random process does not interference effect. <po= k 2 · r 2 - k 1 · r 1 and 6<p is some phase difference
have starting and ending time, it is not sensitive to the time shift. This related to <po. V is related to the optical intensities at r1 , r2:
process is known as a stationary process, which satisfies
_ 2J(J(r1, t))(J(r2, t))
(Yx(t1 + T)Yx(t2 + T) ... Yx(tk + T)) = (Yx(t 1)Yx(t 2) ...Yx(tk)). (1.32) (1.35)
V = (J(r1, t)) + (J(r2, t))'

1.5 Classical Coherence Theory of Light where (J(ri , t)) = (IE(ri, t)l2) (i = 1, 2) . ,12 is the spatial coherence
function defined as
In the classical coherence theory of light , the field function E (r , t) is a (E*(r1, t)E(r2, t))
random process that satisfies Maxwell equations. It is a function of some ,12 = -'-;======:=======c=
J(J(r1,t))(J(r2,t))
(1.36)
 Historical Development of Q'Uant1mi Optics and A Brief Introduchon 19
18 Quant'Um Optics For Experimentalists

Here, because phases change with time, (E) usually is zero. So, the quantity
')'12above is the same as the correlation coefficient between random variables
E*(r1, t) and E(r2, t) defined in Eq. (1.24). From Eqs. (1.34) and (1.35),
we find that when the two intensities are equal, the absolute value of the
spatial coherence function ')'12 is the visibility of the interference fringes.
Therefore, it describes the coherence property of the optical field at two
points r 1 , r 2 .
E(t)
Fig. 1.8 Sketch of Mach-Zehnder interferomet er.
Next, we take a look at Mach-Zehnder interferometer (Fig. 1.8). We can
show similarly that its interference visibility is given via field correlation
function as
Classical Interpretation of HB T Effect
1.6
The light bunching effect first discovered by Hanbury Brown and Twiss
is about the auto-correlation function gC2 ) ( T). Figure 1.9 shows the re-
sult measured by Hanbury Brown and Twiss in 1958. Normalized correla-
tion function f 2 (v0 , d), which is gC2 )(T) - 1, is plotted against separation
d (equivalent to CT). The bunching effect appears at d = 0 or T = 0 with
c2) (0) = 2, which also indicates that there exists intensity fluctuation for
9 2 2
the detected optical field, for otherwise we must have (1 ( r , t)) = (I (r, t))
gC ) (0) = 1.
2
or
Next, we will explain the HBT effect using what we learned earlier in
statistical optics. First, when T • oo, I( T) and J(0) are completely irrele-
vant to each other or uncorrelated and we have (I (T)I (0)) = (I (T)) (I (0))
or gC 2 ) ( oo) = 1, which is consistent with Fig. 1.9 at T • oo. For the case
of 7 = O, it is related to the probability distribution p(I) of the intensity of
the optical field. To find p(J), we need to establish a model for the optical
field. For the atomic vapor cell, we assume that there are N identical atoms
emitting light at t = 0. So, the observed optical field can be written as
(1.39)

')'(T) = (E*(r, t + T)E(r, t)) _ where Xi = Ai cos c.pi,Yi =Aisin c.pi-Because of the randomness of the time
(1.37) of atomic emission, the amplitude and the phase of each atom at t = 0 are
J(I(r, t + T))(I(r , t))
all random, as shown in Fig. 1.5. This means Xi, Yi are random variables
This is the temporal coherence function which describes the coherence pro-
for each atom.
perty of the optical field at two different times. Due to cosine- or sine -
function, field function E(r, t) is more sensitive to the fluctuation of its
phase than the fluctuation of its absolute value or intensity IE(r, t)l2. So, l·O

what coherence function ')'12 describes is typically the correlation of the

phases of the optical fields.
In HBT experiment, the correlation measured by two photo-detectors
gives the correlation function between the intensities of the optical field at
two locations or two times:

(2)(T) = (J(r,t+T)J(r,t))
9 (1.38)
(I(r, t + T))(J(r, t)) ·
10
Here, when T = 0, we obtain the auto-correlation of intensity: g( 2 ) (0) =
!/
(12 (0) (J(0) )2. Obviously, 9( 2 ) ( T) depends on the fluctuations of intensity
separation of cathodes, d (mm)

(amplitude) of the optical field but is ind ependent of the phase fluctuations Fig. 1.9 The bunching effect observed in Hanbury Brown-Twiss experiment. Reprodu-

of the field, which is closely related to ')'12(T) as we have just seen . ced from [Hanbury Brown and Twiss (1958)].
 ..
Historical Development of Quantum , Optics and A Brief Introduction 21
20 Quantum Optics For Experimentalists

For the atomic vapor , N rv 10 12 "' 14 and we can apply the Central Limit
Theorem to E(0) = X + jY so that
both are Gaussian random variables. Because of the randomness of cpi, we
have (X) = 0 = (Y), (X 2) = (Y2) = a 2 where a 2 is determined by the
int ens ity of the optica l field: (I(0)) = (IE( 0)l2) = (X 2) + (Y 2) = 2a 2. For
t he high er order moment (! 2(0)), we can us e Iss er lis theorem in Eq. (1.28)
of Section 1.4.3 for multiple Gaussian variabl es to ded uc e
(12(0)) = ((X2 + y2)2) = (X4) + (Y4) + 2(X2)(Y2)
= 3a 4 + 3a 4 + 2a 2a 2 = 8a 4 = 2(!(0))2.
Hence, we obtain the HBT bunching effect:
(2) = (I2(0)) -
g (0) - (!(0))2 - 2.
For the dependence on the separation d or delay T of g(2 ) ( T), we will present
a more detailed discussion based on multi-mode description of the optical
field in Section 4.2.4.
The derivation process above shows that HBT bunching effect is simply
a result of Gaussian distribution. We can further show that random variable
I= X 2 + Y 2 has an exponential distribution: p(I) = e- I / Io / I 0 with Io =
(I(0)) (see Problem 1.4). Therefore , we obtain higher order bunching effect:
(n)( ) = (In(o)) - n!Io - '
g O - (I(0))n - Io - n ..
In later chapters on quantum optics, we will give an explanation of the
bunching effect in terms of the photon picture. It will be shown that the
photon bunching effect in Eq. (1.43) is a result of multi-photon quantum
interference (Section 8.3.1).
1. 7 Anti-Bunching Effect of Light and the Need
Quantum Theory of Light
The bunching effect of light is actually the result of the Cauchy-Schwarz
inequality in mathematics when it is applied to statistical optics. For two
arbitrary real random variables X, Y, this inequality states: (XY) 2 ::;
2 2
(X ) (Y ). Taking X = I(0) , Y = 1, we have
(I (0)) 2 ::; (f 2 ( 0)) ,
which gives g( 2)(0) = (I 2(0))/(I(0)) 2 1, i.e., the photon bunching effect .
Therefore , the wave theory of light, after the introduction of random va-
riable descript ion to become statistical optics, must satisfy the bunching
(1.40)
conditions: g( 2 ) (0) > l.
However, if we assume the particle nature for light , as Einste in did
in explaining photo-electric effect, the energy of each photon is already
minimum and cannot be divided further. So when a stream of light in
the form of sing le photons hits a beam splitter, each photon can on ly go
to one of the two sides of the beam sp litt er, as shown in Fig. 1.10. If we
perform a HBT experiment on th is stream of light, eac h photon can on ly
be detected by one of the two detectors while the ot her detector h as none,
2
which leads to zero outcome in photo-current corre lat ion , i.e., (! (0)) = 0
(1.41)
l
or 2 ) ( 0) = 0 < 1. This is the anti-bunching effect and does not satisfy the
bunching condition. So, it cannot b e explain ed by t he wave theory of light.
Th e photon anti -bu nching effect was first observed by Kimble, Dagenais ,
(1.42)
and Mandel in 1977 [Kimble et al. ( 1977)]. To explain this phenomenon,
we must resort to quantum optics where the optical field is quantized.
Fig. 1.10 A stream of light with single photons is separated by a beam splitter: eac h
photon can only go to on e of the two sides. The dash ed lines connect eve nts of the same
(1.43) time (T = 0) and no two photons appear on the two sides at the same time and this
leads to the anti -bun ching effect of light .
1.8 The Topics of Quantum Optics
In th e following chapters, we will study the topics of quantum optics in the
for A form of the following questions:
(1) How to describe an optical field quantum mechanically? We will intro-
duce the quantum state description of light in Chapters 3 and 4.
(2) How to describe the evolution of an optical field? We will discuss the
transformation of quantum state. This is done in Chapter 6.
I 1
(3) What is the relationship between the quantum theory and the wave
theory of light? We will categorize quantum states into classical states and
( 1.44) non-classical states in Chapter 5.
22 Quantum Optics For Experim entalists
(4) How do opt ical fields interact with eac h ot her? We introduce nonlinear
optics in Chapter 6 to generate a variety of quantum states of light.
(5) How to observe the optical field? We will discuss this in the second
part of th is book. We will introduce Glauber's photo-detection theory and
make a connection with the exper im enta l observations . In this part, we will
discuss two basic exper im enta l measurement techniq ues of quantum optics
and find their app lications .
1.9 Problems
Problem 1.1 In Section 1.3 , we presented the first model of light fluctua -
tions . In Fig. 1.4 , t he phases of the optical field jump to another value after
every p er iod of !lT wh ereas the amplitude of th e optical wave is fixed. As-
sume the phase change is completely random . Find the temporal coherence
function 1 (T) and coherence time Tc.
Problem 1.2 Stellar Interferom etry is a technique to find the size of a
star by measuring the spatial coherence property of the light after it travels
from the star to the earth.
(i) Consider two independent point sources of separation a. Draw a line
perpendicular to the line connecting the two points. Now move a distance
D > > a along this line and consider two points with one located on the line
and the other a distance d( < < D) to the line. Assume all the points are
in a plane . Find the normalized second order correlation function 112(d)
between these two points.
(ii) Do the same if the two point sources are replaced by a line source of
length a. Assume the amplitude A(x) for a point on the line source satisfies
(A*(x)A(x')) = A 2 b(x - x') because of the independence of the points on
the line source .
(iii) Do the same if the lin e source is replaced by a two-dimensional rec -
tangular source of size ax b. Assume the amplitude A(x, y) for a point on
the source satisfies (A*(x, y)A(x', y')) = A 2 b(x - x')b(y - y'). Show that
,12(c, d) = sinc(1rac/ D)sinc(1rbd/ D) where c, d are the coordinates in the
observation plane.
(iv) Do the same if the line source is replaced by a uniform disk of radius
R << D. Show that
(1.45)
where J 1 is the first order Bessel function with /3= kdR/ D.
Histori cal D ev elopm ent of Quantum Opti cs and A Bri ef Introdu ction 23
From this problem , we can see that the size R of a star can be found
by measuring 112(d) with interferometric method on earth. This technique
was first proposed [Michelson (1890, 1920)] and implemented [Michelson
and Pease (1921)] by Miche lson.
It is clear from the results that 112 follows the diffraction pattern of the
corresponding opening . This is actually a special case of the far-zone form
of the van Ci ttert-Zernike theorem [van Cittert ( 1934); Zernike ( 1938)] .
Probl em 1.3 Intensity correlat ion measurement can be used to reveal
higher-order interference between independent sources. We will investigate
this in the following.
Cons ider two independent point sources, which are separated by a dis-
tance of d. Let the x-ax is bisect the lin e connect ing the two point sources.
Move a distance D along the x-axis and place two detectors in such a way
that the x-ax is is perpendicular to the line con nect ing the two detectors and
the distance to the x-axis is y 1, y 2 for the two detectors, respectively. Find
(a) intens iti es at t he two detectors an d (b) intensity correlation betw een
the two detectors as a fun ct ion of D , d, Y1,Y2 and wavelength A, assuming
D » d, Y1,Y2-
Probl em 1.4 Assume X, Y are two indep endent Gaussian random varia-
bles, whose averages are zero and standard deviations are CJ'.
(i) Calculate C(r) = (exp[jr(X 2 + Y 2)]) where r is a parameter.
(ii) The characteristic function of random variable I= X 2 + Y 2 is exactly
the function calculated in (i). Find therefrom the probability distribution
p(I) for I.
Pro blem 1.5 Proof of the Central Limit Theorem.
We will provide a non-rigorous proof of the Central Limit Theorem in this
exercise . We start by considering N identical but independent random
variables of xi(i = 1, 2, .. . , N) with an average of (xi) = a and a variance of
((xi - a) 2) = CJ'2 for i = 1, 2, ... , N. Define a new random variable
(1.46)
(i) Show that XN has an average of zero and a variance of var(XN) = CJ'2/ N.
(ii) Use the independent property of {xi} to show that the characteristic
 (J11.nntm11Optics For E :cpcrime11talists

fun ction of X N can be written as

(1.4 7)
Chapter 2
with Ci(r) = (exp[jr(xi - a)/N]) .
(iii) Expand Ci(r) to the first non-zero order of 1/N .
(iv) Show that for large N, CxN (r) exp (-r 2 o-2 / 2N) , which gives the Mode Theory of Optical Fields and
Gaussian distribution with a variance of var(XN) = o-2 / N ind epende nt of
the distribution of xi .
Their Quantization
The result in (iv) shows why N mul t ipl e meas ur eme nts in an exper im ent
will improv e t he precision from single measurement by a factor of 1/.Jiv.

As we found out at the encl of last chapt er , the classical wave theory of
light cannot explain the anti-bunching effect of light. We have to resort to
the photon pi ct ure in quantum th eory of light for explanation. Unlike the
classical wave theory, quantum th eory of light can cover all known optical
phenom ena and is therefore a mor e general theory. But as we will see in
this chapter, the quantum theory of light is constructed with Maxwell 's
classical theory of electromagnetic wave as its foundation.
In a broader perspective , the quantum theory of light is the part about
photon in the general quantum field theory. In this general theory , th e fol-
lowing is how the wave and particl e are unified to give ris e to the duality :
different particles of matter , such as electron, muon, photon, phonon , etc.,
have each a description by a field of its own through its own wave equa-
tion (or the field equation because the spatial spread of the wave provid es
a field for the particle). Different particles have differ ent field equations.
For example, the wave equation for electrons is the Dirac equation and
Schrodinger equation is the wave equation for spinless non-relativistic par -
ticles. "\\Thenwe make a second quantization of the wave equation, we then
create the particle for this field by energy excitation . The reason we call
this process a "second" quantization is because we already call the pro -
cess of going from classical Newtonian equation of motion for a particle to
the wave equation (such as Schrodinger equation and Dirac equation) as a
quantization process, in which energy is quantized , e.g., the energy levels of
hydrogen atom. So, the first quantization process makes a wave (or field)
for a classical particle but the second quantization goes back to particle in
return. However , this cycle does not go back to the old classical particle
but leads to a new duality picture in which wave and particle are unified.
We will discuss this unification further at the end of this chapter after we
introduce the quantum field theory of light .

25
 Mode Theory of Optical Fi elds and Th eir Quanti zation 27
26 Quantum Optics For Exp erimentalists

For light field, its quantization process is a bit more specia l than other 1\Iaxwell equations without charges have the following form:
fields. The classica l theory that we use so often is already a wave theory 18B (2.1)
VxE+--=0.
with Maxwell equations as its wave equ ations . So, quantizat ion of the opti - Cot
cal fields described by l'vfaxwe ll equat ions is equ ivalent to a second quanti- v7 · B = 0, (2.2)
zation process. If we still want to make a comparison with other particles in 1 BE
the quantization process as a whole, we may cons ider the process from New-
v7 x B - -;,,at=0, (2.3)

v7 · E = 0. (2.4)
ton's corp uscu lar theory of light or geometric t heory of light to Huygens'
wave t heory and further to Maxwell's electromagnetic theory as the first Here, we used the cgs unit system, in which electric field and _ma~netic field
quantization process for light . But it seems unclear where the quantization have the same unit and the Maxwell equat ions are symmet ric with respect
occurs in this process since there is no concept of quantization in Maxwell's
to these two quantities .
electromag net ic theory. It turns out, as we will see in this chapter, that From Eq. (2.2),we can introduce a vector potential A: B = v7 x A.
as we solve Maxwell equations with boundary cond it ions, the solutions for Subst ituting into Eq . (2.1), we obta in
the electromag neti c waves only allow discrete values for the frequency of
1 8A (2.5)
light, which are similar to the energy eigenvalues obtained when solvin g t he v7 x (E + --;,,
7Jt) = 0.
Schrodinger equation. This is the mode theory of electromag netic waves,
which is the foundation for the quantum theory of light. The above equ at ion allows us to introduce a sca lar potential ¢: E + ¾ = ft
8

In this chapter, we will start by reviewing systematically Maxwell's -V ¢ . Hence, the electric and magnetic fields can be exp ressed in terms of
electromagnetic wave theory with emphasi s on the mod e theory of light the sca lar and the vector pot enti als ¢, A:
fields. Based on this, we can then qu antiz e light fields with the model of 18A (2.6)
E = --;,,at - v7¢, B = v7 x A.
simple harmonic oscillators. In fact, the mode th eory of light is very us eful
even for just und ersta nding t he wave aspect of light . Subst itutin g Eq. (2.6) into Eqs. (2.3) and (2.4), we obtain

32A - v72A + v7
c2 at2
(v.A + !c 8¢)
Bt
= 0, (2.7)
2.1 The Classical Theory of Light
2 1 BA_ O (2.8)
v7 ¢+--;,,v7·8t-.
2.1.1 Maxwell Equations
But ¢ and A are not unique for E and B. For example, th e following
Light fields consist of electric and magnetic fields. In Maxwell 's electro- transformation
magnetic theory of light , the electric and magnetic fields ar e generated by
charges. But electromagnetic fields and charges are totally differ ent types ¢' = +
</> t A' = A - Vx (2.9)
of matt er. So , we must separate the two in th e quantization processes. will leave E and B unchanged. So, we can freely choose x to obtain differe~t
To do so, let us only consider pure electromagnetic fields without charges , ¢ and A. They give the same E and B. Every differ ent set of</>and A 1s
or the free fields. Charges can be considered as another field of matter called a "gauge" in field theory and E and B are unchanged under gauge
that are totally independent of the electromagnetic fields . The int erac tion transformation. Henc e, electromagnet ic fields have the property of gauge
between electromagnetic fields and matter fields with charges gives rise to invariance or gauge symmetry . They are therefore called gauge fields. Two
electromagnetic interaction and leads to the generation and absorption of
commonly used gauges are
electromagnetic waves. The related phenomena are discussed in the the-
ory for the int eract ion between light and matter and are not the sub j ect (1) Lorentz Gauge:
of the current chapter (see Chapter 6). So, here we only deal with free 18¢ (2.10)
electromagnet ic fields without charges. V. A +--;,,Bt = 0,
28 Q'Uant'Um Opti cs Fm· Exp erimenLa lists l\Jodc Theory of Optico .l Fi elds and Th e·i,· Quo :11/iz ai'ion 29

(2) Coulomb Gauge: as the cige11-cquatio11 for operator '\72 . By operator theory, for a certain
bonnclary condition (e.g. A(r) = 0, r = boundary), we can always find a
'v · A =0. (2.11)
set of values for k and a group of orthonormal eigenfunctions {A>-(r)}, such
Obviously, Lorentz ga uge sat isfies the Lorentz invariance of relativity and
that they satisfy :
is suit ab le for the electromagnetic fields of high energy, such as ,1-rays.
'\72 A >-(r) + kiA>-(r) = 0. (2 .18)
Coulomb gauge is more suit ab le for the electromag netic fields of low energy,
such as the commonl y used visible band of light , for which the photon
'v · A >-(r ) = 0. (2.19)
energy is around a few eVs. Th e speed of electro ns at this energy level
is non-relativistic. Th erefore, we will in general us e the Coulomb gauge
J d3 rA ~(r ) • Aµ (r ) = 6>-µ- (2 .20)

in quantum optics. Furthermor e, we lear ned from electromagnetic theory The solut ion for the temporal part (Eq. (2 .17)) is that for a simpl e harmonic
oscillator . It has two ind ependent solut ions: qi±) (t) = q>-(0)e±iw>-t.Here,
that the scalar potential ¢ is generated from charges, which are absent for
a free field . So, we can set ¢ = 0 for free fields. In this case, Lor entz gauge we use comp lex number to represent the harmoni c motion with its real
and Coulomb gauge are the same and Eqs. (2.6) - (2.8) change to part corr espo nding to the form of motion for the harmonic oscillator, i.e.,
cosine-fun ctio n . In the two ind epend ent solutions , qi-) (t) = q>-(0)e-iw>-t
'v2A-
c2 fJt2
32A = 0, (2 .12) corresponds to the plane wave solution that will be discussed later while
qi+)(t) = q>-(0)eiw>-tis its complex conjug ate, which will be included when
'v ·A= 0, (2 .13)
converting back to the representation in real numbers. In the following we
E = - , B = 'v x A. ( 2 .14) only use qi-\ t) = q>-(0)e-iw>-t as the solution to the tempora l p art.
Combining the spatial and temporal parts , we arrive at the special
Equation (2.12) is in the form of wave equation that must be satisfied for any
eigen-solutions for Maxw ell equations: A>-(r, t) = q>-(0)e-iw>-tA>-(r). Since
classical waves. From Eq. (2.12), we conclude that electromagnet ic fields
Eqs. (2 .12) and (2.13) are linear and homog eneous , any superposition of
can propagate in the form of waves and th e speed of the electromagnet ic
their solutions should also be a solution. From op era tor theory , we can
wave is the speed of light c. This is what led Maxwell to postulate the
prove further that the eigenfunctions of the operator '\72 are complete, i.e.,
existence of electromagnet ic wave , which was later confirmed by Hertz , and
any arbitrary solution of Eqs. (2.12) and (2.13) can be written as a lin ear
light wave as a form of electromagnetic wave. We now call Eq. (2.12) as
superposition of the eigenfunctions :
Maxwell wave equation or simply Maxwell equation for short.
A(r , t) = L q>-(t)A>-(r) = L
q>-(0)e-iw>-tA>-(r) , (2.21)
,\ ,\
2.1.2 The Eigen-Solutions of Maxwell Equation or, if we use real numbers to represent optical fields , we have
With wave equation , let us find the exact expressio n for the light fields by A(r , t) = L [q>-(t)A>-(r) + qt(t)A~(r)]
solving it. We will use the method of separation of variables in solving ,\

partial differential equat ions: Eq. (2.12) has separate spatial and temporal (2.22)
differentiations. So, let us write separately the space and temporal part of ,\

A(r , t) as A(r , t) = q(t)A(r). Equations (2.12) and (2.13) become where q>-(0)is some constant. Since A>-(r) is normalized, the size of q>-(0)
2
'\7 A(r) +k 2
A(r) = 0, (2.15) will then determine the strength of the optical fields.

'v · A(r) = 0, (2.16)

2.1.3 Special Eigen-Solutions: Box Model and
d2q(t) 2
d(2 +w q(t) = 0. (2.17) Plane Waves

Here k and w are some constants to be determined and they satisfy k = w / c. As we hav e discussed in the previous section, the solution of the Helmholtz
Equation (2.15) is also known as Helmholtz equation. It can be considered equation is determined by the boundary conditions. Let us start with
30 Quantum Opti cs For R1:perimenta lists Mode Theory of Optical Fields and Their Quantization 31

the simp lest border , i.e., a cubic box. We use the continuity boundary Notice that the discr ete ness of the wave vector k leads to the quanti zed
conditions in which the field function A(x, y, z) is continuous from one side angular frequency w = ck = (2?Tc/L) Jn~, + n~ + n ; . This is analogous to
to the opposite side of the box, that is, th e energy quantization from Schrodinger equation in the first quanti za-
A (O,y , z) = A(L, y. z), A( x, 0, z) = A(x , L , z),
tion process. So. Maxwell wave equation is equivalent to Schrodinger wave
equation for the first quantization of optical fields , as we discussed in the
A(x, y, 0) = A(x , y, L). (2.23)
beginning of this chapter.
Her e, L is the side length of the cubi c box. We can easi ly obtain the plane
wave solution s for the Helmholtz equat ion under the continuity boundary 2.1.4 Special Eigen-Solutions: Continuous k-Space and
cond ition in Eq. (2.23) Plane Wave Solutions for Three-Dimensional Free
A (d) (r) = E eik-r/L3/2 (2.24) Space
k ,s s '

where Es is the unit vector and the eigen wave vectors are k = When L goes to infinity, the finite box space becomes infinite three-
2?T(nx, ny, n z)/ L. nx, ny, n z are int egers. So, the eigen wave vectors k dimensional free space and the original discrete k-space is transformed into
cover the whole k-space but they hav e discrete values and the minimum a continuous k- space. For this transition, let us re-define the spatial part
separat ion between two values is 6-.k = 21T/ L. Since the eigen wave vector of the field function in Eq. (2.24):
k is discrete, we place a superscript "d" in field function to distinguish AS~~ Ak,s (r) = (L/21T)312AL~~(r) = Ek,seik-r / (27T) 312, (2.26)
from the continuous case to be discuss ed in the next section .
Unit vector Esdetermines the direction of the vector function A(r). For and in the meantime, the orthonormal relation is changed to
a 3-dim vector, it has three independent directions, that is , Es corresponds
to three independent polarization states with subscript s = 1, 2, 3. But the
r d 3rA~,s(r).
jL3
3 3
Ak, ,s (r) = c5s,s'c5k,k (L/21T) = c5s,s'c5k,k'/(6.k) . (2.27)
1 1

ASd~
field funct ion must sat isfy the Coulomb gauge condition in Eq. (2.19). Here, 6-.k = 21T/ L is the minimum separation between two adjacent discrete
AS~~
0

Substituting the field function in Eq. (2.24) into Eq. (2.19) , we find wave vectors k = 2?T(nx,ny, nz)/ L. In the limit of L • oo and 6-.k • 0,
that Es must satisfy the transverse wave condition: k · Es = 0, that is, the we obtain the following transition:
polarization of the optical field must be perpendicular to the wave vector
k. So, Escan only take two indep endent and orthogonal polarization states 1= L 6.k c5k,k'/ (6.k )
3 3
• J 3
d3 kc5() (k - k') =1 (2.28)
such as linearly polarized x, f) or circularly polarized E± = (x ± if))/~ - k

Because of the transverse wave condition, the polarization of an optical
field is related to the wave vector: E = Ek,s(s = 1, 2) . Let k = k/ k be
the unit vector for wave vector k. Since k,Ek,l, Ek,2 are three unit vectors
perpendicular to each other, they form the base vectors of the 3-dim vector
space .
It can be easily proved that the set of eigen solutions in Eq. (2.24)
satisfies the discrete orthonorma l relations:
Hence, the discrete c5-function is changed to a continuous c5-function:
(2.29)
We will use again later the above transition from discrete to continuous
cases.
We then obtain Eq. (2.26) as the plane wave solution for three -
dimensional free space with orthonormal relation:

r d rAS~~*(r)· AS~\,(r) =
jL3
3
c5k,k'c5s,s' · (2 .25) J d3 rA~,8 (r) · Ak 1,
8
1 (r) = 6s,s'c5(3 \k- k'), (2.30)

Combining with the temporal part, we obtain the plane wave solution where the wave vector k = (kx, ky, kz) is an arbitrary vector in a three-
for Maxwell equation in Eq. (2.12): dimensional continuous k-sp ace and the transverse wave condition is the
A(r , t) = Ek,sei(k-r-wt) / L3/2 . same: k · Ek,s = 0.
 Mode Theory of Opt ·ical Fields and Their Q'Llantization 33
32 Q'Llant'Llm Opt'ics For Experimentalists

2.1.5 The Concept of Modes and the Decomposition in To prove the above, we start from the electromagnetic theory and write the
Terms of Modes expression for the total energy of the electromagnetic field as

Because of the orthonormal relation in Eq. (2. 20), the eigen-sol u tion of Utot = -l
81T
1· d3 r[E(r. t) · E(r , t) + B(r, t) · B(r, t)], (2.34)
l\Ia.xwell equation has its special meaning. First, the general solution of where E , B are electric and magnetic fields and must have real values.
l\Ia.xwell equation can be written as the superposition of the eigen solutions, So, they must be derived from Eq. (2.14) via the real expression of A
that is, in the form of Eq. (2.21) or (2.22). When we obtain a complete set in Eq. (2.22). Let us first calculate the first term in Eq. (2.34), that is,
of eigen solutions, we will have the arbitrary solution. Secondly, although the energy of electric field. From Eq. (2.14) and orthonormal relation in
an arbitrary solution is a superposition of a number of eigen-solutions,
Eq. (2.20), we obtain
the orthonormal relation of the eigen solutions ensures the independence
among all the eigen solutions and their addition will not lead to the energy Ut(!) =
8
~ J d3 rE(r, t) · E(r, t)
exchange among them, as we will see in the following decomposition of
energy in terms of the energy of eigen-solutions. 1 Let us start with the
total intensity of the field. In the second part of this book on experiment,
= ~Jd
81TC
3
rL
.\,N
[<J>-(t)A>-(r)+c.c.]· [<JN(t)AN(r)+c.c.]

we will discuss about intensity of an optical field whose expression in terms

of field amplitude is J(r, t) ex E *(r, t) · E(r , t), where E can be obtained
= L [<J>-(t)q;,(t)+ q;(t)qN(t)]5>-,>-'
81TC .\,N
from Eq. (2.14) with A given in Eq. (2.21). Then the total intensity is

ltot = j d rl(r, t) j
3
ex 3
d rE*(r, t) · E(r , t). (2.31)
1
+- -
81Tc26
j
[<J>-(t)qN(t) d3 rA>-(r) · AN(r) + c.c.]
.\ ,A'

Using Eqs. (2.14) and (2.21) together with the orthonormal relation in = s12L [<J>-(t)q;(t)+ q;(t)q>-(t)J
Eq. (2.20), we obtain ?TC .\

1
j d r[ ~<i'.::(t)A),(r)] · [ pv(t)Av(r)]
hot ex : 2
3 +-81Tc2
-
6
.\,A'
(t)h>-1 + c.c.].
[<J>-(t)q>-1 (2.35)

Here, h ,>-'= J d3 rA>-(r) • AN(r). For the convenience of introducing ope -

= c\Lq;(t)qN(t) j d 3
rA~(r) · AN(r)
t) , t) in
rator in the future, we kept the original order of product for Q>-( q;(
.\,N
the express ion above. The evaluat ion of the energy of magnetic field needs
= :2Lq;(t)qN(t)5>-y = c12Ll<J>-(t)l2 the following equation of vector differentiation:
.\,N .\
(V X A)· (V X A')= L [(VAm) · (VA~)
- (VAm) · (8mA')]
=Lh- .\
(2.32) m

Here, I>-= l<J>-(t)l2/c 2. m

Similarly, we will prove in the following that the total energy Utot of the
optical field can also be decomposed into the sum of the energy of individual m
eigen solutions: Here {A, A'}= {A>-, An.After using Eqs. (2.18) and (2.19) in the equa -
(2.33) tion above, we obtain

1
(V x A) · (V x A')
One may argue that int erference effect due to superposition is energy re-distribution .
But that is a result of int eraction between optical fields and detectors and no energy =V· (LAmVA~)- V · (LAm8mA')+ k A-A'. 2
(2.37)
re-distribution occurs for superposit ion of free fields. m m
 34 Quantum Optics For Exp erim entalists Mod e Th eory of Opt'ical Fields and Th eir Quanti zat ion 35

Using the Gauss divergence theorem: Now. we just proved the decomposition of total energy in Eq . (2.33), where
3
/ d r(V · F ) = f dS(n · F ) (2.38) U>,.=
1
47rC 2
w;(q>..(t)q;(t) + q;(t)q,>-(t)] =
1
27rC 2
w;[q>..(0)[2 . (2.43)

and boundary cond iti on A (r ) = 0, we change the spatial integral of We see from Eqs. (2.22), (2.42), and (2.43) that the optical field can be
Eq. (2.37) to 2 decomposed into the sum of the eigen solution s and is determined by them

J 3
d r(v' x A )· (v' x A' )= k 2 / d3 rA • A'. (2.39)
independently. This is very sim ilar to the normal mode description in a me-
chanica l oscillator system . The eigen solution s of the optica l field is sim ilar
to the normal modes of the oscillator system. We likewise define each eigen
Now we can calc ulat e the magnetic part in Eq . (2.34): solution as a mode of the optical field which can be described comp lete ly
(B) 1 / 3
by all the modes. Compared to quantum mechanics, the modes of the op-
Utot = S1r d rB(r , t) · B (r. t) tical field is analogou s to the eigen energy states for Schrodinger equation

= 8~ J d r
3
L [q>-(t)V x A,>-(r) + c.c.]· [qv(t) V x Av (r) + c.c.]
>-,V
in mathematical formalism. So, solving the modes for Maxwell equat ion is
equival ent to the first quantization pro cess involving Schrodinger equation.
We can also find from Eq. (2.43) that the energy in each mode is only
= 8~ L [q>-(t)q;,(t) Jd 3
r(V x A,>-(r))· (V x A;,(r)) + c.c.] determined by the initial condition q,>-(0)but is completely unr elated to the
mode function A,>-(r). So, quantity q,>-(0)is also known as the excitation
>-,V
of mode A. Since q,>-(0)is independ ent from other differ ent modes, each
+ 8~ L [q>-(t)qv(t) Jd
>-,V
3
r(V x A,>-(r))· (V x Av(r)) + c.c.] mode can b e excited separate ly in an indep endent way. If th ere is only
one mode being excited, we hav e a single-mode field. If more than one
1
= 81r L k; [q>-(t)q;(t) + q;(t)q,>-(t)] mode is excited, the optical field is of multi-mod e. We will see later in
>- the coherenc e th eory of light , the number of excit ed modes will affect the
coherence property of the optical field.
(2.40)
On the other hand , the mode function A>-(r) is ind epe nd ent of the mod e
excitation q,>-(0).As can be seen from the process that the mode function
Here, we used the orthonormal relation in Eq. (2.20). Combining Eqs. (2.35) A>-(r) is derived, it is only determined by the boundary conditions which
and (2.40) and using k,>-= W.>-/ c, we obtain th e total energy of the optical are usually related to the geometry of the border. So, it is unrelated to
field (electromagnetic field) as
whether the mode is excited or not and it exists even for vacuum where

Uto, = S;cI:>- {[,i,(tM:(t) + <i\(t)<i>.(t)]+ wI[q,(t)ql(t) + q;(t)q,(t)]}

2
there is no mode excitation at all (q.>-(0)= 0) . However , when the boundary
conditions or the geometry of the border change, it may result in the change
in the mode function as well as the eigen-values k.>-, which will lead to field
+ s:c2 L {[<i>-(t)qv (t) + w,>-wvq.>-(t)qv(t)] h ,v + c.c.}- (2.41) energy change even for vacuum if it has non-zero energy as in quantum
>-,>-' theory of light. This is the famous Casimir effect (see Section 2.4).
We next use the solution for the temporal part in Eq. (2.17): q,>-(t)= q,>-(0) In general , mode function A,>-(r)of the optical field can be written as
= -iw,>-q,>-(t).Substituting into Eq. (2.41), we have
e- iw >-.t to obtain Q.>-(t)
(2.44)
1
Utot = 47rc 2 Lw;(q,>-(t)q;(t) +q;(t)q,>-(t)] = LU>-. (2.42) It can be determined by the following quantities of independent degrees of
>- >- freedom:
~------------
2
It can be shown directly that the plane wav e so lution s in Sections 2.1.3 and 2.1.4 also (1) Polarizations of light wave, such as Es = x,y. Different orthogonal
sat isfy Eq. (2.39). polarization states correspond to different polarization modes.
36 Qiwntum Opt ics For Experimenta lists J\Jodc Theory of Optical Pields and Their Quantizati.on 37

(2) Frequencieo of light wave W>, = cJkJ. Different color s of light corr espon d For this . we make a translation along the ;:;-ax is: ;:;' = z - zo. It is easy to
to different frequency modes . see that Eq. (2.49) is un changed . There is a simple ph ysical picture for thi s
(3) Dir ect ion s of light wave k = k / JkJ or spatial fun ct ion s u k.s(r ). Different tran sformation: th e center of the sp herical wave is moved to ;:;= zo by the
directions of propagation or sp at ial fun ctio ns give rise to different spatia l transformation. Now, let us mak e a trans lation of an im aginary number:
modes. ::,
11
= z - ib and the Fresnel form of th e spher ica l wave io changed to
In t he next sect ion, let us intro du ce some common ly used modes of 2
optical fields. A ( x + y2 ) (2.50)
uc( r ) = q(;) exp ik 2q(z) .

2.2 More Modes of Optical Fields Her e we only write down t he part of u-function with q(z) = z - ib. Since
ib is a constant, a simple derivation shows that this transformation doe s
2.2.1 Paraxial Rays and Gaussian Beams - the Modes of not cha ng e the form of Eq . (2.49), eith er. So, t he transformed u-function
Optical Resonators in (2.50) is st ill a solution of Eq. (2.49). A simple substitution can verify
t his.
When kx, ky « k 2 , we can make the paraxial approximation. As an exam -
In order to find the physical meaning of Eq. (2.50), we chang e its form
ple , let us first look at a sph erical wave:
to :
A ( ) _ €Ao ikr
(2.45) 1 z . b 1 . ,,\
sph r - -r- e ' -- = --- +i--- 2 = - - +i---, 2 (2.51)
q(z ) 2
z + b 2 z + b2 R( z ) nw (z )
where r = J x 2 + y 2 + z 2. In paraxial approximation, the field does not where
leave far from the z-axis: (x 2 + y 2)/ z 2 = 02 « 1. So, we can make the 2
b
following approximation:
R(z)=z[i+(;) ], w(z)=wo[ l+ ( bz )2]1 ;2 , (2.52)
2 2 2
0
r = z 1+ x ; Y = z y'l+02 "" z ( 1 + ;) = z + x y' (2.46) with,,\ as the wavelength and w5= >.b/n.Defining (( z) = tan- 1(z/b), 3
we
hav e q(z) = -iJq(z)Jei((z). Equation (2.50) then becomes
Then, we obtain the paraxial approximation
Fresnel form:
of the spherical wave in the
' wo [ x2 + y2 l [.
x 2 + y2 .
uc(r) = A 0 w(z) exp - w 2(z) exp ik 2R( z ) - i((z)
l . (2.53)
€Ao eikz exp ( ik---
A sph(r):::::::-
. x2 + y2 ) .
(2.47) Compared with (2.47), the wave front of uc(r) is the same as a spherical
z 2z
For arbitrary waves, we can write the vector wave function A(r) as wave with radius R( z), but the transverse distribution of its amplitude or
intensity is a Gaussian distribution:
A( x,y,z) = fo(x,y ,z )eikz,
where u(x, y, z) satisfies the conditions for paraxial approximation:
(2.48)
2 2P [ 2(x2 + y2)
Ic(r) = JA(r)J = nw2(z) exp - w2(z ) '
l (2.54)
J8u/ 8z l « ku , J82u/ 8z 2 I « k 2 u. Then Helmholtz equation in Eq. (2.15)
where P is the total power of the beam P = J Ic(r)d xdy . This beam
can be approx imated as
is called "Gauss ian beam" because of the Gaussian distribution for the
82 u 82 u au transverse intensity, whose cross-section is a circle of radius around w( z ).
ax~') + 8y 2
+ 2ik- z = o.
8
(2.49)
Figure 2.l(a) shows w( z ) as a function of z, where it reaches minimum
It can be easily shown that the Fresnel form of the spherical wave in value of w 0 at z = 0 and the divergent angle is 0o = tan - 1(wo/b). wo
Eq. (2.47) satisfies the paraxial Helmholtz equation in Eq. (2.49). is also known as the "waist " of the Gaussian beam . When 0o is sma ll,
It is not so easy to find a general solution for Eq . (2.49) but we can obtain 00 rv w 0 /b = ,,\/nw 0 , which correspo nds to the diffraction angle by a circular
a special solution, that is , the Gaussian beam, by a simple transformation . 3 Phase (( z) cha ng es slowl y with z from -1r / 2 to 1r/ 2 and is known as Guoy phase .
 Mode Theory of Optical Fields and Their Quanti za tion 39
38 Quantum Optics For Experimentalists

\\'(,::)
The Gaussian modes discussed above can be produced from optical re-
sonator formed by spherical reflection mirrors. The Gaussian modes so
211·0 bea111
\\'Gist produced are determined by the geometry of the optical resonator (see
(a) \\'o/
Fig. 2.3). The boundary conditions require the wave fronts of th e Gaussian
,- ,· ,- . 00
beam coincide with the shape of the two reflectors , that is.
-b 0 b
b2
R2 = z2 + - , L = z2 - z 1, (2 .57)
z2
where L is the length of the resonator, z 1, z2 are respectively the locations
of the two reflectors relative to the waist of the Gaussian beam. Then we
have
Fig. 2.1 (a) The radius w(z) of th e Gaussian beam cross-section vs. the propagation L(L + R1) (2.58)
distance z . (b) Th e wave front of th e Gaussian b ea m vs. the propagation distance z .
z 2 = 2L - R2 + R1 '

opening of radius wo. So, we can say that Gaussian beams are diffraction-
2
L(R2 - L)(R1 + L)(L + R1 - R2) (2.59)
limited. As can be seen, the parameter b uniquely determines the waist b = (2L + R1 - R2) 2 .
and the divergent angle of the Gaussian beam and thus defines the whole
Gaussian beam. Figure 2 .1(b) shows the change of the wave front of the So, the waist of the Gaussian beam wo = and locations z1, z2 are
Gaussian beam as propagation distance z changes. In the interval (-b, b), uniquely determined by the radii R 1, R2 of the two reflectors and their
the wave front is nearly plane wave-like. This interval is also known as the separat ion L. From Eq. (2.57), the signs of R1, R2 depend on z1, z2.
Rayleigh region.
.-------L------ 1
Equation (2.53) is a special and most simple Gaussian beam solution to
the paraxial Helmholtz equation in (2.49). The paraxial Helmholtz equation /

also has higher order Gaussian beam solutions in a general form [Siegman Ml'Ri ;
(1986) l /

Fzm(x,y,z) 2 2 Z=O
Uzm(x, y, z ) = q(z) exp[ik(x +y )/q(z)], (2.55)
Fig. 2.3 The Gaussian modes produced by the optical resonator consisted of two sp-
where herical reflectors.

Fzm(x, Y, z) = uoHz(x/w(z))Hm(y/w( z))e-i(l+m)((z). (2.56) The eigenvalues kzm are determined by the resonance condition:
Here Hz(x) is the Zth order Hermite function. All the Gaussian beams with (2.60)
different l , m form the Gaussian modes of the optical field. Figure 2.2 shows
the images of the spatial distribution of a number of Gaussian modes. Then the resonant frequencies of the optical resonator are
Vzm= ckzm/211:= N !.iv+ (l + m + l)bv, (2.61)
HG 1
where !.iv = c/2L, bv = !.iv[(( z 2) - ((z 1)]/1r = l.ivcos- (Jgl9i)/1r with
(/,m) (0,0) (1,0) (2,0) (I, I) (2, I) (2,2)
=
9 1 1 + L/ R 1, 92 =
1 - L/ R 2 . The first term in Eq. (2.61) corresponds to
the fundamental modes given by Eq. (2 .53) and is known as the longitudinal
Fig. 2.2 Spatial distributions for a number of different Hermite-Gaussian modes. mode while the second term is for the transverse modes corresponding to
 Quantum Optics For E:rperimcntalists
higher-order Gaussian modes in Eq. (2.55). Cavity stability requires a real
or O :::;9192 :::; 1.
To analyze the mode structure of a resonator experimentally, we direct
a beam of single-frequency laser onto the resonator. When the laser fre-
quency coincides with one of the resonant frequencies of the resonator , the
correspond ing Gaussian mode is excited and built up. Some of the excited
field can also leak out of the resonator and be detected by a photo-detector.
According to Eq. (2.61) , Fig. 2.4 shows the leaked field int ens ity as the la-
ser frequency is scanned : all the resonant modes can be scanned out. The
height for each mode depends on the mode match of the in cident laser
beam to each transverse mode of the cavity. On the other hand, if there is
a gain medium in the resonator and for some modes, if the gain is higher
than the losses, laser oscillation can start and a laser beam is produced.
Since only Gaussian modes can be excited, the output modes of the laser
beam are simply the Gaussian modes of the resonator. Which mode will
lase is determined by the gains and losses of the modes. In general, several
modes can be excited and have laser output. This leads to a multi-mode
(multi-frequency) laser. To obtain a single -mode (single-frequency) laser,
we need a Fabre-Perot etalon to select mode (frequency).
---.1v--
l+m= o 1 2 3 4 0 I 2 3 4
0 I 2 3 4
V
N N+J N+2
Fig. 2.4 The fr eq uencies of the modes of the resonator are obtained as the laser fre-
quency is scanned.
For more discussions about optical resonator cavit ies and Gaussian mo-
des, the book by Siegman is very comprehensive and covers many of the
topics [Siegman (1986)]. Besides Hermite-Gaussian modes discussed here,
Laguerre-Gaussian modes have spatial profiles which are circularly symme-
tric and are used to describe orbita l angular momentum (OAM) of optical
fields [Andrews and Babiker (2013)]. OAM of an optical field describes
structured and twisted light such as optical vortices which have a high
degree of freedom for information carry ing.
In addition to the Gaussian modes, other frequently used spatial modes
are those from optical fibers, which are widely used for optical communi -
Mode Theory of Optical Fields and Their Quantization 41
cations. Their mode structures are similar to the Gaussian modes so they
are commonly used in the lab to clean up spatial modes of optical fields.
For details of optical fiber modes , refer to [Buck (1995)].
2.2.2 Temporal Modes and Generalized Definition of
Single-Mode Fields
What we discussed in the previous section is the spatial mode. It gives rise
to the spatial distribution of the optical field in the transverse direction
that is perpendicular to the direction of propagation. In this section, we
will discuss the longitudinal distribution of the optical field along the di-
rection of propagation. For one-dimensional field (see Section 2.3.5), the
longitudinal direction is equivalent to the temporal degree of freedom. In
the mode decomposition of the optical field discussed earlier, the tempo-
ral part is e-iwt . So, a single frequency field corresponds to a frequency
mode and spectral analysis is equivalent to decomposition into longitudi-
nal modes . On the other hand, in the application to ultra-short pulses,
it is more convenient to work in temporal domain. Consider a pulse of a
one-dimensional field:
E(z, t) = J dw[(w)ei(k z -wt). (2.62)
If [ (w) is a well-behaved function such as a Gaussian function, the spectral
bandwidth 6.w of [(w) and the pulse width 6.t of E(z, t) satisfy the relation
for Fourier transformation:
6.w6.t = 2n. (2.63)
Such a pulse is known as the transform-limited pulse. It can be genera-
ted from a mode-locked laser where each frequency mode has the same
phase. So, such a pulse has really good coherence and its waveform is a
well-defined function of time. From function theory in mathematics, we
learn that for a well-defined temporal function J(t), we can always con-
struct a set of functions 9 1 ( t), 9 2 ( t), ... so that the whole function set of
{J (t), 9 1 ( t), 92 ( t), ... } forms a complete and orthonormal base. This set has
the same properties as the mode function set {u,A.(r)}discussed earl ier. The
function set {J (t), 9 1 ( t), 92 ( t), ... } then forms a set of temporal modes and
function f (t) is one of the modes. Hence, a transform -limi ted pulse is a
sing le temporal mode.
Notice that the single mode defined by Eq. (2.62) is a superposit ion
of many frequency modes. When [(w) is a well-defined function, E(z, t)
 Mode Theory of Optical Fields and Thefr Quanti z ation 43
42 Quantum Optics For Experimental'ists

energy expression in Eq. (2.43). It is the starting point for the quantization
in Eq. (2.62) is a solution of Maxwell equation. From this property, we
may define a generalized single-mode field as any one of the solutions of of the optical field. The coefficient 1/ 4nc 2 in the expression is a fixed
constant . In order to be consistent with the formulae after quantization,
Maxwell equation with a well-defined function of (r, t). It can be a linear
superposition of many modes: we combine it with q>,(t). So, Eq. (2.43) changes to
A(r , t) = L C>,A>,(r)e-iw»\ (2.64) U>,= wl [q>.(t)q:(t) + q:(t)q>,(t)]. (2.65)
>, The field function in Eq. (2.22) becomes
where {C>,}is a set of definite constants. The generalized field mode so
defined will be applied to the quantization of optical fields in the next A(r , t) = ~L [q>,(t)A>.(r) + q:(t)A~(r)]
>,
section and as we will see later (Section 8.4), this becomes important when
we discuss the photon indistinguishability: if a group of photons can be = ~L [q>,(O)A>.(r)e-iw»t + q:(o)A~(r)eiw»t]. (2.66)
described by the same well-defined solution of Maxwell equation, they are >,
labeled as in the same mode and are completely indistinguishable. Photons Because q>,(t) is a complex number, the expression in Eq . (2.65) for single
having this property will exhibit the largest effect of quantum interference. mode energy is not in a form that looks familiar to us. Let us use the real
We can also understand the concept of generalized mode from the per- and imaginary parts of q>,(t) as new variables:
spective of linear algebra. A comp lete set of modes is like a complete base Q.>-= q>,(t) + q:(t)
vector in a vector space. One mode corresponds to a base vector . A gene - P>,= W>,[q>,(t)- q:(t)]/i, (2.67)
ralized mode can be considered of as a vector that is a linear superposition
where we add a factor of W>,for the imaginary part P>, so that it has the
of the base vectors. This vector can form another complete base set with
meaning of a generalized momentum for the generalized coordinate Q >..
other vectors orthogonal to it so that it is one of the base vectors in the
Equation (2.67) can be rewritten as
newly formed complete bases. In th is ana logy, the generalized mode is also
a base mode and belongs to a single mode in a new set of complete base q>,(t) = (Q>.+ iP>,/W>.)/2. (2.68)
modes. Take an examp le of polarizations: x,i) form a base set for the po- Substituting Eq. (2.68) into Eq. (2.65), we have
larization vectors of light field in which i; describes a single polarization
U>,=2 1(P.>-+w>.Q>-.
2 2 2) (2.69)
mode. But polarization vectors in 45 degree and 135 degree can also form a
comp lete set of polarization base vectors. So, even though polarization vec- = -iw>,q>,(t), we obtain from Eq. (2.67)
Since CJ.>.(t)
tor in 45 degree is a superposition of x,i), it actually is a single polarization P>.=Q>.- (2. 70)
mode. Other example such as circularly polarized light E± = (x ± ii))/../2 The energy expression in Eq. (2.69) is exactly the same as that for
is also a single polarization mode. In fact, any ellipti ca lly polarized light a simp le harmonic oscillator of mass m = 1. The momentum P of the
E = x cos 0 + i)eicpsin 0 is also in a single polarization mode. oscillator is given by Eq. (2. 70). Indeed, if we treat Eq. (2.69) as the
The concept of single mode will become crucial in und erstand ing and Hamiltonian of the oscillator, i.e.,
describing multi-photon indistinguishability. We will come back to this
1(P>.+w>.Q>-,
H>,= 2 2 2) ( 2.71 )
lat er in Chapter 8. 2
with Q >-as the generalized coordinate and P>. as the generalized momen-
2.3 Quantization of Optical Fields tum, Hamiltonian mechanics gives rise to an equat ion of motion same as
Eq. (2.17), i.e.,
2.3.1 Description of Modes by Simple Harmonic
Oscillators

When the modes of an optical field are fixed, the time evolution of the field (2.72)
is determined by the mode excitation q>,(t). Let us now go back to the
 Mode Theory of Optical F-ields and Th eir Qua'lltization 45
44 Quantum Optics For Experimentalists

c Hamiltonian of the optical field with all modes included takes the
The last line of the above equations is the same as Eq. (2.70). anCltl 1
It is well-known that the time evolution of a system is determined by forlll of
A A
H = LH>-. = L
~lt; (At A + a>-.a>-.
,iw>-. a>-.a>-.
A At) , (276)
.
the Hami ltonian of the system. So, each mode of the optical field can 2
be described by its corresponding simple harmonic oscillator with mass >-. >-.
m = 1 and its oscillating frequency as that of the mode. In fact, many where fI >-.
's are ind ependent of each other for different ,\ and the commu -
forms of wave motion in nature can be described by harmonic oscillation of tation relation of operators CL>-.,al can be obtained from Eq. (2.73):
some particles such as water waves, the stand ing wave on a violin string. al,l = 5>-.Y·
[a>-., (2.77)
However , different from these waves that are formed by real particles , the The 5-function in the expression above is the result of the independence
harmonic oscillators for the description of the modes of opt ical fields are between different modes. With the commutation relation above, the Ha-
not real obj ects but virt ual oscillators. So are the ir positions and momenta. miltonian in Eq. (2.75) is now in a form familiar to us:
The classical description of t he virtua l harmonic oscillator is given by ,>,.+ nw>-./2.
fI>-.= nw>,.ala (2.78)
Eq. (2.72) . Next, we will present its quantum description. Since the optica l
2':0 for
Since operator ala>-.is a pos itive definite, meaning that ('lj;\ala>-.\'1/J)
field is decomposed int o harmonic oscillators of each modes, the quantum
any state \'lj;),t he energy E>-.= ('lj)\H>-.\'1/J)
has a minimum value of nw>-./2.
descr iption of the harmo nic oscillators will lead to the quantum description
Later in Sect ion 3.1, we will see that nw>,./2is the energy of vacuum for
of the whole field.
mode,\ _
In the Heisenberg picture, the equat ion of motion for a>,.is
2.3.2 Quantization of Simple Harmonic Oscillators
da>-._ ~[ A fI l (2. 79)
dt - in a>,., A '
The quantization of harmoni c oscillators is a typical example that can b e
found in almost every t ext book of quantum mechani cs . Here , we will only or
da>,. .
--= -'lW>,.a>,.. (2.80)
A

present the results. In the quantum description of a harmonic oscillator, dt

the generalized coordinate Q and mom entum P be come operator Q, P with It has a soluti on of a>-.(t)= a>-.(0)e-iw>J_
a commutation relation
[Q,P] = in. 2.3.3 Field Operators for an Optical Field

Sinc e differ ent mod es of the optical field are ind epend ent of each other, An nihilation and creat ion operators are the basic operators for the descr ip-
the operators corresponding to differ ent modes commute with each other. tion of the optical field. Any other physi cal quantity can be expressed in
Hence, the commutation relations for all th e operators of the modes are terms of them. From Eq. (2.66), we have the operator form for the vector
potential:

Introduce the annihi lation and creation operators :

(2.73)
A(r , t) = L
>-.
r:::
v~ e-iw,>JA>,.(r)+ aleiw>,tA~ (r)]. (2.81)
[a>-.

A - (QA
a>,.=
+ .A)V2n'
>-. iW>-.
;w; At - a>-.=
(QA .A)V2n'
;w;
>-.-'lw>-. (2.74)
Here, a>,.= a>,.(0)is the op erator in Schrodinger pictur e. From Eq. (2.14),
we obtain the operator forms for the electri c and magn et ic fields:
These a-quantities are proportional to the q-quantities in Eq. (2.68): a =
qJ2w>-./n. We will later use this relation to estab lish the correspondence
E(r, t) = ;ff
v'41rL
>-.
a,e-iw;l A,(r) + h.c.
between quantum and classical descriptions. The Hamiltonian in Eq. (2. 71)
changes to B(r , t) = ~L
>-.
r:::
v~a>,.e-iw>,tvx A>,.(r)+ h. c. (2.82)

(2. 75) Here, h. c. sta nd s for Hermitian conjug ate .

 46 Quantum Optics For E xpe1·im entalists Mod e Th eory of Optical Fi elds and Th eir Quanti zation 47

2.3.3.1 Field Operators for Discrete k-Space or

In the box model, A>-(r) is given by Eq. (2.24). We then obtain the operator (2.88)
form of the vector potential for the discrete k-space:
Hence , the annihilation operator a,3 (k) defined in Eq . (2.87) for the conti-
{n ei(k-r-wt)
~L nuous k-space satisfies the commutation relation for continuous variables
A

A(r, t) = V ak.sEk,s £3 / 2 + h.c. (2.83)

in Eq. (2.88). Equation (2.86) then changes to the vector potential for the
k,s
continuous k-space:
Similarly, the operator forms for the electric and magnetic fields are
rn;;; ei(kr-wt) A

A(r, t) = s~
J 3
Jh
as(k)ik,s
ei(k-r - wt)
( 1r)3 / 2 + h.c. (2.89)
L V2 ak,sEk,s £3/2 + h.c. d kv
A

E(r, t) = i~ 2
2
k,s
Similarly, the electric and magnetic field operators in Eq. (2.84) change to:
rn;;;
A

B(r , t) = i~
•

L
k ,s
y2
ei( k-r -wt)
A

ak,s(k
A

X
A

Ek,s) £3 / 2 + h.c. (2.84) A

E (r, t) = i~ L
J 3
d kv
{fC; ei( k-r - wt)
2 as( k )Ek,s (21r)3/ 2 + h.c.

LJ
s=l ,2
2.3.3.2 Field Operators for Continuous k-Spac e A

3 ("fC; A ei( k-r- wt)

and 3-Dimensional Free Space B(r , t) = i~ d kv 2 as( k)(k X Ek,s) (21r)3/2 + h.c. (2.90)
s=l,2
When L goes to infinity , the discrete k-space changes to continuous k-space. Note that the mode function is now given by Eq. (2.26) with an orthonormal
However, the mode function A>-(r) given by Eq. ( 2. 24) will go to zero and relation given by Eq. (2.30) for the continuous k-spac e .
the sum in Eqs. (2.81) -( 2.84) will change to integral. So , this process cannot
be trivially done. We will handle this process in the following.
2.3.4 The Quasi-Monochromatic Field Approximation
In the discrete k-space, the smallest volume is b. 3k = (21r/ L ) 3 . When
L goes to infinity, b. 3 k = (21r/L) 3 -+ d3 k, i.e., 3-dimensional differential° When the spectral width b.w F of the exc ited modes of the optical field is
volume element for integration. So, the transition from a summation to an much smaller than the central frequency w0 of the optical field, i.e., b.wp «
integral is: w 0 , this optical field is called quasi-monochromatic field. On the other

For this transition,
(~)3L = Lb.3k-+
k k
J d3k. (2.85)
hand, there is a finite detection bandwidth b.wD for any photo-detector.
This means that detectors will not respond to the spectral modes of the
optical field outside the spectral bandwidth of the detectors. So, we can
we rewrite the vector potential in Eq . (2.83) as ignore the frequency modes outside the detection spectral bandwidth in our

A(r, t) =
~(21r)3
41rc~ L
{n
Ly~ A

ak ,s
( L)! Ek,s ( 1r) /
7r
A ei(k-r-wt)
+ h.c. (2.86)
calculation. In general, the spectral bandwidth of the detectors is usually
quite narrow so that b.wD « wo. Therefore , the optical fields seen by the
2 2 32
k ,s detectors are simply quasi-monochromatic fields.
But there is still a factor of (L/21r) 312 in the sum that will diverge as From Eqs. (2.81) and (2.82), we find that the operators of the optical
L -+ oo. To solve this difficulty, we introduce the annihilation operator for field can all be written in the form of
the continuous k-space: F(r , t) = L l(w) ak,sEk,sUk,s(r)e-iwt + h.c. (2.91)
as(k) = ak ,s(L/21r) 312 = ak,s/(b.k) 3 2 k ,s
1 . (2.87)
From the commutation relation in Eq. (2. 77), we obtain where l(w) = J21rnc 2 /w for A and l(w) = J21rnw for the electric and
magnetic fields . A quasi -monochromatic field has b.wp « w0 for which
[as(k),a!,(k')] = 5s,s 5k,k /b. 3k-+ 5s,s/5(3)(k-k')
1 1
the change in l(w) is much slower than e- iwt_ In this case, we can make
JS Quantum Optics For E.1:perimentalists Mode Theo7·y of Optical Fields and Their Quantization 49

a quasi-monochromatic approx im ation: l(w) ;:::::

l(w 0 ). Th en Eq. (2.91) ca n th e b eam size of the one -dim ensiona l field. Our selection of the coordinate
b e approximated as system is such that z-dir ect ion is the direction of propagation of the b eam
F (r , t);:::::l(wo) L ak,sEk,sUk,s(r )e-iwt + h.c. (2.92)
5 = k = k/ k. The boundary cond ition is that A is contin uou s only on the
two oppos it e sides in z-direction:
k ,s

So, different field operators on ly differ by a constant . We can then define a A(O) = A(L ). (2.97)
new field operator: The orthonorma l field function satisfying th is condition is then
V(r , t) =L ak,sEk,s'Uk,s(r) e-iwt + h.c. =y (+)( r , t) + v H (r, t) , (2.93) A( z) = Eseikz/ Jsi , (2.98)
k ,s with k = 21rm/ L(m is int eger ). So, the one-dimensional form of the vector
where we write Vin terms of the sum of the positive and negative frequency potential is
parts with the positive negative and frequency parts as
A( z, t) =
LY~ r,; ak 'sEsei(kz-wt) /\/SL+ h.c. (2.99)
y (+) (r,t) = L ak,sEk,sUk,s(r)e -iwt = (2.94)
k,s
In the photo-d et ection theory discussed later , we will learn that the output
k ,s
photo-electric signal from the detector is usually proportional to the square
From this, we may defin e the operator
of the field function , i.e. , IAl2 and the photo-electric signal is the contributi-
fi(r, t) = vH(r, t). y(+)(r , t). (2.95) ons from all points on the cross-section of the detector, i.e., an integration
Its physical meaning can be obtained from the following calculation: over the cross-section of the detector. But the contribution only comes
from the part with the illumination of the optical field. So, the area of
/d 3
rnA(r ' t) = ~~
L L ak
k,s k' ,s'
A ,s' ei(w-w')t(A*
At ,sak' Ek,s
A ,s' )jd3
. Ek' ruk* ,sUk' ,s' integration is the beam size S and the photo-electric signal is proportional
to IAl2 S. From Eq. (2.99) , we find that the cross-section area Sis canceled.
(2.96) Hence , we can drop out S from the expression for the vector potential in
k ,s k,s Eq. (2.99). Setting L • oo, we make a transition to a one-dimensional
form with a continuous k-value:
=
From Section 3.1.2, we find that aL,sak,s
operator of mode {k, s }. So, Nror is the total photon number operator
and fi(r, t) = y(-) · y(+) is the photon number density operator of the
fik ,s is the photon number
A(z , t) = L
s=l,2
J dk~ a.,(k)E.,ei(kz wt)/(21r)'l 2 + h .c.

optical field.
= \f'4rrI:: /
s=l,2
dw[f!
a,(w)t,e-iwt' /(21r) l 1 2
+ h.c. (2.100)
2.3.5 One-Dimensional Approximation of Optical Fields
where one -dimensiona l sum is changed to one -dimensional integral by
In the future chapters, we will encounter numerous situations where the (21r/L)x Lk = Lk l::::.k• J dk and as(k) = ak,s(L/21r) 1 ! 2 = ak,s/f::::.k
1 2
1,
optical field only propagates in one fixed direction, that is, only modes with which satisfies the commutat ion relation for continuous variables:
one fixed direction are excited while modes in other directions are otherwise [as(k),a!,(k')] = (L/21r)[ak,s,aL,l = 6s,s'bk,k'/!::::.k
in vacuum . In this case, the three -dimensional free field operators can be
• 6s,s 6(k - k') .
1 (2.101)
further simplified to one-dimensional field operators. To do this, let us go
back to the box model of Section 2.1.3 . The difference is that the cubic box In the integra l of the second line in Eq. (2.100), we made a change of
is replaced by a rectangular box with length L and a cross-section S and variable: k • w/c and a 5 (w) = a 5 (k)/vc, which satisfies
the field function A(r) only depends on one coordinate z along the length [iis(w),a!,(w')] = (l/c)[ak,s,aL,l = 6s,s (l/c)b(k-
1 k')
of the box: A(r) = A( z). Here, the cross -section S is usually chosen as = 6s, 6(w - w').
5
1 (2.102)
50 Quantum Opti cs For Experimentalists Mode Theory of Optical Fields and Their Quantization 51

Note that in the second lin e of Eq. (2.100) , the time variable is changed to
t' = t - z / c. So, in one-dimensional case, we only need the time variable.
Spatial translation is equivalent to a delay or advance in time: flt = -llz / c.
Similar to Eq. (2.100) , we have the one-dimensional expressions for the
electric and magnetic fields:
A
section, we will demonstrate how this effect arises from vacuum energy
change and derive a formula for the Casimir force.
Consider the box model for mode structure. But different from the
periodic box in Section 2.1.3, the box here is an L x L x L cube with
perfectly conducting walls , in which the electric field is zero. To change the

E(t ) = iJ4; L /

dwv
r,;;;;;
as(w)Es
e-iwt
+ h.c.
mode structure, we insert another perfectly conducting plate of size L x L
2 (2.103) and negligible thickness at x = a(« L) and parallel to the y z plane. We
_
12 v2nc
8- )
compare the total energy of the system with and without the plate.
A

B(t) = i~ L /

dwy
r,;;;;; A

as(w)(k x Es)
e-iwt
+ h.c. (2.104)
For the case without the plate, the simple boundary condition of E = 0
s-- 12)
2 v2nc at the walls (x, y, z = 0, L) gives rise to a mode function of the electric field
as
For the quasi-monochromatic field, we have the one-dimensional expres-
sion for the field operator Ek(r) = Ek(f )3sink xx sinkyysink zz (2.107)

y (+\t) = - -
1
L J dw as(w)Es e-iwt_ (2.105) with k = (kx, ky, kz) and kx = nxn / L, ky = nyn / L , k z = nzn / L (nx, ny, nz
s=l,2 = positive integers). Polarization vector Ek satisfies the transverse wave
Since there is only time variable, the physical meaning of the quantity condition Ek · k = 0, which leads to two independent polarization modes.
yH (t) · y( +) ( t) = R( t) is no longer the photon number density but the Furthermore, the mode function of the form
photon number rate in time. This can be seen from the time integral of
.R(t): (2.108)

J dtfl(t) = J
I:,
s,s'
dwdw'ii!(w)ii,,(w')(i'; • €,,) LJ dtei(w-w')t is allowed so long as Ek = z because E ..1 may not but E11must be zero at the
conducting walls due to the general boundary conditions of electromagnetic

LJ
fields and likewise,
= dwa!(w)as(w) = Nror- (2.106)
s
x( ; 3 ) sin kyy sin kzz, i) ( ; 3
) sin kxx sin kzz (2.109)
We will use the one -dimensional expression in Eq. (2.105) for the field ope -
rator quite often in the later chapters . are also allowed.
With the modes given above and w = ck, we find the vacuum energy of
the system without the plate as
2.4 Further Reading: Casimir Effect, A Quantum Effect of
Vacuum due to Mode Change = L nw>,/2 = L cnk>,/2
>- >-

{n,,n;,~len2+ (¥)2 + (¥) 2

From Eqs . (2 .76) and (2.78) in Section 2.3.2, we find the minimum energy of
the system is Evac = L>- nw>,/2, which is also the energy for vacuum from = C: 2

+n,t
Section 3.1 and hence is labe led with a subscript "vac". From Sections 2.1.3
and 2.2.1, we find W>,depends on the mode structure. So, the energy of
vacuum depends on the mode structure . If we change the mode structure,
the energy of vacuum will change. Energy conservation means that some
3 e~7rn (2.110)

work needs to be done for the change. This leads to a force . This effect Here NP denotes the case without plate . The factors of 2 and 3 in front
of vacuum energy is known as the Casimir effect [Casimir (1948)] . In this of the summations are for the two independent polarizations and 3 special
 >
52
Quantum Optics For E.1:perimcntalists
/I.lode Theory of Optical Fields and Their Q11antizatio11 53

s~lution~, respectively. The three special modes have the same contribu- the rema ining two terms du e to the spec ial solutions arc the same and we
tion._ With the transition from discrete to contin uou s k-space discussed in make a transition to the continuous k-space:

= 1=
Sect10n 2.3.3. we have for L -+ oo
en{2 (£)2
EPuac = -2 - "°"' dk dk (nr1r
-·- )2+ k . + k . 2 2
E{:,~= 2 2(~)3
c/t { { "dk xdkydk , /ki +kB+ k]_ L....t
7f
n~-=l
O y . z a y ;;

L
+3(;) r=
Jo dkydk /kB +k ;_} 2
2 (2.111)
Note that 6.k = 1r/ L here.
(2.113)
The ex~ress ion in Eq. (2.111) diverges so we need to compare it with
the case with plate, where there are two regions with different modes· ot e that the integration differential is dk x = ~k x = 1r/ (L - a) for the right
the
.
left side has k x. = n x 1r/a , k y -- ny1r/L , k z = n z7r/ L while
· the nght
· · side of the plate and we only keep the terms up to L 2 , which will be the
side has k x = n x 1r/ (L - a) , k y -- / L k /
ny1r , z = n z1r L, except the mode largest non-zero contributing terms to the energy difference (see below).
x( _i3)
sin kyy sin kzz, which is the same for both regions. So, we have for The energy difference between the cases with and without the plate is
the case with plate then

E!:ac
= c2/t{ 2 n,.JLI
(~) + (¥) + (¥) 2 2 2
M = en
(~)'{ f dkydk, k]_+ n~I
+2 f (t~"J+er'r +err
(2.114)
n x, ny,n z =l Now making a change of variables: kx = n1r/ a, ky = u1r/ a, kz = v1r/ a, we

nxt
+2 +en +2 /(t~"J+ en
2

nxt 2 2
arrive at
(2.115)

n,t +2 /(¥)'+(Tr}
with

A= 1= f dudv ( ++ 1u2 v2 -
00

++ u2 v'). (2.116)

= c2/t { 2 n,,n,t~o
e:")'(¥)' + + (¥)'
O n=(0) ,1 O

Here (0) denotes the extra factor of 1/2 for the term of n = 0 in the sum.
In practice , there is a high frequency cut-off because for high frequency,

+2 f
say, ')'-ray, the conducting metal plate becomes transparent and will not

nx,ny=l ,nz=l
(t~"J+err+(n'")'
L
,-------=------
support those modes. With this, the integral in Eq. (2.116) has an upper
cut-off limit and it can be shown that A = -1r /720 [Casimir (1948)]. So ,

+ err
f /(Ln~"J the energy difference per unit area by introducing the plate is
+2 2
nx,ny=l ~E / L 2 = - crm3 (2.117)
720a '

(2.u21 which increases with the separation a betw een the plate and the wall. For
this increase, an external force has to do the work. This leads to an at-
tractive force per unit area between the plate and the wall:
~ere_ in the second equation, we absorb the second sum from the first equa - 2
t1011mto the first sum as the term with n z -- O. With a fi111
·t e an d L -+ oo, F = ~(~E/L 2
) = crm4 . (2.118)
aa 240a
 54 Quantum Optics F01· Exp erim entalists
As has been demonstrated, this attractive force stems from t he energy of
vacuum due to the ch ange of the mode struct ur e so it is independent of the
material of the plates.
The effect of mode structure change of vacuum can also modify the rate
of spontaneous emission of atoms. This is b eca use atoms radiate ·light into
t he vacuum modes of the surround ing optica l fields and when the ava ilab le
modes change du e to the modification of the surrounding geomet ry, the
rate of radiation will chan ge . The mode st ructur e ca n be eas ily modified in
a cavity environment and this gives rise to the cav ity QED effect for ato ms
[Haroche and Klepp ner ( 1989)].
2.5 Some Remarks about the Unification of Particles and
Waves in Quantum Theory of Light
In the quantum th eory of light , the particle and wave pictur es are unified
in th e expressions of the field op erat ors in Eqs. (2.81) and (2.82), which
have a general form of
V(r, t) = LUA (r, t)aA + u~ (r, t)al. (2.119)
A
Here, uA(r, t) is the mode function satisfying Maxwell's wave equat ion s and
thus has all the properties of waves . On the other hand , the creation and
1,
annihilation operators a aA concern the particles of photons as we will see
in the next chapter. The classical wave phenomena such as interference
can be explained through the mode functions which are related to spatial
and temporal behaviors of the optical field, whereas the quantum behaviors
associated with particles is attributed to the quantum mechanical operators
of al, a,A via the expectation (average) values of the operators. Notice that
during the second quantization process, the classical quantities qA, ql in
Eq. (2.22) or (2.66), which determine the mode excitation strength, are
replaced by operators al, aA. So, operators al, aA will likewise determine in
quantum theory the excitation and thus the strength of the optical field via
their expectation (average) values but in a way that the energy is quantized
as particles of photons.
Although it seems that the quantum behaviors of the optical field are all
1,
borne in operators a aA, the mode function uA ( r , t) nevertheless plays an
important role in quantum interfer ence via the concept of photon indistin-
guishab ility thanks to the complementarity principle of quantum mechanics
(see Sections 8.2.3 and 8.4). When photons are created in a common mode
Mod e Th eory of Opti cal Fi elds and Th ei1' Quanti zal.ion 55
described by u A(r, t), they are comp lete ly inclistinguishable from each ot-
her and will give rise to t he maximum effect in quantum interference. On
the other hand, if two photons are respectively in two orthogonal modes
of u A(r, t), u N(r, t) with J dtd 3 r u A(r. t) · u ~, (r . t) = 0, they become com-
pletely distinguishable an d produce no quantum int erference effect, as we
will demonstrate in Section 8.4. They will have partial indi sting uish ab ility
and some quantum interference effect if J dtd3 r uA(r , t) · u ~, ( r , t) #-0 but
uA(r , t) #-u N( r. t). i.e., t heir modes are partially overlapped. Since int erfe-
rence effect is characte ri zed by coherence function, the mode fun ctio ns will
also be import ant in determining the coherence function (see Section 8.4.4).
All t hese discussions app ly equa lly well to the genera lized mode where the
genera lized mod e fun ct ion is a superposition of a set of orthogonal mod es:
u (r, t ) = LAcAuA(r , t).
From the discussion abov e, we find that the mode fun ct ion is basically
the identit y of the pho to n and we cannot talk about the photon without
its mode function , or the wave aspect of the optical field. This was first
pointed out by Lamb in a pap er with a provocativ e title of "Anti-Photon"
[Lam b (1995)], where he argued that "Photons cannot b e localiz ed in any
meaningful manner, and th ey do not behav e at all like particles, whether
described by a wave function or not. " So in this sense, Lamb considered the
word "photon" as a bad description of the quantum radiation field which
does not resemble a particl e at all but is actually a wave in the classical
limit, i.e., the mode function of waves.
To make a more close connection with our daily experience, in some
sense, the mode function resembles a house that p eople live in and photons
are the people in it. Houses, whose shapes are highly dependent on the
environment, are fixed in locations with physical addresses while people
move in and out of the houses. Modes of the optical field exist even in
vacuum just like empty houses without people living in. The difference is
that there are still activities for modes in vacuum due to (real) quantum
fluctuations and the activities in empty houses can only be (unreal) ghostly.
Since we cannot discuss photons without its wave description and waves
are intrinsically of nonlocal nature - mode function u(r , t) is spread in both
space and time , it is not surprising that the particle of photon - if we insist
using this term to describe the radiation field as Newtonian particles - will
be nonlocal and the violations of the Bell's inequaliti es [Bell (1964)], which
are satisfied by local realistic theories, are inevitable.
 M ode Th e01·y of Opt ica l Fi elds and Th eir Quanti zati on 57
56 Quantum Opt ics For E :rp eri m ental ists

2.6 Problems (iii) Prove

Problem 2.1 Any arbitrary physical quantity can be expressed in tenm,
of the creation and annihilation operators a8 (k). a!(k) , for example, the
energy expression in Eq. (2.78). Starting from the 3-dimensional free space
electromagnetic field operators in Eq. (2.90) . derive the total momentum
of the optical field:
(2.120)
L sh~'ja! (k)as (k). (2 .125)
s=+,-
where I,:= Ei x E2 is the direction of the propagation of the optical wave.
So, th e spin of photon only takes two values of ±ri with the sign determined
by the state of the circularly polarized photon (left or right). The photon
s~in states have S = l , m = ±1. But the state with m = 0 does not exist
because of the transverse property of light waves.

Problem 2.2 Prove that the three -dimensional free space electromagnetic
field operators in Eq. (2.90) have an equa l-time commutation relation:
A A , • a 3 ,
[Ej( r , t), Bk(r, t)] = -4rnriEjkl-a c5 (r - r ), (2.121)
rz
where

l
l, if i, j, k are even permutation of 1, 2, 3;
Eijk = -1, if i, j, k are odd permutation of 1, 2, 3; (2.122)
0, if any two indices are equal.

Problem 2.3 Define the spin operator for the optical field as

As l /
= --41rc 8.Ak
d 3 rE kz--Az A

(2.123)
[2 .
J J at
(i) Use the operator form of the vector potential A to prove that the spin
of the optical field takes the form of

nJ= iriL
s ,s'
J 3
d k a!(k)as,(k)(Ek,s' x Ek,sk (s,s' = 1,2) (2.124)

(ii) Prove that, if we define E+a+(k) + E_a_( k) = E1a 1(k) + E2a 2(k) where
E± = (E1± iE2)/-/2, we hav e
a±(k) = [a1(k) =t=ia2(k)l/V2,
[a±(k), a~(k')] = 0, [a±(k), a~(k')] = c5(k- k').
This means that a±(k) is the annihilation operator for the circularly pola-
rized light E±.
•

Chapter 3

Quantum States of Single-Mode

Fields

We start with the simplest case, i.e., a single -m ode field to discuss about
the description of the quantum states of an optical field. In this case, on ly
one mode of the optical field is excited while all other modes are in vacuum.
This is of course an idea l case . It is hard to produce a purely single-mod e
field in exper iments. In general, we mostly use a multi-mode description for
the optical field in exper im ents . Bu t und er some spec ial circum sta nces, we
can treat the optical field in the experiment approximately as a single- mode
field, when, for exa mpl e, the spectral width of the field is much narrow er
tha n that of the detectors. In this case, we will find that the single-mod e
description gives ris e to the same result as the multi-mode description.
Here, we may regard t he field approximately as a single-mode field with
only one frequency component. A transform-limited pulse, even though its
bandwidth is much wider than that of the detectors, can be regarded as a
single temporal mod e field. The single -mode approach is of course much
simp ler than the multi-mod e treatment.

3.1 Energy Eigen-States and Number States

3.1.1 Energy Eigen-States of a Simple Harmonic Oscillator

and the Concept of Photon

In pr evious chapter, we learned that a single- mode field can be descri-

bed as a simple harmonic oscillator. Quantization of the field is basically
the quantization of the harmonic oscillator. This subject can be found in
any standard textbook on quantum mechanics. Here, we will only present
the results. The Hamiltonian of a simp le harmonic oscillator is given in
Eq. (2. 78) and together with the commutation relation in Eq. (2.77) for

59
 •
Q1ia11tmn Stat es of Si11gle-Mod e Fi elds 61
GO Quantum Optics For Er:pcrirnen talists

one >..we can derive the energy cigen -states JE 11 ) of the Hamiltonian as: see that \11) is the cigen-statc of operator ata with an cigcn -value of n:

HIEn) = '7w(a.ta + 1/2)1En) = nw(n+ 1/2)IE,J. (3 .1)

ata\n)= 11\n). (3.3)

where we drop the mode-labeling sub scr ipt ·')." for the single-mod e field Therefore, 77= ata is called the photon number o~erator . ,
and n = 0. l, 2, ... . The energy of t he h armoni c oscillator takes discrete From the commutation relation (sing le-m ode) m Eq . (2.77) of the oea-
values with equal spacing betwee n adjac ent energy levels. When the os- tion and annihilation operators at. a. we can deduce (see Problem 3.1)
cillator absorbs energy of nw,it will jump to the next hi gher energy level atn
\n) =
c.\0). (3.4)
becoming mor e excited wher eas wh en it lowers it s energy by jumpin g from vn!
a higher energy level to the next lower level. it will release energy of nw. From the above and the commutat ion relation of at, a, we can eas ily obta in
So, the minimum amount of energy that can be absorbed or release d by
the oscillator is nw."Energy quanta" is the ea rly nam e that Einstein gave
atin)= Jn + l In+ 1),
to the minimum amount of energy nw[Einstein (1905)]. It was later com- a\n)= fa \n- 1). (3.5)
monl y known as the energy of a "photon ". Wh enever the oscillator absorbs Th e exp ression s above show that the action of operators at, a on the number
or releases nw, we will say in terms of the photon languag e that the field state will increase or decrease the photon numb er by one. So, they are also
acquires or loses a photon. known as the photon number raising or lowering operators.
From Eq. (3.1), we find that IEo) or the stat e with n = 0 is the state
of the harmonic oscillator with the least energy, or the ground state. It 3.1.3 q-Space Representation of Photon Number States:
corresponds to the case with no photon, or the vacuum state of the field. Wavefunction of a Single-Photon State
But from Eq. (3.1), we see that the vacuum state of the field has a non-zero
energy of nw/2 . This is one of the major differ ences between the quantum We learn ed from Section 2.3 .2 that for the virtual harmonic oscillator in
and classical theory of light . The energy of the vacuum state of the field the description of a single-mode field , its generalized coordinate (positim~)
will show up in th e form of Casimir effect [Casimir (1948)] (see Section 2.4). operator is Q and th e corresponding generalized momentum operator 1s
p = -ind/dQ. From quantum meC;._hanicstextbooks, we know that the
3.1.2 Photon Creation and Annihilation Operators and eigen-states \q) of position operator Q form a set of complete ba s:s f~r the
Photon Number States state space of the oscillator. The wave function 'ljJ(q) is the proJect10n_ of
an arbitrary state \w)on the \q)-base: 'ljJ(q) = (q\'ljJ),i.e. , the repr esentat10n
When the harmonic oscillator is in the first excited state IE 1), the ene rgy of of state \w)in the q-space. Then, what is the wave function Wn(q) for the
the field is one photon's energy more than the ground state or the vacuum number state \n)? We will derive it in the following.
state. Then, the field has an occupation of one photon. The first excited First of all, for the vacuum state of n = 0, we have a\0)= 0. From
state JE1) is the single-photon state of the field. When the field is in the Eq. (2.74), we obtain
n-th excited state !En), the field has the energy of n photons more than
the vacuum state. This is the n-photon state of the field. Since number a= ffn(Q+if), at= lfn(Q-i~) (3.6)
n uniquely determines the state, we use In) instead of !En) to represent
the n-photon state. This is also known as the number state of the field. Then aJ0)= 0 becomes
So from the description above , we see that a photon is simply an energy p
(3.7)
A

(qlQ + i-l0) = 0
excitation of nw.In terms of the new lab els, Eq. (3.1) can be rewritten as w

iIJn) = tiw(ata + 1/2)ln) = nw(n + 1/2)ln), (3.2) in the Q-space. Acting operators Q, P on (q I, we have
where I0) is the ground state of the field and satisfies al0) = 0. It corre- nd (3.8)
(q + ;: dq) (q\0) = 0.
sponds to the vacuum state of no photon excitation. From Eq. (3.2), we
 •
Quantum States of Single-Mode Fields 63
62 Quantum Optics For Experimentalists

furthermore, since the energy eigcn-states { IE11)} are non-degenerate, that

Introducing the dimensionless quantity x = q/q 0 with q0 = ~, Eq. (3.8)
is, its eigcn-value uniquely determines the eigen-state, this set of eigen-
is changed to
states forms a complete base set for the quantum state space of the harmonic
(X + d~ ) ?/Jo(x) = 0. (3.9) oscillator. Hence. the photon number states {In)} satisfy the completeness
relation:
The solution of this differential equation is
2
L ln)(nl = i, (3.15)
1/Jo(x)= ce-x !2 . (3.10) n

C is a constant determined by the normalization condition. So, the nonna- where i is the unit operator.
For any operator p, we have after using the completeness relation in
lized wave function for the vacuum state is
Eq. (3.15)
'l/Jo(q)= 1 e-q2/2q5 (3.11) p = ipi
Qo = is the amplitude of a classical oscillator whose mass is 1 and
= L \n)(nlp L \m)(m\
n m
total energy is nw/2 or the energy of the vacuum state.
For then-photon state In), using Eqs. (3.4) and (3.6), we have its wave = L Pnmln)(m\, (3.16)
m,n
function as
where Pnm = (n\fJlm). So, in the number state representation {In)},
operator p can be uniquely determined by matrix {pnm}- For example,
the position and momentum operators have their matrices as

n 0 1 0 ..
l O ../2···
·1
(3.12)
{Qnm} = ffw (
O J2 0 : '
(3.17)

d~:
1
2 2
where Hn(x) = (-1rex (e-x ) is the nth order Hermite polynomial.
Particularly for n = l, we have the single-photon wave function as
0 -i O .. ·
nw i o -i../2. ••
?/J
i (q) = ../2q e-q2 /2q5 (3.13) {Pm,,}= ff (
O iJ2 0 · : ·
(3.18)

In Section 10.4, we will discuss quantum tomography technique for the

J (l1L}
measurement of the quantum state of the optical field. When the field is For the creation and annihilation operators, we have

(~~0
in a single-photon state, Eq. (3.13) is the wave function of the field to be
measured.
(3.19)
{anm}= {a~m}=
3.1.4 Photon Number States as the Base States of the State
Space
The relation in Eq. (3.15) allows an arbitrary quantum state of the
Because the Hamiltonian of the harmonic oscillator is a Hermitian operator, harmonic oscillator or a single- mode field to be expressed in terms of the
its energy eigen-states {!En)} or the photon number states {In)} are ortho- superposition of the photon number states {\n) }:
normal:
\?/J)= i\'lj)) = L \n)(n\'lj)) = L cn\n), (3.20)
(3.14) n n
 •
Q'Uani'l.lm States of Single-Mode Aelds 65
64 Quantum Optics F01· Experimentalists

where Except a constant phase term, c0 ;=(0lcv)can be de~ennined by the nor-

(3.21) .. t· 011 condit ion as co = e-lal 12 . Then we obtam the coherent state
lll8- 1l Zcl 1 ·
ext, we will discuss a specia l sup erposition state of number states - in the number state representation:
2 O.n
the cohe rent state. \a) = e-lal /2 -\n). (3.25)
L...,
n
v1nf
3.2 Coherent States In) \i\Tith Eq . (3.4) for the number states, the express ion above is changed to
The exp ression for the cohere nt state was first given by Schrodinger in 1926
[Schrodin ger (1926)]. He prov ed that the coherent state gives the minimum
\a) = e-lal2 /2 L a~
n.
a,tnIO)= e-lai2 /2+aat \0)
n
value in the Heisenb erg un certa int y inequ alit y an d used it to describe the (3.26)
==D(a)\0),
classical trajectory of a quantum mechanical harmoni c oscillator. But later, A 't *'
it was Glauber who discovered that the coherent state is the best quantum where D(a) is the displacement operator. Its final form D( a) = eaa -a a

state to descr ibe the coherence properties of an optical field and developed can be obtained from Eq. (3.120) by a and a,t algebra in Section 3.6.
therefrom the qu antum theory of optica l coheren ce, which was the founda -
tion of quantum optics [Glaub er (1963a,b, 1964)]. 3.2.2 Photon Statistics Distribution and Photon Number
Fluctuations of a Coherent State
3.2.1 Definition of the Coherent State and Its Number
The coherent state is a superposition state of many number states. The
State Representation
probability for ea ch photon number state, that is, the photon numb er sta -
The definition of the coherent state given by Glaub er is tistical distribution is
ala) = ala), (3.22) (3.27)
That is , th e coherent state is th e eigen -state of th e annihilation operator a.
But since the annihilation operator a is not a Hermitian operator, its eigen- This is the Poisson distribution with an average of fi = \a\ . From the
2

value a may not be a real number. The coherent state is thus not a pro- above, we can find the variance of the photon number distribution as
jection state of any physical measurement. But this is not what this state
(3.28)
is meant for.
n
From the definition of the coherent state in Eq. (3.22) , we can deduce
its specific express ion in terms of the superposition of the photon number On the other hand , we can also use the method in quantum mechanics
states. There are many ways for this. Among them, the most direct way is to calculate the expectation value of an operator. For photon number
to use the expression of photon number states in Eq. (3.4), the definition operator n ==at a, we obtain
of the coefficients of a superposition state in Eq. (3.20), and the definition (3.29)
of the coherent state in Eq. (3.22). The derivation is as follows:
la) = :I:Cnln ), (3.23) and

with
n
(n2) = (a\(ata)2\a)
= (a\ at aata\ a) = (a Iat (at a + 1)aIa)
= \a\4 + \a\2 = fi2 + fi
= (n) 2 + (n) . (3.30)

(3 .24) The expression above is exact ly Eq. (3.28).

 66 Quantum Optics For Experimentalists Quantum Stat,es of Single-Mode Fields 67

3.2.3 Classical Trajectory and Quantum Uncertainty
Simple Harmonic Oscillator
We mentioned in the introduction of the coherent state that Schrodinger
derived the expression of the coherent state in order to find a quantum
state that gives a description that is closest to the classical trajectory of a
harmonic oscillator. That is, the position average (x(t)) as a function of
time is consistent with the classical trajectory, and its quantum uncertainty
is the minimum. In Problem 3.2, we will prove that the number state is
not such a state. But the coherent state is. To prove it, we start with the
Hamiltonian in Eq. (2. 78) and obtain the time evolution of the coherent
state :
of Therefore, the quantum harmonic oscillator in a coherent state has its
average trajectory same as the classical trajectory: in the x - p phase
space, the trajectory of the virtual harmonic oscillator, a~tcr s~me proper
normalization to each quantity in Eqs. (3.32) and (3.33), 1s a circle. From
the result above, we obtain the equivalent energy of the classical oscillator
as
u= [(fa)2(t) +w 2 (Q)2 (t)]
= nwlal 2 = nnw. (3.34)
which, apart from a constant of vacuum energy, is the same as the quantum
counterpart.
As for the proof of minimum uncertainty for the coherent state, let us
first consider the dimensionless quantities of quadrature-phase amplitudes.

3.2.4 Quadrature-Phase Amplitudes and Quantum Noise

For the virtual harmonic oscillator in the description of a single-mode field,

the position Q and momentum P do not have a straightforward physical
meaning. But let us check out the following dimensionless quantities:
(3.35)
(3.31) These are the quadrature-phase amplitudes of the optical field. The phrase
So, apart from a general phase, the harmonic oscillator initially in a cohe- "quadrature -phas e" comes from the decomposition of the Hermitian field
rent state will still be in a coherent state later , but the value a is multiplied operator E(t) = ae-iwt +at eiwt in Heisenberg picture in orthogonal phases:
by a phase factor e-iwt : a(t) = ae-iwt_ Now we find the average value for E(t) = ae-iwt + at eiwt = x cos wt+ Y sin wt. (3.36)
the position (x(t)). For the virtual harmonic oscillator corresponding to a
X, Y represent the amplitudes of the phase-orthogonal terms of
single -mode field, x = Q = ~(a+ at). We then hav e
cos wt, sin wt, respectively. For a quantized harmo:1ic 9scillator, they corre-
(Q)(t) = ~(a(t)l(a + at)la(t)) spond to the position and momentum operators Q, P after some normali-
= ~[(a(t)lala(t)) + (a(t)latla(t))] zation: X = Q~, Y = P~.
For a coherent state, the average of these two quantities are
= ~[a(t) + a*(t)] = lal cos(cp0 - wt). (3.32)
(3.37)
We find from the above that the average is similar to the expected classical
oscillatory trajectory of the position of a harmonic oscillator. Similarly, we
In order to find the fluctuations (L1 X), (L1 Y) of
2 2
X, Y, we may first cal-
culate (X 2 ), (Y 2 ) as follows:
obtain the oscillation of the velocity of the harmonic oscillator:
(X 2 )a = (al(a + at) 2 la)
(Q)(t) = (P)(t) = ~(a(t)l(a - at)/ila(t)) = (ala 2 + at2 + aat + at ala)
= ~[a(t) - a*(t)]/i
= (ala 2 + at2 + 2at a+ Ila)
= lal sin(<p wt).
0 - (3.33) = (a+ a*) 2 + 1. (3.38)
68 Q'Uant'Um Optics For Experimentalists Q,uanhlm St,ates of Single-Mode Fields 69

Similarly, we have .6.X .6.1· 2: 1. Therefore, since ~X ~y = 1 for a coherent state, it is

(3 .39)
Hence,
(3.40)
The quadrature-phase amp litude for arb itrar y phase angle is defined as
(3.4 1)
It is easy to see that .X = X (0), Y = X (7r /2) . Actually, X (cp) is relate d
to X, Y by a rotation of ang le cp: _f(' = X cos <p + Y sin cp = X (cp) (see
Fig. 3.1). Th e average of .X(cp) is
(.X(cp))a = (.X)a coscp + (Y)a sincp = 2lcxlcos(cp - 'Pa)- (3.42)
Simil ar to Eq. (3.40), it is easy to prov e
(3.43)
Notic e that the result above is independ ent of th e rotation angle cpas well as
ex. In the phase space of X-Y (similar to the classical x-p phase spac e, i.e. ,
the Wigner phase space, see Section 3.7.2) , the coherent state is described as
a vector with a circular pattern of radius of 1 at its tip, as shown in Fig. 3.1,
wh ere the unit size of the circle represents that the varianc e of .X(cp) is 1
as given in Eq. (3.43) , and the vector from th e origin of the coordinates
to the center of the circle represents the comp lex number (.X)a + i(Y)a =
2cx = 2lcxleicp
°'. Note that this coordinate system rotates clockwise with an
angular speed of w so that the time-varying quantity 2cx(t) = 2lcxlei(cp°'-wt)
is changed to time-independent quantity 2cx = 2lcxleicpa.. From Eq. (3.35) ,
we have [.X, Y] = 2i. This leads to the Heisenberg uncertainty relation:
the state that Schr odinger required to give rise to the minimum value in
Heisenberg uncertainty inequality.
\Yhen o. = 0, the coherent state becomes the vacuum state, which is
represented by a circle with its center at the origin and its radius as 1,
as shown in Fig. 3.1. The variance of X, Y in Eq. (3.40) represents the
fluctuations of vacuum field, i.e., the vacuum noise. In classical physics,
vacuum has nothing in it . So, vacuum noise is caused by the quantization
of the electromagnet ic fields and is a kind of quantum noise . The squeezed
states that we will discuss later in Section 3.4 are a kind of states in which
this type of vacuum quantum noise is squeezed. From Eq. (3.26), we find
that the coher ent state can also be cons idered as being displaced by 2cx
from the vacuum state (see Fig. 3.1).
3.2.5 Non-Orthogonality and Over-Completeness of
Coherent States
For two different coherent states lex),l,8), we obtain their inn er product from
Eq. (3.25) as
(3.44)
or
(3.45)
So, different coherent states are not orthogonal. This reflects the non -
Hermitian propert y of the annihilation operator a. But when lcx-,812 >> 1,
we have (cxl,8) 0.
On the other hand , using Eq. (3.25) and the completeness relation in
Eq. (3.15) for the bases of number states, we can directly prove

coherent
2
) (cxl =
/ d cxlcx 1r i. (3.46)
y state
The expression above indicates that the whole set of the coherent stat es is
overcomplete. This can be viewed as a direct result of the non-orthogonality
vacuum
state""- 2a of the coherent states demonstrated in Eq. (3.44). Nevertheless, we can
L\X
rewrite Eq. (3.46) as

(3.4 7)

and use it for the expansion of an arbitrary quantum state of a sing le-mod e
Fig, 3.1 The repr ese ntation of a coherent state la) and the vacuum state IO) in X-Y field in terms of the coherent states:
phas e space . The circle of radius 1 represents the size of th e quantum noise, The coherent
state is obtained by displacing the vacuum state with a comp lex valu e 2a, I'!j;) = i I'lj;) = / d2 exIex)(exI'lj;) = / d2 cx'lj;(ex)Iex), (3. 48)
70 Quanfom Opt ic s For E xperim entalists Qua n tum Stat es of S ingl e-M ode F ie lds 71

where 'lj;(a) (al i/J)/n. So, the whole set of the coherent states forms consid ered as two classical macroscopic states. Th e sup erposition stat e of
an overcomplete set of bases with which any quantum state can be th ese two states is the Schroding er cat state of an optical field:
expressed. But the non-orthogonality of the coherent states in Eq. (3.44)
makes function 'lj)(a) in the expansion in Eq. (3.48) non-unique. This can t = N ( In) + ei ¢ 1/3)) .
l'I/J)ca (3.49)
also be considered as the result of the overcompleteness of th e coherent Her e, In ), 1;3) ar e nearly orthogonal: if la - ;312
» 1, th en I(n 1/3)12 =
2
state bases. e-lo-/31 0 and th e normalization constant N - = 2[1 + R e( (nl;3)eicp)]
2

The consequence of non-uniqueness of the coherent state expansion is 2. The commonly discussed Schrodinger cat state has ;3 = a ei 0 , i.e. , stat e
not all negative. We will show later in Section 3.7.1 that the overcomplete- 1/3)is rotated from la ) by an angle 0. So, la - ;312 = 4lal 2 sin 2 0 / 2. When
ness of the coherent state bases makes it possible to express the density ope- 0 » 1/ lal, we have la - ;312 » 1.
rator of any quantum state in Glauber-Sudarshan P-representation. This is Normally , th e consequence of quantum superposition is interference phe-
the basis for Glauber's quantum coherence theory as well as the foundation nomena . But different from the interference phenom ena du e to the super-
for quantum optics. Later in Chapter 5 when we introduce Glauber 's quan- position between two optical fields, the quantum superposition state in
tum coherence theory, we will discuss more properties of coherent states. Eq. (3.49) does not give rise to the traditional interference fringes in which
the intensity (photon number) changes with the phase difference. A direct
3.3 Further Reading: Schrodinger Cat States calculation gives the average photon number for the Schrodinger cat state
in Eq. (3.49) as
The Schrodinger cat state is a quantum superposition of two completely
exclusive classical macroscopic states (such as the dead and alive states of (3.50)
a cat). Even though the state concerns two classical states, its existence is It is independent of the phase difference between the two states. This is
allowed in quantum mechanics. Since it has the superposition property of because the two states la), l/3) are nearly orthogonal.
the microscopic world, it leads to the famous Schrodinger cat paradox in The consequence of the quantum superposition in Eq. (3.49) is exhibited
the exp lanation of quantum mechanics [Schrodinger (1935)]. So, whether in the interference effect in the probability distribution of the quadrature-
or not we can create it in the lab is a test for the suitability of quantum phase amplitude X. To see this, consider the following Schrodinger cat
mechanics for the macroscopic world. It is just because we usually do not state:
find Schrodinger cat states in the macroscopic world that our macroscopic
world can be safely described by the classical theory of physics most of l'I/J)cat= Nr(I - ir) + lir) ), (3.51)
the time. This also shows on the other hand that it is extreme ly hard to 2
where r is real and Nr- = 2(1 + e- 2 r ) so that l'I/J)cat
2
is normalized. Since
realize a Schrodinger cat state. Nevertheless, scientists recently produced a X = a+at = vl2Q / q0 with q0 = ~' the distribution of X is that of Q.
Schrodinger cat state in the lab [Brune et al. (1996); Monroe et al. (1996)] For this, we calculate the wave function '1/Jcat(q)
for state l'I/J)cat:
and therefore proved the suitability of quantum mechanics in macroscopic
world and the feasibility of quantum computers. More importantly in their = (ql'I/J)
'1/Jcat(q) cat = ((qi - ir) + (qlir) )Nr. (3.52)
exper iment is that the scientists observed the process of de-coherence of From Problem 3.3, we find the wave function for the coherent state la) as
the Schrodinger cat state, that is, the transition process from quantum
superposition to classical mixture states [Myatt et al. (2000); Deleglise et 'Ip (x) = _l_e-Im 2
(a)-(x-av'2)
2
/2 (3.53)
a (n)l/4
al. (2008)]. This exp lained why there are very few quantum phenomena in
our macroscopic world. with x = q/qo. Substituting into Eq. (3.52), we then obtain the wave
We mentioned earli er that a coherent state is a quantum state that function for the Schrodinger cat state :
is closest to a classical description of the electromagnetic waves. So, two
_ 2Nr -x2/2
coherent states with large separation and thus very small overlap can be '1/Jcat(x)- (n)l/ 4 e cos v 2rx. (3.54)
72 Quantum Optics For ExperhnC'n/ ;al'ists Quantum States of Single-Mode Pi.clds 73

So, the probability distribution for the position Q of the virtual harmonic inequality whereas from Section 3.2.4, we proved that the coherent states
oscillator in the Schrodinger cat state is (including the vacuum state) has the minimum value of the Heisenberg
2N; _ 1;2 1n
inequality. i.e.,
Pcat(x)= foe· (l+cos2v2rx). (3.55)
(3.57)
It can be seen that this probability distribution shows an interference fringe Therefore, it seems we already reached the minimum value of the Heisen-
as the position x changes. berg inequality with the coherent states and cannot go even further. This
The visibility of the interference fringe in Eq . (3.55) is 100%. Classically, - argument is correct if both (6 2X) and (6 2Y) are reduced together.
the two states can only be in a statistically mixed state which we will discuss However, the inequality in Eq. (3.57) is about the product of the two
later in Section 3.5. It gives no interference, i.e., the visibility is zero (see quantities, that is, quantum mechanics only sets the limit on the product of
Section 3.5.1). The quantum superposition in the Schrodinger cat state is the two uncertainties but not on individual quantities. So, we can rearrange
extreme ly sensitive to losses. In Chapter 6. we will treat the effect of losses the two quantities in such a way that only one of them is reduced at the
quantum mechanically and prove that when the loss is 'Y, the distribution expense of the increase in the other while keeping the product unchanged,
in Eq. (3.55) is changed to [Walls and Milburn (1985)] (see Problem 6.9) i.e., Eq. (3.57) is unchanged.
2 Suppose some state Ir) satisfies the requirement above, that is,
Pcat(x) ~e-x ( 1 + V cos 2V2rx~), (3.56)
(62 X)r = (62 X)ae2r = e2r, (62Y)r = (62Y)ae-2r = e-2r, (3.58)
where V = e- 2,,r is the visibility of the fringe. From the above, we find that
2

where r is a positive number to determine the degree of squeezing. In the

for a macroscopic Schrodinger cat state with r 2 » 1, when ry » 1/r 2 ,----, 0, following, we will derive such a quantum state Ir).
V 0, that is, a very small loss can make the Schrodinger cat state de-
cohere to the classical mixture state. The larg er the macroscopic quantity
3.4.2 Squeezing Operators
is, the smaller the required loss is for de-coherence. This exp lains why it
is hard to find a quantum superposition Schrodinger cat state in a real Suppose we can obtain Ir) by app lying a unitary operator Son the coherent
macroscopic world. state:
Ir)= Sia). (3.59)
3.4 Squeezed Vacuum States and Squeezed Coherent States
Substituting into Eq. (3.58), we have
As we discussed in Sections 3.1.1 and 3.2.4, vacuum fluctuations of elec- (St6 2XS)a = (6 2X)ae 2r, (St6 2YS)a = (6 2Y) 0 e- 2r. (3.60)
tromagnetic fields is a kind of quantum noise , i.e. , it is the inevitable con-
sequence of quantization of the optical fields. Although it is impossible The sufficient condition for the above is that the unitary operator S must
to avoid the quantum noise, can we rearrange and redistribute it so that satisfy the following operator relations
some of it will be reduced while other part will incr ease? The squeezed sta- stxs= fcer, st-vs=
P-e-r. (3.61)
tes that will be discussed in this section are such kind of quantum states,
which have some of their noise below vacuum noise level. Let us see how Notice that the operator transformation above keeps commutation relation
to squeeze the quantum noise in the following. [X, Y] = 2i unchanged. From Eq . (3.61), we can find the transformation
for a,a,t:
3.4.1 Squeezing of Quantum Noise stas= acoshr + at sinhr, stats= at coshr + asinhr. (3.62)

From Problem 3.2, we find that the photon number state has the uncer- Similarly, the commutation relation [a, at] = 1 is unchanged for this trans-
tainties (6 2 X)n, (6 2Y)n larger than the minimum value of the Heis enberg formation.
74 Quantum Optics For E:rper-imentalists Quantum States of Sing le-Mode Fields 75

From the operator algebra for ii and at (see Section 3.6) , we can deduce Using the above and the noise ellipse of the squeezed vacuum state. we
from Eq. (3.62) the unitary operator S as [Stoler (1970)] obtain in Fig. 3.2(a) the representation of the squeezed coherent state in
the X-Y phas e space. The squeezed vacuum state is represented by the noise
(3.63) ellipse with its center at the origin. It can be seen that the noi se along Y-
This is the squeezing op erator. Its action on the vacuum state gives the direction (i.e., (~ 2 Y)r) is smaller than that of th e vacuum state (the circle
squeezed vacuum state : v.rith its center at origin and radius of 1). The squeezed coherent state is
shifted away from center by the amount given in Eq. (3.67) . Note that the
Ir) = S(r)IO), (3.64) original coherent components are squeezed in Y-direction and stretched in
and when it acts on the coherent states, we obtain the squeezed coherent X-direction just like the noise part.
states: From Eq. (3.63), we obtain §t (- r) = S. Hence,

Ir, ex) = S(r)lex). (3.65) Sast = ii cosh r - iit sinh r. (3.68)

For the quadrature-phase amplitude X(zp) = iic icp+ ateicp with an If we define operator b = ii cosh r - at sinh r, then from Eqs. (3.68) and
arbitrary angle zp, we can use Eq. (3.62) to prove (3.22), we obtain

(~
2
X (zp))r = e 2r cos 2 zp+ e- 2
r sin 2 zp. (3.66) blr, ex) = sast Slex) = Sal ex) = exlr,ex), (3.69)

In the phase space of X-Y as shown in Fig. 3.2, the quantum fluctuation that is, the squeezed coherent states are the eigen-states of operator b with
of X(zp) is represented by an ellipse, which is the area in which the Wig- an eigen-value of ex. Similar to Eq. (3.22), Eq. (3.69) can also be considered
ner quasi-probability density is appreciably non-zero (see Section 3.7.2 for as the definition of squeezed coherent states [Yuen (1976)].
more). This is the noise ellipse. For vacuum state and coherent states the
ellipse becomes a circle of unit radius. ' 3.4.3 Coherent Squeezed States and Squeezed Coherent
Using Eq . (3.62), we have States

(r, exjYlr, ex) = (ex- ex*)e-r/i. (3.67) From Fig. 3.2(a), we find that when the squeezing operator acts on the
coherent state, not only its noise in Y or X is squeezed or amplified but
coherent also are its averages. This is because the squeezed coherent state can be
squeezed
coherent state written as

.,.-
y-,--,,,,/statea~> Ir, ex) = S(r)D(ex)IO), (3. 70)
vacuum 2
state".. where D(ex) is the displacement operator defined in Eq. (3.26) . So, the
-
--\--t----¥'--+---l----- ~ X squeezed coherent state is subject to displacement first and then squeezing
"--- squeezed
vacuum state so that the displacement amount is also squeezed. If we switch the order of
(a) (b)
the squeezing and displacement operators, we obtain the coherent squeezed
state:
Fig. 3.2 (a) The representation of squeezed coherent states and squeezed vacuum state
in the X-Y phase space. The squeezed ellips e represents the size of the quantum noise lex,r) = D( ex)S(r) I0). (3.71)
in the polar coordinates. Quantum noise is squeezed in Y-direction and is smaller than
the vacuum noise but is increased in X-direction. The coherent components are also This is different from the squeezed coherent state Ir, ex) defined in Eq. (3.70).
squeezed and stretched in Y- and X-directions, respectively. (b) The representation of From Section 3.6 on the algebra of ii, iit, we can prove
the coherent squeezed states in X- Y phase space: coherent squeezed states is shifted by
2a from the squeezed vacuum state. b\ex)iiD(ex) =ii+ ex. (3.72)
 76 Quanf'llm Optics For Experimentalists Quantum Stat es of Singl e-1\Iode Fields 77

With t he above, we have y la, r \ - " coherent

"-.... ~. squeezed state
(et, rlalo, r) = (01st(r)bt(a)ab(a)S(r)I O) \.._. ,- ''·__)
~-
= (0/S\r)aS(r)IO) +a vacuum state 2a C--~
la. rr/ 0 '
=a. (3. 73)
-+---1----¥'---+--+----- -x
Similarly, "- squeezed
vacuum state
(3.74)
Hence. Fig. 3.3 The representation in the X-Y phase space of the coherent squeezed state wit h
an arbitrary squeezing parameter = reie: t h e noise ellip se is rotated counterclockwise
(a, rlXla, r ) =a+ a*, (a. rlYla, r) = (a - a*)/i . (3.75) by 0/2.

So, differe nt from the squeezed coherent state in Eq. (3.67), the averages of
X , Y for the coher ent squ eezed state are the same as t ho se of the coh ere nt
Using this, we can prove that for a coherent squeezed state la,~) =
state . Using Eq. (3.72), we can prove b(a)S(~)IO),
2
(a, r/6 XI&, r) = e
2
r, (a, rill 2
Y I&,r) = e- 2 r, (3. 76) (3.83)

that is, the fluctuations of the coherent squeezed stat e are the same as
those of the squeezed coherent state . Therefore , as shown in Fig. 3.2(b) , a
coherent squeezed state is displaced by 2a from a squeezed vacuum state .
Although the squeezed coherent state is obtained by squeezing the co-
h~rent _state , we can see from Fig. 3.2(a) that it can also be obtained by
d1splacmg the squeezed vacuum state. The displacement amount is not 2a
but depends on both a, r. Using Eq. (3.62) , we can prove
Comparing with la, r) (0 = 0) , la ,~) simply rot ates counterclockwise the
noise ellipse by an angle of 0/ 2 in the X-Y phase space (see Fig. 3.3).
Next, let us see how the electric field strength changes with time for an
electromagnetic field in a coherent state or a coherent squeezed state. This
allows us to have a dir ect picture of the quantum field in a similar way to
the picture for a classical electromagnetic wave (see Fig. 1.3 in Section 1.3).
From Eq. (2.84), we find th e electric field operator as

st (±r)b(a)S(±r) = b(a=F), (3.77) E(r, t) = Ciae-iwt' + h.c. = CX(wt' - 1r/2). (3.84)

where a±= acoshr ± a* sinhr. Using S(-r) = st(r) = s- 1 (r), we have where t' = t - k • r / c and C = J 41rnw/ 2£ 3 is a constant.
For a coherent state la)(a = lalei'Pa), we have the average value of the
D(a)S(r) = S(r)D(a_) , (3.78)
electric field as
S(r)D(a) = D(a+)S(r). (3.79)
So, the relation between a squeezed coherent state and a coherent squeezed
(E(r, t))a = Ciae -iw t' + c.c. = 2Clal sin(wt' - <pa)- (3.85)
state is This corresponds to the change in space and time of the electric field
strength for a classical field, as shown in Fig. 3.4(a). In the meantime,
la, r) = Ir, a_), Ir, a)= la+, r). (3.80)
the fluctuation of the electric field is
For an arbitrary squeezing parameter = rei 0 , we can genera lize the
squeezing operator in Eq . (3.63) as (3.86)
S(~) = e(~at2_Ca2)/2_ So, the uncertainty, i.e. , fluctuation, of the electric field at every moment is
(3.81)
the same, i.e., C. Figure 3.4(b) shows the change of the electric field with
Making a transformat ion of a'= ae-i 0 ! 2 , Eq . (3.81) is then changed back
time for an electromagnetic field in the coherent state . The fluctuation of
to Eq. (3.63) . So, we can easi ly obtain
the electric field is represented by a band with a width of C in Fig. 3.4(b).
st (~)aS(~) = a cosh r + i 0 at sinh r. (3.82) For comparison, Fig. 3.4(a) shows the change of the electric field with time
 •
78 Qua11tum Optics For Experimentalists Quantum , States of Single-Mode Fields 79

fluctuation of the field in the coherent sqncczed state jilnl, r ) is larger than
a coherent t,tatc.
We find from Eq. (3.83) that the squeezing position for the fluctuation
of the electric field can be adjusted by the phase 0 of l- So, for the coherent
squeezed state Ii\o_ \, - r ) with 0 = 1r, we can use a similar analysis as before
and find that the phase fluctuation of the electric field is squeezed while
2Cla, the amplitude fluctuation becomes large, as shown in Fig. 3.4( d). That
£(/')
is why jila\, -r) is called the phase-squeezed state whereas Ii\al, r) is the
- 2Clal amplitude-squeezed state. For an arbitrary phase cpa of the coherent com-
TC/2 TC 3TC/2 '2TC STC/'2 3TC 7TC/2 4TC 9TC/2 ponent, we find that the coherent squeezed state la , -re 2icpCl) corresponds
{u/'-qJa
to the amplitude-squeezed state and \a, re 2 i'Pa.) is for the phase-squeezed
state. From the diagram of the amplitude -squ eezed state \a,-re 2icpCl)in
Fig. 3 .4 The wave of the electric field as it changes with time: ( a) a classical field: (b)
a field in a coherent state; ( c) a field in an amplitude -squeezed state; ( d) a field in a the X-Y phase space [Fig. 3.5(a)], we can see that as the phase 'Pa of a
phase-squeezed state. increases from zero, the tilt ang le of the noise ellipse also changes . The
noise squeezing direction always keeps at the amplitude direction (radial
for a classical wave, i.e., Eq. (3.85), which has a definite value at every direction). Figure 3. 5 (b) shows the diagram for the phase-squeezed state
moment. la , re2icp"').
For the coherent squeezed state la, r)(a = laleicp"'), the average value
of the electric field strength is the same as Eq. (3.85) but the fluctuation of r- y lna1.
-r; /
/
the electric field is , ·-- la, re2ilfJ.
,)

Ju~2 E(r,t))r = cJ(!:i2 X(wt -1r/2))r 1

2a
'"\ ~Ila\, -r )
2a

= cJe r cos 2(wt' - w/2) + e- 2r sin 2(wt' -

2 w/2). (3.87) _jL___J_ ')_(f) -----I---+--
O=-'.l X

It is a function of time and changes as the wave propagates. Set ({Ja = 1r/2 \
or a = i Ia I. The peak of the electric field occurs at wt' - <pa= wt' - 1r/ 2 = (a) (b)

Nw ± 1r/2(N = integer). At this moment, the fluctuation of the electric

field given in Eq . (3.87) is minimum, i.e., ce-r.
Figure 3.4(c) shows the
Fig. 3.5
la, -re
The diagrams in the X-Y phase space for (a) the amplitude-squeezed
2 i'Pa.)and (b) the phase-squeezed state la , re 2 i'Pa.).
state

evolution of the average value and the fluctuation of the electric field as
time changes. We find that the uncertainty of the electric field at the peak
of the electric field is smaller than the corresponding part in Fig. 3.4(b)
3.4.4 Further Reading: Photon Statistics of a Squeezed
for the coherent state. So, the amplitude fluctuation of the electric field is
State and the Oscillation Effects
squeezed for a field in the coherent squeezed state jilal , r). On the other
hand, when wt' - ({)a= wt' - 1r/2 = N 1r, the electric field is zero, i.e., at We have just seen that the squeezed coherent state can squeeze the quan-
the trough of the wave. At this moment, the fluctuation of the electric tum fluctuation in the quadrature-phase amplitudes. This is what we get
field is the maximum, i.e., Ger
by Eq. (3.87). We know that for a sine if we make a homodyne detection (see Chapter 9 for more). On the ot -
function, it is most sensitive to phase change at zero crossing so its zero her hand, if we measure photon number as in photon counting technique
position is used to determine the phase. But the phase uncertainty at zero (see Chapter 7), the behavior of the squeezed coherent state is completely
position in Fig. 3.4(c) is larger than that in Fig. 3.4(b). Therefore, the phase different.
80 Quantum Optics For Expe,-i.nientalists Quantum States of Single-Mode Fields 81

In order to find the photon number behavior , we need to write the with ,8 = a/~- For moderate r ?:, 1. the shape of P11 (/3) depends
squeezed coherent state in the number state base: mostly only on /3 and n. In Fig. 3.6(a), we plot P11 as a function of ,8 with
I - r, a) = S(-r)D(a)IO) = L SnIn), (3.88) r = 2 for n = 5(i), lO(ii), 20(iii). It shows an oscillatory behavior with the
71 number of oscillation given by [(n + 1) /2]. This behavior is attributed to a
where Sn= (nlS(-r)D(a)IO) = (nlSt(r)la). Here , for the ease of calcula- multi-photon interference effect that will be the subject of Chapter 8 (See
tion and interpretation, we consider the squeezed state with squeezing in Eq. (8.122) of Problem 8.5).
X-direction. This coefficient can be calcu lated from the following quantity Further oscillatory behavior occurs in Pn as a function of n, as shown in
Fig. 3.6(b), where we plot Pn as a function of n for r = 5 and ,8 = 20. This
/2(alS(r)l/3) = el/312
el/312 /2(alS(r ) (e- 1/312
/2 " ,Bn In)) oscillatory behavior is a result of quantum int erference in the X-Y phase
L..,,
n
JnT
space [Schleich and Wheeler ( 1987)]. Interestingly, the interference effect
(3.89) in Fig . 3.6(a) can also be exp lained as interference in the X-Y phase space
[Schleich (2001)].
by expa nding the left-h and side in the series of f3. \a IS(r) I/3) is obtained
from Eq. (3.139) in Section 3.6 with the operator algebra method as
1
(alS(r )l/3) = - exp [~( a* 2 - /32) + a*,8 - ~(la l2 + 1/31
2)]. (3.90)
Jµ 2µ µ 2
Using the generating fun ct ion of Herm it e function:
tn
exp(2xt - t 2 ) = "Hn( x )- , (3.91)
L...,, n!
n
we can write
Fig. 3.6 The oscillatory behavior of photon number probability Pn of the squeezed
exp [- ~ /32+ a*,8] =
2µ µ
L ~Hn(~)
n n!
(!3V~)n
2µ
(3.92)
coherent state: (a) Pn(/3) as a fun ctio n of /3 = a/~ for,= 2 orµ= 3.76, v = 3.63
with n = 5(i) , lO(ii), 20(iii) a nd (b) Pn(/3) as a function of n for /3 = 20 and,= 5.

Combining Eqs. (3.89), (3.90), and (3.92) , we have

2
1 ( v ) 1
+ va2µ ) Hn ( a ) 3.5 Mixed Quantum States
n/2 ( 2
(3.93)
Sn=~ 2µ exp - 2 ial ,
Stoler was the first to derive this coefficient in unnormalized form [Stoler In classical physics, if we know the initial states of a system, we can find
(1970)]. The absolute value square of this gives the probability of n photons the evolution path for the system through its Hamiltonian. Similarly in
in the squeezed coherent state: quantum physics, the initial states of a system uniquely determine the
2 evolution of the system via its Hamiltonian. The same determinism and
2 e-1°1 (v)n
Pn = lsnl = µn! 2
µ
2
2 2
exp [v µ (a + a* ) Hn a . (3.94) ]I ( )l2 causality principles apply to both classical and quantum systems. The
difference between quantum and classical physics is in the description of
The photon number distribution of the squeezed coherent state shows some
the states of the physical system . On the other hand, if the system is very
interesting oscillatory behavior as a function of both n and a. To see this
large and contains many degrees of freedom, it is impossible for us to know
and for the ease of numerical calculation, we take a as real and positive.
every detail. Particularly when a large system contains many identical small
Then Eq. (3.94) becomes
systems such as an ensemble of gas system consisting of many identical

Pn = ---
2
e!3 /(µ+v)
2
(v)n
- e-/3
2 1
Hn( /3)
12
(3.95)
molecules, we are unable to know the exact states of each molecule . But
µn! 2µ since the number of the small syst em (the number of molecules for gas
 >
Quantum Optics For Experimentalists
82
system) is usually quite large (rv 1023), we can use statistical methods
to describe them. The identical nature of the small systems allows us to
describe each of them as a sample in statistics. The whole collection of the
samples is known as an ensemble in statistics and it is just the large system
(the gas system). Although the states of each sample are undetermined or
unknown to us , we can use probability distribution to describe the ensemble.
The state of the large system can be described as the statistical average
of the ensemble of the small systems. This is classical statistica l physics.
The statistical fluctuations of the states of the small systems may lead
to the uncertainty of the state of the large system . This is the origin of
the classical uncertainty. For a large quantum system, we will encounter
similar situation, that is, due to the largeness of the system size, we are
unable to obtain the quantum state for the whole system. In this case, we
can treat the problem by using statistical methods simi lar to the classical
statistical physics, i.e., using the probability distribution of the small system
to describe the large system . This is quantum statistical physics. Here the
uncertainty of the quantum system due to probability distribution is the
same as that in classical statistical physics. But different from a classical
system, a quantum system has intrinsic quantum uncertainty in addition
to the classical uncertainty. So, a large quantum system has both classical
uncertainty due to its many degrees of freedom and the intrinsic quantum
uncertainty.
In Chapter 2, we find an optical field has many modes and thus many
degrees of freedom. If many modes are excited but we are unabl e to know
exact excitat ion for each excited mode, this can lea d to classical unc ertai nty ,
as described above. This is the basis for statistical optics to be discussed
in Chapter 4. Even for a single mode field, the unc erta inty in the em ission
process leads to uncertainty in phase and amplitude of an optical field and
statistical description is needed, as we found in Section 1.3. When the
optical field is treated quantum mechanically, we need to use a quantum
statistical method to describe it. This is the mixed state description of
optical fields.
3.5.1 Density Operator for the Description of a Mixed
Quantum State
Before we discuss mixed states in quantum statistical description , let us
introduce the density operator representation of a quantum state. When a
system is in a definite quantum state l?/J), we say the system is in a pure
Quantum States of Single-Mode Fields 83
quantum state. We can assoc~ate it with ~n operator: Pw = l?/J)('lfJI,~1ai~ed
as the density operator. Tl11s operator 1s also known as the proJection
erator because its action on any state projects it onto this l?/J)state as
0
/result: p.~Jlc[>)= ((?J;lw))l?J;)with l(?J;lw)l2 as the projection probability.
Since it is a projection operator, it must satisfy fJ!
= Pw· This is the
cwl l ·acteristic of the density operator of a pure state. For the system in this
ft
state, the measurement of a physical observable with operator O will result
in an expectation value as
(O)w = (?J;IOl?/J)= Tr((?J;IOl?J;))= Tr(l?/J)(?J;IO)= Tr(fJwO). (3.96)
So, any quantum state l?/J) can be equivalently represented by a density
operator Pw= l?/J)(?/JI-
Now let us discuss mixed states. Consider two quantum states l?/J1),l?/J2)
for a system. Suppose that we do not know due to some reason which state
the system is in but we know that the probability is P1 for the system in l?/J1)
and p 2 in j?J;2) with p 1+p 2 = 1. Taking the polarization states of light as an
example, let us consider two independent beams of light with polarizations
of x and i), respectively. Now combine the two with a 50:50 beam splitter,
as shown in Fig. 3.7. What is the polarization of the combined beam? It
cannot be :i; or i) for sure. Can it be 45°-polarized? But then why can't it
be 135°-polarized? We know that both 45° and 135° polarization states are
superposition of x and i) polarizations and require a fixed phase between
:i; and i) polarizations: E45 = (x + i))v12, E135= (x - i))vf2. It turns out
that all other polarization states such as circu larly polarized states require
fixed phase relation between x and i) polarizations. But there is no fixed
phase relation between two independent fields. Therefore, the combined
beam after the beam sp litt er (BS) does not have a definite polarization
state. Since we use a 50:50 BS , we can claim that it has 50% of probability
of being x-polarized and 50% of probability of being i)-polarized. This
is a statistical mixture of x and i) polarizations and is a mixed state of
polarization.
Fig. 3.7 Mixed polarization states: ind epe ndent i:-polari zed and y-polarized fields are
comb ined with a 50:50 beam splitter.
 84 Quantum Optics For Experimentalist ,s Quantum States of Single-1\Iode Fields 85

Let us next look at the result of a polarizer on this light field of mixed li/'1 ). 14,2 ) and give rise to quantum int erference phenom ena. The mixe_d
polarization. Assume the polarizer has a passing direction of 0 relative state Prndoes not produce such coherence terms. So, the pure state Ii/!) 1s
to .r and the intensity of the combined field after the BS is Io. After the called a quantum superposition state of 1'1/;i), l?h) whereas the state Pmis
polarizer, the contribution from the i:-polarization is Io cos 2 0 while that a classical statistical mixed state of Ilf!1).Ii/J2).
from they-polarization is Io sin 2 0. Since both have 50% probability due to A typical example is the Schrodinger cat state discussed in Section 3.3.
the BS , the total contribution is In this case . l'lf1), l1P2) are two mutually exclusive macroscopic classical
states, i.e., ldead) or !alive). As we just mentioned , the cat state l'lfJcat)=
l out = 10 cos 2 0 21 + . 2
JOSlll 0 21 = JOI 2. (3.97)
X X
(!dead) + !alive)) /-/2 will produce some interference phenomena whereas
Hence, the int ensity after the polariz er is independent of the ang le 0 of the the mixed state ,Om=(ldead)(deadl + lalive)(a live l)/2 will not. Mixed state
polarizer. This result is the same as that for a natural light. On the other is what we see in the classical world since we do not see interference
P1n
A

t
ha~d, for 45°-polari zed light , the output intensity is J4 = J0 cos 2(0 - 45), effect in our daily life.
which depends on the angle of the polarizer. The transition from a quantum Schrodinger cat state l1Pcat)to the clas-
Now we come back to the case of the mixed state of two quantum states: sical mixed state Pmis a de-coherence process caused by the interaction of
I1P1), i'1/J2). Similar to the case of polarization measurement, the result of the system with the env ironment. We will see in Section 6.3 of Chapter 6 an
the measurement on the physical observable 6 is examp le of photon dissipation in a single cavity mode through its coupling
to an outside vacuum environment with a continuous spectrum of modes.
(0) = (0)1 1 X Pl+ (0)1 2 X P2 This dissipation mechanism is what causes the loss of the coherent cross
= Pl (1P1IOl1j;1)+ P2 (1P1IOl1j;2) terms in the pure state density operator in Eq. (3.99). The more degrees
= Pi Tr(,81 1 0) + p2Tr(,81 2 0) of freedom the system is coupled to, the faster the de-coherence is. So,
= Tr(,omO) , the largeness of a macroscopic system makes it coupled to more degrees of
(3.98)
freedom and thus more likely to lose the quantum coherence than a small
where Pm= P1l1P1)(1P1I+ P2l1P2)(1P2Iis the density operator for the mixed microscopic system. Eventually, a quantum system quickly de-coheres and
state of I'lj;1) , I'lj;2) .
becomes a classical system.
On the other hand, for a pure state: j?j;) = ffel?j; 1) + v'P2i1P2), it also In the more general case, a quantum system may be in any one of n
has P1 probability in state l1P1)and P2 probability in state j?j;1). What is the different quantum states. Suppose the n different states are labeled as
difference between this pure superposition state and the mixed state Pm= l'lfJ
1), l?j;2), ... , l1Pn)- Note that these states may or may not be orthogonal
P1 l1P1)(1P1I+ P2 l1P2)(1P2I? For comparison, we write the density operator for to each other. If we are not certain which state the system is in , we can
the pure state:
describe them with probability: the probability is P1 in state l1P1), P2 in
P1 = (5i(1P1I + v'P2(1P21)(5il1P1) + v'P2i1P2)) state l1P2), ... , Pn in state l1Pn) with P1 + P2 + ...+ Pn = 1. If we make a
measurement of physical observable 6 in this system, the expectation value
=Pm+ vPlP2l1P1)(1P2I+ vPlP2l1P2)(1P1I- (3.99)
of the measurement is
It has two more cross terms than the mixed state Pm· If we make a mea-
(0) = Pl (1P1IOl1j;1)+ P2 (1P2IOl?j;2)+ ···+ Pn \1PnIOl?j;n)
surement of physical observable 6, the results for the two states are n

(0) P,µ = Pl (1P1IOl1j;1)+ P2 (1P1IOl1j;2) = LPj\ 1PjlOl?j;j), (3.101)

j=l
+vPlP2( 1P2IOl1P1) + vPlP2(1P1IOl1P2) With the introduction of density operator :
(O)Pm = P1(1P1IOl'lj;1)+ P2(1P1IOl'lj;2). (3.100) P= P1l1P1)(1P1I+ P2l1P2)(1P2I+ ···+ Pnl1Pn)(1Pnl
n
Again, the pure state l1P)has two more cross terms known as the quantum = LP}l1Pj)( 1Pjl, (3.102)
coherence terms, which keep a fixed phase relation (coherence) between j=l
•

86 Quantum Opti cs For E.Tperim ental ist s Qu antum Stat es of S ingl e- Mod e F ie lds 87

Eq. (3.101) becomes Substituting the number state representation of the coherent state as in
Eq. (3.25) into the expression above and carrying out the phase integral ,
(0) = Tr(pO). (3.103) we obtain
The mixed polarization state discussed earlier can be obtained from the
/y) )/ v0. If the two incident fields
general polarization state /E¢) = (/x) + ei<P
Pla se r = I: Pm /m) (rn / , (3.106)
1n
of x and y polarizations are independent , the phase ¢ is random , i.e., it has
2
a uniform probability density p( ¢ ) = 1/ 21r in [O,21r] range. So, Eq. (3.102) where Pm = (/0:/2 )m e-la 1 / m! is the Poisson distribution, which is the same
gives the density operator for the mixed polarization state as as the photon number distribution of the coherent state.
(2-rr
Pp= lo d¢ p(¢)/t:¢)(t:¢/ = (lx ) (xl + /y)(yl) / 2, (3.104) 3.5.3 Density Operator for a Thermal State

which properly describes the respective 50% probability for x and y pola- Thermal light sources are the most common light sources in nature. The
rizations for the combined light field. quantum state describing this kind of sources is called a thermal state. Any
Next, we will discuss some commonly encountered mixed state of light. object with a finite temperature is a radiation source of electromagnetic wa-
ves and the emitted electromagnetic waves have the same photon statistical
3.5.2 Density Operator for Lasers with Random Phases property as the thermal state.
For a simple harmonic oscillator corresponding to a single-mode optical
The coherent state is a quantum state derived from theory. Besides quan - field, when it reaches thermal equilibrium with a heat reservoir of tempera-
tum noise, it has a well-defined amplitude and phase, which is the closest ture T, it is in the thermal state. The density operator for the description
quantum counterpart for a classical monochromatic infinite electromagne- of the thermal state can be derived from statistical quantum mechanics as
tic wave train. The output of a single-mode laser is a lab -produced light
source that is closest to a coherent state. But the phase of a single -mode
A
Pth = e -{3iI;z , (3.107)
laser is not completely fixed. Due to the inevitable spontaneous emission
where f3 = 1/kT with k as the Boltzmann constant, fI = nw(a)a+ 1/2),
with a random phase, the phase of a single -mode laser can diffuse with
and z = Tr(e-f3iI). In the number state representation, we have
time. The characteristic diffusion time determines the upper bound on the
coherence time of the laser and leads to the Schawlow-Townes line width of
a single-mode laser [Schawlow and Townes (1958)], which is of quantum na -
Pth= e-~H ( L ln)(nl)/z = L (e-~flln)(nl)/z
n n
ture .1 The other technical imperfection can result in broader line width and
shorter coherence time. In a time interval much smaller than the coherence
= ½I: e-f3nw(n+l/2) /n) (n/
n
time of the laser, its phase diffusion is negligibly small and the output of the
laser can be approximated as a coherent state . But for a time interval much (3.108)
n
longer than the coherence time, the diffused phase ranges from Oto 21r, i.e.,
the phase of the laser output can be comp lete ly random with a un iform with ( = e-n/kT . The average photon number of the thermal state is
probability distribution in the range of [O,21r]: p(<p)= 1/21r(O :s;<p:s;21r).
Therefore, similar to Eq. (3.104), we have its density operator as n = (n) = Tr(atCLPth)= 1/(enw/kT - 1). (3.109)

(2-rr
d From Eq. (3.108), we obtain the photon statistical distribution of the ther-
Plaser = lo 2
://a/ei<p)(/a/ei'P/. (3. 105) mal state :
--:--------------
1 In Section 8.3.1, we will present a simple derivation of the Schawlow-Townes line width P = (l _ ()(n = (l _ e-n/kT)e-nn/kT = fin (3 .110)
based on photon indistinguishabi lity in multi -photon interference. n (1 + n)n+l
 >
Q'llantum Optics For E:cperimenta Lists
This is exactly the Bose-E instein statistie:; of the blackbody radiation. From
this, we CRn find
(3.111)
71
which can be rewritten as
(3. 112)
Here a,t2a 2 = : fi,2 : is the normal ordering operation on n 2. In Chapter
5 when we discuss about photodetection, we will find the intensity of the
optica l field is proportional to the average photon number: (I) = rJ(n) (ry
is the quantum efficiency of the detector) and the auto -corr elation of the
intensity is proportional to(: n 2 :) , i.e., (12 ) = rJ2(: fi 2 :). From Eq. (3.112),
we obtain
9 (2) = (12)1(1)2 = (: n2 :)/ (fi)2 = 2 , (3.113)
2 between quantum and classical worlds.
where g( ) is the normalized second -order intensity correlation function.
It is the intensity correlation function measured in Hanbury Brown -Twiss
experiment. So, the value of 2 for g( 2 ) in Eq. (3.113) gives a direct quantum
explanation of the photon bunching effect observed in Hanbury Brown -
Twiss experiment. In Chapter 1, we provided the classical wave explanation
of the photon bunching effect. The explanation here is directly from the
photon statistics of the thermal state and is the most direct explanation of
the photon bunching effect.
From Eq. (3.111), we can find the photon number fluctuation of the
thermal state:
(!).2fi) = (n2) - (n)2
2
=fi+n . (3.114)
The expression above was first derived by Einstein in 1909 [Einstein ( 1909)].
The formula he derived is about the energy fluctuation of blackbody rad i-
ation :
(3.115)
If we use Planck equation E = nnwfor the energy of blackbody radia -
tion, the two terms in Eq . (3.115) correspond exactly to the two terms in
Eq. (3.114) (the extra factor Z is due to multi -m ode cons ideration) . In
discussing the physical meaning of the two terms in Eq. (3.115), Einstein
discovered that the first term is due to the particle nature of the radiation
whereas the second term originates from its wave nature. He derived the
Quantum States of SingLC'-l\Jode Fields
first term by using Poisson statistics of random particle. Ind eed. Poisson
distribution of the photon number of the coherent state leads to this term,
a , shown in Eq . (3.28). For the second t erm, Einst ein was able to deriv e it
by using the plane wave mod el of th e radiation. Indeed , we ca n see from
Eq. (3.112) that this term gives rise to the photon bunching effect in HBT
xperirnent and as we showed in Section 1.6 of Chapter 1, wave nature al-
one can explain the photon bunching effect. Furthermore , we will show in
Section 8.3.1 of Chapter 8 that photon bunching effect can also be explai-
ned by multi-photon constructive interference principle of quantum waves,
which indicates the wave origin of the second term.
From Eq. (3.114:), we can see that, when the average photon number is
much smaller than 1, the first term dominates. In this case, the optical field
exhibit s particle nature (quantum). But when the average photon number
is much larger than 1, the second term dominates , leading to a classical
wave field. This is consistent with our intuition about the relationship
The thermal sources from blackbody radiation have a wide spectrum.
Besides blackbody radiation, other methods to obtain a thermal source in-
clude the light emitted from ionization processes in hot vapor of atoms such
as hydrogen lamps and mercury lamps and the spontaneous emission from
optical amplifiers such as erbium-doped fiber amplifiers and parametric am-
plifiers. In ear ly experiments, atomic ionization is very common such as the
source used in the HBT experiment. After the invention of lasers, a com-
monly used thermal source is a kind of quasi-thermal light source produced
by passing a laser beam through a fast rotating grounded glass plate, as
shown in Fig. 3.8. As the grounded glass plate rotates, the laser beam hits
different positions of the grounded glass. But the surface of the grounded
glass is very rough and scatters incident light to all directions randomly.
So, the amplitude transmissivity coefficient t at different place is a random
complex number, i.e., its amplitude and phase are both random variables:
rotating
grounded
glass
~-- -tQuasi-thermal
Source
Fig. 3.8 The schematic for producing a quasi -therma l light so urce by rotating a groun-
ded gla ss plate with a coherent state illumination.
 •
90 Quantum Opt ics For Experirnental is ts Quantum Stat es of Singl e-Mod e Fie lds 91

t = itlei'Pt. \i\Then the rotating sp eed is fast , th e obs erv ed transmi ssivity is Appl ying th e abov e to a and at , we hav e
an averag e of a large number of ran<lom variables from different locations . D( a) = ea.at -0, * a = e- 0,* aea.at ef0,f2/ 2 = ea.at e-n*a e-lal2 /2 . (3.120)
We learned from the central limit theorem in Section 1.4.3 of Chapter 1
that the probability distribution of averag e transmissivity t is approxima- Thi s is th e normal ord ering form of th e displac em ent op erator D (o,).
2 Th e mor e gen eral theor em is
ted by a Gaussian distribution: P(t) = e-l t l / a; f'rra'f. If the incid ent laser
beam is described by a coherent state , Iao), the instantaneous state of the 8
1
e A e- B =A+ [B , Al + [B , [B, Al]
transmitted light is also a coherent state of !tao ) . But for a tim e period 21
1
long enough, we take t as random variabl e with a Gaussian distribution + ...+ -dB. [B, · · · [B, A]··· ll + .... (3.121)
n.
and the transmitted state is a mixed state of different coherent states with
a density op erator: Applying the above to D(a) and ii , we hav e

(3.116) bt (a)ab(a) =a+ a. (3.122)

2
Applying to the squeezing operator S(r) = exp[r(at2 - ii )/ 2l in Eq. (3.63)
where d 2 t = dRe(t)dlm(t). Substituting the number state expansion of the
and ii , we have
coherent state in Eq. (3.25) and carrying out the integration oft. we have
(a;laol2 )n st aS = a cosh r + at sinh r'
A

PQth = L
n
(l + a;laol2)n+l ln)(nl. (3.117) stat S = at cosh r + a sinh r, (3.123)

The state above is in the same form as the density operator in Eq. (3.108) which is just Eq. (3.62). So, we proved that the squeezing operator S(r)
for a thermal state with an average photon number (n) = a;la0 1 2
. The in Eq. (3.63) can give rise to the required relations in Eq. (3.62). But the
reason we call this type of source a quasi-thermal source is because in the argument in Section 3.4.1 requires the reverse, i.e., derive the squeezing
integration in Eq. (3.116), the range of !ti is from O to oo but in reality, operator S(r) in Eq. (3.63) from the relations in Eq. (3.62). For this, we
!ti only takes values from O to 1. Only when at « 1, this approximation need a different approach.
is true. Moreover, the transmissivity is periodic for a uniformly rotated Suppose operator S(r) satisfies relations in Eq. (3.123). We see that
grounded glass. The average transmissivity at depends on the roughness of S(O) = i, the identity operator. For infinitesimal r = 8r « 1, we can make
the surface and the speed of rotation. The coherence time of the thermal a linear expansion of S(r) in terms of 8r:
source is determined also by the speed of rotation. s(r) i + i8rf(a,at) (3.124)
where f(a, at) is some function of ii, at to be determined. Since S(r) is
3.6 Further Reading: The Operator Algebra of ii and at
unitary, i.e., st(r)S(r) = S(r)St(r) = i, substituting Eq. (3.124) into
All the physical observables of optical fields can be written in terms of the these and keeping only the linear terms in 8r, we have ft = f, i.e., f is a
operators ii and iit . Since a and iit do not commute, the commutators of Hermitian operator.
the physical quantities of the optical fields highly depend on the behavior Next , write Eq. (3.123) for infinitesimal r = 8r « l:
of ii and iit. In this section, we will discuss about how to handle functions (i - i8r f)a(i + i8r f) =a+ 8rat, (3.125)
of ii and at.
or,
Let us start with Campbell -Baker -Hausdorff theorem. For two non -
commuting operators A,B with a+ i8r[a, fl =a+ 8rat, (3.126)
[A,[A,Bl] = [B,[A,Bl] = o, (3.118) where we only keep up to linear terms in 8r. Comparing the two sides of
then the above, we have
(3.119) [a, fl = -iat. (3.127)
92 Quantum Opt ics For Expe1·inientalists Qnantnm States of Singl('-J\Jode Fields 93

Similarly , Herc. we divid e both sides by (al,8). Take a trial solution of the form:
[aAt, fl -- -1 •A
.a. (3.128) 5( 11)(0,*./3,r) = eG(a* ,(3,r) with G(a*.,8.r) = A(r) + B(r)a *2 + C(1·),82 +
1, D (r)o_*,8. Here , A(r). B(r) , C(r). D(r) are to be determined. Substituting
In Problem 3.1 , we can prove [a, &,tn] = n&,tn- [at, a ] = - n&,n- i _ It 11

into Eq. (3.135) and carrying out the derivatives , wc hav e the right side as
is stra ightforward to extend these two relations to any ordered function of
a,&,t: f (at' a) = L c1;;,a)k&,l as follows
A !( a,a
At A)] --,:i of [At of
right = t[ 2
c1:* 2
- ,8 - 4B ,80:* - 2D ,8
2

[a, a , !( a.a
At A)] -- - (3.129)
-2B - (2B0: * + D ,8)2] s C11 \0:*. ,8, r).
A ' ' ,:iA·
uaT ua (3.136)
Hence, Eqs. (3.127) and (3.128) become
of _ •At of _ •A Comparing both sides of Eq. (3.135), we have
oat - -ia ' 8a - ia . (3.130)
Apart from a c-number, we have the solution off as dA = - B dB = - 2B2
dr ' dr 2 '
f(at ,a) = i(a 2 - at2)/2. (3.131)
dC = -~(D + 1)2, dD = -2B - 2BD. (3.137)
So, S(or) = i - or( a2 - at2) /2. For a finite r, we can divide r into N small dr 2 dr
or's with Or = r IN: The initial conditions are A(O) = B(O) = C(0) = D(O) = 0. We can solve

S(r) = [S(r/N)]N = lim

N-+=
[i- !.._(a
N
2
- &,t2)/2] N
B immediately as B(r) = (1/2)tanhr,
D(r) = 1/coshr-
then A(r) = -(1/2)ln(coshr), and
l, and C(r) = -B. The final form of s(n)(0:*,,8,r) is
--+ exp[-r(a 2 - &,t2)/2], (3.132) 1
s(n\0:*,,8,r) = - - exp [~(0:* 2 - ,82) + (~ -1)0:*,8], (3.138)
which is exactly the squeezing operator in Eq. (3.63). fo 2µ µ
The relations in Eq. (3 .129) are especially useful to calculate the nor-
mally ordered form j(n)(&,t ,a ) of some function f(at,a) of at,a. Here whereµ= cosh r, v = sinh r. With this, we have
normal ordering is such that at is moved to the left of a. For exam- (0:IS(r)l/3) = s(n)(0:*,,B)(0:l/3)
ple , the normally ordered form of fi 2 = (ata) 2 is fl, 2 = &,t2&,2 + &,ta.
1 [ V 0;* /3 l 2 + l/312)] . (3.139)
We will discuss more of normal ordering and its applications in Chap- = - exp -(0:* 2 - ,82) + - - -(10:1
fo 2µ µ 2
ter 5. The usefulness of the normally ordered form is its action on co-
herent states: (0:lf(n) (at, a) 10:) = j(n) (0:*, 0:). To see how we can make
3. 7 Glauber-Sudarshan P-Distribution and Wigner
use of Eq. (3.129), we consider the squeezing operator S(r) and calculate
Distribution
(0:IS(r)l/3) = s(n)(0:*,,B)(0:l/3).
Take a partial derivative of S(r) with respect tor:
As we have seen in Section 3.2.5, the coherent states form an over-complete
8S(r) -- -l( aAt2 - aA2)SA(
-- r )-- -a
1At2sA()r - -a
l A2SA()
r . set of bases with which any arbitrary state can be represented. We can
(3.133)
or 2 2 2 do the same for any operator 6 by using the completeness relation in
Using relations in Eq. (3.129) twice, we have Eq. (3.47):
as( s(
A2 A A A2 r) A 8 2 r)
a S(r) = S(r)a + 2 oat a+ oat2 . (3.134) 6 = J d2,8l,8)(,BIO
n
J d20:10:)(0:I= / d20:d2,8
n 2
(,81O10:)l,8)(0:I. (3.140)
n
Now take the normally ordered form s(n)(&,t,a) for S(r) in Eqs. (3.133)
and (3.134) and apply (0:I from the left side and 1/3)from the right side: (,8IO I0:) is then the coherent state representation of operator 6. For ex-
ample, the coherent state representation of the squeezed operator S (r) is
as(n)~r*,,8,r) = 1(0:*2-,82-2,38!* - a!:2)s(n\0:*,,8,r). (3.135) derived in the last section and is expressed in Eq. (3.139). The connection
94 Qu an tum Optics For Experi m,Pntal-ists Quantum States of Single-Mode Fields 95

bet ween th e coherent st at e repr esenta ti on and the numb er st ate repr esen- T he act ual form of fun ct ion P (Cl) ca n b e derived direct ly from t he
tat ion can be eas ily deriv ed as follows: density op erator ,o [l\Iehta and Sud arshan (1965)] . Consider t he funct ion

= (/31( L In) (nl ) 6 ( L Im) (ml ) In) (- /31,ol/3)of compl ex vari able /3. From E q. (3.142), we have

J
(/310 la)
n m 2
(-/31,ol/3) = d a P (o-)(-/31n) (al/3)

ni,n
(3.143)
= e-(lal2+l,8l2)/2 LO ri:m, j3*nam
or
m,n vm1nT
(3.144)
= e-(lal2+l,Bl2)/2fo (/3*,a), (3.141)
wh ere 0 nm = (n lOlm ) is th e number stat e repr esent ation.
, i\Trit ing out t he compl ex vari abl es in ter ms of th e real vari ables : /3 = u +
iv, o = x + iy , we h ave e,Ba*-,B*a= e2i(xv-yu) and Eq. (3.144) b ecom es
If 6 is a tra ceabl e, non-neg ativ e definite Hermitian operator , IOnml
is bounded. Th en th e double infinit e series fo( /3*,a ) in Eq . (3.141) is
absolutely convergent for all /3* and a and is thus an anal ytic function of
(- /31,ol/3)e1?+v2 = J dx dyP( x, Y)e-(x2+ y2)e2i(xv-y u). (3.145)
2 2
the complex variables /3* and a . Indeed , if we writ e op erator 6 in th e So, function J( u , v) = (-/3 l,ol/3)eu +v is just th e Fouri er transformation of
2 2
normally ordered form of 6 (n ) (at, a), then (/310la ) = o (n ) (/3*, a) (/31a) and funct ion g(x, y) = P( x, y) cCx +Y ). The revers e transformation gives
fo( /3*,a) = Q(n)(/3*, a) e,B*a,which is an analytic function of /3* and a for
a well behaved Q(n)( /3*,a) .
g(x , y) =
7r
J dud v f(u , v )e- 2i(xv-y u) (3.146)
According to the theory of complex analysis, the analytic function or
fo(/3 *, a) can be determined everywhere from some arbitrarily small range (3,147)
of /3*, a. Particularly, if we know the value Jo(a *, a), fo( /3*, a) is th en en-
tirely determined. Therefore , (/310la ) can be uniquely determined by th e We have then expressed P(a) in terms of the elements (- /31,ol/3) of the
diagonal elements (a IO Ia) . density operator ,o.
Furthermor e, since the average of any normally order ed op erator mo-
3.7.1 Glauber-Sudarshan P-Representation of Density ment : a,na,tm : = a,tma,n has the form of
Operator (: an a tm :) = (a tm an) = Tr [,oatm an]
We can apply the above to the density operator for the description of the
quantum state of a system. From Section 3.5, we find that the density ope-
= J d 2aP(a)a*rnan = (a *rnan)P , (3.148)

rator satisfies Tr,o = 1 so it is traceable and it is a non-negative Hermitian we can define the characteristic function C N ( u , u *) for the normally ordered
from its definition in Eq. (3.102). Therefore, we should be able to write it moment : a,na,trn : = a trna,n as
in the form of the diagonal coherent state representation: CN(u,u *) = (: eua, t - u*'a:)= (eua-t e- u*'a)
p= J d 2aP(a)la)(al , (3.142) = Tr[,oeuat e-u *a] = J d2aeua *-u* aP(a). (3.149)

where P(a) is some real function of complex variable a because ,o is a The average of any normally ordered moment (: a,na,trn:) = (atma,n) can
Hermitian operator. This is the Glauber-Sudarshan ?-Representation first be expressed as
introduced by Glauber [Glauber (1963c)] and Sudarshan [Sudarshan (1963)]
independently. It plays a key role in the quantum coherence theory to be (3.150)
discussed in Chapter 5.
 9G Qu ant um Opti cs Fo1· Ex pe1·i m ent alist s Quantum S tat es of Sin gle- l\l ode Fi elds 97

where we treat u, u * as t,vo independent variables. Since e 1rn *- l! *ct is the (c) muber states
kernel of the two-dimensional Fourier transformation, the inverse-Fourier
transformation of Eq. (3.149) has the form of For th e number state of p = Jn )( nl ,
2
(-8J p J,8) = (- ,Bln )( nl ,8) = e-l!Jl (-181
2
)71/ n!.
Substituting it into Eq. (3.14 7) , we obtain
(3.151)
P(a) e-lal2 = -.1-jd2 ,B( -l,B J2)11e/3*a- 1fo* , (3.155)
7r2n!
It is easier to use Eq. (3.147) for P(a) than Eq. (3.151) because th e calcu-
lation of (- ,Blpl,8) is easier than CN(u, u*) in most cases. which is divergent in the normal sense. But we can express it in terms of
Let us find P(a) for some commonly encountered states: the special <5-
function as
2
(a) Coherent state P(a) = _e lcxl
____
32n 1 / ef3*a-B a * d2 ,B
n! 8a *n3eyn 1r 2
For a coherent state of Iao),
2 lcxl2 32n
(- ,Blpl,8) = (- ,Blao)(aol ,8) = e- lcxol-l !Wef3cx*- (3*a_ e ___ 5(2)(a). (3.156)
n! 8a*n3eyn
From Eq. (3.147), we have
So, P(a) is the 2n'th derivative of a <5-function, which is even more singular
P(a)e-lcxl
2
=
7r2
1
d2,Be-lcxol2ef3*(cx-cxo)- (3(cx* -cx~)
than a <5-function. This will lead to nonclassical behavior for the number
states (see Section 5.2 .4).
= e-lcxol2 / d2,Bef3* (cx-cx0 )- (3(cx* -cx~)
7r2 3. 7.2 Wigner W-Representation of Density Operator
= e-lcxol252(a - ao). (3.152) The Wigner distribution (also called the Wigner function) is a quasi-
2(a - a ) for the coherent state la ). probability distribution for position and momentum of a particle. It was
So, P(a) = <5 0 0 introduced by Eugene Wigner in 1932 to study quantum corrections to clas-
(b) Thermal state sical statistical mechanics [Wigner (1932)]. Specifically, it is defined for a
particle in a pure state of wave function 'l/J
(x) as
For a thermal state of average photon number n, Pth = Ln Pn In) (nl with
Pn = nn/(n + 1r+ 1 . So, we have W(x,p) = -1 /_00 .
dy'lj)*(x+ y)'l/J(x - y)e- 2ipy/n
1rn -oo
( -,Bl Pth l,8) --
A nn
(n + l)n +l
(-1r
n!
2n -1/312
l,BI e = _1 /_00 .
dqcp*(p+ q)cp(p_ q)e-2ixq/n, (3.157)
n 1rn -oo
where cp(p) is the Fourier transformation of 'l/J(x) and x,p are the position
and momentum but could be any conjug ate variable pair. For a mixed state
of p, it becomes
= ~- 1/312
n+l
exp (- nl,812).
n+l (3.153)
W(x,p) = - 1 loo
1rn -oo
dy(x + YIPlx - .
y)e- 2ipy/n
Substituting
we obtain
the above into Eq. (3.14 7) and carrying out the integration ,
= ]__1
1rn
00

-oo
dq(p + qlplp _ q)e-2ixq/n_ (3.158)

P(a) = ~e-lcxl2 /n_ It is a quasi-probability distribution of the quantum system in x - p phase

(3.154)
Hn space in the sense that W is normalized to 1: J dxdp W (x , p) = 1, which
 98 Quant'llrn Optics For Experim ,entalists Quantum States of Single-Mode Fields 99

can be shown from Eq. (3 .158): The second one is that the "\i\Tigncr function can become negative for

I
.
dxdpTV(x,p) = J dxdy(x + YIPl~r- y)
wh
j dpe-2ipy/ri
some values of x, p for some states, which can be shown from Eq. (3.162)
for orthogona l states Tr(,8 1,02) = 0:

= J dxdy(x + yjpjx - y)5(y) 2wn I dxdpHl,oi(x,p)W,a2(x,p) = Tr(p1,02) = 0. (3.164)

= f dx(~rl,ol.i:)= Tr,o = 1, (3. 159)

which means eith er TV,a1 (x,p) or Wh(x,p) must be negative for some x .p in
order for the expr ession above to stand. So, like the Glauber P-distribution,
and the marginal distributions are x and p probability distributions, which the Wigner distribution cannot be a true probability distribution.
can be shown from Eq . (3.157): The third one is that unlike the Glauber P-fun ct ion , the Wigner function
is bo und ed, which can be shown from Eq. (3.157):
j dpW(x,p) j dy,j,'(x + y),j,(x - y)tJ(y)
= = l'P(x)I'
lw(x,p)I::; wln I dyl'lj;*(x + y)'lj;(x - y)e-2ipy/hl
I dxW(x,p) = I dq,j,'(p + q),j,(p - q)tJ(q) = l,P(p)I' (3 .160)
= wlnfdyl'lj;*(x+y)'lj;(x-y)I
for pure states and from Eq. (3.158):

(xlplx) = j dpW(x,p), (PIPIP) = j dxW(x, p) (3.161)

:S :n f dyl 'lj;*(x+ Yf f dyl'lj;(x -
2
y)l
1
for mixed states. (3.165)
There are some interesting properties for the Wigner function. The first
wn'
where we us ed the Cauchy -Schwarz inequality for the last inequality above .
one is the overlap of quantum states by the product rule:
The Wigner function has practical significance in quantum optics . As we
Tt(,01p2) = 2wn I dxdpWPl (x,p)W,a2 (x,p), (3.162)
will show in Section 10.4 , homodyne detection techniqu e can b e applied for
the m eas urem ent of the Wigner function of an optical field by the method
which can be shown from Eq. (3.158) : of quantum state tomography. This allows us to measur e the complete
quantum state for an optical field.
Tt(p1,02) = 2 / dxdy1dy2(x + Y1IP1lx -y1)(x + Y2IP2/x -y2) To prepare for this, let us apply the Wigner function defined in
Eq. (3.158) to the virtual harmonic oscillator for a single -mod e field. We
x - l / dpe- 2ip(y1+y2)/ri change x,p to the dimensionless variables X = xJ2w/mn, Y = pJ2m/wn
wn with m = 1. We need to choose the correct coefficient so that W is norma-
= 2 / dxdy1dy2 (x + Y1/,01/x-y1) (x + Y2/h/x -y 2)5(y 1 + y2) lized: J dXdY W(X , Y) = l. The exp licit form of the Wigner function for
an optical field is then given by
= 2 / dxdy1 (x + Y1!P1/x - Y1)(x - Y1/p2/x + Y1)
W(X , Y) = 1
du(X + u/p/X - u) e-iu Y , (3.166)

= J dz1dz2(z1/P1/z2)( z2/p2/z 1)
2w
which can be expressed in a more familiar form in quantum optics:

= f dz1 (z1 IP1P2/z1) = Tr(i31P2), (3.163)

W(X Y) = -
'
1
(2w)2
-f
dudvC (u v) e-iv X+iuY
w '
(3.167)

where we changed integration variables x + y 1 = z 1, x - y 1 = z 2 and us ed with

relation J dz2 /z2) (z2 / = 1. (3.168)
 Qnanlnm Optics Fo'I· E:1:perimentalists Quantum States of Single-Mode Fields 101
100

where 17= u + i V and X = a + at, y = (a - a))/i. nncl calculate the variance (6. 2.ri) = (I·7) = fd.r1d.r2I}TFCf'1.,r2) = l.
To prm'e the expression in Eq. (3 .16 7) is the same as Eq. (3 .166), we first So, the distribution has a standard deviation of 1 and the contour of one
express C1du, v) in the base of the eigen-states {I~)} of X (XI~)= ~I~)): standard deviation for the distribution is I·t
+ :f~ = 1, which is a circle of
C1i-('u, v) = J d~(~lpeiv,Y-iu .f'I~) = J d~(~lpeiuue-iu)'eiv_\"I~)
radius 1 centered at (2a 1 , 2a 2 ). We take this circle as the representation
of the size of the quantum noise when depicting the coherent state in X-Y

= J d~ eiv(Uu) (~lpe-iu) ' I~)

phase space (see Fig. 3.1).

(b) Coherent squeezed state ( css)

= J d~ eiv(Uu) + 2u),
(~IPI~ (3.169) A coherent squeezed state defined in Eq. (3.71) has the form of la, r) =
.D(a)S(r)I0) = S(r)la-) given by Eq. (3.80) in Section 3.4.3. Here, a_ =
where we used Campbell -B aker-Hausdorff theorem in Eq. (3.119) with
a cosh r - a* sinh r = a 1 e-r + ia 2er. So, its characteristic function can be
[X,Y] = 2i, and eivX I~)= eiv~I~),e-iuY I~) = I~+ 2u). Substituting the
calculated as
above into Eq. (3.167), we have

W (X Y)
'
= _1
(21r)2
/ dudveiuY -ivX J d~ eiv(~+u)(~IPI~ + 2u)
C vV(u, V) = Tr ( Pcsse ivX -iuY)
A

= (a_ 1st(r)eivX -iuY S(r) la-)

=
2
~ J Jdl(lll'll + J
dueiuY 2u) dv e'"(';:-X)
= (a_ lest (r)(ivX -iuY)S(r) la-)
= (a_ I exp[iver X - iue_ .Y]la-) 1

= ~
2
J J + +
dueiuY d~WPI~ 2u)5(~ u - X)
1
= (a_ leiv X-iu Y la-)
1

J _ -(u 12+v12)/2e2iv1a~
-2iu a;'
1

-e (3.174)
= 2_ +
du(X - ulplX u)eiuY, (3.170)
21r where we used relations in Eq. (3.61) and u
1
= ue-r, v
1

which gives Eq. (3.166) after a variable change of u -+ -u. With a 1 e-r, a; = a 2 er. So, the Wigner function is
Eq. (3.166), Eq. (3.165) becomes
_ -- 1 / d d i1ix2-ivx1
, · al- 2iu
, e- (U 12+v12)/2e2iv · a2 I I I I

1 1 W css (x1, x2 ) - (21r)2 u ve

jw(X, Y)j:; 2~
for pure states.
or - - < W(X Y) < -
21r - ' - 21r
(3.171)
= ( : )
2 2
J du'dv' eiu'.,;-iv' ,; e-(v" +v")/2 e2iv' a'.,-2iu' a;

Now let us find the Wigner function for some known states.
1
= 21r [ 1
exp - 2
(-12
X1
-12)]
+ X2 ' (3.175)
(a) Coherent state

For a coherent state la) with a= a 1 + ia2,

we have
2 2
Cw(u, v) = _T/*a,)= (a le-(u +v )l 2eT/ate-T/*°'la )
Tr(p ae T/a,t (3.176)
= e-(u 2 +v2)/2eT/a*e-T/*a = ei2va1-i2ua2-(u 2+v2)/2 . (3.172)
The contour of one standard deviation is iie- 2r + x~e2r = 1, which is an
Taking a Fourier transformation, we have ellipse centered at (2a 1 , 2a 2) with major and minor axes as er and e-r .
This ellipse depicts the quantum noise for the coherent squeezed state in
W (x 1 , x2) = 11r exp [ - 1( x-21 + x-22)](3 .173)
2 2 X-Y phase space (see Fig. 3.2). Compared with the coherent state, its noise
with Xi = Xi - 2ai (i = 1, 2). Here, we changed the variables from X , Y to part is squeezed along Y but stretched along X, but the displacement of
X1,x2. Since W(x 1,x2) > 0, we can treat it as a probability distribution (2a 1 , 2a 2) from the origin is the same .
 102 Quantum Optics For Exp e1·imentalists
Quantum Stat es of Singl e-Mod e Fields 103

(c) Single-photon state /1) (sps)

which is the same as the distribution in Eq. (3.55) but with .r 1 = :r ./2,.
The characteristic function for a single-photon state is In Section 10.4, we will discuss the quantum state tomography teclmiqne
Cw(u, u) = Tr(Pspse"lat-77*a) = (lle - (u2+v2)/ 2e77ate-T/*all) for the measurement of the Wigner function. An essential part of this
technique is the measurement of the probability distribution for a rotated
= e- (u2+v2)/ 2(l - /rJ/2) = (1 - u2 - v2)e-(u2+v2) / 2_ (3.177) quadrature-phase amplitude Xe = ae-i0 + at ei0 = X cos 0 + Y sin 0. Let
Substituting into Eq. (3.167) , we have the Wigner function for a sing le- us now express it in terms of the Wigner function 1V(x 1 ,x 2). For this, we
photon state: need to make a change of variables from (x 1 , x2) to another set of canonical
variables of (xf, x~):
Wsps(x1,x2) = ~(xi +x~ - l)exp [-t(xi +x~)]. (3.178)
2 X1 = X0
1 COS 0- 0 Sln
X2 . 0 , x2 = xg cos 0 + xf sin 0. (3.183)
Note that when X1 = 0 = X2, Wsps(0, 0) = -1/21r, i.e .. the minimum value
allowed by Eq. (3.171). The negativeness of the Wigner function is an Then we have
indication of nonclassical behavior of the optical field in the single-photon
We(xf, xg) = W(xf cos 0 - xg sin 0, xg cos 0 + xf sin 0). (3.184)
state.
Since (xf, x~) is a set of canonical variables, the marginal probability P(xe)
(d) Schrodinger cat state
is then
For the Schrodinger cat state of
P(xf) = j dxg We(xf, xg)
/1/J)cat= Nr (/ - ir) + /ir)) with Nr- 2 = 2(1 + e- 2r 2)
from Eq. (3.51), the characteristic function is
(3.179)
= J dx~ W(xf cos 0 - x~ sin 0, x~ cos 0 + xf sin 0)
Cw(u,v) = cat(?/Jle-(u2+v2) /2e 77ate-77*0,l?/J)cat
= N;:e-(u2+v2)/2 (e-ir77(ir/ + eir77(-ir/) ( e-ir77*/ir) + eir77*/ - ir))
= J 0
dx~ dx~ 5(x~
0
- xf)

2 x W(xt cos 0 - x~ sin 0, x~ cos 0 + xt sin 0). (3.185)

= N .;e-(u +v2)/2 ( e2iru + e-2iru + e-2rv e-2r 2 + e-2rv e-2r 2)
Making a change of variables from (x~0 ,x~) back to (x 1 ,x 2), we have
2 2
= 2N;:e-(u +v )! 2 [ cos(2ru) + cosh(2rv)e- 2r 2].
Then the Wigner function is
(3.180)
P(xf) = J dx1dx26(x1cos0+x2sin0-xf)W(x 1,x2)- (3.186)

2 In Section 10.4 , we will make an inverse of the relation above to express

Wcat (x 1, x 2) = :;. {e-xi/2 [e-(xr2r) /2 + e-(x 2+2r) 2/2]
the Wigner function in terms of P(xf) so that W(x 1, x 2) can be obtained
2
+e-x~/2e-2r [e-(x1 -2ir) 2/2 + e-(x1 +2ir) 2/2]} by measuring P(xf) with homodyne detection.

2
= :;. {e-xi/2 [e-(xr2r) /2 + e-(x 2+2r) 2/2] 3.8 Problems

+2e -(xi+x ~)/2 cos 2rx1 }- (3.181) Problem 3.1 Expressing the number states in terms of the creation ope-
rators
It shows a strong oscillation around the origin. The marginal probability
Peat= /1/Jcat(xi)/2 is then (i) Using the commutation relation [a,at] = 1 and the method of induction ,
Pca1(x1) = J dx2W cat(X1,x2) = /4 e-xl/ 2 (1 + cos2rx 1 ), (3.182)
prove
A Atn] -_ naAtn-1 ,
[a,a (3.187)
 Quantnm Optics For E.rperimc'/ltalists Quantum States of Single-Mode Fields 105
10.J.

(ii) Using Eq. (3.187) and aI0) = 0, prove (i) Using the definition of the coherent state in Eq. (3.25) and the method
in Section 3.1.3, prove the wave fun ction of the coherent state is
ata(at Io))= nat n Io).
11
(3.188)

that is, a,tnl0) is the eigen -state of t he photon number operator Fi= ata ,!, (::r)= _l_e-Im
2
(o.)-(x-o v0)2 /2 (3.192)
lfo, (w)l/4 ,
with an eigen-va luc of n.
(iii) Usi ng Eq. (3.187) and al0) = 0, prove
where x = q/qo(qo = /nlw).
aATJAtnl
a O) -_ naAn-l aAtn-110) = ... = n.'I0), (3.189) (ii) Using the definition of the squeezed state in Eq. (3.69) and t he m et hod
of Section 3.1.3, find the wave function of the squeezed state.
and prove further (0la11 a,tnl0) = n! so the normaliz ed form of the numb er
state is Problem 3.4 The phase un certaint y of the coherent state and the phase-
In) = ~a,tnlO). (3.190) squeezed state.
vn!
The uncertainty of the phase of the electric field can be roughly de-
termined from the width of the zero crossing in the electric field function
Problem 3.2 The field amplitude fluct uation of the numb er state and in Fig. 3.4. The width in Fig. 3.4(a) is zero, indicating that the phase
Heisenberg uncertainty relation. is certain at any moment for classical waves. But the widths are finite in
Since the photon number state is the eigenstate of the photon number Figs. 3.4(b )- (c), indicating the uncertainty in the phase of quantum electro -
operator n, the photon number of the optical field is completely determined, magnetic waves. Using the relation between the phase and the electric field
that is, /:).n = - (n)) 2 ) = 0. But we see from Section 3.1.3 that the strength in Eq. (3.85) and the uncertainty of the electric field in Eq. (3.86) ,
corresponding harmonic oscillator's wave function '1/J n(q) has a wide range, find the phase uncertainties of the coherent state and the squeezed state.
which indicates that the position of the harmonic oscillator has a large
uncertainty. We will work out its position and momentum uncertainties Problem 3.5 Photon number fluctuation of a phas e-squeezed state:
here. la,re2ir.p°').
From Eq. (3.6) , we have For the displacement quantity lal >> the squeezing quantity r, calcu-
late the variance (!:).2n) of the phase-squeezed state of la,re 2i'P a ) by using
(3.191)
Eq. (3.82).

For the number state In), prove

Problem 3.6 Find the photon statistics for the Schrodinger cat state in
(i) (Q)n = 0, (F)n = 0.
Eq. (3.51). Calculate the variance (!:).2 n) for this state.
(ii) (Q2 )n = (n + 1/2)11/w , (F 2 )n = (n + 1/2)nw.
So, !:).Qn = + 1/2)11/w, !:).Pn = + 1/2)nw or !:).Xn = !:).Yn =
+ 1. Then, l:).Q71 !:).p71 = (n + 1/2)11 > n/2, or !:).Xn!:).Yn = 2n +1> 1. Problem 3. 7 Find the Wigner function for the thermal state.

Problem 3.3 The wave functions of the coherent state and the squeezed Problem 3.8 Photon statistics for squeezed vacuum states.
state. The general formula for the photon statistics of squeezed states is gi-
In Section 3.1.3, we derived the wave functions for the number states. ven by Eqs. (3.94) and (3.95) through some rather cumbersome derivation.
We can us e the same method to find the wave functions for the coherent We will pres ent the outline of a simpler derivation for the special case of
state and squeezed state . squeezed vacuum states.
106 Quantum Optics For Experimentalists

For the squeezed vacuum state with maximum squ eezing in X = a+ at,
the state has the form of

1-r)=S(-r)IO)=St(r)IO) with r>O Chapter 4

= L Cn In) with Cn = (nl - r). (3.193)
n Quantum States of Multi-Mode Fields
(i) By using Eq. (3.62), show

(µa+ vat)I - r) = 0, with µ = coshr , v = sinhr. (3.194)

(ii) By applying (nl to Eq. (3.194) from the left , show the recursive relation:

(3.195)
Wh en multiple modes of an optical field are excited, we must describe it
with Cn defined in Eq. (3.193). A special case of n = 0 gives c1 = 0.
with a multi-mode quantum state. If all the modes are independent like
(iii) By using the result in (ii), show what is given in the Hamiltonian in Eq. (2.76) , we can use single-mode
quantum states described in the previous chapter to describe each mode
Cn = { 0 for n = 2k - 1 separately. The quantum state of the whole field can be represented by
(-lrco
k (2k)!
(2kk!)2
(v)k
µ for n = 2k
(3.196)
a direct product of all the single -mode states. However, if the modes are
correlated, the quantum state for the whole field cannot be written as a di-
where lcoJ2 = 1/µ is determined by normalization with the identity rect product of the single -mode states but must be the superposition of the

(
l _ V2
µ2
)-1/2
=
L (2k k!)2
(~)
µ
2k
(3.197)
different direct product states. This leads to the concept of entanglement.
This is the main emphasis of this chapter.
k

Pn = Jcnl2 gives Eq. (3.94) with a = 0 for squeezed vacuum state . Based 4.1 Multi-Mode Coherent States of Independent Modes
on this result and the principle of mu lti -photon interference, we will show
later in Prob lem 8.5 that the case of a #- 0 (Eq. (3.94)) is the result of Let us start with the simplest case of a multi-mode field with all modes
multi -photon interference between a coherent state and a squeezed vacuum independently excited. Since modes are independent, we can use the single -
state . mode quantum states for each mode and the quantum state for the whole
is a direct product of these sing le-mode states:
Psys = IT®p>., (4.1)
>-
where h is the density operator for describing mode ,\_ For examp le, if
the optical is in coherent states with its modes excited independent ly, the
quant um state of the field is descr ibed by a mu lti -mode coherent state wh ich
is a direct product of the coherent states of each excited mode:
=
l?JJsys) IT®Ja>-) = J{a>,})•
>- (4.2)
>-
For the two -mode case, we have
(4.3)

107
 Q'Uant'Um Optics Fm · Exp erim entalists Q'Uantwn Stat es of A!'Ulti-l\fod e Fi elds 109
108

The operators of each mode on ly act on t he state of its own mode: 4.2 Classical Description of Multi-Mode Optical Fields
iii\a , µ) = (h1\a)i) 0 \µ)2 = (a\a)i) 0 \µ)2 = a\01,µ). (4.4) In SecLion 1.5, we menLioned thaL Lhe ilucLuaLiorn-;of a11 electromagnetic
Simila rly, field can be represented by a random process E (r, t) and a random process
(4.5) is a time function of some random variables {Xj}: E( r, t) = f ({ Xj}, t).
For a more comp licated field operator such as the electric field operator in But no matter what it is, this function must satisfy Maxwell equations for
Eq. (2.90): electromagnetic fields. So, we can express it in terms of the modes of the
electromagnetic field:
E(r, t) = L E k,sEk,sUk.s(r)e-iwt + c.c. (4.11)
k ,s

Compared with Eq. (4.9), Ek,s corresponds to the coherent state

excitat ion s et 5 (k) of the modes. From Eq. (4.11), we find that only Ek,s can
be arb it ra ry. If the electromagnet ic field is to be described by a random
process, only Ek,s can be the random variables. Hence, the classical des-
criptio n of a multi-mod e field is based on the description of the random
variables Ek,s. This is the starting point of the classical coherence theory.
For simplicity, let us only consider one-dimensional scalar fields with a
comp lex repr esentation:

E(r , t) = J dwE(w) ei(kz-wt)

where
....,=
£(r, t) i~ L
J 3
d kv
/,i;;; ei(k-r-wt)
2 Ek,slts(k) (2n)3/2 (4.9) = J dwE(w)e-iw(t-z/c)' (4.12)
s=l,2
where k = w / c. Defining T =t - z / c, one-dimensional scalar field is only a
The expression above is the same as the classical expression of the electric
function of time:
field obtained from Eqs. (2.14) and (2.66) with qk,s(0) = et 5 (k) ~- This
relation for qk,s is the same as Eqs. (2.68) and (2.74) after the quantization. E(r, t) = E(T) = l J dwE(w)e-iwT , . (4.13)
Here we make an exchange of a +-+a.
where E(w) is a random variable similar to Ek ,s and it can be expressed
In the photo -detection theory of Chapter 5, we have the probability of
by E(T) as
finding photo-electrons in optical detectors proportional to the intensity of
the field: I= (i) with the intensity operator as i = t(-)(r, t) · E(+)(r, t). E(w) = l J .
dwE(T)eiwT_ (4.14)
For the multi -mode coherent state, we can easily obtain
E(w) is the frequency component of the optical field, whose probability
I= (i)a = ( {aA}IE(-\r, t) · E(+)(r , t)l{aA}) = [*(r, t) · f(r, t). (4.10)
distribution P({E(w)}) = P(E(w 1 ),E(w 2 ), ... , E(wj), ...) determines the
But in classical wave optics, the intensity I of the field is defined as the statistical properties of the field.
absolute square of the amplitude of the electric field. So, £( r, t) corresponds
exactly to the amplitude of the electric field in classical optics. Therefore ,
4.2.1 Continuous Waves (CW) and Stationary Processes
we also claim that a multi -mode coherent state is a quantum state whose
description of the field is closest to the description of classical optics. Next, Consider the correlation function of the optical field:
based on this connection, let us discuss some more about the description
of a multi -mode field in classical wave optics.
r(t , T) = (E*(t)E(t + T)) = J d{E(w)}P({E(w)})E*(t)E(t + T), (4.15)
110 Quantum Optics For E:1:pe1·imentalists Quantnm States of l\folti-Mod e Fields 111

where random process E(t) is related to the random variables {E(w)} Similarly, the correlation function in Eq. (4.16) becomes
through Eq. (4.13). For continuous waves, the starting time is irrelevant,
so, f(t, T) = r(T) is independent of time t ancl continuous waves are usually
f(T) = J dtE*(t)E(t + T)
described by stationary processes. Furthermore, continuous waves are of-
ten ergodic, that is, in a sufficiently long time, they include or run through
=
2 7r
J dtdw 1 dw2 E *(w1 )E(w 2 )ei(w1- w2 )t c- iw2T

all possible allowed values. Therefore, the average over the probability
distribution in Eq. (4.15) can be replaced by an average over time:
= J dw1dw2E*(w1)E(w2)6(w1- w2)e - iw27

I'(T) = (E*(t)E(t r
+ T)) = lim !__ dtE*(t)E(t + T).
T-+oo T Jr
(4.16) = J dwjE(w)l2e-iwT

One special case of Eqs. (4.15) and (4 .16) is T = 0: f(O) = (E*(t)E(t)) .

This is the intensity I of the optical field. Using Eq. (4.13), we obtain
= J dwS(w)e-iwT , (4.22)

where we used 2~ J dt eiwt = <5(w).The last line is the same as Eq. (4.19).
r(T) = 1dw1dw2(E*(w1)E(w2))ei(wi-w 2
)te-iw T.
2 (4.17) So, the spectral relation in Eq. (4.20) stands for pulsed light field, too.
2w
Since r (T) is unrelated to time t for continuous stationary waves, we must 4.2.3 Coherence of Optical Fields - Phase Correlations
have
In the coherence theory, the normalized field correlation function
(4.18)
ry(T) = r(T)/f(O) = (E*(t)E(t+T))/(IE1(t)l2), (4.23)
where S(w 1 ) is some function of w1 . The expression above shows that there describes the correlation of the phases of the optical field at different times.
is no phase relation between different frequency components of a continuous The degree of correlation of the phases determines the visibility of inter-
stat ionary field since (E*(w 1)E(w2)) = 0 for w1 -/- w2. Hence, Eq. (4.17) ference fringes and therefore the coherence of the optical field, as we have
changes to seen in Eq. (1.37) of Section 1.5. The coherence time Tc of an optical field

f(T) = J dwS(w)e-iwT (4.19)

is defined as the range within which ry(T) is appreciably different from zero
(see Fig. 4 .1) .
or

S(w) = 17r / dTI'(T)eiwT_

. (4.20)
2
From Eq. (4.19), we obtain I = I'(O) = J dwS(w). So, S(w) is regarded
as the spectral function of the optical field. The statistical relation in
Eq. (4.18) is a signature of continuous stationary fields.

4.2.2 Pulsed Waves and Non-Stationary Processes

For a pulsed field , we can define an instantaneous intensity: I(t) =

E* (t )E( t). But if the detector is not as fast as the pulse , it is impossi- Fig. 4.1 Correlation function 11'(
T) I and the coherence time Tc.
ble to observe the shape of the pulse and all it sees is an integration over

1:
time, i.e., the total energy of the pulse: We can also define the cross-correlation function between two optical
fields:
I= dtE"(t)E(t). (4.21)
(4.24)
112 Qua:nt'Llm Optics For E:cpcrimentalists Quantmn States of Multi-Mode Fields 113

and the normalized cross -corre lat ion function:

(d.25)
( E1 ( t) 2 ) ( E2 ( t)
I 1 I 1
2
) 2.0

It corr esponds to the phase corr elat ion between two optical fields and gives
the visibility of the int erfer ence fringes for the interfer ence between the two 1.0 - - - - - - - - - - ----- -- - ·- •- -,- - ·--
fields. When t he two fields correspond to the same optical field but at
two different loc ations , we obtain the spat ial coh erence function, as we first
0.0 ._____ ...__ ____.
__ __.__ ___.._~
introduced in Eq. (1.36) of Section 1.5.
2T0 r
4.2.4 Hanbury Brown and Twiss Effect - Intensity
Fig. 4.2 g (2) (T) as a function of the time delay T. It shows the photon bunching effect.
Correlations

A more complicated correlation function than Eq. (4.15) is the higher order coherence function fr(T)Idepicted in Fig. 4.1, the expression in Eq. (4.29)
intensity correlation function of the optical field: gives g( 2 ) ( T) as a specific function of T shown in Fig . 4.2, which agrees well
2\T) with the experimental observation shown in Fig. 1.9.
r( = (I(t)I(t + T)) = (E*(t)E*(t + T)E(t + T)E(t)). (4.26)
In general, applying the Cauchy-Schwarz inequality:
For a continuous stationary optical field, it is independent of time t. It
corresponds to the correlation of the intensity of the optical field at different (4.30)
times. For a thermal source with Gaussian statistical distribution, we can to arbitrary optical fields described by classical wave theory, we obtain
apply the Gaussian moment theorem (see Isserlis theorem in Eq. (1.28)):
(4.31)
(ABCD) = (AB)(CD) + (AC)(BD) + (AD)(BC). (4.27)
Hence, the classical wave theory always gives rise to the photon bunching ef-
Then Eq. (4.26) becomes
fect. In the multi -mode description of a thermal field, which has a Gaussian
r(
2\T) = (E*(t)E*(t + T))(E(t + T)E(t)) statistical distribution for all the modes, the size of the photon bunching
+(E*(t)E(t + T))(E*(t + T)E(t)) effect, which is characterized by the excess correlation g( 2 ) -1, is usually re-
lated to the average number of modes M of the detected field (see Eq. (4.84)
+(E*(t)E(t))(E*(t + T)E(t + T))
of Problem 4.4).
2 2
= Ir(T) I + Ir (o)I , ( 4.28)
On the other hand, quantum fields can exhibit photon anti-bunching
where, due to the randomness of the phase of the field, (E*(t)E*(t + T)) = effect, i.e., g( 2 )(0) < 1, as we have seen in Section 1.7. We will discuss more
0 = (E(t + T)E(t)). Hence, the normalized intensity correlation function is about this in Section 5.3 of Chapter 5.
For two optical fields, the correlation function between the intensities
(2)(7) = (E*(t)E*(t + T)E(t + T)E(t))
g - + T)E(t + T))
(E*(t)E(t))(E*(t of the two fields is

= r( \T)/lr(o)1 = 1 + fr(T)l2 ,
2 2
(4.29)
where --y(
T) is the temporal coherence function given in Eq. (4.23). This gi- The corresponding Cauchy -Schwarz inequality for this quantity is
ves the complete explanation of the Hanbury Brown -Twiss photon bunching
effect by the classical wave theory, which gives not only g( 2 ) (0) = 2 > 1
[ri;) (o)]2 ri~)(o)rg)(o). (4.33)
for photon bunch ing effect, as we have done in Eq. (1.42) of Section 1.6, Later in Sect ion 7.4 of Chapte r 7, we will show that this inequality is
but also its temporal dependence on the time delay T. For the tempora l violated by two optical fields with quantum corre lation.
 114 Q·uanturn Opt'ics For Experimentalists Quantum States of Multi-Mode Fields 115

4.2.5 Transform-Limited Puls es - Mode-Locked Optical Bence. the field within the optical pulse is completely coherent. "\i\Tefind
Fields from Eq. (4.35) that the temporal function £ (t) of the field and its spectral
As we discussed ear lier, the frequency components of a continuous statio - function £ (w) are a Fourier transformation pair. So, we call this type of
nary field must sat isfy Eq. (4.18), indicating that there is no phase correla - field a transform-limited optical pu lse, whose spectral width .6.w and pulse
tion between them, t h at is, different frequency components will not interfere width satisfy .6.T.6.w = 21r. Therefore, Tc= 1/.6.v = 21r/.6.w = .6.T, that
with eac h other. 1 Bu t what will happen if all the frequency components of is, the coherence time Tc is equa l to the pulse length .6.T for a transform-
an optica l field have phase cor relat ion . limited pu lse. This is cons istent with the fact presented in Eq . (4.37) t h at
any two points inside a transform-limited pulse are comp lete ly coherent .
f--2rr/Q~ On the other hand, a non-transform-limited optical pulse h as its coherence
time Tc always smaller than the pulse length .6.T. Figure 4.4 shows the
A physical picture for these two cases .
frequency comb pulse train
pulse enve lop
non-transform-limited
(a) (b) (a)
T,< .1T

Fig. 4.3 (a) A comb -lik e frequency distribution; (b) A pulse train in the time domain.
(b)
T,=.1T
For simplicity, let us consider a comb -like frequency distribution , as
shown in Fig. 4.3(a) . Suppose that there are 2N + 1 frequency components
Fig. 4.4 (a) A non-tran sfo rm-limit ed pulse with its coherence time Tc < the pulse
of equal height and they are separated by the same quantity n. If they all lengt h 6.T ; (b) A transform-limit ed pulse with its co herenc e tim e Tc = the pulse length
have the same phase, the optical field has the form of 6.T .
N
E(t) = L E oe-i(w ot+nO)t = E oe-iw o t ---
sin NDt
sinnt
. (4.34)
n=-N
4.3 Multi-Mode Single-Photon States - Entangled States
Eo = IEolei'Po is the complex amplitude of each frequ ency component.
of Single Photon
Figure 4.3(b) shows the field function in time domain: it is an infinite pulse
train with equal spacing of 21r/D. Now let us come back to the description of the quantum states of a multi-
For an optical field with a continuous spectrum, if the relationship mode field. "\i\Tefirst consider the single-photon case. This photon may be
among all frequency components is fixed, that is, E(w) = A 0 £(w) where in different modes. Because there is only one photon , this photon must be
£(w) is a definite function but Ao may be a complex random variable , then in a superposition state of the single -photon states for different modes. It
the optical field has a definite pulse shape £ (t): has the form of
E(t) = Ao J dwE(w)e-iwt = AoE(t). (4.35) (4.38)
Its correlation function is
r(t1, t2) = (E*(t1)E(t2)) = (IAol2)£*(t1)£(t2) (4.36) where Ili) = IO)i ... IO)i-i Il)i IO)i+l ··· is the quantum state with only mode
i having one photon excited and other modes in vacuum. For the modes
so the normalized correlation fun ct ion is
in vacuum, we can omit them in writing the state in Eq. (4.38). Since this
(4.37) state cannot be written as a direct product of the states for each mode, i.e.,
1
Interference between fields of different frequencies shows up in the form of t emporal IL @l?j;)i, it is a single-photon entangled state. We will exam ine a couple
beating.
of specific multi-mode sing le-photon states next.
 Qua11tum Staf<'s of Mull i-Mode Fields 117
llG Quantum Opti.cs For E:rper-imentalists

4.3.1 Two-Mode Single-Photon States the contin uou:::; variables (see Section 2.3 .5). A special case is c(w)
cp(w)eiwT(</>(w)= real). It corresponds to a single-photon wave packet:

J
Two-mode :::;i11gle-µl10to11:::;tate is the simp lest entang led state of photons.
There are two modes of t he fields being excit ed but in the meantime there IT)¢= dw</>(w)eiwTat(w)\
O) = Al(T)\O), (-1..-13
)
is only one photon. It can be written a:::;

(4.39) where Acp(T) = J dw<J>(w)e-iwTa(w) and satisfies [A¢(T). Al(T)] = 1 du e

to normalization J dw\<J>(w)l 2
= l. The reason that Eq. (4.43) describes
When a single-photon in a single- mod e is incid ent on a lossless beam a single-photon wave packet is because th e probability density of photon
splitter, the output state is a two-mode single-photon state (for quantum det ect ion at time t is (see Section 2 .1. 5)
tr eatme nt of a beam splitter, see Section 6.2 in Chapter 6). Another exa m-
pl e of two-mode single -ph oton state is a photon in an ar bit rary but defined (4.44)
polarization state. From classical optics , we know that an arbitrary pola-
rization state of light is describ ed by E = x cos 0 + f;ei 5 sin 0. In quantum where
optics, we can associate it with an annihilat ion operator in the similar form:
aE = ax cos 0 + ayei 5 sin 0 (see Chapter 6). When the corresponding cre-
E(t) = l J dfl a(fl) e-i ·nt
'" ) g(t) = 1 / dWty,-1.,
( w ) e -iwt . (4.45)

ation operator acts on the vacuum state , th e single-photon state in this g(t - T) is a wave pack et centered at
For a well-b ehaved function <J>(w),
polarization state is generated:
time T (see Fig. 4.5).
(4.40)
(a) (b)
Setting 5 = 0 gives a single-photon state with a lin ear polarization along 0-
direction: ll)e = cos0ll) x +sin0ll)y- Here, we do not write out the modes
in vacuum. 5 = ±1r/2, 0 = 1r/4 corresponds to a single-photon state in left
or right circular polarization I1)± = ( I1)x ± i I1)y) / v'2.

4.3.2 Multi-Frequency Single-Photon States - Fig. 4.5 (a) Wave packet for a photon in a single temporal mod e; (b) N-photon single
Single-Photon Wave Packets t emporal mode wave pack et.

Another multi-mode single-photon state is the multi-frequency single- As discussed in Section 2.2.2 , a generalized mode is a linear combina -
photon state. Assume all the excited modes have the same polarization and tion of several different modes. Although it consists of multiple modes,
spatial mode function and the only difference is their frequencies. Such a its mode function is still a solution of Maxwell equat ions. Hence, it re-
photon state has the form of presents a generalized mode and the annihilation operator A. or A.cp (T) is
(4.41) the annihilation operator for the photon in this genera lized mode after
quantization. It gives a quantum description for this genera lized mode.
The quantum state generated by the action of the correspond ing creation
where A= Li C:aw
i . Since ll)m,J is normalized , i.e., Li \cil 2 = 1, we hav e
operator on vacuum is thus a singl e-mode single-photon state in this gene -
(4.42) ra lized sense. A.cp(T) describes a single temporal mode with a pulse shape
i ,j i,j of g(t) , similar to th e operator a,Efor the polarizatio_n mode with ~;1arb~-
trary polarization E. The coherent state correspondmg to mode Acp(T) 1s
For a continuous spectral distribution, we hav e ll) m,J = J dw c(w)aT(w)IO)
with a,t(w) satisfying commutation relation [a(w), at(w')] = 5(w - w') for la ¢ (T)) = b A q,\0) (DA q, = eaA:( T) -a* A q,(T)), which describes a laser pulse
 •
118 Qva.nt-um Optics For Experinientalists Quantum Stat es of Multi-Mode Fields 119

of shape g(t -T). An N-photon state of single temporal mode is described
by the multi-photon state (see more in Section 8.4. l of Chapter 8):
IN, T )cp= [A~(T)] NIO). (4.46)
Because the N photons are in one temporal mode, they are complete ly
indistinguishable and will produce a maximum multi-photon interference
effect (see Section 8.4 of Chapter 8). This N-photon state can be produced
probabilistically from single -pho ton states with the help of beam splitte rs
(see Section 8.4.1).
There are two practical ways for the generation of single-photon states.
The first one is to use the commonly available two-photon states generated
from spontaneous parametric processes. One of the two correlated photons
becomes a temporal sing le-photon state when heralded on the detection of
the other photon [Hong and Mandel (1986)]. In this case, we can use the
temporal single-photon state in Eq. (4.43) to describe it with T determined
by the detection time of the other photon. The second method is to use
a short laser pulse and coherently drive a single light emitter such as an
atom, an ion, and a quantum dot. This is possible because of the photon
anti-bunching effect for the single light emitter, which cannot emit another
photon right after the emission of one photon [Kimble et al. (1977); Die-
drich and Walther (1987)]. The emitted field can also be described by the
temporal single-photon state in Eq. (4.43) but with T determined by the
short laser pulse. The shape of g ( t) is given by Eq. (4 .45) from the spectral
function ¢( w) of the single light emitter.
4.4 Multi-Mode Two-Photon States - Two-Photon
Entangled State
Even though the multi -photon state in Eq . (4.46) is written in the form
of multi -fr equency mode, it is, as we mentioned earlier, still describes in
essence a single-mode field, i.e., a single temporal mode field. In this section,
we will discuss some genuine multi-mode multi -pho ton states, that is, no
matter how it is written, the photon state must be described by multiple
modes. On the other hand, if we can arrange the mode structure so that
each mode is independently excited, we can use single -m ode method to treat
each mode individually and the whole system is a direct product of all the
single -mode states, as we did in Section 4.1. But if a multi-mode state is in
some form of superposition, it usually is an entangled state, which cannot
be written as a product of the states of independent modes. In this case,
the modes are not independent but correlated. "\Vemust treat all mod es
together. So, the multi-mode photon states in this section are usually a
superposition state of multi-mode multi-photon states. Let us consider a
t,vo-mode case first, say, a tvm-photon state of polarization.
4.4.1 Two-Photon Polarization States
For the case of two photons in two polarization modes of x,y, we have three
base-states: l2x, Dy), IDx,2y) , llx, ly)- The superposition state of the three
states with equal weight is
(4.47)
This is so-called W-state [Dur (2001)]. Another two- mode two-photon state
is the superposition state of maximum photon number:
jNOON(2)) = (l2) xl0)y + IO)xl2)y)/V2. (4.48)
This is the so-called two-photon NOON state [Ou et al. (1990b); Ou
(1997a); Boto et al. (2000)], which has important application in precision
phase measurement (see Chapter 11). These two states cannot be written
in the form of Eq. (4.46) so they are some sort of entangled states.
4.4.2 Two-Parti Two-Photon States of Polarization - Bell
States
Next, we extend the two -photon polarization state to two-parti system of
A and B. Now there are four modes: xA, yA, xB, yB. Consider only the
cases of one parti each having one photon. We have four possibilities of
llx)Allx)B, jly )Ally)B, llx)Ally)B, jly)Allx)B- A set of superposition sta-
tes of the four are:
IW±) = (llx)Ally)B ± lly)Allx)B)/V2, (4.49)
l<P±) = (llx)All x)B ± jly)Ajly)B)/V2. (4.50)
Here the absolute value of every superposition coefficient is the same. They
form the base states for the sub -space of two-parti two-photon polarization
states with A and B each having a photon . The four states in Eqs. (4.49)
and (4.50) are known as the Bell states. They can violate Bell 's inequalities
and play an important role in the test of nonlocality of quantum mechanics
[Bell (1987)] . The four Bell states are also the basis for the teleportation
of a single-photon state of arb itrary polarization [Bennett et al. (1993);
Braunstein and Mann (1996); Bouwmeester et al. (1997)] . Problem 8.6
discusses more interesting properties of Bell states (Eq . (8 .124)).
120 Quantum Optics For E.rperimentalists Quantwn States of Multi-Mod e Fi elds 121

4.4.3 Multi-Frequency Two-Photon States - Frequency- 4,5 N-Photon Entangled States

Entangled States and Time-Entangled States
Next, we consider multi-photon states in multiple modes. Here, we assume
Consider now the frequency modes. An interesting one is the two-frequency the total number of photons of the optical field is a fixed number of N.
two-photon state of two-parti system of A and B: \Ye already encountered the case of N = 2 but we will consider the case of
(4.5 1) N > 2.
Again. each parti has one photon. This is a frequency-entangled state.
4.5.l Two-Mode N-Photon Entangled States - NOON
When making a direct time-resolved two-photon coincidence measurement,
States
we can observe a two-photon beating effect [Legero et al. (2004)]. Injecting
it into a Hong-Ou-Mandel interferometer (Section 8.2.1), we can observe The simp lest is an N-photon state in two modes. Consider the following
a spatial beating effect [Ou and Mandel (1988); Li et al. (2009)]. For a N-photon state:
two-photon state with a continuous spectrum, we have
!NOON)= (IN)AI0)B + IO)AIN)B)/v0. (4.54)
I\J!2)= r dw1dw2\Ji(w1;w2)a
}6.w
1
(w1)a (w2)I0); (4.52) This state is known as the NOON state [Ou (1997a); Kok et al. (2002)].
which can be obtained from spontaneous parametric down-conversion pro - In this state, all N photons are either in mode A or mode B. Since all N
cesses (see Section 6.1.5). When the pump field to the parametric pro- photons are always together, we can equivalently regard the N photons as
cess is a continuous wave with single frequency, we have \.II(w 1 , w2 ) = one entity with a total energy of Nnw 0 , where wo is the frequency of one

Vp'IJ;(w1)6(w1+ w2 - wp) (see Section 6.1.5). Hence, Eq. (4.52) is changed photon. The equivalent de Broglie wavelength of the combined entity is
to then >.0 / N. This state can be used in the precision measurement of phase
and reach the Heisenberg limit in phase measurement (see Chapter 11) [Ou
IW2(CW)) = / dw1dw26(w1+ W2- wp)'IJ;(w1)a~(w1)ak(w2)IO) (1997a)].

= J dw1dw2Vp6(w1 + w2 - wp)
4.5.2 N-Parti Polarization Entangled States - GHZ States
and W-States
X j'I/J(wp - w2)a~ (w1)ak(w2)IO)

=CJ dw1dw2dTei(wi+wrwp)T ¢(w1)<p(w2)a~(w1)ak(w2)IO)

=CJ dTe-iw pTIT)¢,AIT)cp,B,
where IT) is a single-photon wave packet given by Eq. (4.43). Equation
(4.53) indicates that two wave packets are produced simultaneously at time
T but with T undetermined. The physical picture of this state is straight -
forward: the monochromatic continuous pump field means that the pump
is an infinitely long wave train with pump photon appearing at any time,
but whenever down-converted, it converts to two wave packets simultane -
ously. Since the time for the pump photon is uncertain, so is the generation
time for the two-photon wave packets. The state in Eq. (4.53) is a time-
entang led two-photon states and can be used to generate time -bin entangled
two-photon states (see Section 8.2.2).
Earlier, we mentioned that the Bell states of two-parti system can be used
to test nonlocality of quantum mechanics. For a multi-parti system, the
nonlocal effect of quantum mechanics becomes more apparent. For a three-
(4.53) parti system, the three-photon Greenberg -Horne-Zeilinger (GHZ) state has
the form of
(4.55)
It was proved that the GHZ state of three -p arti system can demonstrate
locality violation of quantum mechanics without the need of a Bell-like
inequality [Greenberger et al. (1989)]. The GHZ state in Eq. (4.55) is a
three -pho ton entangled state. Moreover, the following W-state has similar
property and can be used in quantum information [Dur (2001)]:
IW(3)) = (llx)Ally)Blly)c + lly)Allx)Blly)c + lly)Ally)Bll x)c )/vf:3.
(4.56)
 •
122 Quantum Optics For Expcrirnentalists Quantum States of Multi-Mod e Fields

4.6 Two-Mode Squeezed States - Photon Entangled States This shows that the photon numbers of the two modes are exactly the same.
of Continuous Variables \\iiting i1Jab)in the number state representation in the general form:

The multi-mode states discussed in previous sections are all written in the (4.61)
number state representation. In these states, photon numbers are counta - n1,n

ble. They are suitable for quantum information applications with discrete we have straightforwardly
variables where information is encoded in each photon. Besides discrete
N_l77ab)= (ata- btb)l77ab)= Ldmn(m - n)lm)aln)b- (4.62)
variables, another coding method is to work on the continuous variables. n1,n
It encodes the information on the amplitude or phase of an optical field.
This type of applications of quantum information requires entangled states But N_l77ab)= 0, so we obtain dmn(m-n) = 0, i.e., dmn = Cm<\nn·Hence,
with continuous variables. The earliest and most commonly used photon Eq. (4.61) becomes
entangled states of continuous variables are generated from spontaneous (4.63)
parametric amplifier (see Section 6.1.5 or 6.3.4). They have the form of
hab) = Sab(rJ)IO) (4.57) From Problem 4.1, we can prove Cm = gm;cm+l so we obtain the final
where the two-mode squeezing operator Sab(rJ)is form of the two- mode squeezed state in number state representation:

Sab(rJ)= e17a,tbt_11*ab, (4.58) lrJab)= L (gm;cm+l) lm)a lm)b- (4.64)

ni=O
which is similar to the squeezing operator S(r) in Eq. (3.63): when a= b,
Since each term in the sum has identical photon number for the two modes
i.e., the two modes are the same, operator Sab(rJ)becomes the single -mode
of a, b, this state is also called "twin beams state".
squeezing operator S(r). So, the state in Eq. (4.57) is also called two-
From the way Eq. (4.64) is written, it is impossible to write the twin
mode squeezed state. It was first introduced and studied by Caves and
beams state in the form of direct product states of any two modes. So,
Schumaker in 1985 [Caves and Schumaker (1985)]. Operator Sab(rJ)has
twin beams state is an entangled state. Interestingly, if we only look at one
similar properties as S(r) (see Problem 4.1):
mode, say, mode a, the state is a mixed state with the following density
A= SliLSab =Ga+ gl) , B = Gb + gat (4.59) operator:
where G = =(
cosh IrJI, g 17/ IrJI) sinh IrJI- Next, let us find the properties Pa= Trbhab)\rJabl
of the two -mode squeezed state lrJab). But the specifics depend on what
physical quantity we measure.
= L (lgl m;c
2 2 2
m+ )lm)a(ml = L Pmlm)a(ml, (4.65)
m=O

4.6.1 Twin Beams
When the physical quantity we measure is photon number or intensity, the
photon numbers of the two modes are strongly correlated. Define pho -
ton number difference operator: N_ = at a - bt b. Then we have in the
Schrodinger picture
N-lrJab) = (ata - btb)Sab(rJ)IO)= Sabslb(ata - btb)Sab(rJ)IO)
= Sab(A_tA - iJtB)jO),
where we used Eq. (4.59). It can be easi ly shown from Eq. (4.59) thAatwith
G 2 -g 2 = 1, Jitfi_iJtiJ = ata-btb. So, N-lrJab) = Sab(ata-btb)IO) = 0.
where Pm= lgl2 m;c 2 m+2 = nm/(n + l)m+l (n = lgl2 ). This is the photon
statistica l distribution of a thermal state (see Section 3.5.3) and the density
operator in Eq. (4.65) is exactly the same as that in Eq. (3.108) for a thermal
state . Hence, each field of the twin beams, when viewed individually, is in
a thermal state. This was first pointed out by Yurke and Potasek [Yurke
and Potasek (1987)] and demonstrated by Ou et al. who showed g(2 ) • 2
for only one field [Ou et al. (1999a,b)]. Nowadays, the measurement of g(2 )
is used to determine the number of modes for the fields out of a parametric
(4.60)
amp lifier [Liu et al. (2016)].
For the two -mod e squeezed state lrJab)in Eq. (4.57), although the photon
numbers of modes a, b are perfectly correlated, the average photon numbers
124 Q'Ua'lli'Um Optics For E:rpe1·im e11talists Quantum Stales of !lfolt-i-Mode Fields 12fi

are JgJ2 = sinh 2 J17J and are relatively sma ll 'Nith a finite ga in for direct the two-mode squeezed state a.re perfectly correlated and this is trne si-
photo-detection. The average photon numbers can be boosted up by acting mnlt aneonsly for two conjugate physical obscrvables X, Y. This is the
Sab(17) on a cohere nt state at mode a.: Sab(17)Ja)a, which h as an average famous Einstein-Podolsky -R oscn(EPR) correlation and leads to the EPR
photon number of (N;i) = C 2 JaJ2 + [g[2
2 2
C 2 [a[2 , (NB)= [g[ J(a[ + 1) paradox for quantum mechanics [Einstein et al. (1935)]. So, the two-mode
JgJ2 [a[2 for Ja[2 » 1. On the other h and, since A_tA - f3t B = at a - bt b,we squeezed state is also called an "EPR-entang led state'·. The idea of using
two-mode squeezed states to demonstrate EPR paradox was first proposed
have
2 by 1\1. Reid [Reid (1989)] and demonstrated by On et al. exper im enta lly
(6.2(NA - NB)) = (~ 2 (Na - Nb))= JaJ2 + 1 JaJ [Ou et al. (1992b)]. Here the two corre lated particles arc the two virtu al
= ((NA)+(NB))/(C 2
+[g[ 2 ) harmon ic oscillators for the two modes of t he electromagnetic fields where
«(NA)+(NB) for c2 »1. (4.66) X, Y corr espond to the position an d momentum operators. respect{vely.
So, the photon numb ers are still corr elated. Even for finit e C, we hav e
( ~ 2 (NA - NB)) < (NA)+ (NB)= (~ 2 (NA - NB))cs where cs sta nd s for 4.6.3 Squeezed State in Multi-Frequency Mode - Spectrum
the cohere nt state average. of Squeezing

In the dis cussion abo ut EPR correlation, we measure separately the

4.6.2 Two-Mode Entangled States of Continuous Variables quadrature-phase amplitudes of the two modes. But if we measure the
with Einstein-Podolsky-Rosen Correlation two modes all together, we will hav e completely different outcomes. This
In the previous section, we examined the intensity correlation of th e two is the case wh en the two modes are two frequency modes. We will obtain
modes, which corresponds to the amplitude corre lat ion of the two modes. the spectrum of squeez ing of the optical field.
But phase and amp litude are a pair of conjugate observables, so we are also
interested in the phase corre lation. For this, we use the quadrature-phas e
amplitudes X, Y introduced in Section 3.4 for a sing le mode field. For the
two -mod e case, we have
xa = a+ at, Ya= (a - at)/ i
(a) (b)

xb= b+ bt, Yb= (b- bt)/i. (4.67) Fig. 4.6 (a) Input-output of a deg enera t e opti ca l parametric
t hr es hold ; (b) th e correlat ed fr equen cy components.
oscillator (OPO) below

In the previous section, our concern is the quantum states of the field so
we used the Schrodinger picture. Now we will use the Heisenberg picture.
Then the evolution of the field operators is given in Eq. (4.59). For the
evolution of the quadrature-phase amplitudes , we have
XA = GXa + g.Xb, YA= GYa - gYb
XB = GXb + gXa, YB= GYb- gYa
with XA =A+ A_t,YA= (A-At)/i, XB = B + f3t, YB = (f3-f3t)/i. Here ,
we assume g > 0 for simpli city of argument. From Eq. (4.68), we obtain
XA - XB = (G - g)(Xa - Xb) = (Xa - Xb)/(G + g)
This shows that when g -+ oo, XA - XB -+ 0, YA + YB -+ 0, that is,
XA = XB , YA = -YB- Therefore, the quadrature-phase amp litud es of
Suppose the two modes of the two-m ode squeez ed state belong to two
different frequency modes of a one-dimensional optical field, we may rewrite
Eq. (4 .59) in terms of the frequency modes as
a(wo + D) = G(D)ao(wo + D) + g(Sl)al(wo - Sl), (4.70)
(4.68)
with [ao(w) , al(w')] = 5(w - w'). So, the frequency components of w 0 ± S1
of the field are coupled, as shown in Fig. 4.6. Later in Section 6.3.4, we
will prove that the operator evolution in Eq. (4.70) can be achieved in
a degenerate optical parametric oscillator (OPO) below threshold, where
(4.69) a pump photon of frequency Wp = 2w0 is down -converted to two lower
frequency photons of frequencies w 0 ± S1satisfying the energy conservation
wo+D+wo-Sl = 2wo = wp. This conversion couples the two lower frequenc y
 •
126 Quantum Optics For Experimenta lists Quan.turn States of l\lulti-1\Jodc Fields 127

ibed by Eq. (4.70). T he exp licit forms of G(O),g( rt)

pl_1oto m, in _a w,-:Ydet:5c1:
Here . we assume no(w ± f2) are in vacnu m so t hat (X~?) (rt)X~?)t (D')) =
will b_: clen__ved m Sect 10116.3.4. Bu t by appl )~ing [a(w), at(w ' )] = 5(w -w ') ()
2
0
1
and [a(w), a(w )] = 0 to Eq. (4.70). we obta m some genera l pro p ert ies as c5(D- D'). Hence, S,-...: 00
(D) = Sx( D) = (IG(D)I - /g(D)/) = 1/(IG(D)I +
follows: /g(n)/) 2 < 1 and Sx 00 + n / 2 (D) = Sd D) = (/G(rt)/ + /g(D)/) 2 > 1.
Notice t hat Sx (D)Sy (D) = 1 and for the vacuu m state . g(D) = 0 and
2 2
/G (rt )/ - /g (rt )/ = 1, G(rt )g(- rt ) = G(- rt )g(rt ). (4.71) G(D) = 1 so t hat sxac(n ) = 1 = src( n ). T herefore. S~( rt ) <ASx0 c(n )
2 2
So. /G(f2)/ = 1 + /g(rt )/ > 1. Wi t h E q. (4.71). it is stra ight forward to
but S,,( D) > Sf!1C (D) . Sx (D), Sy (rt) are simil ar to 2
u~.
X), \~ 2 Y) of the
prove single-mo de field, describin g t he qu adrat ur e-ph ase ampli t ud e fluct uat ions
of a mult i-frequency field . Sx( D) < Sxac(rt ) indi cates t hat t he squ eezed
/G(- rt )/ = /G(rt )/, /g(rt )/ = /g(-rt )/, state in multi-fr equ ency mode has one of its qu adra t ur e-ph ase ampli t ud e
fluct uati ons smaller t han t hat of t he vacuum state, achievin g vacuu m qu an-
<pt+<p;='Pc+ <pt, (4.72)
tum noise redu cti on .
where <p~= arg[G( ± rt )], 'Pi= arg[g (±rt )].
In Section 9.5 , we will discuss the homod yne detec ti on measur ement 4.7 Problems
who se output curr ent sp ectrum is relat ed to th e quantit y Sx'P(rt) defined
through th e following expression: Problem 4.1 Th e derivati on of Eqs. (4.59) and (4.64).

(4.73) (i) Set 6 = r,afbt - r,*ab, prove th at [a, O] = r,bt , [bt, O] = r,*a.
where (ii) Usin g the result of (i) and Eq. (3.121 ), prov e Eq. (4.59) and the following
A A At - A At
SabaSab - Ga - gb . (4.77)
(4.74) (iii) Using the left- hand side of Eq. (4. 77) and definition of I'rlab ) in
with 'P,_.as th e phase of the local oscillator in the homodyne detection. Notice Eq. (4.57) , prov e
that XJ(rt) = X cp(-rt). This quantity is similar to the quadrature -phase ( 4 .78)
~mplitud e defined in Eq. (3.41) for a singl e-mode field . We can regard (iv) Substitute l'rlab) in Eq. (4.63) into Eq . (4.78) t o show Eq . (4.64) , i.e.,
it as the multi -mode equivalence [Caves and Schumaker (1985)]. Using the final form of I'rlab).
Eqs. (4.70) , (4.71) and (4.72), when 2<p= 200 = 'Pc ++<p-+1r= rn-+rn++1r
. . g re rg ' Problem 4.2 Heralded singl e-photon state
we obtam the two conJugate quadrature -phase amplitud es:
For the multi -frequency two-photon state /'112)in Eq. (4.53) , the de-
x (rt) = X0 0 (rt) = (/G(rt) I - /g(rt) /)X~?\rt) ei<t>
o,
0 t ection of a photon at time t = T 0 in field B will project the state /w2) into
Y(rt) = X0 +-n-;2(rt) =
0 (/G(rt)/ + /g(rt)/)X~f ~71"
12(rt)ei <t>
a (4.75) A) = B\To/'lF2) for field A where
/'l/J

with 0b= 1r /2 + ((f); - 'Pc)/2 and ¢0 = ( (f)t -
defined in Eq . (4.74) but with a0 (w0 ± rt) .
From Eq. (4.75) , we have
'Pc)/2 . x~?\rt) is simi larly /To)B = E1(To)/va c) with EB(To) = j dwe-iwToa,B(w). (4.79)
0
Show that the projected state /?j,,A) is a single -photon state of the form in
Eq . (4.43). Find Tin term of To.

(X(rt)xt(n ' )) = (/G(rt)I - /g(rt)/) 25(n - rt ' ) P robl e m 4 .3 Using Eq. (4. 78) and the operator algebra in Section 3.6,
= Sx(D)5(D - rt') prove that the coherent state representat ion of the twin beams state is
given by
(Y(D)Yt(n ' )) = (/G(D)/ + /g(rt)/)25(O - rt')
= Sy(rt)5(rt - rt ' ). (4.80)
(4.76)
 Q·uanlum Opt ics For E xprr im cntal ist s

Problem 4.4 Photon bunching effect for pulsed multi-mode thermal stat e.

Consider a pulsed thermal field described by J,.J temporal modes

{.fj(t)}(j = L 2, .... M): Chapter 5
l\I
E(t) = LEj]j(t) , (4.81) Theory of Photo-detection and
j= l

where {EJ}(j = 1.2, ... , J,.I) are independent complex Gaussian random
Quantum Theory of Coherence
variab let-5with the same average absolute square: (IEJl2 ) = 10 , which are
similar to the comp lex random variable introduced in Section 1.6 , and the
mode functions {Ji (t)} (j = 1, 2, ., . , M) satisfy the orthonormal relation:

J dtJJ(t)fk(t) = bjk · (4.82)

Now we have learned how to describe a complicated system quantum me-
chanically in terms of quantum states. This description is developed from
For fast pulse detection, the pulses are usually much faster than the re- the fundamental theory of quantum mechanics and is thus logically self-
sponse of the detectors so that the observed intensity is a time integral: consistent. However , how do we know this description is correct? In phy-

J
sics, we know the ultimate test is the experiments. As we will see in this
2 chapter , the photo-detection theory is the bridge between the theory and
f = dtlE(t)l . (4.83)
experiments in quantum optics. It connects the experimenta lly measura-
(i) Evaluate (I) and (12 ). ble quantities with the quantum states that we use to describe the system.
(ii) Show that the normalized intensity correlation function These measurable quantities are the subjects of the coherence theory of
(J2) 1 light, which is about the fluctuations of optical fields in both space and
g(2) = -
- (1)2
= l+-
M. (4.84) time. The coherence theory uses statistical methods to describe an optical
field and this description is closely related to experimentally measurable
This is generally true for arbitrary thermal fields with Gaussian statistics
quantities through the photo-detection theory . We will examine these in
and M is the average number of modes of the field [Goodman (2015);
this chapter.
de Riedmatten et al. (2004)]. We will come back to this when treating
photon bunching effect in spontaneous parametric processes in Problem
7.1 of Chapter 7 (Eq. (7.64)). 5.1 Classical Theory of Coherence and Semi-Classical
Theory of Photo-Detection

5.1.1 Classical Coherence Theory

The classical theory of optical coherence was developed around the end
of 1940's by Emil Wolf and was first documented in his classic textbook
"Principles of Optics" with M. Born [Born and Wolf (1999)]. It mostly
deals with optical interference effect . It is based completely on the wave
theory of light and treated electromagnetic fields as random variables. The
t ime evolution of the fields is thus described by stochastic processes. In
this aspect, coherence theory of light is also known as statistical optics in
ana logy to stat isti cal mechanics. In comparison, classical Maxwell electro -

129
130 Q'llantnm Optics For Experimentalists Theory of Photo-drtcction and Quantum Theory of Coherence 131

magnetic theory of light is equivalent to classical Newtonian mechanics and where I(r. t) = IE(r , t)l 2 is proportional to the intensity of the field. If we
is detenninistic. It explains well the phenomena such as interference and write E = IEie 1
<P, then Eq. (5.5) becomes

diffraction of light waves. Similar to statistical mechanics, coherence theory

rg· 1\r1. t1; r2. i2) = (IEi(r1, ti)Ej(r2. t2)lei[<p(r:d2 )-<p( r i,tl )l)p. (5.7)
of light uses statistical methods to study the fluctuations of electromagnetic
fields through such quantities as correlation functions.
Consider the general solution in Eq. (2.22) for the l\Iaxwell wave equa-
Since ei'P is a fast varying function of sp, r;J·
1
) is more about phase corre-

lation whereas r( 2 -2 ) is only related to intensity corre lation.

tion with A.-\(r) in Eq. (2.44): Another quantity deals with fourth-order coherence and is associated
A(r , t) = L qk,s(O)Ek,sUk,s(r)e-iwt + c.c .. (5.1) v:ith two-photon interference phenomena:
k ,s
rir 2\r1, t1; r2. t2) = (E7( r1, t1)E;(r2, t2)Ei(r2, t2)Ej(r 1, t 1))p. (5.8)
where uk,s (r) is the spatial mode function, Ek,s describes the polarization
mode, and w = lklc. qk,s(O) gives the mode excitation. The electric field 5.1.2 Semi-Classical Theory of Photo-Detection
can be obtained from Eq. (2.14) (we will not give the magnetic field because
To see how the correlation functions are related to what we can measure in
intensity is related to the electric field only):
the lab, we look into the detail of the photo-detection process.
E(r , t) = L Ek,sEk,sUk,s(r)e-i wt + c.c., (5.2) In experiments, we measure optical fields with photo-detectors via the
k ,s photoelectric effect where light fields interact with photo-sensitive media
to produce electric currents and optical energy is converted into electric
where Ek,s = ikqk,s(O) describes the electric field excitation for mode k , s.
energy. However, as is well-known, classical wave theory has difficulty in
In coherence theory of light, {Ek,s} are a set of random variables with
exp laining photoelectric effect and the concept of photon was introduced
a known probability density P( { Ek,s}). So, the electric field is a random
process. The measurable quantities (also known as observables) are the by Einstein to understand it [Einstein (1905)]. On the other hand, Mandel,
Sudarshan and \Nolf in 1964 introduced a semiclassical theory of photo-
correlation functions:
detection [Mandel et al. (1964)] and successfully explained with classical
r(N ,J\J\ r1, t1; ... ; rN, tN; rN+l, tN+l; ... ; rN+f',J, tN+M) waves of light basically all aspects of photoelectric effect including the cut-
= (E; 1 (r1, t1) ... E; N (rN, tN )EJN+i (rN+l, tN+i) ... EJN+r>I(rN+f\I, tN+M )) p off frequency n_w> rLwc= Wo (Wo is the work function of the medium).
In the semiclassical theory, even though light fields are described as
(5.3)
electromagnetic waves , atoms of the medium are treated quantum mechani-
Here, cally. The photoelectric process is modeled through the interaction between

= L Ek ,s [Ek,s]jUk,s(r )e-iwt
atoms and light waves via the interaction Hamiltonian
Ej (r, t) (5.4)
A e
k,s H1(t) = --p(t) · A(r , t), (5.9)
m
is the _j-th component of the positive frequency part of the electric field in
where p(t) is the momentum operator for electrons in the atoms and A(r, t)
Eq. (5.2) and the average in Eq. (5.3) is with respect to the probability
is the vector potential in Eq. (5 .1) for the electromagnetic waves. Using the
distribution P( { Ek,s}). The most commonly used correlation functions are
perturbation theory in quantum mechanics, we can calculate the probability
the second -order field correlation function:
of ejecting (ionizing) an electron, which is a photoelectric event that gives
(5.5) rise to the photoelectric current for measurement. In this calculation, we
have to first assume the field is deterministic or it is one of the realizations
and the fourth-order field correlation or the intensity corre lation function:
in the statistical ensemble of its stochastic process. With some reasonable
r( 2,2) (r1, t1; r2 , t2) = (IE(r 1, t1)12IE(r2, t2)12)P = (I(r1, t1)I(r2, t2))P, (5.6) assumpt ion about the bandwidth of the field and the time response of the
 ...
1:32 Quantum Optics For E.rperi111entalists Th eory of Photo-d et ectio'II and Qua11tum Theory of Cofi c,.e11cc

medium, we may find that the differential probability !::..Piof obtaining one For a fluctuating field, we must treat the field quantity E(r. t) as a
photoelectron within a relatively small t im e interval !::..tis given by random variable and make an ensemble average of all the expressions above
!::..Pi= 17I(r,t)!::..t, (5.10) over the probability distribution P( {Ek.s}). We then obtain probability
densities:
where I(r. t) = IE(r, t)l2 is proportional to the intensity of the field and 17
is some proportional constant related to the atoms. P1(r. t) = !::..Pi/1::..t= 17(I(r.t) )p. (5.15)
We can also find the joint differential probability !::..P2 (r. t: r', t') for two p 2 (r. t: r'. t') = f::..P2(r.t: r'. t') / !::..t!::..t'
= 1717'
(I(r. t.)I(r', t') ) p. (5.16)
photoelectric events to occur. Assuming the two events are independent if p N (r 1 , t l : ... ; r N , t N) = 7}l ... 17N (I (r l · t 1 ) · · .J ( r N , t N)} P · (5.17)
atoms are loc alized and separated, the joint probability is the product of
The same is true for the probability P(n, T) of n events in a finite time
t he probabilities of the two separate events at r , t; r', t':
interval T:
= 1717
.l l
1
f::..P2(r,t; r', t') I(r, t)I(r', t')!::..t!::..t'. (5.11)
This can be extended to N multiple photoelectric events: P(n, T) = ~! ( [TJ I(r, t)df exp [ - T) I(r, t)dt]) P. (5.18)

f::..PN(r1,t1; ... ; r N, tN) = 1]1---17NI(r1,

t1) ...I(rN, tN )!::..t1
...!::..tN-(5.12)
If atoms are separated within a distance that is much smaller than that
over which the field changes, the formulae above are correct even for the
same location: r 1 = ... = r N = r.
So far, we only dealt with photoel ectr ic events in one or more diffe-
rential time interval !::..tand we find the differential probability of finding
one photoelectron in !::..tis given in Eq. (5.10), which is much smaller than
unity due to short time interval. Then, this differential quantity is also
the expected average number of photoelectron within !::..t.1 So, for a finite
interval T, the total average photoelectrons is simply the sum of all the
events in the interval:
(n)r = 11 l I(r , t)dt.
If all the photoe lectric events are independent of each other, it is a good
approximation that they follow the Poissonian distribution with an average
number (n)r in Eq. (5.13). Then we obtain the probability P(n, T) of n
events in a finite interval T:
P(n, T) = ( (n)~ )n e-(n)r
n.
Later on, we will present a similar expression for quantized fields from Glau-
ber's full quantum photon-detection theory. But before going to quantized
fields, let us first look at the application of the classical coherence theory
to Hanbury Brown-Twiss effect.
5.1.3 More on Classical Explanation of Hanbury
Brown- Twiss Effect
In Section 1.5, we discussed in detail the second-order coherence function
and its relation to the phase correlation of optical fields. We will not repeat
it here except that we need to point out that the basis for that discus-
sion is Eq. (5.15) for photoelectric events in a single detector. In Sections
(5.13) 1.6 and 4.2.4, we discussed the intensity fluctuations of the optical field
with an implicit assumption that detectors measure directly the intensity
of the optical fields. Now with the semiclassical theory of photo-detection in
Eqs. (5.15) - (5.17), we can explore what the experimental observations will
be and how they are connected to various correlation functions of the op-
tical fields. Without the involvem ent of quantum theory of light , Hanbury
Brown -Twiss exper iment is the perfect platform to discuss the semiclassical
theory of photo-detection and most of the discussion can be applied to the

= ~! l
[ry I(r , t){ exp [ - 7/ l I(r , t)dt] (5.14) quantum theory of photo-detection.
In experiments, intensity correlation can be measured by two detector s
Notice that the formula above for Poissonian distribution is true only if the through coincidence measurement (see Chapter 7 for more). We will count
field is deterministic and the randomness is purely from the probabilistic coincidence of two photoelectric events within a time int erva l of !::..Tfor a
nature of quantum mechanics for the photoelectric events. time period of T. Take t' = t +Tin Eq. (5.16) . Then P2 (r, t; r', t + T)!::..t!::..T
1
The expected value for the number of photo electrons is ~n nP n P1 because higher
is the joint probability that one photoelectron appears at time t within f::..t
order probabiliti es such as P2 rv (6.t) 2 is much smaller than Pi rv 6.t for short 6.t . and another one at time t + T within 6.T. Similar to Eq. (5.13), we find
134 Quantum Optics For Experimentalists Theory of Photo-detection and Quantum Theory of Coherence 135

the average rate R e of coincidence count within a finite coincidence window for time variable only. But here we include spatial variables. So, the ex-
6-T to be CC'SS intensity corre lation function A(r1, t1; r2 , t2) is directly related to the

R c(t)
= (n c(6.T) )1;,.t=
6-t
J~ 6.T

6.T
,+
dTp2 (r, t; r. t T) exp lains well the HBT effect (Section 1.2): whenever r1 .r2 are within co-
2
second-order coherence function: A(r1, t1; r2 , t2) = h( r 1, t1; r2, t2)1 - This

-~
herence area of the field, there is an excess intensity fluctuation, otherwis e,
= w/ !~ -~
6.T

6.T
dT(I(r , t)I(r' , t + T)). (5. 19) there is none. It is interesting to note that although I is best to characterize
phase corre lation, it nevertheless is related to excess intensity corre lation
Normally, photo-detectors and the subsequent cir cuit have some intrin sic A, wh ich h as nothing to do with phase. This is all because of the Gaussian
time resolution TR. Then 6-T 2 TR. Equation (5.19) connects the intensity nature of the probability distribution.
corre lation function with the meas urable quantity R e. For the HBT exper im ent with a beam splitter and time de lay, we take
Let us write r = r' in Eq. (5.21) and assume the field is stationary. Then we can write

(I(r, t)I(r', t + T)) = [l + A(r, t; r' , t + T)](I( r , t))(I(r', t + T)), (5.20)

(5.23)
where quantity A characterizes the correlation in intensity fluctuations of
the fields. Then Eq. (5.19) becomes where A(T) is related to the excess intensit y fluctuations of the incoming
2 field and A(T) = b( T)l2 for the thermal light. When the resolution time TR
R e= R1R2TR [1+ _2_
TR
JTRdTA(r,
_Tf
t; r', t + T)], (5.21) of the detection system is much larger than the coherence time Tc of the
field, b (T)I rv 0 for T = TR/2 » Tc so that we can replace the integration
where, according to Eq. (5.13), R 1 = TJ(I(r, t)), R 2 = TJ1 (I(r' , t + T)) are limits with ±oo:
the average rates at which photoelectric events arrive at the two detectors ,
respectively. Here we take 6-T = TR and assume R 1 , R 2 do not change (5.24)
significantly within TR so that we can pull them out of the integral. The
first term in Eq. (5.21) stems from purely accidental coincidence event from So, Eq. (5.23) becomes
the two detectors as if the fields are uncorrelated. The second term is from
the excess contribution due to correlation in intensity fluctuations. R, = R1R2TR[1 + ;J (5.25)

Hanbury Brown-Twiss Effect

For a thermal field , the probability distribution P( { Ek,s}) for the field in
Eq. (5.4) is Gaussian . We then use Isserlis Theorem in Eq . (1.28) for multi-
variate Gaussian distribution to obtain for polarized field
(I(r1, t1)I(r2, t2))P = (IE(r1, t1)12IE(r2, t2)12)p
= (IE(r1, ti)l 2)P(IE(r2, t2)12)p
+ (E(r1, t1)E*(r2, t2))p(E*(r1, t1)E(r2, t2))P
+ (E*(r1, ti)E*(r2, t2))p(E(r1, t1)E(r2, t2))P
= (I(r1, t1))(I(r2, t2)) [1 + h(r1, t1; r2, t2)12],
where the last term in the second equation is zero because of random phase
fluctuations. The equation above is the same as Eq. (4.29) in Section 4.2.4
Experimentally , we can define a measurable quantity g(2 ) = R e/ R1R2TR.
Then we have g( 2 ) = 1 + Tc/TR = 1 + 1/ M with M = TR/Tc. In the lan-
guage of temporal mode, Tc is roughly the size of a single-temporal mode.
Then NI = TR/Tc is the number of distinguishable or non -overlapping (ort-
hogonal) temporal modes detected by the detectors within the detector's
resolution time TR. This relation is the same as that in Eq. (4.84) of Pro-
blem 4.4 for pulsed fields. Here, we proved it for the CW field.
The result above can be applied to stellar intensity interferometry where
intensity correlation at two different locations is measured for the light from
a distant star to determine hd via A(r1, r2). We learned from Eq. (1.45)
(5.22)
of Problem 1.2 in Chapter 1 that the spatial coherence function 1 12 can
be used to measure the size of a star by the method of stellar interfero -
metry [Michelson (1890, 1920); Michelson and Pease (1921)]. But here, we
136 Quantum Optics For Expc '1"i'lncntalists Theory of I'hoto-detectfon and Quantum Theo1·y of Coherence 137

can measure , 12 wit hout observ ing interference fringes as in the method of is the intensity operator of the optical field. 1\Iore generally for N photoe-
1\Iichclson, which can be very unstable due to atmospherica l fluctuations. lectric events, we have
This is a new type of ste llar inte rferometry based on intensity correlation
6.PN(r1.t1: ... ;rN ,t N) ex:(: i(r1.ti) ...i(rN.tN) :)1p6.t1 ... 6.tN. (5.30)
techniq ue [Hanbury Brown and Twiss (1956b)]. Ste llar intensity interfero-
metry was t h e first application of the intensity corre lation techn iqu e imme- Equations (5.26), (5.28) and (5.30) arc simil ar to the semiclassical equiva -
di ate ly after it was invented in the famous Hanbury Brown-Twiss photon lents in Eqs. (5.15) (5.17) except that the average is now over the quantum
bunching experiment [Hanbury Brown and Twiss ( 1956a)]. state 'lj;.
HBT effect with thermal light can also be app lied to optica l imaging Not ice t hat Eqs. (5.26), (5 .28) and (5.30) are for differential probabilities
based on int ens ity corr elation through a strange ··ghost im age" phenomenon valid on ly for infinit esimal time int erva ls and cannot be normalized by time
[Pittman et al. (1995); Bennink et al. (2002) ], which is a kind of nonlocal integration. Based on Glauber's quantum theory of photo -d etect ion , Kelley
fourth-order classical wave interference effect [Gatt i et al. (2004)]. and Kleiner [Kelley and Kleiner (1964)] derived the probability of counting
n photoelectrons in a finite t im e int erva l T. The result is
5.2 Glauber's Photo-detection
Theory of Coherence
Theory and Quantum
P(n, T) = ~! ( : [7J £ i(r, t)df ex p [ - 7) £
i(r, t)dt] : ),,- (5.31)

This formula is similar to Eq. (5.18) for the classic al wave theory .
5.2.1 Photo-Electric Measurement and Normal Ordering

In parallel with the work by Mand el, Sudarshan , and Wolf [Mandel et al. 5.2.2 Glauber's Quantum Theory of Coherence
(1964)] on the sem iclassical theory of photo-d etect ion , Glauber [Glauber
From the results of the quantum theory of photo-detection , Glauber then
(1964)] developed a full quantum theory of photo-detection, in which both
defines the general quantum correlation functions [Glauber (1963a)]:
atomic media and optical fields are described quantum mechanically. The
result is very similar to the semiclassical theory. In particular , Glaub er eva- G (N,l\J)(x 1 ·, X 2 · ··, · X N+M )
-
luated the differential probability of one photo elect ric event in a relativ ely = (.. EA(-) (xi) ... EAH (xN )EA(+) (xN+1) ...EA(+) (x N+M) .-)vi (5.32)
small time interval 6.t as
with xi = ri, k These quantities are directly related to the measurement
6.Pi(r, t) ex:(EH(r, t) · E (+\ r, t) )vi6.t , (5.26) of photoelectric events and relevant coherence phenomena. For example,
wher e the average is over the quantum state 'lj; of the optical field and the second-order quantity
E (+\r, t) = L l(k)ak ,sEk,sUk,s (r)e-iwt
= [E(-\ r, tW (5.27) c( 1,1\x1; x2) = (i;( -) (x1 )E( +) (x2) )vi (5.33)
k,s
has it s normalized quantity
is the positive frequency part of the electric field operator in Eq. (2 .82)
or Eq. (2.90) in Section 2.3.3. The joint differential probability of two 91 = c(l ,l ) (x1; x2) / G(l ,l) (x1; X1 )G(l ,l) (x2; x2)
photoelectric events is given by
directly related to the visibility of interference fringe and describes phase
6.P2(r1 , t1; r2 , t2) corre lation of the optical fields. The fourth-order quantity
A(-) A(-) A(+) A(+)
ex:L)E i (r1,t1)Ej (r2,t2)Ej (r2 , t2)Ei (r1 , t1)) vi6.t16.t2 c( 2,2\x1; x2) = (i;(-\x1)i;(-) (x2)E( +) (x2)E( +) (xi) )vi (5.34)
i ,j
is related to the intensity corre lation and coincidence measurement (see
(5.28)
more in Chapter 7).
where :: denote normal ordering of the creation and annihilat ion op era tors For a multi-mode coherent state given by
at, a and
i(r , t) = E C-\ r , t) • E (+)( r, t) (5 .29)
l'lj;) = II l{o:k,s}), (5.35)
k ,s
138 Q'Uant'Urn Optics For Experimenta lists Theory of Photo-detec tion and Q'Uant'Um Theory of Coherence 139

we have On the ot her hand , we lea rn ed from Sect ion 3.7.1 th at any arbi tra ry
quant um state of light can be describ ed by a density op erator whi ch has t he
(5.36) Glaub er-Sud arsh an P-r epr esent ati on or t he coherent repr esent at ion [Glau-
with E(r , t) = Lk ,s l(k) a k,sEk,seik-r-iw t b eing a compl ex numb er and nor- ber (1963b ); Sud arshan (1963)]:
mal ordering leads to p= J 2
d ala)(al P c(a) (5 .42)
(5.37) for a single-mod e field or
and
(5.38)
P= JITk ,s
d2 { a k,s} I{ a k,s}) ( {a k,s} IPc ({a k,s}) (5.43 )

for a multi-mode field. Here fun ctions P c (a ) and Pc( { a k,s}) are som e
For the multi-mode coherent state of Eq. (5.35) , the correlation function in
funct ion of a 's with normaliz ation:
Eq . (5.32) then becomes
c(N ,M\x1;x2 ... ;xN+M) = E*(x1) ... E*(xN)E(xN+1) ... E(xN+M)- (5.39)
J d2 a Pc (a) = l and / ITd
k ,s
2
{ ak ,s}P c ( {ak ,s}) = 1. (5.44)

If we let {ak,s} fluctuate and become random variables , then E(r, t) =
E({ak,s},t) is a random process. Suppose {ak ,s} are described by a pro-
bability distribution: P( {ak ,s}) = P( a 1, a2, ... ) , which is the same as
P( {Ek,s}) given earlier in classical coherence theory of light. Then we
need to average Eq. (5.39) over this probability distribution and Eq. (5.39)
becomes
c(N ,M) (x1; X2... ; XN+M) = (£* (x1) ... E* (xN )E(xN+1) ... E(xN+M )) P , (5.40)
which is in the same form as the quantity r(N ,M) in Eq. (5.3) for classical
coherence theory with E (x) replaced by E (x).
5.2.3 Connection between Quantum and Classical Theory
and Optical Equivalence Theorem
Notice that when {ak,s} is a set of random variables, we are not certain
about which coherent states the system is in and the uncertainty in de-
termining the quantum state leads to a statistical description of quantum
state for the system . In this case, we use a statistical mixture of coherent
states or a mixed state given by the density operator:
Comparing Eqs. (5.41) and (5.43) , we find that if Pc = Pel, then quan-
tum and classical theories of light are equivalent. This is the so-called
Optical Equivalence Theorem [Sudarshan (1963)]. So, it seems that there
is no need for a quantum theory of light and the classical wave theory can
cover the quantum theory if Pc = Pel. However , this is a big "if'' and
this is where the difference lies between the classical wave theory and the
quantum theory of light.
5.2.4 Classical and Non-Classical States of Light
Let us examine the difference between the two distributions of Pc and Pel.
We start with Pel, which we know better. It is a probability distribution so
it is normalized to one: J rrk ,s d2 { ak ,s} Pel ( {ak ,s}) = 1 and Pel must not
be negative: Pel 0. Now let us look at Pc: it satisfies the normalization
equation in Eq. (5.44) just like Pel. But does it satisfy the non-negative
condition like Pel for an arbitrary quantum state? It turns out that the
answer is "NO".
To prove it, we just need to present a counter -example. Consider the
single -pho ton state 11) for one mode. We showed in Section 3.7.1 that it

Pel = JITk,s
d { ak,s} I { ak,s}) ({ak,s} IPcz
2
({ak,s}) (5.41)
has a Glauber P-representation of

P 11\a, a*)=
C oaoa*
ecw* ~c5( 2)(a) . (5.45)

for the quantum state of the system. The subscr ipt "cl" means that this In terms of real variables x, y with a = x + iy, a* = x - iy, we have the
quantum mechanical description by the density operator Pel gives exactly change of variables:
the same result as the classical wave theory, i.e., Eq. (5.40) for the quantum a a ax a ay 1( a .a) a 1( a .a) 5 5
description is identical to Eq. (5.3) of the classical description. Ba = ox oa + oy oa =2 ox - i oy ' oa* =2 ox + i oy . ( .4 )
1'10 Quantum Optics For Experimentalists Theory of Photo-d et ection and Quantum Theory of Cohc1·c11cc 1-ll

Then Eq. (5.45) b ecomes c1nd we must hav e Po(P) < 0 for some ,8. It is easy to show that the
Schrodingcr cat state in Eq. (3.51) satisfies this.
1 ex2+Y2 ( 32 32 )
Pb\x, y) = -4- 8:r2 + 8y2 5(x)5(y). (5.4 7) Since Pc is not always non-ne gat ive, in th e quantum descr iption of light ,
we can categorize the quantum states of an optical field into two types: (1)
Although PJ}-) is written in terms of the special functions 5(x). 5(y), we can those with Pc 2::0 for all and (2) those with Pc < 0 for some values of
take it as a limiting case of a Gaussian function: o_. The former ha s a classical wave correspondence of Pc1 = Pc and can
be described by the classical wave theory whereas the latte r does not have
1 . 2/2 2
5(x) = lim --e-x a- (5.48) such a correspondence and cannot b e described classically. In this sense,
a--+ 0 CJ .J2ir we call the form er "classical states" and the latter "non-classical " .
and An example of classical states is a thermal state whose Glauber-
2 Sudarshan P- representation is derived in Section 3. 7 .1 and has following
d 1 2 2 -x 2 ;2 0-2
- 5(x) -_ hm.
f<L ( x - CJ ) e . (5.49)
dx 2 a---+O CJ5 v 21r form
In Fig. 5.1, we plot a Gaussian function of width CJ and its second derivative. (5.52)
We find that the second derivative is negative for lxl < CJ. So, Pf ) is not
non-negative for all values of a. This is not a surprise since a single-photon with n as the mean photon number. This function is always positive and can
state describes the optical field as a particle of photon which does not have serve as a true probability distribution Pct for the classical wave description.
a counterpart in wave theory. So, the quantum exp lanation of the HBT experiment is exact ly the same
In fact, if two states are orthogonal with Tr(,81 ,82 ) = 0, then we have as the classical exp lanation. Besides thermal states , coherent states are
classical state. So, the light fields from lasers can be described classically.
2 2
Tr(,01,82) = Tr [ / d aPg\ a) la) (al / d ,BPg\ ,e)l,8)(,81] As shown earl ier, an example of non-classical states is a single-photon

= J d2 ad 2 ,BPg\a)Pg\,e) I(al,8)12 = 0. (5.50)

Since l(al,8)12 > 0, one of Pg\a) and Pg\,e) must be less than zero
in order for the equation above to stand . Especially if a state ,o =
J d2,BPc (,8)l,8)(,Blis orthogonal to a coherent state lao) with (aol,olao)= 0,
Eq. (5.50) becomes
state. All photon number states with a non-zero number of photon are
non-classical states since their P-functions are higher order derivatives of
5-function. On the other hand, there are many other non-classical states
which may not have an explicit form of Glauber P-function. But how do
we know these states are non -classical? As we will see, there are some
non -classical phenomena that we can observe to find out if a light field is
non-classical.

(5.51)
5.3 Anti-bunching Effect
1u~-- ----- --~ 0.1 ~----------
The first observed non-classical phenomenon is the photon anti-bunching
effect of light [Kimble et al. (1977)].
f(x):::::1 / \ I .f"(x)",
:'I Consider int ensity correlation function
0.1 -0.115

0.05 0.1

0
.(, 6
.(),J'ic__
.(,
____ ___ __ _
2 2
f ( , \t, t + T) =(: i(t + T)i(t) :)p
X
(a) (b)
X
= \j;(-)(t)ft(-\t + T)E(+)(t + T)ft(+)(t))p
F ig. 5.1 (a) Gaussian function of width CT = 1.5 as a limit for a o-function; (b) its
= ([*(t)[*(t + T)[(t + T)[(t))p 0
second derivative. Note that there is a region where/' (x) < 0, leading to Pc < 0. = (J(t + T)f(t))p 0 , (5.53)
 ..
1<!2 Quantum Optics For Experimentalists Theory of Photo-detection and Quantum Theory of Coliernice 1<!3

where I(t) = IE(t)l2. If Pc is a true probability distribution, then I(t), I(t+ 5.4 Photon Statistics and Photon Correlations
T) must satisfy the Cauchy-Schwarz inequality:
Another nonclassical phenomenon occurs in photon statistics. As we will
(5.54)
see in the following, classical states and some nonclassical states will give
For a stationary process, we have (I 2 (t + T)) = (I 2(t)) = (! 2(0)). Then rise to completely different statistical behavior in the number of photons in
Eq. (5.54) becomes those states.
Given an arbitrary state of light, what is the probability of finding the
(5.55)
system in { lni)}? Or we make photon number measurements of all the
after dividing both sides by (!(0))2. Since g(2)(T) • 1 for T • oo, we modes N1, N2,..., what is the probability of finding outcome n 1, n 2, ...(=
also have g( 2 ) (0) 2:: 1. Therefore, all classical states of light must satisfy {n})?
g(2 )(o) 2:: 1 and g( 2 )(T) :; g( 2 )(0). We learned that for a thermal state, For a pure state l?J;)= L{n} c{n}l{n}), we have
g;~\O) = 2 > 1 and this is the so-called photon bunching effect.
On the other hand, if Pc is not a true probability distribution, the
P({n}) = ic{n}l2 = l({n}l?J;)/2 = ({n}l?J;)(?J;l{n})
Cauchy-Schwarz inequality may be violated and we may not have the pho-
ton bunching effect. Therefore, a violation of either of the inequalities = \ {n} /p?µ/{n} ) . (5.59)
g(2 ) (0) 2::1 and g( 2 ) ( T) :; g(2 ) (0) means that the optical field is in a non-
classical state of light. For example, a single-photon state will have zero
intensity correlation function:

(: i(t + T)i(t) :)11) = (llE(-\t)EH(t + T)E(+\t + T)E(+\t)ll) P( { n}) = ~Pil({n}/?J;i) /2

=0 (5.56)

because there are two annihilation operators in E(+)(t + T)E(+)(t). So, we

= Pi ( {n} I?J;i)(?J;iI{ n})
have gfJ;(O) gfJ;(T)
= 0 = for any value of T.
This leads to the photon
anti-bunching effect gfJ;
(5.60)
(0) < 1.
The photon anti -bunching effect was first observed by Kimble, Dage -
nais, and Mandel in the resonant fluorescence from single atoms of Sodium Let us write p in Glauber-Sudarshan P -representation: p =
[Kimble et al. (1977)] (see Section 7.3.2) and later by Diedrich and Walther J d2 { a} Pc( {a}) I{a}) ({a}/. Then we have after substituting into Eq. (5.60)
from single ions [Diedrich and Walther (1987)].
Another violation of Cauchy-Schwarz inequality is by the intensity cross -
P({n}) = Tr(p/{n} )( {n}/)
correlation function

(5.57) = j d {a}Pc({a})1r(/{a} )( {a}/{n} )( {n}/)

2

which, for classical states with Pc 2::0, satisfies = Jd {a}Pc;({a}) ITla~'.~n;

2
e-1°,I' (5 .61)
l(Ii(t)h(t))l2 :s;(If(t))(I?(t)) or [gg)]2 :s;gii)gg) _ (5.58)
i

This inequality is violated by the two -photon state from parame- Next we use the property of normal ordering : f.lm := a,tma,m_ This
tric down -conversion . We will discuss this non-classical phenomenon in gives (al : f.lm : la) = (a/atma,m/a) = a*mam and (al : ef3N : /a) = ef31al2,
Section 7.4. which leads to (lal 2 m/m!)clal
2
= (al : (f.lm/m!)e-f.l : la). Then we can
 Quantum Opti.rs F01· E :tperimcntalists Theory of Photo-detection and Qua'llfum Theory of Coherence l.J.G

rewrite Eq. (5.Gl) as For an arbitrary state described by a density operator p. we can
once again write it in the Glauber-Sudarshan P-rcpresentation: p =
Jcv·J2n; . J
P( {n}) =
J d2 {a}Pc( {a}) II ~e-lQ,I '2 2
d {a} Pc ( {a}) I {a} ) ( {a} I and the photon number probability is

=j
1

J d {a} Pc ( {a}) ( {a} I • :~ e- JV• I{n})

2
P ( n)
= d2 {a} pc ( {o}) II / Cl i \ : N~;e- N; : \n i) un
=L
. \ n1.
1

(5 .62)
=
Jund2 {a}Pc( {a} )~e-u (U JaiJ2)

= \ -;Je-u) Pc . (5.67)
Here {n} = {n1,n2, ... }. It is straightforward to see that (n) = L nP( n) = (U) Pc and (n( n - 1)) =
In practice , it is hard to count photon number in each individu al mode. En(n - l)P(n) = (U2 )Pc· Thus, we have (n 2 ) = (U(U + l) )Pc and the
But we can easily measure the total number of photons in the whole field: variance of photon number
Ntot = L i ni· We will then calculate the probability P(n) of finding the
2n) = (n2) - (n)2
(t:i..
total photon number Ntot = n:
= (U) Pc + (U 2 ) Pc - (U)ic
= (n) + (!:::i..
2
U)Pc· (5.68)
The second term in the last lin e of the equation above gives the extra
photon fluctuations that deviate from a Poissonian distribution. For a
classical state of light , Pc is a true probability distribution and Pc 2: 0.
= ~!(:(LNire-I:iNi:
\· (5.63) So, we have
2 2
i
(!:::i..U)Pc = ((U - (U)) ;Pc
Here we used the multinomial expansion identity:
= J 2
d {a}Pc({a})(U - (U))
2
2: 0. (5.69)
(5.64) 2
Hence , we have (t:i..n) 2: (n) and this leads to a super-Poissonian or Poisso-
nian distribution for photon number. Therefore, all classical states always
Defining the total number operator N= Li Ni, we have have super -Poissonian or Poissonian photon number statistics.
On the other hand, nonclassical states have Pc < 0 for some values
j\Tn , ) (5.65) of a's and this may lead to (t:i.. 2
U)Pc < 0 and (t:i..2
n) < (n), which gives
P(n) = ( : - e-N : _.
n.1 P the sub -Poissonian distribution. So, a sub -Poissonian photon distribution
always means nonclassical for the state of the optical field and this type of
For a multi-mode coherent state lw) = I{a}) such as the state from a
field cannot be described by classical wave theory. A simple example is the
laser , we obtain the Poissonian distribution: 2n) = 0 < n = (n). The amplitude -squeezed
number state In) for which (t:i..
un (5.66) state depicted in Fig. 3.5( a) also exhibits sub -Poisson photon statistics (see
P(n) = - e-u,
n.1 Eq . (9.80) in Problem 9.2)
where U = Li JaiJ2 = (n) is the average total photon number. From this, Although the photon concept is quantum mechanical and requires the
2
n) = U = (n), which is typical for
we have the photon number variance (!:::i.. quantization of the optical field, the photoelectrons produced in photo-
a Poissonian distribution. detection process can be described by both the semi -classical wave theory
146 Quantum Optics For Experimentalists
and the quantum theory of light as seen in Sections 5.2 and 5.3. So, expe -
rimentally we can count photoelectrons and measure their statistics totally
independent of how we describe the optical field. The counting proba-
bility distribution for photoelectrons is given in Eq. (5.18) by the semi-
classical theory of photo-detection, and in Eq. (5.31) for the full quantum
theory. It is straightforward to see that these two formulae are the same
if P({Ek ,s}) = Pc({ak ,s}), i.e., the light field is a classical field. Further-
more, we find Eqs. (5.31) and (5.67) are the same if we set U = TJfr dtI(t)
with I(t) = IE(t)l2 and E(t) given in Eq. (5.36). So, the variance in coun-
ting photoelectrons is the same as that in Eq. (5.68) and the conclus ion
about photon statistics is applicable to the statistics of photoelectrons that
is observable experimentally.
For example, for thermal fields, which can be described as classical
waves, from what we discuss in this section, we should observe a super-
Poissonian statistics for counting the photoelectrons. The effect of sub-
Poissonian photon statistics in counting photoelectrons was first observed
by Short and Mandel for the light from resonant fluorescence of Sodium
[Short and Mandel (1983)] .
5.5 Quantum Noise and Its Reduction by Squeezed States
and Twin Beams
The phenomena of quantum noise reduction by the squeezed states and twin
beams are also non-classical phenomena that cannot be explained by the
classical wave theory. We will prove their non-classicality in the following.
For simplicity of argument, we will only consider single -mode cases. The
multi-mode cases are discussed in Problem 5.1.
In the homodyne detection of optical fields (see Section 9.3), the out-
put photo-current fluctuation (!~.2i)is directly proportional to the quantity
X'P) = 1 + (: fj. 2x'P :) with x'P = ae-i'P + atei'P_Here, the first term
(C:!.2
of 1 corresponds to the shot noise contribution in photo-detection. With
Glauber-Sudashan P-representation and x'P = ae-i'P + a* ei'P, we have
2 2 2
(: C:!..X.'P
:) = (C:!.Xcp}Pc= ((x'P - \XcpJ)JPc
= j d {a}Pc(a,a*)(x'P
2
- (x'P)/ 2 0 if Pc 2 0. (5.70)
2Xcp}cl 2 1, or homodyne detection of
So, classical states always have (C:!.
classical fields have the shot noise as the lower limit in the photo-current
fluctuation. On the other hand, we know from Section 3.4 that squeezed
Theory of Photo-detection and Quantum Theory of Coherence 147
2X'P) < 1 for some rp. For the vacuum state, (C:!.
states have (C:!. 2x~ac) = 1.
Thus, squeezed states are non-classical states of light that have noise below
the vacuum quantum noise level. We will discuss more about the detection
of the squeezed states in Sections 9.5 and 10.1.2.
Violation of Cauchy-Schwarz inequality in intensity cross-correlation
function in Eq. (5.58) is an indication of non-classical correlation between
two optical fields. We will show in Section 7.4 that a two-photon state from
spontaneous parametric processes at low average photon level can give rise
to the violation. However, when the light intensity is high, the equality
in Eq. (5.58) is approached, leading to small or no violation. In this case,
the nonclassical behavior is exhibited in intensity difference between two
beams.
From the photo-detection theory in Section 9.2, the fluctuation in the
difference of the photo-currents of two detectors is proportional to the in-
tensity difference between fields being detected:
2i_) ex (C:!.
2(11 - 12))
(C:!.
2(11 -12) :),
= (11) + (12) + (: C:!. (5.71)
where 11 = a}a1 (j = 1, 2). The first two terms are again the con -
tributions from the shot noise of the two detectors, which are uncorre -
lated. For classical fields with Pc 2 0, it is straightforward to show
2(11 - 12) :) 2 0. We thus have the shot noise as the limit for in-
(: C:!.
tensity difference measurement of two classical fields. But for twin beams
with coherent state boost discussed in Section 4.6, we have from Eq. (4.66)
2(11 -12))
(C:!. = ((11) + (l2))/(G 2 + lgl2) < ((11) + (12)) = (C:!.
2(11 - l2))cs
( "cs" denotes coherent state). So, the non-classical correlation between the
intensities of the twin beams gives rise to noise reduction in intensity diffe-
rence that leads to detected noise below the classical shot noise limit. We
will discuss more about noise reduction in twin beams in Section 10.1.4.
5.6 Remarks about Normal Ordering and Its Relation with
Classical and N onclassical Phenomena
In most of the discussions about the difference between classical and quan-
tum theories, the emphasis is on the non-commutation of operators in quan-
tum theory. However, as can be seen in our discussions in the previous
sections about the nonclassical phenomena, we almost always end up with
some normally ordered quantities in our calculat ion . This reveals directly
the similarity and the subtle difference between the classical and quantum
148 Qnantnm Opt.ics Fo'I' Expc1·imenta.lists T/1rory of I'/1oto-detection and Q1rn11tum The01·y of Coherrncc

theory of light. The normal ordering seems to eliminate the difference of where Pt1,(Cl.>..17>.Jis gi,·cn in Eq. (5.52) ·with fl>. as the m·cragc photon
ordering of the operators and thus leads to the equivalence between the two number in each mode.
theories. that is, the classical and quantum theories provide the same ex- (iY) Assume {,8>.} have a common phase <p for all A. Find (6.d 2 ) for the
planations for a large class of optical phenomena. including the well-known 11rnlti-mocle squee'.0ed state:
Hanbury Brown-Twiss photon bunching effect. (5.76)
On the other hand, there indeed exists a difference between classical and >.
quantum theories, which lies in the calculation of these normally ordered where ll/J>.)is given in Eq. (3.64) of Section 3.4.2 with r = T>. for each mode.
quantities with the employment of the Glauber-Sudashan P-representation Find the minimum of (6.d 2 ) as the phase cp is changed. Compare it with
for taking average. The possible negativeness of the P-function rules out RHS of Eq. (5.74) and discuss.
its interpretation as a classical probability density and leads to no classical
equivalence. Since the P-function is a quasi-probability density, which is Problem 5.2 The photon number uncertainty of the amplitude-squeezed
normalized to 1, most of its values must be positive and negativeness of state.
the P -fun ction is thus rare. This means that the nonclassical states of light
are not common in optics. As we will show in the next chapter, nonclas - Prove, when the displacement quantity lal » the squeezing quantity
sical states cannot be produced from classical states by linear interaction r, the amplitude-squeezed state la, -re 2i'Po.) has an uncertainty in photon
(see Section 6.2.4). Nonlinear optical processes must be involved for the number as 6.n = J
(6. 2 n) with 6.n = e-r < ffl, J7l0
that is, the
generation of nonclassical states. amplitude-squeezed state has a sub -P oissonian photon number distribution.

2
5. 7 Problems Problem 5.3 g( ) for the modified coherent state.

A modified coherent state is a coherent squeezed state la,~) with small

Problem 5.1 Consider a multi-mode operator
« 1.
excitations, that is, !al, 1~1
* t 2 2
d = ~(,8>.a>. + ,B>.aJ. (5.72) (i) Calculate g( ) = (: n :)/ (n) 2 and expand it in the first two orders of
A

>.
lal, 1~1
-
Given the multi-mode density matrix (ii) Find the condition for which g( )
2
is minimum. What is the minimum

P= j d {a,}
2
l{a,})({a,}I Pc({a,}) with
2
/ d {a,}Pc({a,}) = L
value of g( 2 )?

(5.73) Problem 5.4 g( 2 ) for the number state.

2
(i) Calculate (d) and (d
Prove that g( 2) = (: n 2 :)/ (n) 2 is 1 - 1/ N for the number state IN).
).

(ii) Prove that

This result shows that the criterion for claiming that a state contains a
(6.d2) L ,8>.,B; (5. 74) single-photon state is g( 2 ) < 1/2.
>.
for any classical field with P( {a>.}) 0 at any value of {a>.}- (Hint: try Problem 5.5 Another Cauchy-Schwarz inequality for classical fields.
to write (6.d 2 ) in terms of the quantity (not operator) d = ~>,(,8>.a>.+
,a;at) and its fluctuation average (6.d2)p and prove (6.d 2 ) p > 0 for the We have seen in Section 5.3 how photon bunching effect for classical
?-distribution given). fields arises from the Cauchy -Schwarz inequality: (AB) 2 :s;(A 2 ) (B 2 ) when
I 1

(iii) Calculate (6.d 2) for thermal state with we set A= I(t), B = I(t+T). Now by setting A= 6.I(t) = I(t)-(I(t)), B =
6.I(t + T) = I(t + T) - (I(t + T)), show the following inequality
P({a>.}) = ITPth( a>.,fi>.), (5.75)
>.
lg( 2 )(0) -11 lg( 2 )(T) - 11 (5.77)
150 Qnant'Urn Optics For Exp e'rim ent al'ists

for stationary classical fields [Rice and Carmichael (1988)], with gC2 )(T) :::=::
(I(t + T)I(l))/(I(t + T))(I(t)) = (I(T)I(0))/(1(0)) 2 . This inequality was
violated by the fields emitted from cavity QED system [Foster et al. (2000)] Chapter 6
and optical parametric oscillator with a coherent state injection [Lu and Ou
(2002)], demonstrating another nonclassical effect.
Generation and Transformation of
Quantum States

The quantum states discussed in Chapters 3 and 4 are generated by nonli-

near coupling and linear transformation of different modes of optical fields.
Optical resonators (cavities) are special linear devices which are often used
to enhance nonlinear interaction between different modes of the fields. They
are also spectral filters to shape the frequency modes of the optical fields.
The combinat ion of linear and nonlinear devices can produce some quite
exotic quantum states with interesting properties. These are the topics of
this chapter.

6.1 Generation of Quantum States: Nonlinear Interactions

between Light Fields

The interesting quantum states encountered in the previous Chapters are

described in the bases of the eigen -states of the free field Hamiltonian in
Eq. (2.76). However, they cannot be generated from that Hamiltonian. To
produce those quantum states, we need nonlinear interactions, which lead
to coupling between different modes of the optical fields.
The most commonly encountered optical fields in the lab are those in
the coherent states, which are produced from lasers , and those in the ther-
mal states, which can be obtained from blackbody radiators or discharged
atomic vapor gas. Because of the favorable properties of lasers , they have
become popular sources of light to replace the thermal sources ever since
their invention in the 1960s. In this section, we will see how to produce
interesting quantum states from the coherent states via nonlinear optical
processes, which are the subjects of nonlinear optics.

151
 Quantwn Optics For E:rpc1·imentalists Gen r ration a11d Tm11s.for11rntio11of Quantum Statrs
152

6.1.1 A Brief Introduction to Nonlin ear Optics: where

Thr ee- Wave Mixing and Four- Wave Mixing pNL = \ (2 ) : EE+ \ (3 ) : EEE+ ... (6.4)
According to the electromagnetic theory, the total energy of an optical field In the interaction picture. the quantum state of the system evolves und er

c~
in a medium is given by 1 the unitary operator

U=
81r
J d3 r [D (r, t) • E(r. t) + B (r . t) · H (r . t)]. (6.1) ii= exp JdtiI,n} (6.5)

l\Iost of the optical media are not magnetic but are dielectric , so that B = H The time integra l in Eq . (6.5) and the spat ial integral in Eq. (6.1) will give
and D = E + 41rP with the electr ic polarization P given by rise to energy and momentum conservation, respective ly. The latter is also
known as phase matching in nonlinear optics. These conservation laws will
p = l 1) : E +l 2
) :
3
EE + :\_'() : EEE + ... (6.2)
restrict the number of modes that can effective ly coupl e to each other so
Here, f(l), _x'(
2 ), \ ( 3 ), etc . are tensors and : denotes the tensor product. The that we only need to cons ider a few modes of the optical fields. Let us start
first term of the express ion above leads to lin ear optics best described by with the simpl est case of three modes for three-wave mixing (TWM). The
the indices of refraction for isotropic media or birefringence for an isotrop ic most general form is
media such as optical crysta ls. It can be treated as a free field in the
(6.6)
formalism of Chapter 2 with some modifications.
where h.c. denotes the Hermiti an conjugat e. Energy conservat ion requires
A A a
A as signal
" w3 = w 1 + w2. Likewise, the simplest form for four-wave mixing (FWM) is
~-~-{j signal
pl
--,
I,~
X
A
1--- a ')
a.
121 }:: :,

idler
(6.7)
pump a 1 idler fl-
' with w1 + w 2 = w3 + w 4 . Note that the following form of int eraction
(a) (b)
iif;/:'J\I' = in11a!
a!a1a4
+ h. c. (6.8)
Fig . 6.1 (a) Three -wave mixing; (b) Four-wave mixing. is also possible from four-wave mixing but we do not consider it because in
practice , energy conservation requires w 4 = w 1 + w2 + w3 so that a4 field
The higher order terms in Eq. (6.2) are responsible for the pheno- has a much high er frequency. Furth ermor e, its effect can also be too weak
mena in nonlinear optics. The i 2 )_term corresponds to thre e-wave mixing to implement expe rim entally (see the end of Sect ion 6.1.4).
(Fig . 6.l(a)): two waves come from each factor of the product of EE and
the third is the generated wave , or the revers e process. Likewise, the i 3 )_ 6.1.2 Two-Photon Processes: Parametric Processes
term leads to four-wave mixing (Fig. 6.l(b)). i 2 Lnonlinearity normally
dominates in bulk solid materials such as crysta ls whereas i 3 )_term is the Normally, th e nonlinear coefficients x( 2 ) and x( 3 ) are very small, requiring
lea ding term in gaseous atomic and mol ecular media because x( 2 ) is zero strong coupling fields to produce significant nonlinear effects . So far , at
due to central symmetry of randomly oriented atoms and molecules. These least one of the waves in three-wav e mixing and two in four -wave mixing
nonlinear wave mixing processes give rise to coup ling and interaction among must be strong to have an observable quantum or classical effect in th e
the modes of the optica l fields. Th e interaction Hamiltonian stems from lab. Let us mak e the a3 field in Eq. (6.6) for three-wav e mixing or a3 , a4
these interactions has the form of in Eq . (6.7) for four-wave mixing strong so that we can replace them with

Hint = 1
2
J P
A NL
(r, t) · E(r , t)d r,
A 3
(6.3)
c-numb ers A 3, A 4 , respectively. Th en we obtain th e Hamiltonian for the
parametric processes:

1W e are using cgs unit h ere . (6.9)

15--1 Qu.antum Optics For E :r:
perim ental-ists Generation a11d Trnn!:ijonnation of Quantmn Stales 1.ss

where ( = 17A 3 for three-wave mixing or ( = 17A3A.Jfor fom -wave mixing.

2 6.1.2.2 Generntion of squee zed states
This is a two-photon process since two photons in a1, a2are sinmltaneous ly
YVhen the two quantum fields become the same field, i.e .. ih = a2 = a, the
annihilated or created. This Hamiltonian will lead to an evolution operato r
interaction Hamiltonian in Eq. (6.9) changes to
for the state of the fields:
(6.10) s = l(l'f=.(/'At2
H mt
A

..,a - ',/'* aA2). (6.15)

with~ = -(*t . If initially on ly a 3 , a<-.1
are excited with lasers and a1, a2 are
in vacuum, the state of the output field is This is the reverse process of second harmonic generation . The state of the
IW) = S'12(~)lvac) = e,;<'harCaia1 lvac). (6. 11) system evolves from vacuum to
This state is exactly the two-mode squeezed state discussed in Section 4.6.
In some cases, it is more convenient to consider the evolution of the (~ = -(*t): (6.16)
operators defined as
which is the squeezed vacuum state discussed in Section 3.4.
(6. 12)
Other states such as coherent squeezed states can be produced by su-
This is sim ilar to the Heisenberg picture but U is only related to the in- perposing a coherent state and a squeezed state with a beam sp litt er. See
teract ion Hamiltonian Hint. On the other hand , with the final state I\fl l) Section 6.2 for more.
related to the initi al state I\fl i) by I'11l) = u_P
IWi), we have the expectat ion
value (O)(t) = (w 1101'111) = (wi1u_Pt6 u_P1wi) = (wilO(t)lwi)- So, it is
meaningful to evalu ate O(t). 6.1.3 One-Photon Process: Frequency Conversion
For 6 = a1 , a2 , we obtain (Section 4.6)
When a 1 field in Eq. (6. 6) and a 1 , a4 fields in Eq. (6. 7) are strong and can
= Ga1 + gei a,; (6_13)
0
~ 1= a1~P be replaced by c-numbers, we obtain the following Hami ltoni an :
{ A2 = uPr a2Up = Ga2 + geio aI

with G =
cosh l~I, g =
sinh l~I and eio = Ul~I - This is also the evolution (6.17)
equation for a parametric amplifier.
where ( = 77A1for three-wave mixing and ( = 77A1A 4 for four-wave mixing.
This is a one-photon proces s in which only one photon is annihilated or
6.1.2.1 Gen eration of two-photon and four-photon states
created at a time. It annihilat es one photon in mod e a3 but in the meantim e
For l~I « 1, we make an expansion of the expo nential in Eq . (6.11) and creates another in mode a2 or vice versa. If mod es a2, a3 have different
obtain the state in numb er state base as frequencies, this process realiz es a frequency conversion of photons. Th e
(6.14) first frequen cy up-conversion of a quantum field was realized experimentally
With small, th e second term dominates and this is a two-photon state by Huang and Kumar [Huang an d Kumar (1992)] for a squeezed state. It
that is widely used in two-photon int erfer ence. See Section 8.2 for more. was also applied to convert photons at optcom wavelength of 1550nm to
For four-photon coincidence measur ement , the first two terms hav e no con- photons of 800 nm at which th e photon counting technique is more matur e
tribution and the dominating term is the four-photon state l2)i 12)2 with [Vandevend er and Kwiat (2004)] or the reverse proces s in which a photon at
two photons in each mode of 1 and 2. the atomic transition wavelength around 800 nm to optcom wavelength of
Other states such as entangl ed states can be produc ed by using beam 1550nm for quantum communication [Ding and Ou (2010); Takesue (2010)].
splitters on the two-photon and four-photon states in Eq. (6.14) . See As we will see in Section 6. 2, the Hamiltonian in Eq. (6 .17) gives rise
Section 6.2 for more. to an evolution operator of uF = exp(~ala2 - Ca3a!), which is for a
2 Hfn':1
M' in Eq. (6.8) can a lso give the same Ha miltonian but requir es much higher linear lossless beam splitter with amplitude transmissivity t = cos l~I and
frequency field in 0,4 than all other fields , which is not practical. reflectivity r = (Ul~I) sin 1~1-
156 Q'Uantum Optics For Experimentalists Ge11emtiori a11d Tmnsformahon of Qnanf'U,117States 157

12)p
As
;;1
(I

x,Ol
8
I 1; >.
I
11 >;:::;.,
As
xf3l
I ,'~r p
On the other hand, the condition of 0 = 1r /2 or = 1r /2/2 is harder
to achi eve than the parametric process in Eq. (6.9) from four-wave mixing
because ex:17A.1so that much higher power at A.1 is needed.

(a) (b)

6.1.5 Multi-mode Parametric Process

Fig. 6.2 A photon number conve rt er.
6.1.4 A Photon Number Doubler
In the Hamiltonian for four-wave mixing in Eq. (6.7), if we only let one of the
fields, say a 4 be strong and replace it with a c-number A4 and furthermore ,
we let a 1 , a 2 be the same field: a 1 = a2 , we obtain the Hamiltonian for a
photon number converter (Fig. 6.2):
HANint = in·tcc(
;-A AP ;-*At A2)
..,a3a2 - ',, a3a2 (6.18)
with ( = 17A4 . The time evolution operator has the form of
2
i)(N\~) = e(a1a;-Ca3a1
with = -(*t. This process converts one photon in a3 to two photons in
a 2 . This is somewhat similar to the parametric process in Eq. (6.9) but can
do it in an efficient way if~ rv 1, as we will see in the following.
Consider the case when a3 is in a sing le-photon state and a 2 is in va-
2
cuum: l'11)in = ll)3I0)2- Defining A= ~ala~-(*a3a~ , it is straightforward
to show that All)3I0)2 = -C\/210)312)2 and AI0)312)2 = ~\/211)310)2- So,
using these two identities and after expanding the exponential in Eq. (6.19)
in an infinite series, we obtain the output state as
with 0 = \/21~1 and eJ5 = Ul~I- If 0 = 1r /2, we can achieve 100% conversion
from I1) 3 to I2) 2 . The process is reversed if the initial state is IO)312)2:
6
U(N\~)10)312)2 = cos0I0)312)2 - ej sin0ll)3I0)2.
This reverse process with 0 = 1r /2 can be used to take out the two -photon
2
state in a weak coherent state (la)2 ~IO)+ all)2 + (a /2)12)2 + ... ):
U(N\ 0 = 1r /2) IO)3la)2
2
I0)3I0)2 + al0)3ll)2 - ej<p(a /2)11)310)2 + ...
Note that for the a 2 term, a 2-field has no two -photon state, leading to
photon anti -bunching. 3
3 A two-photon destructive interference effect can also take out the two-photon
from a weak coherent state for photon anti-b un ching effect. See Problem 6.2.
In prev iou s sections, we treated the nonlinear processes with few-mod e mo-
dels , placing each wave for wave mixing in singl e mode. Though simpl e
and straightforward in demonstrating the physics, it is far from exper i-
mental rea lity where multi-mode excitat ion s are common. On the other
hand , it is also not practical to consider all the modes in the optical fields.
In exp erim ents, the spatial mod es are usuall y well-defined because most
lasers are from optical cavities which produce Gaussian spatial modes as
we discussed in Section 2.2.1. So, it is a reasonable assumption that all
the fields involved in wave mixing are in single spatial modes. Henc e, we
(6.19) can use one-dimensional approximation for each field , which only involv es
frequency and temporal modes. Furthermore , phase matching conditions
and energy conservation in nonlinear wave mixing [Boyd (2003)] place re-
striction on the polarization states of the optical fields so that only one
specific polarization state is allowed for each wave . This means that we can
treat the optical fields as scalar fields. Optical filtering , either spectral or
spatial, can further restrict the directions of propagation and the frequency
bands of the optical fields.
Since both three-wave mixing and four-wave mixing lead to the same
Hamiltonian for parametric interaction, we will only consider three-wave
(6.20) mixing due to its simplicity. Because of the restrictions above, we can
write the electric field operator in Eq. (6.4) in terms of separate waves:
(6.23)
(6.21)
for three-wave mixing. Here Ei =EH+ E(+) for i = 1 2 3 with
i i ' '
[E~-r = E~+) = ~2i
V L-7r
r dwai(w)ei(k iz -wt) '
j l:i.wi
(6.24)
(6.22)
where ki = niw / c and ni is the index of refraction which is a function of
w due to dispersion. Here we take the field operator in the wave propaga-
tion form of Eq. (2.105) in the one-dimensional and quasi -monochromatic
state
approx im ation with a fixed polarization state.
158 Quantum Optics For Experhnentalists Generation and Transformation of Quantum States 159

Substituting Eq. (6.23) into Eq. (6.3) and considering only the relevant For spontaneous processes, we have the output state
terms, 4 we arrive at
(1- 21(1) lvac) + ( Jdw1dw2<I>(w1,w2)lw1)ilw2)2
2
l<I>)
L
A]\J
Hint = in17
•

1 0
A(-)A(-)A(+)
dzE 1 E 2 E 3

with 77= 6E3 · (x( 2) : Ei E2 in, and L as the interaction length. The letter
)/
+ h.c.
AI denotes multi-mode here. Substituting Eq. (6.24) into the expression
above and carrying out the spatial integral and time integral , we have
(6.25)
+ e;Jdw1dw2dwidw;<l>(w,,w2)<l>(wi,w;)]w1,wi)i h, w;)2. (6.32)
Here lw1)ilw 2) 2 = a!(w 1)a!(w2)lvac) is a two-photon state and these-
cond term in the expression above is the multi-frequency two-photon state

1 1=
in -= dtHint
A ]\J
in Eq. (4.52) that we first encountered in Section 4.4.3. The A, B sy-
stems over there correspond to fields 1 and 2 here. lw1,w~)i lw2,w;)2 =

= J dw1dw2<p(w1,w2)a!(wi)a!(w2)a3(w1 +w2) + h.c. , (6.26)

aI (w1)a! (wDa!(w 2)a!(w;)lvac) is a four-photon state. We will discuss more
of the four-photon state in the third term above in Section 8.3.
When the peak power of the pump laser is high , as usually in the case of
where we used J eiwtdt = 2?Tc5(w)and
pumping by short pulses, ( becomes large and we cannot make an expansion
77£ sin /3 -i/3 of the exponential and write explicitly the output state. In this case, it is
<p(w1,w2) = vf2irTe (6.27)
easy to work with the evolution of the operators as in Eq. (6.13).
with (3 = 6.klw3 =wi+w2 L/2 and 6.k = k1 + k2 - k3 as the phase mismatch. In general, the evolution is very complicated and we have
When the a 3-field is in a coherent state from a strong laser (usually
known as the pump field), we can replace the operator with a c-number:
b1(w) = uta1(w)U = rG1(w,w')a1(w')dw'
11 + r 91(w,w')a!(w')dw'
111
a3(w1 + w2) • ap (w1 + w2) where ap (w) is the spectral profile of the strong (6.33a)
pump laser. Then Eq. (6.26) becomes
r G2(w,w')a2(w')dw' + 11
rg2(w,w')a!(w')dw',
1
1=
in -= dtHint
AM = ( J dw1dw2<.Q ( )
At ( W1) a2
W1,W2 al At ( W2) + h.c. (6.28)
b2(w) = uta2(w)U =
111
where I, I I represent the bands of the two fields of
(6.33b)
a1, a2, respectively and
with
G 1 , 2, g 1,2 have very complicated dependence on ( and i.Q(w1,w2). Notice
17L sin /3 -i/3
_
!cL-(3 e
i.Q(w1,w2)= ap(w1+w2), (6.29) that b1,2(w) is related to all frequency components of a1,2(w') in general.
(v2?T On the other hand, when the pump field is of single frequency at Wp,

J:
where ( is such that i.Q(w1, w2) satisfies the normalization condition that is, ap(w) = a 0 o(w - wp), Eq. (6.28) becomes
J 2
dw1dw2li.Q(w1,w2)1 = 1. (6.30) i~ dtiI[:',, = I; j dw,.f (w,)ii.\ (w,)ai(wp - w1 ) + h.c. (6.34)
( is usually proportional to Lv1P with P as the peak power of the strong where f (wi) is some spectral function centered around w10 at the center of
laser. i.Q(w1, w2) is also known as the joint two-photon spectral function. band I of the a 1-field. The frequency component of the a2-field is completely
When the power of the pump laser is relatively small, ( is small and correlated with the a 1-field by w2 = wP - w1. Notice that in this case, the
we can expand the exponential in the unitary operator in Eq . (6.5) in an joint spectral function i.Q is not normalizable because of the single frequency
infinite series and take the first few terms: nature of ap(w) but we can always choose ( so that J dw1lf(w1)12 = 1.

U = exp (
-=
1=~:/it) 1=
dt
in
1+
-=
dt ~/;;,t + [
in 2 -=
dt
in
2
(6.31) 1=~i~tl Now unlike Eq. (6.33), it is straightforward to find that b1,2(w) is only
related to the frequency components of a1,2(w) and a2,1(wp - w):
4 Terms such as :fuC+):fu(+):fu(-)
are equivalent to :fuC+):fu(+):fu(-)
if we interchange the
b1(w) = G(w)a 1(w) + g(w)a!(wp - w)
3 2 1 1 2 3
indices. Terms such as Ei+) Ei+) Ei-) are special cases of Ei+) E~+) Ei-) by setting 1=2. b2(w) = G(w)a2(w) + g(w)aI (wp - w) (6.35)
lGO Qu an tum Op t'ics For Br pc'l·im cn talis ts Gc 11em ti o11 and Trans f or mat io n of Qn antum S ta tes 161

with G(w) = cosh[~lf(w)I], g(w) = sinh[~lf(w)I]- This is the relation for a Becaus e of th e ortho-normal relations for ¢ 1,:(wi), l/J1,:(w2) and the commu-
parametric amplifier with e
as th e gain parameter and f (w) as the gain ih satisfy
tation relations [a 1 (w), a}(w')] = 61j6(w - w') , .A.1,:.
spectral profile. \i\Then the two fields of a1, a2become one field: 01 = a2= A At
[A1,:,A1,:
-
,] - 61,:
A
,k' , [B1,:,B1,:
At
,] - 6/;:,k
-
',
A At -
[A1,:,B/;:,] - 0 , (6.39
)
a0 , we arrive at the evolution equations in Eq. (4. 70) in Section 4.6.3 for ih} ar e th e annihilation operators for two sets of
which means th at {.A.1,;,
th e squeezed state in multi-frequency mode with a spectrum of squ eezing
orthogonal modes. As a matter of fact, from Sections 2.2.2 and 4.3.2 on
in Eq. (4.76).
temporal mod es, we find that {Ak, Bk} in Eq. (6.38) define two sets of
Single-frequency pumping corresponds to continuous-wav e (cw) opera-
orthogonal temporal modes. Mor eover, these are two sets of normal modes
tion of lasers. However , cw lasers usually have a finite bandwidth 5wp. But
for Hi[t because they have a one-to-one coupling between the two sets but
as long as 5wP is much smaller than the bandwidth of sin (3/ (3 in Eq. (6.27),
are decoupled within each set.
our treatment here is still valid (see [Ou (2007)] for details).
With the interaction Hamiltonian in the decoupled form of Eq. (6.37),
we can rewrite the evolution operator as
Normal modes and complete temporal modes
UA = exp ( 'G k
At At
+ h.c.)
rkA1,;B1,; (6.40)
For pumping by ultra-short pulses, we have a large 5wp that is comparable
to the bandwidth of sin /3/(3 in Eq. (6.27). However, instead of working with
the complicated expressions in Eq. (6.33), we resort to a technique called
singular value decomposition [Gentle (1998)] to rewrite the joint two-photon (6.41)
spectral function <I>( w1, w2) as So, each pair of {Ak , Bk} follows the single-mode model of the parametric
<I>(w1,w2)= Lrkcf>k(w1)'1j;k(w2) (k = 1,2, ... ) , (6.36) process discussed in Section 6.1.2 and produces a two-mode squeezed state.
k The materials covered in this section will be the basis for the experi-
where ¢k(w 1) and 'lj;k(w2) are the complex functions which satisfy the ortho - mental study of pulsed entangled quantum states in Part 2 of this book.
normal conditions J ¢k 1(w)¢k2(w)dw = 6k1,k2 and J 'lj;kl (w)'lj;k2(w)dw =
6ki,k2, respectively and the parameters rk(k = 1,2,3, ... ) are real and 6.2 Linear Transformation: Beam Splitters
non-negative: rk 0. From the normalization relation in Eq. (6.30) for
<I>(w 1,w 2) and ortho-normal relations for ¢k(w1),'1j;k(w2), we find {rk}(k = 6.2.1 General Formalism
1, 2, 3, ... ) satisfy the normalization condition Lk r~ = 1. {rk}(k = Optical beam splitters are important devices in optical interference: they
1, 2, 3, ... ) are referred to as the mode amplitudes. For the sake of clarity,
are usually used to split the amplitude of an incoming wave or combine two
the mode index k are arranged in a descending order, so that rk rk+1 waves for interference. Therefore, there are four ports: 2 input ports and
fork 1. For this arrangement, the mode functions ¢1 (w8 ) and 'lp1(wi) are 2 output ports and a beam splitter then couples these four modes. The
referred to as the fundamental mode. wave behavior of beam splitters is well-documented in the classical electro-
With the mode decomposition in Eq . (6.36), we can rewrite Eq. (6.28) magnetic wave theory [Born and Wolf (1999)]. Quantum mechanically, the
as
i~ 1: dtiJ.f/,,= l rk J (w1)ai(w2)+ h.c.
dw1dw2¢k(w1)'Pk(w2)&!
relationship between the four input and output operators has been studied
by a number of researchers [Zeilinger (1981); Yurke et al. (1986); Prasad
et al. (1987); Ou et al. (1987); Fearn and Loudon (1987, 1989); Campos et
(6.37) al. (1989)] and there is a simple relation for the field operators of the four
modes in Heisenberg picture :
where
~1 = ta1 + ra2, (6.42)
(6.38) {
b2 = t'a2 + r'a1.
162 Quantum Optics For Experirnental'ists Generation and Transformation of Quantu771 States 163

t, r, t' , r' are the comp lex amp litud e transmissivity

and reflectivity of the All the properties at the output can be calcu lated by averag ing t he opera-
beam splitter, respectively. From the comm utation relations to rs b1, b2 in Eq. (6.46) over the state in Eq. (6.-17).
On the other hand, all t hese properties can also be equ ivalent ly calcula-
[bk,bl]= Ok/, (k, l = 1, 2) (6.43)
ted in the Schrodinger picture in which the output operators b1, b2 are the
we obtain itl 2 + irl 2 = 1, it'l 2 + ir'l 2 = 1, and t*r' + r*t' = 0, which leads scnne as the inpu t operators a 1, a 2, but the output state is connected to the
to itl = it'I, lrl = lr'I, and
input by
<pt - <pr+ <pt' - <prt = 1f. (6.44) l'1')out = UI <P)in· (6.48)
The phase relation ab ove is universal and ind ep end ent of the specifics of the Generally, in order to find the output state, we need the exp licit form
beam splitter. This relation can also be derived via input-output energy of (J (see Appendix A). In many cases, however, this is not necessary, as
conservation from classical wave theory [Ou and :t\Iandel (1989) ; Smiles- illustrated below. For simp licity , let us first consider a single photon state
Mascarenhas ( 1991)] . input at port 1:
In general, t and rare complex numbers. However, by ca refully choosing (6.49)
the reference point, we may arbitrarily change <pt, <pr, <pt', <pr' within the The output state then becomes:
restriction in Eq. (6.44). For simplicity, we choose <pt, <pr, <pt' to be zero AAt - AAt At A
l'1')0 ut = Ua 1I0)i ® I0)2 - Ua 1U UI0)i ® I0)2, (6 .50)
then <pr' = -1r according to Eq. (6.44). Therefor e, Eq. (6.42) becomes:
where we insert the unitary relation (Jt(J = 1 between a1 and I0)i.
~1 = t~1+ r~2, (t,r>0) (6.45) It is easy to see that UI0)i ® 10)2 = I0)i ® I0)2, that is, vacuum input
{
b2 = ta2 - ra1. gives vacuum output for a beam spliAtter. (This can also be easily Acon~rmed
directly from the exp licit form of U in Appendix A.) To find U a1 ut , we
6.2.2 State Transformation of Number States through a invert Eq. (6.45) to relate ch,a2 in terms of b1, b2:
Beam Splitter a1 = Ub1 (Jt = tb1- rb2,
The operator relationship in Eq. (6.45), together with the input state, is a2 = Ub2ut= tb2+ rb1. (6.51)
usually enough to determine the properties at the output ports. However, So, the principle of reversibility gives:
the approach above with operators lacks the visual connections to such ua1ut = ta1- ra2, (6 .52 )
interesting phenomena in quantum information as quantum entanglement { Ua2UT = ta2+ ra1.
and other nonclassical effects in the output ports. So, to see what emerges
The relation above can also be derived directly from the explicit form of U
from the beam splitter, it is better to work in the Schrodinger picture and
in Appendix A. Therefore , we have, for the output state:
find the output state of the beam splitter. For this purpose, a unitary
evolution operator is needed to connect the input and output states. l'1')0 ut = (tbI- rb!)IO)i ® I0)2 = ti 1, 0) - rl0 , 1). (6.53)
We start in the Heisenberg picture, in which the output operators are Note that we replaced the input operators a1, a2 by the same output ope-
connected to the input operators by a unitary transformation: rators b1, b2 in the Schrodinger picture.
Likewise, we can find the output state of an input state of I1, 1) in the
~1 = = t~l + T~2 ,
Hong-Ou-Mandel interferometer [Hong et al. (1987)] (see Section 8.2.1):
(6.46)
{ b2 = uta2U= ta2- rn1.
l'1')out = Ull)ill)2 = Ua!IO)ia!I0)2 = Ua!a!IO)
Here, U is a function of a 1 and a 2. (In Appendix A, we will discuss a simple = ua!utua!utu10) = (Ua!ut)(Ua!ut)IO)
derivation of the explicit form of U using the operator algebra presented in
Section 3.6.) The state is unchanged and is the same as the input state in
= (ta!- ra!)(ta!+ ra!)IO)
2 2
this picture: = (t 2 - r 2)a! a!I0) + tr(a! - a! ) I0)
(6.4 7) = (t 2 -r 2)11, 1) + htr(l2,0) - 10,2)), (6.54)
16J Quantum Opti.cs For- Experimentalists Ge11endion a:nd Transfonnation of Q'llanf'Um States 165

and for a 50:50 beam splitter, we obtain the two-photon NOON state: and c0 = -J(N + 1)/2N+l = -CN+ l· Taking n1 = k,112 = N + 1 - k as
l\_[J)out = (12,0) - IO,2))/V2.
(6.55) th e photon number s of the two outputs. the output state b ecomes
The disappearance of the 1, 1) state above is the result of two-photon Hong- N+l.O
Ou-Mandel interference (Section 8.2.1).
For a general input state of l.l\I, N), we find the output state as
) out =
I\_[J L (6.60)

l\_[J)oul = Ul.l\J)i IN)2 = uaIJ\IIO)ia!N 10)2 It is interesting to note that c1,: = 0 for k = (N + 1)/2 or n 1 = n 2 when
v AI!N! N = odd. This is a multi-photon destructive interference effect which is
l lfa1M lJ1lfatNlf1lfl0) a generalization of the two-photon Hong-Ou-Mandel effect [Ou (1996)]. A
J .l\J!N! 1 2 multi-photon bunching effect shows up in the probability of finding IN+
l (Ua1ut)J\I(Uat ut)N I0) 1, 0) at the outputs clue to multi-photon construction interference. We will
J .l\J!N! 1 2 discuss more about th is in Section 8.3.
l (tat - rat)M(ta 1 + ra 1)NIO)
J.l\f!N! 1 2 2 1 6.2.3 State Transformation of an Arbitrary State
lvf N (-l)mJM!Nftl\I+N-m-nrm+n
· · Atl\I-m+nAtN-n+ml ) The technique used in Section 6.2.2 can also be applied to an arbitrary
L L (M _ ) , ,(N _ ) 1 1 al a2 vac
m=O n=O m .m. n .n. input state in the Glauber-Sudarshan P-representation [Glauber (1963b);
M N ~------------- Sudarshan (1963)] to derive a general relation between the input and the
= - m + n)!(N - n + m)!
output states. In the Glauber-Sudarshan P-representation, the input and
LL (M - m)!m!(N - n)!n!
m=On=O output states are described by the density operators as
xtM+N-m-nrm+nl.l\1 - m + n, N - n + m),
which can be regrouped as
(6.56)
Pin= J2 2
d a1d a2Pin(a1, a2)la1, a2)(a1, a2I, (6.61)

k=O
=
l\_[J)out
M+N
L Cklk, M +N - k), (6.57) Pout= J2 2
d a1d a2Pout(a1, a2)la1, a2)(a1, a2I, (6.62)

where ck collects the coefficients of the common terms of lk, 1\,1+ N - k)
in Eq. (6.56) and is in a very complicated form for the general case. But
for some special cases, we can derive its explicit form. For example, for a
50:50 beam splitter with M = N, the above is simplified as
\[!)out -_ N!1 N (At2 At2)NI 0 )
where Pin/out(a1, a2) is a quasi-probability distribution and can completely
describe the incoming/ outgoing fields of the beam splitter. la 1, a2) is the
coherent state base. Our goal is to find the connection between Pin and
Pout· From Eq. (6.48), the output density operator is given by
t
A A

I a 1 - a2 Pout= UpinU
2 2 2 A At

=
l
2N E(-
N

l)N-k
At2k At2(N-k)
a~'(;:J
- k)! IO)
= d a1d a2Pin(a1, a2)Ula1, a2)(a1, a2IU .
/
(6.63)

Obviously, Ula 1, a 2) is the output state corresponding to a coherent state

1 N N-k J(2k)!(2N - 2k)! input state of la 1, a 2), and, from classical optics and Eq. (6.45), we know
= 2N E(-l) k!(N _ k)! l2k, 2N - 2k). (6 .58) the output is also a coherent state of the form:
(6.64)
Another special case is when M = 1 with t = r = 1/v'2. In this case,
the coefficient in Eq. (6.57) has an explicit form of with
2k- N -1 N!
Ck = ---- ------ for 1 <_k _<N (6 59) (6.65)
2N+l k!(N - k + l)! .
lGG Qllanlum Optics For Expe.rinientalists Generation and Transformation of Quantmn States 167

The relation above can also be derived by using the method discussed in show up after some linear transformations on these states. \\ re will see this
Section 6.2.2 and writing the coherent state in the form of the displacement in the next section.
operator (see Eq. (3.26)): The format discussed here for ?-representation can be applied to other
phase space function such as \;\Tigner function. We will leave the derivation
la)= D(n)IO ), (6.66)
to Problem 6.4. and present the result as follows. For an input ·Wigner
with function 1Vi11 (X 1, Y1; X2, Y2)- the output Wigner function is
b(a) = exp(aa - a*at). (6.67) Waut(X1, Y1;X2, Y2)
Then, we have: = Win(tX1 - r X2, tY1 - rY2; tX1 + r X2, tY1 + rY2), (6.72)

Ula1, a2) = Ob1(a1)D2(a2)IO. o) which is similar to Eq. (6.71). Here t , r must be real.

= Ob1(a1)D2(a2)ut10,o). (6.68)
6.2.4 State Transformation of Squeezed States
But, with
To further demonstrate the usefulness of the simple technique in Section
Ob1(a1) b2 (a2)ut 6.2.2, we consider the homodyne (mixing) of a coherent state and a squeezed
uexp(a1a1 - aiat + a2a2 - a;a1)ut vacuum state by a beam splitter. From Sections 3.2.2 and 3.4.2, we find
that the coherent state and the squeezed vacuum state can be expressed as
exp(a10a1ut - aiOatut + a20a2ut - a;ua1ut)
exp[a1(ta1 - ra2) - ai(tat - ra1) la)i = D1(a)IO)
1()2 = S2(()IO), (6.73)
+a2(ta2 + ra1) - a;(ta! + rat)]
exp[(fo1 + rn2)a1 - (tai + rn;)at where
+(fo2 - rn1)a2 - (ta; - rni)a1J
b1 (/31)b2 (/32), (6.69) (6.74)

we have Eqs. (6.64) and (6.65). The output state is then

Substituting Eq. (6.64) into Eq. (6.63) and making a change of variables 1'1f)out= UD1 (a,)S2 (()IO)
from a, to (3 by Eq. (6.65), we find the output state as
= Ob1(a)S2(()utu10)
Pout= J 2 2
d /31d /32l/31,/32)(/31,/32IPin(t/31-r/32,t/32 +r/31). (6.70) = (UD1(a)ut)(US2(()ut)IO). (6.75)

Therefore, we have: It is easy to show that

ub1(a)ut = exp(aua1ut - a*uatut)
(6.71) 2
us 2(c)ut = exp((ua~ut - cua1 ut). (6.76)
This relation between input and output ?-functions was first derived in
[Ou et al. (1987)]. It shows that if the input state is a classical state From Eq. (6.69) by setting a,1 = a, a2 = 0, we hav e
with Pin 2:: 0, the output state must also retain this property. Thus, it u b 1(a)ut = exp(taa1 - t*a*at - rna2 + r*a*a1)
is impossible to generate nonclassical states from classical states by linear = b1 (ta)D2(-rn). (6. 77)
transformations . On the other hand, linear transformation will keep the
nonclassical properties of the input states. In other words, if certain non- So the output state becomes
classical properties are not exhibited for some nonclassical states, they may l'11)out = D1(ta)D2(-rn)[US2(()Ut]IO ). (6.78)
 168 Quantum Optics For Expe1·imc11talists Gf.'nC'ration and Transformation of Qna'lltum States 169

If we are only interested in the output sLate of one output port, say, port mcl the output sta te is
1
2, we may trace ouL the staLe in port 1:
IW)out = US1(-()S2(()IO)
fh = Tr1 I'Viou t ( '11/out = US 1(-()S2(()utu1 0)
t
= D2(-rcx.)p~D 2(-rn) ,
A A

(6.79)
= (US1( -()ut)(US2(()ut)I O) . (6.87)
wher e

p~= Tr1[US2(()ut1o)(O/USJ(()Ut] (6.80)

is the output state at port 2 with only squeezed state input. D 1 is traced
out in Eq. (6.79). Furthermore ,
US2(()ut = exp[((ta2 + rai) 2 - (*(t*al + r *a!) 2]
and
~s2(() for t • l, r • O. (6.81)
2 2
So in the limit of transparent beam splitter (t • 1, r • 0) but with {3 = US1(-()ut = exp [ - ~(a2 - a1) + (2* (al- aD ]
-rcx. = finite , the output state in port 2 is
= exp [ (* (At2
a 1 + aAt2)
- ( (A2
a 1 + aA2) ( A A (* aAt1aA2t] , (6 . 89)
2 + a1a2 -
(6.82) 2 2
2
or a pure state of where we set t = r = 1/J2. It is straightforward to show that if A
2 2
(*(a! + al )/2 - ((ar + a~)/2 and B = (a1a2 - (*a!at then
(6.83)
[A.,iJJ=0,
which is a coherent squeezed state discussed in Section 3.4.3. We have
shown in Eq. (3. 77) of Section 3.4.3 that so that we can write exp(B ± A)= exp(±A) exp(B) in Eqs. (6.88), (6.89).
The output state then becomes
sJ(()D2 (f3)!h(() = b2 (/3µ- /3*v*), (6.84)
where /w)out = exp[2(a1ct2 - 2(*a!a!J, (6.90)

v =((//(/)
sinh /(/, µ =)1 + /v/2 = cosh /(/.
So the output state is
/?/J2)out= S2(()D2(f3µ - {3*v*)/0),
which is the squeezed coherent state discussed in Section 3.4 .3 and can
also be produced by directly injecting a coherent state into a degenerate
parametric amplifier [Yuen (1976)]. Experimentally, the technique of using
a highly transmissive beam splitter to combine a squeezed vacuum with
a strong coherent state is quite popular for the generation of a coherent
squeezed state.
Next we look into the case of two squeezed states input to the beam
splitter with one from each side . We will consider a special case when the
beam splitter is 50 :50 and the squeezed states have equal strength but are
180 degree out of phase. From Eq. (6.73), the input state is simply
(6.86)
which is a two-mode squeezed state or a twin -beams state that can be
produced from a nondegenerate parametric amplifier (see Sections 4.6 and
6.1.2).
(6.85)
6.3 Optical Resonators: Input-Output Theory of an Open
Quantum System and Model of Decoherence
Optical resonators or cavities are commonly used optical devices that can
clean up the spatial modes, act as bandpass filters, and enhance the field
strength for nonlinear optical interaction. As we have seen in Section 2.2.1,
the field inside an optical resonator has a discrete set of modes . Outside of
the resonator, the spatial modes are matched through Gaussian waves. But
since there is no resonator outside, the frequency is continuous. So, how
do a discrete set of modes and a continuous set of modes couple together?
Since optical resonators are often used as optical filters, this question also
covers the mode transformation of an optical filter.
 Gen eration and Transfonnation of QtLantum Stat es 171
170 Quantum , Optics For Exp erim entalists

Here w 0 is the resonance frequency and we have e21'P = - exp[i2nL(w -

~j---
T, T2
=
wo)/ c] -ei 0 with 5 =
21r(w - wo)/DFsR (DFsR = 21rc/2riL is the free

-- --£
01
( E
0

(fJ
E2 spectral range). If the input fields have frequency near resonance, we have
5 « 1 and e21'P = -ei 0 -(1 + i5). Furthermore, for high finesse cavity,

-1
El ' E3 E6 , E02
- 1- we have T 1 « 1, T2 « 1. With these approximations, Eqs. (6.92) - (6.94)
become
L Eo1ei'Pfti + Eo2vr;_
E5 = ---------
(T1 + T2)/2 - i5
Fig. 6 .3 Inpu t an d output waves coup led to t h e waves inside a cav it y of transm issiv it y
of T1 , T2, respectively. Eo2ei'PJT1T2 - Eo1[(T1 -T2)/2 + i5]
E1=---------'--------
(T1 + T2)/2 - i5
6.3.1 Classical Wave Model E 01ei'PJT 1T2 + Eo2[(T2 - Ti)/2 + i5]
E2=--------------. (6.96)
(T1 + T2)/2 - i5
Since spatial mod es are match ed, we will only consider the simplest case
wh en a monochrom atic plane wave of freque ncy w enters a Fabry-Perot If we only have one input field, say E 01, both the transm itt ed field and the
cavity with two mirrors of transmissivity T 1, T2 r-v 1, respectively, as shown field inside h ave a Lorentzian line shape:
in Fig. 6.3 . 2E 01ei'P~ b.D /2
Referring to Fig. 6.3 , we denote the input fields as E 01, E 02 and outgoing E2 = T 1 + T2 b.D /2 - i(w - wo)'
fields as E 1, E 2 and the fields inside as E3, E4, E5, E5. Treating each mirror 2Eo1ei'Pfti b.D /2
as a beam splitter, we have 6 (6.97)
E = T 1 + T2 b.D / 2 - i(w - wo)
E1 = E3\fIT +E01)l-T1, E4 = Eo1v"I'i.-E3)l-T1 where the full lin e width at half maximum is
E2 = E5#, - Eo2)l -T2, E5 = Eo2#, + E5jl -T2. (6.91) (6.98)
Fields E3, E5 are respectively related to E5, E4 by E3 = E5ei'P, E 5 = E 4ei'P with F = 2w/(T 1 +T 2) defined as the finesse of the cavity and DFsR as the
with zp = kL = nLw/c as the phase due to propagation (Lis the length of free spectral range. Equation (6.98) is often used for determining the finess e
the cavity and n is the index of refraction). Solving first the fields inside of the cavity: F = DFsR/ b.D , with DFsR , b.D measured by recording the
the cavity in terms of E 01, E 02, we obtain transmitted intensity in a scan over one full spectral range of the cavity.
Notice that when the input field is on resonance , the intensity of the
E 01ei'Pj(l - T2)T1 + Eo2vr;_
E5 = ----------;=======--. (6.92) inside field is IE6 2 = 4IE 01 2Ti/(T1 + T2)2, or the enhancement factor is
1 1

1 + e 2i'Pj(l - T1)(l - T2)

B = 4Ti/(T 1 + T2) 2, which can be quite large if T1, T2 are small. For
Substituting back for the output fields, we have example, let T 1 = 0.02, T2 = 0.01. Then we have B = 89: the power inside
the cavity is nearly 90 times of the input power.
Ei = Eo1 ( vT=Ti + e 2i'P~) + E 02ei'P
(6.93)
1 + e 2i'Pj(l - T1)(l - T2)
6.3.2 Further Reading: Intra-cavity Second Harmonic
Eo1ei'PJT1T2 - Eo2( JI - T2 + e 2i'PJl - T 1) Generation
E2 = -----------;::=======----. (6.94)
1 + e 2i'Pj(l - T1)(l - T2)
In quantum optics, many of the interesting non-classical states are gene-
Due to a 1r-phase shift at one of the two reflecting surfaces, the total round rated through nonlinear interaction such as three - and four-wave mixing
trip phase is 1r+ 2zp. We then obtain the resonance condition: discussed in Section 6.1.1. Three-wave mixing, because of the high nonli-
2zp+ 1r= 2N 1r or wo = (2N - 1)1rc/2nL. (6 .95) near coefficients in some nonlinear crystals such as LiNb0 3 and KNb03,
 172 Quantum , Optics For Experimentalists Gen<'rahon and Transformation of Qua11/u111States 173

has become popular in quantum optical experiments for generating two-
photon states and squeezed states. It requires a high power pump field at
twice of the frequency of the quantum fields. This field is usually obtained
through second harmonic generation (SHG) via nonlinear crystals, which
requires high power for good conversion efficiency. For pulsed fields, this is
not a problem since short optical pulses have extremely high peak power.
for EN LPi « l. EN L can be obtained directly by measuring the single-
pass non-saturated P2 and the input P 1 and is a fixed number for a certain
focussing geometry and crystal length (Typically, EN L rv 10 4 w- 1 for
1-cm long LiNbO 3 and rv 10- 3w- 1 for 1-cm long K bO3).
Referring to Eq. (6.97), we have E 6 as the fundamental field inside the
cavity and Pi = IEG 2 , T 1 = T as the input coupling at one end of the
1

For cw fields, however, this becomes a challenge.
Optical cavities were first proposed and tested for the enhancement of
the nonlinear conversion [Ashkin et al. ( 1966)]. The large enhancement
factor for the field insid e cavity is important for intra-cavity second harmo-
nic generation [Kozlovsky et al. (1988); Polzik and Kimble (1991)], where
high power is a must for efficient nonlinear frequency convers ion. Vve will
consider this technique in this section.
cavity and T2 = L + Ln at the other end but with nonlinear loss included.
The input fundamental field is on resonance and the input power is related
to E 01 : Pin = IE0112 . With these, Eq. (6.97) can be re-written as
(6.101)
The overall SHG conversion efficiency is defined as T/ = P2/ Pin
ENLP'f / Pin· Then Eq. (6.101) can be re-written in terms of T/ as

(6.102)
P.111
or

(6.103)

where Po = L 2 / EN L is a characteristic power for the device defined by the

Fig. 6.4 Cavity en hancem ent of second harmonic generation in a bow-tie cavity confi-
guration. HR: high reflector for the fundamental field. round-trip intra-cavity loss L and nonlinear conversion coefficient ENL·
This is a cubic equation for T/· Figure 6.5(a) shows T/ as a function of the
Figure 6.4 shows a schematic for intra-cavity second harmonic genera- dimensionless quantity J
Pini Po for a set of parameters of T / L. We find
tion with a bow-tie shaped enhancement cavity for the fundamental field that there is an optimum input power for the conversion efficiency TJat each
at wo. A second harmonic field at 2w0 is generated inside the nonlinear value of T / L. Or, in other words, with each input power Pin, the input
crystal in a single-pass fashion. 5 Suppose the cavity has an input coupler coup ler coefficient T can be optimized for impedance matching T = L + Ln
of transmissivity T and an intra-cavity loss of L, which can be modeled as
a two-ended cavity with T1 = T and T2 = L. Moreover, there is a nonli- (iv)
(b)
near loss Ln = P2/ Pi due to nonlinear conversion of the intra-cavity power
17 77op
Pi to second harmonic power P2. This loss is proportional to intra-cavity
fundamental power P 1 as

(6.99)
where effective nonlinear coefficient EN L is defined as the single-pass SHG P!i~------------
1

conversion efficiency:

P2 = ENLP{ (6.100) Fig. 6.5 (a) Intra-cavity second harmonic conversion efficiency 1J as a function of
--------------
5
Traveling wave cavities are preferred over standing wave cavities because of the sim - JPin/Po for various values of T/L: (i)l;(ii)2;(iii)5;(iv)8. (b) Optimized T/op as a
plicity in SHG geometry. function of Pin/Po. Po= L 2 /ENL·
174 Q'Uant-um Optics For Experim ,entalists Generntion and Transformation of Quantum States 175

to reach the maximum conversion efficiency, with as an interaction of the single-mode field with a continuous multi-mode
field outside the cavity described by {b(w)} with a free-field Hamiltonian
Top = (1 + J l + 4hi I Po) Hout= J dwTlW[bt(w)b(w) + ll2] and [b(w),bt(w1 )] = c5(w-w1). Notice that
4PinlPo for different cavity modes, we assume that the eigen-frequencies of the ca-
170 (6.104) vity modes are well separated beyond their linewidths, so the modes can
P = (1 + jl + 4Pinl Po)2
be thought of as independent and coupled to independent groups of outside
In Fig, 6.5(b), we plot the optimized 170 P versus the dimensionless quantity
frequency modes that are only within the linewidth of each cavity mode.
Pini Po and it shows a quick increase up to somewhere around Pini P0 =l We will show at the end that this assumption holds. Here, we also assume
and a slower increase towards unit convers ion after that.
that the spatial mode of the cavity matches that of the outside fields. With
these two assumptions, we can use one-dimensional quasi-monochromatic
6.3.3 Quantum Treatment of Cavity Loss: Decoherence of approximation (Sections 2.3.4, 2.3.5) for the fields outside the cavity.
an Open Quantum System The coupling Hamiltonian between the cavity mode and the field outside
As we discussed in Section 2. 2 .1, an optical resonator (cavity) defines a has the form of
discrete set of modes for the optical waves inside it. If the cavity is ideal
in the sense that there is no loss and interaction, the system is isolated
He= jn J
dwK(w)[bT(w)a- atb(w)]. (6.107)

and each mode will evolve independently. Of course, there does not exist
such an ideal system in reality. For simplicity, let us ignore the interaction
between different modes 6 and concentrate on the losses, which couple the
system to the outside fields and make it an open system.
After the quantization of the cavity modes, we end up with the Hamil-
tonian of the whole system as:
Hcav = L>, H>, with H>,= nw>,(ala>-+ 112).
Here each individual mode ct>,evolves independently in the Heisenberg pic-
This interaction is similar to the beam splitter Hamiltonian discussed in
Section 6.2 and corresponds to linear coupling.
Gardiner and Collett worked out a detailed formalism to connect the
field inside the cavity with the fields outside the cavity, after making some
reasonable Markovian approximation by assuming all outside modes are
coupled to the cavity mode with equal strength: K(w) = [Gardiner
(6.105)
and Collett (1985)]. They derived an equation of motion for a, the field
inside the cavity, as
da A

dt = -(, 12 + jwo)a - v,bin, (6.108)

ture:
where

which leads to a>-(t)= ct>,e-jw;,t_This is of course the free field evolution

(6.106)
bin= J dwbin(w)e-jwt (6.109)

and [a>-,ali] = c5>->-'

· is the input field to the cavity with [bin(w),bJn(w1 )] = c5(w-w 1 ). This is the
On the other hand, when the cavity has losses, the energy of the field
inside will decay and gradually decrease to zero. Energy conservation me-
ans that the energy of the cavity must go outside. So, we need to in-
clude the coupling of the cavity modes to the outside. Consider a single -
mode electromagnetic field described by a with frequency w 0 in an optical
cavity where electromagnetic energy is stored. Note that for the single-
mode field, we have [a,atJ = 1. The loss of the cavity can be modeled
6
Interaction between the modes will lead to a new set of eigen-modes of the system so
one-dimensional quasi-monochromatic field with its spatial mode matched
to the cavity mode. So, the evolution equation above shows that the loss
constant 'r is related to the coupling coefficient K, and the outside field
comes into the cavity as a noise term, which preserves the commutation
relation for a: [a, at] = 1. With a time reversal of the evolution equation in
Eq. (6.108), the coupling in Eq. (6.107) can also give rise to the following
equation for the output field bout [Gardiner and Collett (1985)]:
da A

we can relabel them and treat the new ones as independent modes. See Problem 6 .1. - dt = -(, 12 - jwo)a + v,bout· (6.110)
176 Quantum Optics F01· Expc1·imenta lis ts Generation and Transformation of Quantu111 States 177

Com binin g Eqs. (6.108) and (6 .110 ), we obtain the input-outpnt relation and each set of inpu t/output fields satisfies a boundary condition as in
as related to the inside field: [Gard iner and Collett (1985)] Eq. (6.111 ):
(6. 111) (6.117)
The evolution equat ions in Eqs. (6 .108) and (6. 110) and the inpu t-o utput Similarly we can find
relation in Eq. (6.111) compl ete the inpu t-ou tp ut formalism t hat relates
the discrete mode a inside the cav ity to the outside traveling cont inuous (6. 118)
waves bin, bout·
With the evolution equation in Eq. (6.108) and the bound ary condition and
in Eq. (6.111), we can find the output field bout in terms of the input field A(l) A(2)
b(l) ( ) = bin (w)[w - Wo- _j(1'1- , 2)/2] - jb in (w)0D2
bin. For this , let us write a= (1/ ~) J dwa(w )e- jwt and substitut e it into out w W - Wo+ j(1'1 + ,2 )/ 2
Eq. (6 .108). We have A(2) A(l)
b(2) ( ) = bin (w)[w - wo - j(1'2 - , 1)/ 2] - jbin (w)0D2 ( _
6 119
)
-J2ii-
1 / dw(-jw)a(w)e-JW·t = -(1'/2 + jwo) -J2ii-
1 / dwa(w)e - JWt
· out w w - Wo+ j(1'1 + , 2)/ 2
1 / dwbin(w)e-Jw t , With [b~~) (w), b~~t(w')] = Okzb(w-w') (k, l = 1, 2) for the input fields , it can
-~ -J2ir- A •
(6.112)
be easily check ed that [b~~t(w),b~~J(w')] = Okzb(w-w') for the output fields.
or Notice that the above expressions are similar to those in Eq. (6.96) with a

Jdw[(jw - ,/ 2 - jwo)a(w) e-jwt - ~b(w)Je-jwt = 0. (6.113)

proper phase <pand the coefficients satisfy the conditions for a lossless beam
splitter discussed in Section 6.2. So, a cavity acts as a filter that can be
This gives modeled by a lossless beam splitter with frequency dependent transmissive
and reflective coefficients. Notice that the effect of the cavity is only to
a(w) = .J1bm(w)
j(w - wo) - ,/2
(6.114) the frequency components within the linewidth: lw- woI rv 11 , ,2
so the
assumption of independence about different cavity modes holds.
Substituting into Eq. (6.111) , we obtain Comparing Eq. (6.119) with Eq. (6.97), we can relate the decay con-
w - w0 - Jr /2 stants 11 , 12 to the parameters of the cavity:
J, 12.
A A

bout(w) = bin(w)
w -wo
+ . (6.115)
ri = nFsRTd2w (i = 1, 2) , and ,1+ ,2= nFsR/ F , (6.120)
where the free spectral range Op SR = 2w/ tr = 2wc/2nL with tr and 2nL
as the cavity round-trip time and round -trip optical length, respectively.
We can now answer the question about what happens to a photon when
a
it is filtered by an optical cavity. As we discussed in Section 2.5, we cannot
talk about a photon without its mode . For a field with well-defined spa -
tial modes, we can describe it by one-dimensional approximation with fre-
quency modes (Section 2.3.5). So, the single photon that enters an optical
Fig. 6.6 A two- en d ed cavity with coupling constants , 1, ,2, resp ectively.
cavity filter is a wave packet in a temporal mode described in Eq . (4.43) of
For a two -ended cavity in F ig . 6.6, we denote the two coupling coeffi- Section 4.3.2. The cavity then acts as a frequency -dependent beam splitter
cients as 11 , 12 and the equation of motion for the cavity mode a is with relations in Eq. (6.119) and splits the single -photon wave packet into
two output ports, similar to the case discussed in Eq. (6.53) of Section 6.2.2 :
(6.116)
(6.121)
L78 Quantum Opf/ics For Experimentalists Generation and Transformation of Quantu :111States 179

with
= A(t)e-Jwnntj = B(t)e-jw,,ot with Wno+ Wbo = Wp and write further:
IT)k = Ck Jdw,h(w)¢(w)ejwTaL(w)IO)k (k = 1, 2) (6.122)
(L
A.(t) = (1/ ~) J dDA(D)e-JW, etc. and similar to Eq. (6.114), we can
change Eq. (6.125) to
and fi(w) = [(w - wo) - Jb1 - ,2)/2]/[(w -wo) - Jh1 + "12)/2],h(w) =
2
-)J'Y1'Y2/[(w -wo) - Jb1 + "12)/2]. c,;, = J dwlfk¢12 is for normalization. [ - j(D + ,0.a) + ~] .A.(D)= (iJt (-D) - 01.A.in(D).
The transmitted and reflected fields are still in sing le-photon wave packets
IT)i, IT)2 but their shapes are modified by the filter cavity. t 2 = 1/ci- r 2 == [ - j(D + ,0.b)+ ~] B(D) = (_At(-D) - /73Bi 11 (D), (6.126)
1/ c~ give the probabilities of transmission and reflection. respectively.
= =
where ,0.a Wao- Wa,,0.b WbQ- wb are detuning of the frequency com-
ponents of our interest (Wao,wbo) from the cavity resonance frequencies
6.3.4 Optical Resonators with Nonlinear Interactions
(wa. Wb). Notice that D is the frequency off-set from the frequencies of our
A filter is still a linear device that cannot produce non-classical states from interest: Wao,WbQthat are determined by the detection system (say, homo-
classical states. So, we need nonlinear interaction between different modes dyne detection). Solving jointly .A.(D),B(D) with A.t(-D), iJt ( -D), we
of the cavity. For this, we consider a three-wave mixing described by Ha- have
miltonian in Eq. (6.6). If field a3 is a strong pump field at wP so that it can fo.A.in(D)[/3/2+j(,0.b-D)]
A(D) -
A

be replaced by a field amplitude CY.pe-jwptand the other fields are the two ------=----------=--
- (
2
- [a/2 - j(,0.a + D)] [/3/2 + j(,0.b - D)]
cavity modes denoted by a,b,the interaction Hamiltonian becomes
ih = jn((bt at e-jwpt - abejwpt), (6.123) B(D) _ [a/2 + j(,0.a - D)] + (fo.A.Jn(-D) (6.127)
- ( 2 - [/3/2 - j(,0.b + D)][a/2 + j(,0.a - D)]
with ( = rJCY.p.Here rJ includes the nonlinear coefficient and spatial mode
integral in Eq. (6.3), and other physical constants so that lapl
2 = (ala ) is From the input-output relation in Eq. (6.111), we have
3
the photon number of the pump field a3 . The Hamiltonian in the equat ion A
Aout (D) = G(D)Ain (D) + g(D)Bin (-D)'
A At

above is that for the parametric interaction and the device with the cavity 0
Bout(D) = ej [G(-D)Bin(D) + g(-D).A.Jn(-D)] (6.128)
is known as the optical parametric oscillator (OPO). Now we need to add
fh to Hcav for the evolution equation: with
da 1 [A HA iI l a A r::A
dt = jn a, cav + 1 - 2a - v aain,
_ (
2
+ [a/2 + j(,0.a + D)][/3/2 + j(,0.b - D)]
G(D) = (2 - [a /2 - j(,0.a + D)] [/3/2 + j(,0.b - D)]'
db l A A A /3A /QA (yap
dt = jn [b, H cav + H1] - b - v f3bin (6.124)
2 g(D) = (2 _ [/3/2 + j(,0.b - D)][a/2 - j(,0.a + D)] '
with output coupling decay constants a, f3 for the two cavity modes. Here ,
eje = ( 2 - [/3/2 + j(,0.b + D)] [a/2 - j(,0.a - D)].
for simplicity, we do not include intra-cavity loss other than the output (6.129)
coupling. With the Hamiltonian in Eq. (6.123), we have - (
2
- [/3/2 - j(,0.b + D)][a/2 + j(,0.a - D)]

-da ( a) At .
dt = - 1·wa + -2 a+ (b e-Jwpt - vu
fo-.ain, With the commutation relations [.A.in(D),A.Jn(D')] = b(D D') =
[Bi11 (D), B; 11 (D')] and [.A.i
11 (D), Bin(D')] = 0 for the input fields, it can
dt = - ( )Wb + /3) bA + (a t e-Jwpt
db . be easily checked that the same relations are satisfied by the output fields
- v/QA
f3bin· (6.125)
2 Aout(D), Bout(D).
Normally in an experiment, there are some frequency bands that are The input-output relations in Eq. (6.128) are exactly those for the two-
detected with center frequencies as wao, WbQfor a, b fields. These cen- mode squeezed states discussed in Section 4.6 and are similar to those in
ter frequencies are, for example, from local oscillators of homodyne de- Eq. (6.35) in Section 6.1.5 for the case of single-frequency pumped traveling
tection (see Section 9.3 for detail). Now let us go to the rotating frame: wave three-wave mixing. Notice that the gain parameters in Eq. (6.129)
180 Quantum . Opt.'ics For E :cpcrimentalists Gencnit ion and Transformation of Quant 1w1 States 181

depend on the coupling decay constants which arc important in the case Using the same technique as in solving Eq. (G.125) and A(t) = a(t)eiwot_
of optical cavit ies . The ga in parameters arc the highest when resonance we find the output operator as
condit ions 6a = 0 = 6b are met and we have
2
+ (Cl/2 + jD)(f3/2 - JD) A A At
( Aaut (D) = GD (D)Ain (D) + g D (D)Ain (-D)
G(D) = (2 - (cv./2- jD)((3/2 - JD)' A At
+G L (D)Cin (D) + 9L (D)Cin (-D) (6.135)
(va]J
g(D) = (2 - (o./2 - jD) ((3/2 - JD) (6.130)
with
Note that G(0), g(O) become infinity when ( 2 = a(3/ 4. This means that
a threshold is reach ed for las er -like oscillation which is why the device is _ ( 2 + [(!3+ ,)/2 + j(6 - D)] [(!3- ,)/2 + j(6 + D)]
named as opt ica l parametric oscillator (OPO ). The quantum states to our GD(D) = (2 _ [(f3+ ,)/2 + j(6 - D)] [(f3+ 1 )/2 - j(6 + D)]
int erest are produced below threshold. The sit u at ion above threshold is not
covere d in the treatment here. For more, refer to [Heidmann et al. (1987)]. (D) = ((3
9D (2 _ [(f3+ ,)/2 + j(6 - D)] [(f3+ 1 )/2 - j(6 + D)]'
-
In the case when the two cav it y mod es are the same: a= b, the device
becomes a dege nerate OPO with wa = wb and we choose the detectio n Jlry[(f3 + ,)/2 + j(6 - D)]
frequency at w 0 = wp/2. We can simply replace b with a in Eq. (6.123) and GL(D) = (2 _ [(f3+ ,)/2 + j(6 - D)] [((3+ ,)/2 - j(6 + D)]
the Hamiltonian for degenerate OPO is th en
(D) = (J/ry . (6.136)
gL - (2 _ [(f3+ ,)/2 + j(6 - D)] [(f3+ 1 )/2 - j(6 + D)]
HD=jnf (at2e-jwpt - a 2eJWpt), (6.131)
2
where we replaced ( with (/2 because the nonlinear coefficient for the de- Here 6 = w 0 - Wa is the cavity detuning. The threshold of the degener ate
generate case is half of that for th e non-degenerate case .7 HDis a specia l OPO is l(thl = ((3 + ,)/2, which is the same as the non-degenerate case
case of the Hamiltonian for second harmonic generation: from Eq. (6.130). When there is no loss, or, r = 0, we hav e G L = 0 = 9L
(6.132) and Eq. (6.136) becomes

when the harmonic field has a strong excitation so that we can replace b ( 2 + [f3/2 + j(6- D)] [!3/2 + j(6 + D)]
with its average (b) = ape-jwpt where lapl2 is the photon number of the GD(D) = G(D) = (2 - [f3/2 + j(6 - D)] [!3/2 - j(6 + D)]
pump field and ( = 77ap· The evolution equation for the field in OPO is
((3
changed from Eq. (6.125) to (6.137)
9D(D) = g(D) = (2 _ [f3/2 + j(6 - D)] [f3/2 - j(6 + D)].
da -__ (·JWa + (j_) aA+;-A _ y 1Df3A·
-,,a t e -2}wot µam, (6.133)
dt 2
Then , Eq. (6.135) is exactly in the form of Eq. (4.70) in Section 4.6.3
where (3 is the output coupl in g constant for the OPO field. Now let us
for the squeezed states in multi-frequency mode. It can be eas ily checked
include the effect of loss which can be modeled as the coup lin g to another
that G(D) , g(D) satisfy Eq. (4.71). From Eq. (4.76) of Section 4.6.3 and
outside field Cin with a coupling coeffic ient T Then Eq. (6.133) becomes
Eq. (6.137), we can find the spectrum of squeezing Sx(D) = (IG(D)I -
-da -_ - ( ]Wa
. + -(3 + -r)a+
A -,,a
;-Ate -2 jwot - A - y ;;:::A
y ID(3
µ ain rCin . (6.134) lg(D)l)2 = 1/(IG(D)I + lg(D)l)2 . Figure 6.7 shows Sx(D) as a function of
dt 2 2 2D/ (3 for various parameters of (, 6. The optimum squeezing of Sx (D) = 0
7
This point can be seen from th e x( 2 )_non lin ear term in Eq. (6.4) where if the E-fi eld occurs at threshold ( = (3/ 2 with 6 = 0 and D = 0.
has two mod es : E = E1 + E2, th en P;}l = x( 2) Er + x( 2) E5+ 2x( 2) E1E2. So , the In practice, there will always be some loss es inside the cavity. But we
nonlinear coeffic ients for th e degenerat e cases ( first and second terms) is half of that for
th e non-degen erate cas e (third term). can lock the cavity on resonanc e with wo so that 6 = 0. Th en Eq. (6.136)
 182 Quant.um Optics For Experimentalists Generation and Tmnsformat-ion of Quantum Staf Ps 183

output photons during one round trip. Since squeez ing is due to photon pair
(iii)
correlation between the field components, it is highly sensitive to loss. The
I lost photons become un corre lated. So, the fractional un corre lated photons
(ii) set the limi t of noise level out of OPO.
I I

\ I
\ j (i)
(iii)
----------
2Q/(3

Fig. 6.7 Spectrum of squeezing Sx(D) for va riou s va lu es of(, 6.: (i) ( = (3/2, 6. = 0;
(ii) ( = /3/2, 6. = 5/3/4; (iii ) ( = /3/8,6. = o.

becomes 0.1 0.2 OJ 0.4 0.5 0.6 il.7 O.X 0.9 I

S/srh
GD(O) = (2 + [(,8 + --y)/2 - jO][(,B - --y)/2 + jD]
( 2 - [(,8+ --y)/2 - jOJ2 Fig. 6.8 The amount of sq ue ez in g Sx (0) in log-scale as a fun ction of (/(th for three
values of 'Y//3:(ih//3 =0.11; (ii) 'Y//3= 0.3; (iii ) 'Y//3= l.
gD(O) = (,8 .
(2 - [(,8+--y)/2 - j0]2 , In Fig . 6.8, we plot Sx from Eq. (6.140) in log-scale as a function of
GL(O) = JlFt[(,B +--y)/2 - jD] ( /(t h for three values of --y/,8= L / T . We find that pumping around 90%
(2 - [(,8 +--y)/2 - j0]2 of threshold should be enough to obtain the maximum squeezing given in
Eq. (6.141). Thus, the threshold value is an important parameter in the
gL(O) = ( JlFt . (6.138) experiment for generating squeezed state. We estimate it in the following.
( 2 - [(,8 + --y)/2 - jOJ2 From Eq. (6.138) , the threshold is reached wh en ( = (th = (,8 + --y)/2.
For D « (,8 + ')')/2 and near threshold operation, the quantities above From Eq. (6.120) , we find ,8 and 1' are related to experimentally measurable
have a common phase determined by the common denominator. Then it is quantities so we need to do the same for(. From Eq. (6.132) for degenerate
straightforward to show the minimum value of the spectrum is case (a = b), we find ( = r;ap with iapl 2
as the photon number of the
pump field. But r; is still undetermined since it depends on the nonlinear
Sx(O) = (IGD(O)I - lgD(O)l)2 + (IGL(O)/ - /gL(0)/)2. (6.139) coefficient of the nonlinear medium and the mode overlap of the three waves.
With O « (,8 + --y)/2 or D rv 0, we have In fact, r; is related to the effective nonlinear coefficient EN L that can be
measured with single-pass un -saturated second harmonic generation (SHG)
Sx(0) [( - (,8 - --y)/2]2 + ,8--y
(6.140) (see Section 6.3.2). To find the connection, we use the same OPO cavity
[( + (,8 + --y)/2]2
and nonlinear medium for SHG. From the Hamiltonian in Eq. (6.132) for
When the threshold is approached: ( • (th= (,8 + --y
)/2, we obtain SHG, we find the evolution equation of the harmonic field b as
Sx(0) --
r'
= -- L
,8 + --y T+L' (6.141) d~;)
= ~(ii) 2
, (6.142)

where we used Eq . (6.120) with T as the transmissiv ity of the coupling
mirror and Las the round -trip loss excluding T. This is the best squeezing
that can be achieved for the field generated from the OPO under threshold.
Equation (6.141) has a straightforward explanation: the optimum noise
level out of an OPO is just the fraction of photons lost among the overall
where we assume the fields are all in coherent states and take average of
the fields. For single -pass of the harmonic field, we can approximate
(b) r;t:it(a)2;2 (6.143)
with b.t as the transient time for the harmonic wave passing through the
nonlinear medium. Then, with the power P2 of the generated SHG and the
 G en erntion and Transformation of Qiwnturn Stal es 185
18'1 Q'llantllm Optics For Experimenlal-ists

For a renl birefringent crystal, due to imperfectness in growing process, the

power P 1 of the fundamental wave defined as
optical 1:nis rotates by a very small amount. However, such a rotation 'Nill
P. _ rM2 I(b)12 produce a coupling betv-reen the two polarization modes of x and y. The
2 - 6.t (6.144)
coupling has the form of
we obtain from Eq. (6.143) (6.l<:1:9)

P.
2 6.t
2
= nw2l(b)l = ~nw2 26.
4 TJ t
(Pitr)
nw1
2
_
-
r;2t;6.t
nw2 pl
2
.
(6.145) (i) I\Iake a mode transformation

Here. tr is the round-trip time of the fundamental wave in the OPO cavity A = CLxcos 0 + Clysin 0.
and we used nw 2 = 2nw 1 for SHG. Comparing with Eq. (6.100), we have B = - ax sin 0 + Clycos 0, (6.150)
(see also [Wu et al. (1987)])
which corresponds to a polarization rotation. Find the angle 0 so that
if= Ho+H1 can be written in the free form of Hoand compare the eigen-
(6.146)
frequencies of H with those of Hofor different values of b = w1 -w2. Prove
Now come back to the OPO. From the threshold condition of (th = that operators A, B are annihilation operators by showing that they satisfy
(/3+'"'!)/2 with ( = T]O:p and /3+'"'!
= 21r/trF (Eq. (6.120)) and Eq. (6.146), the commutation relation.
we find the threshold power as
(ii) Consider the state lwA) 10:A)A I0)B where Inode A is in a coherent
2 2 2 state while mode B is in vacuum. Prove that lwA) is an eigenstate of both
p,th = nw2l0:pl = nw2l(thl = 1r
(6.147)
2
6.t ry2 6.t ENLF 2 . ax,ayand therefore IWAJ= ia:1)xla::)y- Find their eigenva lues 0:1, 0::
Here, F = 2w/ (T + L) is the finesse of the OPO cavity at fundamenta l (Hint: use AlwA) = 0:AIWAJ,
BlwA) = 0).
frequency w1 = wo.
(iii) Do the same for the state lwB) = I0)A 10:B)B ·
Experimental detection by homodyne measurement scheme of the squee -
zed states in multi-frequency mode leads to observation of quantum noise (iv) Assume that a light field in coherent state with frequency w and po-
reduction in a wide spectral range and will be covered in Sections 9.5 and larization x is incident into the cavity. What does the output (frequency
10.1.2. and polarization) look like as frequency w is scanned?

6.4 Problems Problem 6.2 Anti -bunching by mixing of a coherent state and a two-
photon state [Shafiei et al. (2004)].
Problem 6.1 Coupling of two cavity modes: frequency splitting .
Consider the following input states for a 50:50 beam splitter. The input
Consider a cavity with a birefringent crystal inside. It can be described mode a is in a two-photon state
by two polarization modes (e-ray and o-ray) of the optical field. The pola-
I0)a + TJl2)a (ITJI<< 1),
ITJ)a;::::; (6.151)
rizations of the two modes are orthogonal to each other and labeled as x, y.
The frequencies of the two modes are w 1 , w 2, respectively, which correspond and the b mode is in a coherent state
to resonant frequencies of the two polarizations, that is, only when the fre-
0:2 0:3
quency of the incident field to the cavity is equal to these eigen -fr equencies, I0)b + 0:ll)b + J212)b + v'613)b (10:I<< 1).
10:)b;::::; (6.152)
there are outputs from the other side. The free Hamiltonian for the two
modes is (i) Find the output state of the beam splitter. Keep only up to the 3-ph oton
(6.148) terms.
 '+'

186 Q'U,antum Opt-ics For E:rperim ,cntalists Generation and Transformation of Q'U,antum States 187

(ii) Calculate up to the second non-zero order the normalized intensity with C2 - g2 + G'2 - g'2 = 1 and
correlation function 9( 2) = (f,2) / (fz)2 for each output field of the BS and
c = [(/32 - ,2)/-1 + 1(12]/1\J. g = (/3/1\J
find the condition for anti-bunching: 9( 2) < 1.
G' = /7F((, + /3)/21\1, g' = ( ~/1\1, (6.154)

Problem 6.3 Three-photon NOON state by mixing of a coherent state and where ( is proportional to the pump amplitude and /3, 1 proportional to
a two-photon state [Shafiei et al. (2004)]. the cavity round-trip output coupling T and loss L, respectively, and JU =
(1 + /3)2/4 -1(1 2. a0 , bo denote the vacuum modes coupled in through the
Use the output state in Problem 6.2 to obtain the condition for which the
losses.
three-photon part of the state is a three-photon NOON state (Section 4.5.1).
(ii) Show that the EPR correlation between the two outputs (see
Problem 6.4 State transformation of the \i\Tigner function for a beam Section 4.6.2), i.e., (.6.2 (Xaout -Xbout)), is 2,/(/3+,) at best when threshold
splitter. is reached: ( =(th= (/3+ ,)/2, similar to Eq. (6.141).

Use the characteristic function in Eq. (3.168) for the Wigner function (iii) Show that the threshold power for the non-degenerate OPO discussed
and the operator transformation in Eq. (6.52) for a lossless beam sp litter here is the same as that in Eq. (6.147) for a degenerate OPO.
to prove the transformation of the Wigner function in Eq. (6.72).
Problem 6. 7 Intra-cavity second harmonic generation with double reso -
Problem 6.5 Conversion between 1-photon and 3-photon for the photon nance condition [Collett and Walls (1985); Ou and Kimble (1993)].
doubler in Section 6.1.4. In Section 6.3.2, we considered the system of intra-cavity second harmo-
For the interaction in Eq. (6.18), if the input state is l'1Iin) = 11)3 11)2, nic generation (SHG) in a single-pass configuration for the harmonic wave
find the output state. What if the input state is I\[r in) = 10)313)2? by using classical wave arguments. We will treat the system with both
waves on resonance and with quantum arguments here using the method in
Problem 6.6 Non-degenerate optical parametric amplifier with losses. Section 6.3.4. We start with the Hamiltonian for SHG in Eq. (6.132) which
is re-written as
When a non-degenerate optical parametric oscillator (NOPO) is ope-
(6.155)
rated below threshold, it becomes a non-degenerate parametric amp lifier
(NOPA). We discussed the case without loss in Section 6.3.4. Now let us Denoting the decay constants as 11 and 12 for the fundamental and har-
include the intra-cavity losses by modeling the losses as the coupling of monic fields, respectively, which are related to the corresponding coupling
the two cavity modes to other outside fields, just like the degenerate case coefficients T 1, T 2 by Eq. (6.120), we have the equations of motion as
in Eq. (6.134) but now for two non-degenerate modes. For simplicity, we
assume the two cavity modes have the same output coupling coefficient of
da (.
dt = - ]Wa + 2
,1)a+
A At
rJ a -
bA
V
~A
,1ain,
a = /3 and the same loss coefficient of , .
db (. ,2 ) A 1 A2 A
dt = - ]Wb+ 2 b - 2rJa - '1fibin· (6.156)
(i) Using the method in Section 6.3.4 to prove that for doubly resonant
condition of .6.a = 0 = .6.b, the output fields from the NOPO have the For simplicity, we will assume double resonance condition Wa - wo = 0
frequency component at S1= 0 in the form of [Ou et al. (1992a)] and wb - 2w0 = 0 with w 0 as the frequency of the fundamental input field
and no intra-cavity losses for both waves. The equations above are cou-
pled nonlinear differential operator equations that are impossible to solve.
aout = cain + gbJn + G'ao + g'bb, However, both the harmonic and fundamental fields are strong in SHG so
A - -A - At - /A -/At
bout - Gbin +gain+ G bo + g a0 , (6.153) we can write b = (b) + .6.b and a = (ii) + .6.a with I(b)12 » (.6.bt.6.b) and
 •

188 Q'Ua11tu111Optics For Experimentali.sts Ge11cratio11 and Trn11sformatio11 of Quantu.m States

I(a) 2 » (6.at 6.a). Substituting these to Eq. (G.156) and keeping only up
1 (ii) Show that the output state for l2)a \2)b input state is
to the lin ear terms in 6.ba.nd 6.a, we can find the linearized equations for 2 2
\W.t)ovt = = TRV6(\4)A\0)B + \0)A\4)B) + (T + R - 4TR) \2)A\2)B
6.band 6.a.
+(T- R)J6TR(\3)A\l)B + \l)A\3)B)- (6.158)
(i) Use the method outlined above to derive equations for (b), (a) (zeroth
Similar to (i), if we make a four-photon coincidence measurement with
order) and for 6.b,6.a (first order).
detection of two photons at both A and B, then the coincidence will be zero
(ii) For second harmonic generation, the fundamental wave has a non-zero when T = (3 ± vf6)/6, R = (3 + vf6)/6. This is a four-photon destructive
coherent state input but the harmonic wave has vacuum input. Find the interference effect [Liu et al. (2007)].
steady state solutions for (b), (a) and (bout).
Problem 6.9 Schrodinger cat state with a loss [Walls and Milburn (1985)]
(iii) Using the method in deriving Eq. (6.147), show that the conversion
efficiency from input fundamental to output harmonic waves follows a cubic We can model the loss with a beam splitter of transmissivity T = l - ,.
equation similar to Eq. (6.102) in Section 6.3.2 but with the effective Teff = Use the result of Problem 6.4 and the Wigner function in Eq. (3.178) for
T1 and EeJL = 4EN L/T2 for this cubic equation. Here, EN L is the single - the Schrodinger cat state in Eq. (3.51) of Section 3.3 to find the Wigner
pass non-saturated nonlinear conversion coefficient given in Eqs. (6.100) function for the Schrodinger cat state with a loss of,. Show the result in
and (6.146). So, the effective nonl inear conversion coefficient is enhanced Eq. (3.56). You need to trace out one of the outputs of the BS.
by a factor of 4/T 2 due to the harmonic resonance cavity.

(iv) Use the method in Section 6.3.4 to solve the linearized equations for
6.b,6.a and find the spectra of squeezing for both the fundamental and the
harmonic waves.

Problem 6.8 Generalized Hong-Ou-Mandel effects for three and four pho-
tons.

Hong-Ou -Mandel two-photon interference effect is for two photons en-

ter ing on a lossless beam sp litter with one from each side [see Eqs. (6.54)
and (6.55) in Section 6.2.2 and more in Section 8.2.1]. Now, consider three
or four photons entering a lossless beam splitter of transmissivity T and
reflectivity R.

(i) Show that the output state for l2)all)b input state is
l\[!3)out = J3T 2 Rl3)AI0)B + J3TR 2 I0)Al3)B
+(T - 2R)VTl2)All)B + (R - 2T)J:llll)Al2)B, (6.157)
where a, b are the input modes and A, B are the output modes of the
BS. If we make a three-photon coincidence measurement with detection
of two photons at A and one photon at B, then the coinc idence will be
zero when T = 2R = 2/3. This is a three-photon destructive int erference
effect [Sanaka et al. (2006)], sim ilar to the Hong -Ou-Mandel effect for two
photons.
 PART 2

Experimental Techniques in Quantum

Optics and Their Applications
 Chapter 7

Experim ,ental Techniques of Quantum

Optics I: Photon Counting Technique

In the classical world where we are, it is usuall y difficult to observe quantum

phenome na. What we measure in experiments, such as current, voltage,
etc ., are all macroscopi c classical quantities. Then, how can we observe the
quant um behavior of optical field s from these measured classical quantities?
It turns out that, as we discussed in Chapter 5, the quantum behaviors
of light are revealed in the statistical correlations among the measured
quantities. The experimental techniques to be discussed in this chapter
and Chapter 9 will be some special experimental methods for observing
thes e statistical correlations that will show quantum phenomena in optics.
To understand these , we must connect what we learned in th e previous
chapters about the quantum theory of light with the physical observables
meas ured in the experiments. In this way, we will use the quantum theory of
light to explain the phenomena observed in experiments and furthermore
to guide the experiments in search of new quantum phenomena. In this
chapter, we will discuss the photon counting technique , one of the two
techniques commonly used in the experimental study of quantum optics.

Spectrum
Optical field : : :
B i Analyzer
---D : !2:J
i i
I Oscilloscope
: ~o :! ~
Theory ------ ----- --------- ---- ---- -- -------- ----------- ------

Photo-detection Experimental observation

Fig . 7.1 The measurement process in optics experiment: from incoming optical field to
recorded outputs of the measurable quantities. Photo -detection process bridges between
theor et ical description of the optical fields and the experimenta l observations.

193
 Quantum Ophcs For Expe '1·ime11talists Experimental Techniques I: Photon Cowiti11g Techniqne 195

7.1 The Process of Photo-Detection
In any one of the optical experiments, one device is essential, that is, the
photo-detector. Although we can see light ·with our bare eyes, they are
incredibly crude and inaccurate devices. Photo-detectors used in the lab are
usually made of photo-sernitive materials. The common materials are sem i-
conductors such as silicon, germanium, and gallium arsenide, etc. They are
made into photo-electric diodes. As shown in Fig. 7.1, photo-detectors are
the bridges between optical fields and the electric currents and voltages we
measure. For optical fields, we describe them with the quantum theory we
established in the previous chapters. The electric currents and voltages are
important parts of the experimental phenomena we can measure in the lab.
The theory of photo-detection connects the theory with the experiments .
In order to observe quantum behavior of light, the photo-detection pro-
cess must be a quantum process. Fortunately, this process exhibits the very
first observed effect correctly explained by the quantum theory of light, that
is , the famous photo-electric effect. Einstein was the first to understand
this effect when he proposed the concept of photon. The process of photo-
detection is also a quantum measurement process. So, the investigation of
this process will help us understand the essence of quantum mechanics.
are separated from each other. \,\'e can then work on each individual pulses
by first performing pulse height discrimination to eliminate the dark counts
and then pulse shaping for furLher operation. During this process, the
detector is quenched and cannot generate another photo-electric pulse and
so, it is not responsive to light. This time period (,.....,10-G s = 1 /LS) is
known as the ·'dead time" of the detector. This limits the maximum count
rate to around 106 counts per second (cps). For this type of photo-detector,
the light intensity must be low enough so that the photo-electric pulses are
resolved in time by the detector. Then, we can count each photo-electric
pulse. This is photon counting technique, as shown in Fig. 7.3.
ll£1>_v~ DA AAAA
Fig. 7.3 Photon counting technique for weak optical fields.
On the other hand, when the light intensity is large and there is no
pulse processing so that there is no dead time limit for the detector, the
photo-electric pulses from different photo-electrons will overlap to form a
continuous photo-electric current. Since the generation of photo-electrons

B
is random due to its probabilistic nature, the photo-current is a random
--- photoelectric process and will fluctuate with time, as shown in Fig. 7.4.
Optical field pulse
,~
~,,.,...._
e
~Amplifier!

PMT,APD G= w+- 106 TR - l ns

(a) (b)
t
Fig. 7.2 The process of photo-electric detection: (a) the generation of photo-electrons;
(b) the a mplifi cat ion of photo- electrons to photo-electric pulses. Fig. 7.4 Continuous photo-current und er relatively strong light intensity.

In the photo-electric process shown in Fig. 7.2, electrons are excited The experimental technique discussed in this chapter and its applicati-
out of the photo-materials by the light field illuminating on them. The ons in next chapter will deal with the case of weak light intensity where we
electrons generated by the illuminating light are called "photo-electrons". can count photo-electric pulses.
After the excitation, the photo-electrons are amplified within an extremely
10- 9 s = lns) to a level of macroscopic elect ric puls es that
short time (TR ,....., 7.2 Detection Probabilities of Photo-Electrons
are measurable by electronic equipments. The generation and amplification
are usually done in one single device such as photo-multiplier tubes (PMT), When the optical field is weak, the photo- electric events are rare. So, within
avalanche photo-diodes (APD). the duration of one photo-electric pulse (,6.t,.....,
TR,.....,1 ns), the probability
When the illuminating light is not so strong, the photo-electric pulses of generating two photo-electrons is very small: P2 « 1, that is, the photo-
 19G Quantum Optics For Experimentalists
Experi111ental Tec/111
:iques I: Photon Counting Technique 197

electric pulses are separated and do not overlap, as shmvn in Fig. 7.5. In this Hence, during a finite time interval of T, the probability of finding a photo-
case, the photo-electric pulses arc countable. We then use digital counters
electron is
to register the number of the photo-electric pulses arnl make a counting
measurement. (7.5)

By the definition of probability, we know that the number of photo-electrons

registered during time T is Ne oc A, that is,

Ne= CoPi = Tl l dt(ft(-)(t)ft(+)(t))w = l dtRe(t). (7.6)

------T----, Co, Tl are all some constants. In the expression above, the counting rate of
photo-electric pulses is then Re(t) = rJ(ftC-)(t)ft(+)(t))w. From Sections
Fig. 7.5 The non-overlapping photo-electric pulses generated by an optical field with 2.3.4 and 2.3.5, we learned that (EH (t)E( +) (t)) w = Rp( t) is the photon
low intensity.
number rate of the optical field in state I W). So, we have Re (t) = TJRp(t)
or Tl = Re (t) / Rp (t), that is, Tl is the ratio between the counted number of
The theory of photo-electric detection was established around 1963 by photo-electrons and the arriving number of photons. So, the physical mea-
Glauber and Mandel independently. Mandel's theory is a detection theory ning of rJ is the quantum efficiency of the photo-detector, i.e., the percentage
with quantized atoms but for classical optical fields (waves) while Glau- of photons converted to photo-electrons.
ber's theory is all quantum mechanical including the treatment of the op-
tical fields. Their results are similar and are all about the probability of
detecting photo-electrons. In Glauber's theory of photo-detection, when an
optical field in the quantum state of Iw) illuminates on a photo-detector, the
probability of producing a photo-electron in the infinitesimal time interval
oft to t + dt is given by I\
E 8 (t)
(7.1)
Here, a is a constant related to the detector but independent of the quantum Coincidence
state 1'11).:fu(+)(r, t) and :fu(-)(r, t) are respectively the positive and negative
frequency parts of the electric field operators of the optical field:
{F;;;; Fig. 7.6 Coincidence measurement between two detectors: only when two pulses overlap
:fu(-)t(r, t) = :fu(+)(r , t) = L J
d 3ky 2
i(k-r-wt)
cis(k) Ek,s e 2w)3/2 . (7.2) within TR , is a co in cidence pulse generated.
s=l,2 (
If we place two detectors at two separate locations denoted as r' and
For an optical field with only one polarization, under the one-
r" shown in Fig. 7.6, we can measure the correlation of the optical field at
dimensional quasi-monochromatic approximation (see Sections 2.3.4 and
these two location s. Actually, what we can really measure is the statistical
2.3.5), Eq. (7.1) becomes
correlation between the photo-electrons generated by the two detectors.
P1dt = a'(ft(-\t)ft(+\t))wdt, (7.3) The simplest method is to use an "AND" gate to measure the coincid ence
count Ne of simultaneously registering two photo-electric pulses within the
where
opening time TR of the "AND" gate, as shown in Fig. 7.6. The connection
[ft(-) (t)] t = ft(+) (t) = - 1- / dwci(w)e-iwt. (7.4) between this measurement and the optical field is given from Glauber's
photo-detection theory: for an optical field in quantum state I w), the j oint
 Quantum Optics For Experimentalists Experimental Techniques I: Photon Count 'ing Technique 199
198

probability of producing two photo-electrons , one by detector A at r' during Here, 7]A , 178 are the quantum efficiencies of detectors A and B , respectively.
time interval t' to t' + dt' and the ot her by detector B at r" during time Rz is the coincidence counting rate for simul taneous appearance of two
interval t" to t" + dt", is photo-electrons in time TR measured in the experiment. As mentioned
before , th is rate can be measured in the exper iment with the ··AND'' gate
P2 (r't', r"t")dt' dt" = ci a" (Ek-)(r", t")E~ - ) (r', t') method for the coincidence of two photo-electric pulses (see Fig. 7.6).
xE~+\r',ntk+\r",t"))wdt' dt". (7.7) For a conti nu ous opt ical field, the ch ange of the field with time is a
stat ionary process. Then, Re(t) of Eq. (7.6) and R2(t) of Eq. (7.13) are
Int egrating over time, we obtain the joint detection probability of two
independent oft an d become constant. The tota l count of photo-electrons
photo-electrons in a finit e t im e int erval T:
and the tota l coincidence count within time Tare Ne = R eT and Ne= R2T,
P2 = £ P2 (t' , t")dt' dt". (7.8) respect ively.
For a non-continuous pulsed optical field (multi -frequency), we need
As shown in Fi g. 7.6, we usually us e the "AN D" gate with a time window to integrate over the response time of the detector to obtain the overall
of TR in the expe rim ent to record the joint prob ability of simultaneous probability of ph oto -electrons. When the pulse duration is much shorter
arrival of the two photo-electric pulses. To interpr et the ph ysical meaning than t he respo ns e time of the detecto r , which is usually the case for ultra-
of Eq. (7.8) as related to the experimental obs ervations, we rewrit e the time short pulses, the time integral will cover the whole pulse and is equivalent
integral in Eq . (7.8) as to int egrat ion over the whole length of the pulse. In this case, we hav e the

1: 1:
probability of generating a photo-electron in one single pul se :
P2 = rdt r P2(t , t + T)dT = JT
JT }TR
rdtRp2(t) , (7.9)
Pi = p 1 (t)dt =a dt(E H(t) · E l+l( t))w. (7.14)
where we make changes of variables: t', t" • t , t + T, and define
In the one-dimensional quasi- monochromati c approximation, the equation
Rp2(t) = r P2(t, t + T)dT. (7.10)

1:
}TR above changes to

From the way Eq. (7.9) is written, we see that Rp 2 is interpreted as the rate P, = 1J dt(El-l (t)FJ(+l(t))w, (7.15)
of simultaneous appearance of two photo-electrons in time interval TR, that
is, Rp2 = dP2/ dt. It is proportional to the measured coincidence counting where T/ is the quantum efficiency of the detector. Similarly , the probability
rate of two photo-electrons in time window TR. for generating two photo-electrons in one pulse is
With the one-dimensional quasi-monochromatic approximation, similar

J
oo A( ) A( ) A(+) A(+) ))
to Eq. (7.3), Eq. (7.7) is changed to P2 = T/AT/B - 00 dt1dt2(E; (t2)EA- (ti)EA (t1)E3 (t2 'II· (7 .16)

P2 (t' ' t")dt' dt" = a' a" ce1-\t")Ei-\t')Ei+) (t')E1+ \ t") )wdt ' dt". (7.11) If the repetition rate or the number of pulses per second is Rp , the count
rate of photo-electric pulses is then R 1 = P1Rp, and the coincidence rate
Here ,Ei~\t) is given in Eq. (7.4). So, similar to Eq. (7.6), the coincidence
count Ne of two photo-electrons in time Tis proportional to P 2 and Eq. (7.9) for two photo-electric pulses is R2 = ARP.
becomes
7.3 Photon Counting
(7.12)
In quantum optics experiments, when the light field is weak, we can use
where R2 (t) ex Rp 2 ( t) and has the form of the photon counting techniques to study the behavior of the optical field ,
so as to obtain the photon statistics, photon correlation , etc. of the optical
R2(t) = r
}TR
T/ATJs(E1- \t + T)E i-\t)Ei+)(t)E1+\t + T))wdT. (7.13) field. These measurements will reveal the quantum behaviors of the optical
200 Quantum Opt'ics For Exp erim .P11I aUsts E:rp erim ental Techn'iqn es I: Photon Comdi11g Technique 201

---- ----- -- --
Photo
''{ PMT some pulsE's with small size. Another role of the discriminator is to shape
Detector : APO Electric Pulse
e- I
the pulses into some standardized digita l pu lses. Nowadays, this part of
Single ~---'~- ---, Photoelectron technology is n -1.ther mature ancl the whole ,process is integrated into a com-
Photon Amp lifier : Macroscopic Counter
Modu le mercially available single-photon module which takes in a photon thro ugh
j\ Electric Pulses
~-~----, I
fiber coupling and directly converts it into a standardized digital pulse for
Discriminator Pulse Shaping
I output. There are tvw types of common ly used digital pulses: one is the
Digital Pulses -from other TTL positive pulse with an adj ustab le pulse width of 100 ns; the other is
detector
the ult ras hort NIJ\1 negative pulse with a pulse width of a few ns. The TTL
TTL Pulse
(positive)
-71- NlM Pulse pulses are used for registering the number whereas the NIM pulses are for
(ns. negative)
time-related meas ur ement . As shown in Fig. 7.7(b), the standardized pul se
(a) (b)
from the output of the single-photon module is input into a "fan -in fan-out"
Fig. 7.7 (a) Commonly use d conve rsion process from photons to sta nd ard ized digita l device and is duplicated to a number of standard ized pulses for further pro-
electroni c puls es in photon co unting technique. (b) Di gita l electro ni c puls e co unting and cessing such as simple number counting and coin cidence measurement with
co incid ence counting . pul ses from other detecto rs.

field and exhibit quantum ph enom ena , such as quantum interference and TR___.
: :--

-u-
: ,
quantum entanglement. in
The centerp iece of the photon counting techniqu e is th e conversion from out , y in 2
LJ
y
optical signal to standardized digital electron ic pulses. Once we obtain
out
these digital pulses, we can app ly the mature digital techniques to process
and analyze the data in these pulses to obtain various statistical proper -
ties of the optical field. Figure 7. 7( a) shows a typical conversion process
Fig. 7.8 Coincidence measurement with an "AN D" gate.
from light to standardized digital electronic pulses. After the illumination
of the detector by the optical field, the generated photo-electron underg-
oes an initial amplification . This function is usually built in the detectors.
Photo-multiplier tubes (PMT) were popular devices in the ear ly days but 7.3.1 Coincidence Measurement with an "AND" Gate
were replaced by avalanche photo -diodes (APD), which output an electric The simp lest method to make a coincidence measurement is to use an
pulse triggered by one photo -electron. The gain of the initial amp lification "AND" gate . It can be general ized to the coincidence of multiple pulses
is around 103 - 4 . The electric pulse after the initial amp lificat ion is still from multiple detectors for the measurement of multi-photon corre lat ion
weak (ie r-v 104 e/lns = 1.6 x 10- 6 Amp) and requires further amplification effects. As shown in Fig. 7.8, a coincidence pulse is generated only when
to become a macroscopic electric pulse. The gain of the initi al amp lification the two or more input pulses have some overlap in the "AND" gate. This
by the PMTs or APDs usually depends on many uncontrollable factors so it requires the difference in arrival times of all the pulses to be smaller than
fluctuates widely , leading to a larg e variation of the size of the electric pulse the pulse width. The coinc idenc e count ing rate is given in Eq. (7.13) where
after the initial amplification. Furthermore, there also exist some thermally TR is determined by the width of the long er pulse of the two input pulses.
excited electrons in the initial amp lificat ion stage. They give rise to electr ic With coincidence counts and the single counts from each detector, we
pulses of relatively small size and are named "dark pulses " because they can ana lyze the data to obtain the stat istical properties of the optical fields.
are there even without the optical field. To eliminat e the dark pulses and In general, the higher the coincidence count is, the stronger the corre lat ion
produce a uniform electric pulse, a discriminator is used to set up certa in is between the optical fields measured by the detectors. However , we need
threshold for the electric pulses. This will cut most of the dark pulses and a reference to judg e how strong the correlation is. This reference is the
 Q'llantmn Opt'i cs Fo, E xp eT'im ental is t s E:cp er-im ent,al T ech11:iqu es I: Ph ot on C011,n tin g Tec hni qu e 203
202

accident al coin cidence. In the following , we will discuss how to defin e and const ant. It can also b e obt ained from Eqs. (7.13 ) and (7.16 ). When A B
measure the accidental coincidence in different circumstanc es. are two ind ep end ent opti cal fields , we h~ve

Accidental coincidence count is th e coincid ence count from two rand- ce t\ t + T)E/A-\ t )E1+\ t )Ek+\ t + T))w
omly generated pulses. It corresponds to th e accidental events between = (E~-)(t )E\+\ t ))(Ek-) (t + T)Ek+)( t + 7)) . (7.1 8)
two photo-detection events of completely uncorrelated optical fields. So, if
Using Eqs. (7.6) and (7.13) , we obtain
the measured coincidenc e count is higher than the accidental coincidence
count , this indicates a positive correlation between the fields, otherwise,
it is a negative correlation. For example, as we will see later, for optical
R ae= r
lrR
RA(t)Rs(t + T)dT. (7.19)

fields in a two-photon state, the probability is very high for detecting two For stationary fields, RA(t) , R 8 (t+T) ar e independent of time and they can
photo- electric pulses simultaneously by photo-detectors. In this case , the be pulled out of the integration in the expr ession above. Then , we arrive at
measured coincidence count is much higher than the accidental count. This Eq. (7.17). Sometimes , due to th e circumstance of the experiment, we need
is the photon bunching effect. On the other hand, for a field in a sing le- to close and open the detectors intermittently. If the intermittent time
photon state, since there is only one photon at a time , two detectors can is much larger than the characteristic time such as correlation time and
only generate one photo-electric pulse. The coincidence count in this case is coherence time of the optical fields , we can still treat them as stationary
much lower than the accidenta l count and sometimes is close to zero. This fields. But Eq. (7.17) needs modification since it only stands for the period
is the photon anti -bun ching effect. when the detectors are open. If the ratio between open and close times
Suppose the window of the coincidence counting is TR . Let us calculate is r, the actually measured count rates are then R~ = rRA / (r + 1), R~ =
the acc identa l coincidence rate Rae as related to the single count rate of r Rs/(r + 1), R~B = r RAB /(r + 1). Using Eq. (7.17) , we obtain the relation
each detector. Suppose we open detector A, B for a time interval T » TR, between all the count rates as
If the pulses from the two detectors are generated randomly, the probabi-
(7.20)
lity density for the two detectors generat ing the pulses is the same, i.e.,
PA = PB = 1/T. Then, the probability of generating a pulse by each of the For non -stationary pulsed optical fields , using Eq. (7.18) for independent
two detectors in the coincidence window TR is PA = PATR = TR/T = PB, fields and Eq. (7.16) for joint probability in the pulsed case, we have
Since the pulses from the two detectors are random, the joint probability of
generating two pulses, one from each detector, is PAE = PAPE = (TR/T) 2 P2 = T/AT/B
J oo A( ) A( +)
dt1dt2(EB- (t2)EB (t2))(EA
A(-) ( ) A( +) (
t1 EA t1
))

I: I:
-CX)

(see Section 1.4.3). Ifwe register NA,NB counts in time period T from the
two detectors, respectively, that is, the sample size is NA, NB, respecti-
vely, the counts in time period TR from each of the two detectors are
nA = NAPA = NATR/T = RATR,nB = NBPB = NBTR/T = RBTR,
respectively, where RA = N A/T, RB · = NB /T are the counting rate for the
two detectors, respectively. The above is for one random variable. With
two together, the sample size becomes NAB = NANB (see Section 1.4.3
for joint probability). So, the coincidence count in the time window TR
is nAB = NABPAB = NANB(TR/T) 2. Writing in terms of rates, we have
the coincidence rate RAB = nAs/TR = RARsTR. Since we are discussing
random events, this is the accidental coincidence rate and is written as
Rae= RARBTR, (7.17)
The express ion above is for stationary fields because p A = pB 1/T
= T/A dt,(E~-)(t,)E~+)(t,))'1B dt2(Ek-l(t2)Elt\t2))
= PAPE, (7.21)
Here, we used Eq. (7.15) for pulsed fields. If the repetition rate of the
optical pulses is Rp, we obtain the accidental coincidence rate as
Rae= P2Rp = PAPBRp = RARE/ Rp, (7.22)
where PA = RA/ Rp, P 8 = RB/ Rp. For intermittent pulsed fields, Eq.
(7.22) needs to be multiplied by a factor of (r + 1) /r, similar to Eq. (7.20~ •
In the experiment, the repetition rate of optical pulses from pulsed opti-
cal fields is easy to measure and it is straightforward to find the accidental
= coinc idence rate from Eq. (7.22) after we measure the count rate from
204 Quan /,um Optics For Experimentalists Experimental Techniques I: Photon Counting Tech11ique 205

each detecto r. For cont inu ous optica l fields , however , we need to m eas ur e u
30000
the coincidence window time TR. This can be done by using background 0

] ,. \ \
light in the lab, which is usually uncorrelated white light, for the illumin a- I
\
\/\
0) I
-0 20000
u
tion of the detectors. We regist er simultaneously the three countin g rates: C:
·o
u /\- /
I
RA, Rs , R AB for this light. Then the coinc idence window time TR can be 10000 I

.,,
C:
0 '-- ,"
calcu lated from Eq. (7.17). TR is usu ally quite sta ble for a coincidence de- 0
.c ', / __ ,
0..
0
vice so we on ly need to measure it once. Then we can use it to calc ul ate the 0
a:: 0 rr' 2 7( 3m2 2rr
E-
acc identa l coin cidence rate from Eq. (7.17) for any light in the expe rim ent. Single-Photon Phase Difference (rad. )
After finding the accidenta l coincidenc e rate, we can compare it with the
coin cidence rate measured from the optical fields of our concern to find the Fi g. 7.9 Application of the simpl e coincidenc e counting method: two-photon int erfe-
rence fringes as phase of some field is scanned. Reproduc ed from [Sun et al. (2006)]
corr elation prop ert ies of the fields. For this, we go back to the coincidence
rate in Eq. (7.13) and rewrite it as
that is, t he rat io of measured coincidence to the accidenta l coinc idence is
the normalized int ensit y correlation fun ct ion at zero time delay.
In the exper im ent , we usually change some physical parameters an d
observe how the coincid ence counting rate varies with these paramet ers.
A typical example is shown in Fig. 7.9 wh ere th e two -ph oton coincidence
counts show interfer ence fringes as the phase of some field is scanned . Notice
where that the period is 1r instead of 21r, which is typical of two-photon interference
fringes. In Chapter 8, we will discuss more about two-ph oton interference as
(7.24) well as multi-photon interference phenomena with more than two photons,
where multi -pho ton coincidence counts change with phases , forming multi-
is the normalized second -order intensity correlation function. For stat io- photon interferenc e fringe pattern.
nary optical fields , RA(t), Rs(t + T), and g~1(t, t + T) = g~1(T) are all
independent oft. Equation (7.23) then becomes 7.3.2 Time-Resolved Coincidence Measurement

RAs(t) = RARE
lrR
r dTg~1(T) , (7.25) In most multi-photon measurement experiments, the simple coincidence
counting method in Fig. 7.8 is enough. But for more complicated optical
which is also independent oft. Using Rae = RARsTR, Eq. (7.25) can be fields, we need to measure the time correlation between photons. It is re-
rewritten as lat ed to the normalized second-order int ensity correlation function g~1 (T).
RAB --
Rae TR
1TR
dTgAB
(2) ( T) . (7.26)
The measurement of this quantity requires the time-delayed coincidence me-
asurement method by introducing a time delay T, as shown in Fig. 7.lO(a).
In this case, the measured coinc idence rate in Eq. (7.25) changes to
Therefore, the ratio between the measured coincidence rate and the acci-
dental coincidence rate is just the average of the normalized second-order
intensity correlation function over the coincidence window TR. When the
RAs(t) = RARE r
lrR
dTg~1(T + T). (7.28)

coincidence window TR is much smaller than the characteristic time of the When the coinc idence window TR is much smaller than the characteristic
optical fields, such as the coher ence time, the optical fields change very time of the optical fields , g~1(T + T) g~1(T) for T within TR. Then ,
little within TR and g~1(T) g~1(o). Then Eq. (7.26) becomes from Eq. (7.28), Eq. (7.27) changes to

RAB/ Rae~ g~1(o), (7.27) (7.29)

20G Quantum Opt 'ics For Expc1·i.menta lists Expe1·imental Techniques I : Photon Counting Technique 207

nant to the atomic transition line is detected by two phototubes after split
by a beam sp lit ter and the time-de layed coinc idences Ne (T) are measured by
a.t ime-to-digital conve rt er (TDC) (Fig. '7.ll(a)). The normali zed int ensity
correlation function g( 2)(T) = 1 + >.(T)can be extracted from Nc(T) by the
(a) (bl formu la g (2)(T) = Nc(T)/Nc(oo), wh ich exh ibits the photon anti-b un ching
(2)
behavior of 9( 2 )(0) < g(2 )(T) (Fig. 7.ll(b)).
Fig. 7.10 (a) Tim e-d elay ed coincidence measurement. (b) l\1easurement of g_~B(T) by
time -to-digital co nvert er (TDC) or time-to-ana log co nverte r (TAC) a nd multi-channel
ana lyzer (MCA). 7.4 Theoretical Description of Experiments

After discussing the experim enta l techniques, we investigate in this section

Hence , as the t ime delay T is scanned, we obtain the normalized se~ond- how we can use the multi-mode description of optical fields discussed in
order int ens ity corre lation function g~1 (T). The non-zero range of 9~1 (T) Chapte r 4 to exp lain the exper imental observations and make prediction
gives the int ens ity corre lation time of field A and field B. for further expe rim ents. Th e comb in ation of exper iment and theory is the
The method above for scanning time delay T is don e one step at a time emp has is of this book. From this section on, we will see how this is done
and is quit e time -consuming. A more dir ect method is to us e a time -to - in detail.
digital converter (TDC) or a time-to -analog convert er (TAC) and a multi- We start with the two-photon correlation measur eme nt expe rim ent for
channel analyzer (MCA), as shown in Fig. 7.lO(b). It can measur e g~1(T) the simple process of spontaneous parametric down-conversion (SPDC) ,
in a single run. An example of time-delay ed coincidence measur ement is where a photon of higher energy is split into a pair of lower energy pho-
the Hanbury Brown-Twiss (HBT) experiment showing the photon bunching tons known as "signal " and "idler " via interaction with a i 2 )_nonlin ea r
effect, where the time delay is achieved by chang in g the position of one of crystal. Th e first description of this experiment was provided by Burnham
the detectors (see Fig. 1.9 in Section 1.6). and Weinberg in 1970 [Burnham and Weinberg (1970)] who observed a cor-
relation time of 4 ns. In 1985, Friberg et al. measur ed the time correlation
function between the two down-conv erted photons with faster modern di-
(a)
gital electron ic devices and obtained a correlation time of 100 ps [Friberg
/
Qptico! Pumping
Beom et al. (1985a)]. However , because the correlation time between the two
down-conv erte d photons (rv 100f s) is much shorter than the response time
- [ll.C~:O°m
Loser
of the detectors (rv lO0ps), the measur ed time correlation function is sim-
ply the time response function of the digital electronic devices. Because
of the large bandwidth from SPDC sources, no progress was made until
1----------L--- Amp.
oodOise~:~~:,,::• Loso,
1999 when Ou and Lu made the first measurement of the true time correla-
+---~~--- Alomic
Boom tion function from a narrow band spontaneous parametric down-conversion
Time 1ntervol r in nsec
inside an optical cavity [Ou and Lu (1999)].
From Chapters 4 and 6, we find the two-photon state from the SPDC
Fig. 7.11 Application of time-delayed coincidence co unting technique for the measure-
process has the form of
ment of /;1(T). (a) Experimental setup . (b) Exhibition of the photon anti -bunching
effect g ( 2 )(0) < g ( 2 )(T). Adapted from [Dagena is and Mandel (1978)]. 1'112)= / dw1dw2'll(w1,w2)a!(w1)a!(w2)IO), (7.30)

A typical examp le of applying modern digital technology for the me- where "s, i " represent the signal and idler fields , respectively. For a SPDC
asurement is the observation of the photon anti-bunching effect shown in process pumped by a CW field, we have
Fig. 7.11, where the fluorescence from single atoms driven by a laser reso- W(w1,w2) = 'l/J(w1)6(w1+ W2- Wp)- (7.31)
208 Quantum Optics FoT Experimentalists ExpeT"im,ental Techniques I: Photon Counting Technique 209

Her e, Wp is the angu lar frequency of the pump field , 'l/J(wi)is the spect ral For thi s, we first calcu late
function for the down- converted photons. For type-I down-conv ers ion pro-
cesses near frequency degeneracy, 'l/J(w i) is a symmetric function with re- Ei+)(t + T)Ei+)(i)l1'2) = ;7r / dwsUWiCLs(ws)e-iw.,fai(wi)e-iw;(t+T)
spect to wo = wv/2: 'l/J(w1)= 'lj)(wp- w1).
Using photo-detectors to make a direct det ect ion of the down-converted X / dw1dw2w(w1, w2)a! (wi)a! (w2)IO)
fields, we first register single detector count rate. From Eq. (7.6), we have = G(t , t + T)IO) (7.39)
the detected count rate for the signa l field as
with
(7.32)
(7.40)
where the field operator has the following form und er the one-dimensiona l
quasi-monochromatic approx im ation: being the two-photon wave function of SPDC. For the SPDC process pum-

[fti-\t)]t = fti+\t) = J dwcis(w)e-i wt_ (7.33)

ped by a CW field , w(w1,w2) is given in Eq. (7.31) . So,

G(t , t + T) = J
e-;:,t ®J21/J(wp - w2)e -iw ,r ""e-iw, (t+T/ 2)g(T) (7.41)
To evaluat e Eq. (7.32) , we first perform the following calc ulation:

fti+\t)l1'2) = J dwci5 (w)e-iwt

with
(7.42)
X / dw1dw2W(w1,w2)ci!(w1)aJ(w2)IO)
Hence , we obtain from Eq. (7.13) the coincidence count rate between the
= J dw1dw2W(w1,w2)e-iwitaJ(w2)IO), (7.34)
signal and idler fields as

wh ere we used the commutation relation [ci5 (w), ci!(wi)] = c5(w-w1). Hence,
R2 = 'rls'rli
lrR
r dTlg( T) 2.
1 (7.43)

(Ei-\t)fti+\t))IJ! = ~ / dw1dw2dw~dw;e-i(wi-w~)t For the single-pass SPDC pro cess in a x( 2 Lnonlinear medium , we usually
2 have TR » Tc = 1/ b.wPD C where 6.WPDC is the spectral width of the
2

x w(w1,w2)1'(w~, w;) (Olai(w;)aJ (w2)IO) SPDC fields, i.e., the width of 'l/J(w) so Tc is the width of g(T). Hence, we
can take the integration range in Eq. (7.43) as (-oo , +oo). Mor eover, since

2=
(7.35)
J_+oodTlg(T)l
-00 47r
J+oo dTdDdD' 'lj)(wv/2 - D)e- if:h'lj)*(wp/2- D')eiD'T
-oo

where we used Eq. (7.31). Finally, = J +oodDdD''lj)(wp/2 - D)'lj)*(wp/2 - D')c5(D- D')

27r -00

R1s = rts l7r / dw1l'l/J(w1)I2 •

2
(7.36) =
27r
J dDl'l/J(wv/2 - D) 2, 1 (7.44)
Similarly, we can find the single detector count rate for the idl er field as we obtain the coincidence rate as
R1i = 'rli l7r/ dw1l'l/J(w1)1-.
') (7.37) (7.45)
2
In order to find the corre lation time between the signal and idl er fields, we Here, we used the relation
need to calculate the intensity correlation function
2\t, t J+ oodTeiD'Te-if:h = b(D - D'). (7.46)
r( + T) = (fti-\t)Ei-\t + T)Ei+)(t + T)fti+\t))1J12
• (7.38) 21r -00
 210 Quantum Opt-ics For Exp erimentalists
Exp<>rimental Techniqu es I: Photon Counting Techn ique 211

Comparing with Eqs. (7.36) and (7.37), we have

~ 1n s) so that the average count for idler field is much smaller than 1,
R2 = T/iRls = 1]sRli· (7.47) i.e., N i (TR) = R1 1TH « 1, we find from Eq. (7.49) that R2 / Rae » l
Especi~lly in the ideal case when the quantum efficiency of the detectors is and it increases as R 1 i decreases . This phenomenon was first observed by
100%, 1.e., ' 7s = T/i = l, we arr ive at Friberg, Hong and 1\Iandel [Friberg et al. (1985b)] . The large coincidence -
to-accidental ratio (CAR) indicates a strong photon bunching effect . So,
R2 = R1s = Rli · (7.48) CAR ==R 2 / Rae » l is anot her two-photon ch aracter ist ic of the SPDC
physical meaning of the expr ession above is stra ightforward: the proba - source. Th is property was used to confirm the two-photon characteristic
bility of two-photon detection is t he same as that of singl e-ph oton detectio for the light source from a four-wave mixing process in optical fibers [Li et
h . n,
t at 1s, whenever we detect one photon, we also detect two photons. So, the al. (2004); Fan et al. (2005)].
detected field must be in a two-photon state . Not ice that even in the non- - The linear relationship in Eq. (7.47) does not describe comp lete ly the
ide al case of T/s, T/i < l due to the exist ence of losses, Eq. (7.47) indicates true situat ions in the exper iment. In fact, because of t he existence of ac-
that the coin cidence count rate is proportion al to the single-detector count cidental coincidences, there sho uld be a quadratic term in Eq. (7.4 7). For
rate. This is the characte ristic b eh avior of detect ing a two-photon-state. SPDC processes, the quadratic term originates from the acc identa l coin-
Figur e 7.12 shows this proportion al relationship , which was measured on cidence of two photon s from two random pairs, one from each pair. The
the SPDC fields with resonant enhanc em ent [Ou and Lu (1999)]. generation of two pairs of photons cannot be described by the state in
Eq. (7.30). We must consider high er-order terms from parametri c down-
conversion. Problem 7.1 discusses this case and gives such higher-order
term in Eq. (7 .57) , from which we can find the extra quadratic t erm (see
0
C
Probl em 7.3) added to Eq. (7.47) for modification. Th e final result is
§
0
(7.50)
u
C
0 where /3ex:TR.
_g
0..
The optical fields from SPDC proce sses are of strong nonclassi ca l nat ure.
This is exhibited mainly in the violation of Cauchy-Schwarz in equality that
stands for the classical sources. From the non-negative P-distribution for
classical fields, we can derive Cauchy-Schwarz inequality (see Eq. (5.58) in
Coincidenc e Count s in 0.1 sec.
Section 5.3):
Fig. 7.12 Th~ c~aracteristic prop erty of a two -ph oton state: the linear relationship
between th e comc 1dence count rate and the sin gle-detector count rate. Reproduced from (7.51)
[Lu and Ou (2000)].
which shows that the cross-correlation between the intensities of two fields is
always smaller than the geometric average of the intensity auto -corr elations
In addition to the proportional characteristic property demonstrated
of two fields. The int ensity cross-correlation of two optical fields can be me-
in Fig. 7.12, another characteristic of the SPDC fields is the high photon
asured directly by registering the coinc idence counts between the detections
correlation between the two down-converted fields. This can be seen from
of the two fields. The auto-correlation of an optical field can be measured
the ratio between the coincidence count rate to the accidental coincidence
rate: by sp littin g it into two with a 50:50 beam splitter and then making a coi-
ncidence measurement between the two split fields. Since a beam splitter
(7.49) reduces light intensity , we need to normalize the inequality in Eq. (7.51).
with Ni (TR) = R1iTR as the average number of counts in time interval Dividing two sides of Eq. (7.51) by (Ii)(h), we obtain
TR for the idler field. When the coincidence window TR is small (typically (2) (2) ( ) (2) ( )
912 (0) :s: 911 0 922 0 ' (7.52)
 E :rperimental Teclmiq'lles I: Photon Counti11g Technique 213
212 Quantum Optics For Experimentalists

where 9g )( O) = (IiI2 )/(I1)( h ), 9g\o) = (Ir)/(Ii)2, 9g)(O) = (IJ )/( h )2. 7 .5 Problems

The average hC're is for the probability distribution of the classical fields. Problem 7.1 Photon bun ching effect for the signal or i<ller field alone in
For quantum states, we use Eqs. (7.24) and (7.27) for cross-correlation and
SPDC.
the beam splitter method discussed above for intensity auto -corre lation .
For th e state of SPDC in Eqs. (7.30) and (7.31) , we have already obtai- After we consid er the high er-ord er term, the quantum state of SPDC in
ned 9g) (0) = r (2) / Us) (I i) ex [9(0) [2 # 0 from Eqs. (7.38), (7.40) and Eq. (7.3 0) changes to (see Sect ion 6.1.5)
(7 .42) . It is also straightforward to show that 9g\o) = 0 = 9g)(O), 2) =
['11 J dw1dw2'1'(w1,w2)a!(w 1)aJ(w2)[0 )
with the subscripts s = 1, i = 2, because the signal field and idler field
each have only one photon, as shown in Eq. (7.30). With these values for
9g), 9i~) , 9g), it is obvious that the inequality in Eq. (7.52) is violated. In
+~
2
J dw 1 dw2 dw~dw; w(w1, w2)'1'(w~, w;)

the act ual exper iment, however, because of the existence of the acciden - xa1( w1)aJ (w2)a1(w~)aJ (w;)[0). (7.57)
tal coincid ences, 9g\o) , 9g\o) # 0. These accidental coincidences stem we will work on the signal field only.
from random two-pair events in SPDC. This situation cannot be described
(i) For the SPDC process pumped by a CW field , E~. (7 .31) gives
by the quantum state in Eq. (7.38) since it neglects the contribution from
w(w1, w2) = 'lj)(w1)5(w 1 + w2 - wp)- Use this and the state m Eq. (7.57) to
higher -order terms (see its derivation in Section 6.1.5). In Problems 7.1 and
prove
7.3, we will consider the case of two-pair generation in SPDC and calculate
the coincidence count rate from each field.
ri~)(t, t + T) = (Ei-\t)Ei-\t + T)E i+)(t + T)Ei+\t)h12
Nevertheless, we can approach this from the experimental point of view. = H 2(0) + H(T)H*(T), (7.58)
In the experiment, we usually measure 9g)(T). For this, the Cauchy - where H(T) = (1/2n) J dw1['l/J(w1)\ 2e-iw 17
From Eq. (7.35), we have
.
2
Schwarz inequality in Eq. (7.51) becomes H(0) = (El-\t)Ei+\t)) ,h- From this, prove 9i;\T) = 1 + h(T)[ with
,1(7 ) = H( 7 )/ H(0). This result is the same as that for a thermal source
(1i (t) h (t + T)) u?(t)) ([d(t + T)) (7.53)
(see Eq. (4.29) in Section 4.2.4).
(2)( ) (2)( ) (2)( ) (7.54) (ii) For the SPDC process pumped by a pulsed field, w(w1, w2~ does not have
912 T 911 0 922 0 ·
Integrating Eq. (7.54) and using Eq. (7.26), we have the frequency correlation in Eq. (7.31). We can calculate with Eqs. (7.15)

(2) (2)
9s s (0)9ii (0) ·
Here, s = 1, i = 2. Rsd Rae is exactly the coincidence -to-accidental
(7.55)
P, =' 1) I:
and (7.16). Prove the single -photon counting probability is
dt(Ej-l(t)El+l(t))<t,

(CAR) measured in the experiment. From Problem 7.1, we have

(2)
rat io
91;\o)
= = Tl J 2
dw1dw2\'1'(w1,w2)\ , (7.59)

I:
2 = 9ii (0). Then, Eq. (7.55) becomes
and the two-photon counting probability for the signal field is
(7.56)
Pss c'c 1)2 dt1dt2(£j-) (t2)Ej- ) (t1)iJj+)(t1)£j+) (t2))<t
As we discussed about Eq. (7.49) for SPDC, Rsd Rae can become very large
for sma ll TR. In this case, the Cauchy -Schwarz inequality in Eq. (7.56) is vi- = 'rl2(A + E), (7.60)
olated for the optical fields from SPDC. This demonstrates the nonclassical
A=Jdw dw2dw~dw;\w(w
where
property for the optical fields generated in SPDC. Experimentally, Clauser
first demonstrated the violation of Eq. (7.52) with atomic cascade em ission
1 1 ,w2)W(w~,w;)l2 = P; (7.61)

[Clauser (1974)]. Zou et al. first observed the violation of Eq. (7.55) for
SPDC [Zou et al. (1991a)] .
E = J dw1dw2dw~dw;w(w 1,w 2)w(w~ , w;)w*(w1,w;)w*(w~,w2)- (7.62)
214 Quantum Optics For Experimentalists Experimental Techniques I: Photon Count-ing Technique 215

Hence , (i) For the SPDC process pumped by a pulsed field, we ne ed to use
Eq. (7 .16) to calcu late the two-photon coin cidence counting probability Psi ·
(7.63) In Problem 7.2, we have calcu lated the contribution from the first term in
Notice that £ ::; A so that g~;) ::; 2. When £ = A , we have g~;) = 2, or Eq. (7.57). Now we can use simil ar method to calcu late the contribution
exact ly the sa me as a singl e-mod e thermal source. The calc ul ation for the from the second term. Prove the contribution to Psi from the second term
idler field is identic al. of Eq. (7.57) is

(iii) For the pulse spontaneous parametric process described by Eq. (6.36)
in Section 6.1.4, us e Eqs. (7.61) - (7.63) to prove
P;;
i = 11,T/i[ : dt1 dt2 (Ef-l (t2)Ej-) (t1)Ej t) (t1)E/+) (t2)) w\''

(2) = 1 + "',,\ 4
= CrJs'r/i(A+ £) 2 = C(l + £ / A) 2 PsPi (7.67)
g ss L J (7.64)
j with C as a constant . Find the constant C.
with
(ii) Do the same as (i) but for the SPDC process pumped by a CW field.
sinh rj~
,,\j = ---:======== (7.65)
/r:,jsinh Tj~
2

For M modes with equal rj = r, we have g~;) = 1 + 1/M . This recovers the
result in Eq. (4.84) of Chapter 4, which is based on classical wave theory.

Problem 7.2 Intensity correlation between the signal and idler fields of
pulse-pumped SPDC.

For the SPDC process pumped by a pulsed field, we have calculated the
single and two -phot on detection probabilities for each field of SPDC in the
pr evious problem. We now continue to calculate the intensity correlat ion
between the two fields from SPDC. Use Eq. (7.16) for pulsed fields to prove
the coincidence counting probability for the two-photon detection of the

I:
signal and the idler fields is

P,i = r,2 dt1dt2(E/-)(t2)Ej - l(t1)El' l(l 1Ji;)+l(t 2 ))w

J
= 'r/s'r/i dw1dw2 l'-lF(w1, 2
w2) = rJiPs = rJsPi.
1
(7.66)

This result is exactly the same as Eq. (7.47) for the case of CW pumping .

Problem 7.3 The contribution to photon coincidence counting from rand -

omly generated two pairs of photons in SPDC.

Using the second term in Eq. (7.57), we can calculate its contribution to
photon coincidence counting. Similar to Problem 7.1, the calculation resu lt
depends on the bandwidth of the pump field for SPDC.
 Chapter 8

Applications of Photon Counting

Techniques: Multi-Photon
Interference and Entanglement

Photon counting techniques are th e most commonly used exp erimental met-
hods for the observation of multi-photon interference effects , which rely
on multi-photon coincidence measurement. There are other experimen-
tal methods developed for coincidence measurement, e.g. , frequency up-
conversion [Dayan et al. (2005); Lukens et al. (2013)]. But because of the
easiness of operation and good efficiency for coincidence measurement, pho-
ton counting techniques are the favorite of experimentalists for observing
some quantum optical and multi-photon interference phenomena. There is
a monograph by the current author with some detailed coverage on vari-
ous multi-photon interference effects [Ou (2007)]. Ten years have passed
since the publication of the monograph and we also introduced the mode
theory for quantum optics in this book. So, in this chapter we will treat
some of the old multi-photon interference effects from a new perspective
of mode theory to establish a good conceptual understanding and then we
will discuss some of the recent developments.

8.1 Multi-Photon Interference in General

8.1.1 Single-Photon Interference and Two-Photon

Interfere nee

When we talk about single-photon and multi-photon effects , one immedi-

ately considers the average photon number, i.e., (n): single-photon case
corresponds to (n) « 1 whereas multi-photon case to (n) » 1. These
are only superficial but not fundamental difference. Consider the famous
Young's double slit and its variations such as Michelson and Mach-Zehnder
interferometers. The interference fringe patterns do not depend on how
many photons there are as long as the exposure time is long enough to es-

217
218 Quantum Optics For E;rperim entalists Applications: Multi-Photon Interference and Entanglement 219

tablish the fringes , even for the case when there is only one photon between A but different phases <p10,<p20-By Glauber's photo-detection theory, the
the source and the observation screen [Taylor (1909)]. This prompted Dirac two-photon coincidence probability is then proportional to
to make the following statement about photon interference [Dirac (1930)]: P2(r1,r2) ex (Et( r1)Et(r2)E(r1)E(r2) )if;
Each photon only inte1jeres with itself. 2
Different photons never interj ere. = IIE(r1 )E(r2) l?j;)I 1

This suggests that it is the single-photon effect that is responsible for in- = ll<D2( r1, r2)l2
r1.r2)l~;)ll2 = l<D2( (8.2)
terference phenomena even at large average photon number: (n) » 1. with
However, the situation changed after intensity correlation technique was = A 2 [ ( eik1·r1eik2·r2 +
invented by Hanbury Brown and Twiss. In 1967, Pfleegor and Mandel de- 2(ri, r 2 )
<}) eik1·r2eik2·r1) ei( cp10+cp2o)

monstrated interference phenomenon in intensity correlation [Pfleegor and +eikl'(r 1+r2) e2icp10+ eik2·(r1+r2) e2icp20
]. (8.3)
Mandel (1967a,b )] . They observed an interference pattern in coincidence
measurement between two detectors in the low photon number limit when Since the two lasers are independent so that <p10- <p20 is random, P2 has
there is no more than one photon at a time between the source and the contributions only from the absolute values of the three terms in Eq. (8.3),
detectors. where the last two terms exhibit no interference and only the first term
There is one thing in Pfleegor-Mandel experiment that is fundamentally contains the addition of two quantities and gives rise to interference effect:
2
different from traditional interference: intensity correlation. From what IA 2(eik1·r1eik2·r2 + eik1·r2eik2·r1)ei( <p1o+cp20)
I
we learned in Chapter 7, both detectors must register a photoelectron re-
spectively in order to obtain a signal. This means that Pfleegor-Mande l = 2A 4 [1 + cos D.k · (r1 - r2 )] = 2A 4 [1 + cos 21r(x1 - x2)/ L], (8.4)
interference effect is a two -photon effect. To see more clearly about th is, where D.k = k 1 - k 2 D.0lk 1In for a small angle D.0 between k1 and
let us examine Pfleegor -Mandel experiment in more detail as follows. k 2 with lk 1 = lk 2 = 21r/>. for wavelength>.. Xj(j = 1,2) = rj · n is
1 1

the location of detector j along the fringe direction n = D.k/lD.kl and

L = 21r/ID.kl >./D.0 is the fringe spacing. Combining the three terms, we
have
n= L\k/lL\kl
tIx~ P 2 (r1 , r2 ) ex 2A 4 + 2A 4 [1 + cos 21r(x1 - x2)/ L]
= 4A 4 [1 + 0.5cos21r(x1 - x2)/L] . (8.5)
This is consistent with the result of Pfleegor -Mandel experiment [Pfleegor
and Mandel ( 196 7a, b)]. If we discard the parenthesis of the first term
Fig . 8.1 Pfleegor-Mandel two-photon interference experiment with two independent but in Eq. (8.3), there are totally four terms in the expression. Figure 8.2
attenuated lasers. graphically depicts these four situations. Figures 8.2( a) and (b) correspond
to the first two terms in which the two detected photons are respectively
In Pfleegor -Mande l experiment, two optical fields from two independent from two different lasers whereas Figs. 8.2(c) and (d) correspond to the last
but heav ily attenuated lasers are allowed to superimpose in an area where two terms in which the two detected photons are from the same laser. Since
two photo -detectors are located at r 1,r2 (x 1,x 2 in Fig. 8.1). Let us write coincidence measurement cannot distinguish the cases in Figs. 8.2(a) and
the field operator as (b), they give rise to interference effect shown in Eq . (8.4). No interference
(8.1) occurs for the cases in F igs . 8.2(c) and (d) because they are from different
lasers and distinguishab le .1
Here we assume plane wave mode for the fields . Let the fields be respective ly 1 Interference could arise from these two terms if there is a fixed phase difference between
in the coherent states : l?j;) = IAeicpio)® IAei'P20) with the same amp litude them, which requires phase correlation between two lasers and erases d istinguishability.
 220 Qiwnturn Optics F01· Experimentalists
(a) (b) (C)
Fig. 8 .2 Four possibilities for the two-photon coincidence detected
two-photon interference exper im ent .
Since int erference effect exh ibit ed in coincidence measurement involves
two photons, it is called two-photon interference. From what we see in
Figs. 8.2(a) and (b), the two detected photons act as one entity and in the
sp irit of Dirac , the two-photon ent it y int erferes with the ent ity its elf. This
cons ideration leads to the concept of two-photon wave function.
8.1.2 Two-Photon Wave Function
From Eq. (8.2), we see that the two-photon detection probability is the
absolute square of the function <I>(r1 , r 2 ) , which consists of four terms cor-
responding to the four situations depicted in Fig. 8.2. Concentrating on
the first two terms, we find that it is the superposition of these two terms
that gives rise to the two-photon interference effect. We can further write
them explicitly as follows:
A2eik1-r1eik2-r2ei(cp10+'P20) = Aeik1·r1ei<p10Aeik2-r2ei<p20
A2eik1•r2eik2-r1ei(cp10+cp20) = Aeik1·r2 eicp10 Aeik2-r1eicp20
Here ¢j(r) = Aeikj·rei'Pj 0
(j = 1, 2) is the one-photon wave function , whose
absolute value square is the single-photon detection probability. Hence , the
two superposing amplitudes are each the product of two one-photon wave
functions, one for each photon. So, in two-photon interference, we need to
consider the wave functions of the two photons together and obtain a new
wave function which we call "two-photon wave function". The absolute va-
lue square of the two-photon wave function gives the two-photon detection
probability in exactly the same way as the one-photon wave function.
Notice that there is an important feature in two-photon interference
that is different from single-photon interference: the interference pattern
does not depend on the phase difference between the two superimposed
fields . This is because the two-photon phase is the sum of the phases of
the two fields: (()2p = (()10+ (()20and appears in both superposing terms
Applications: J,,fulti-Photon Int c'ljerenc e and Entangl em ent 221
and eventually cancels in the final result. Because of this. tvw-photon
interference is phase-insensitive. 2
The concept of two-photon wave function can be generalized to arbitrary
(d)
N-photon cases when an N-photon coincidence measurement is performed.
in Pfleegor -I\Iande l We \Vill discuss three- and four-photon interference later in Section 8.3.
8.2 Various Two-Photon Interference Effects
In Pfleegor-Mandel exper iment, which was done with classical sources of la-
sers, we find from Eq. (8.5) that the visibility of interference is only 50%. It
turns out that this is not uniquely limit ed to Pfleegor-Mandel experiment.
IVIandel was the first to point out that the visibility of two-photon int erfe-
rence with classical sources cannot exceed 50% [Mandel (1983)], which was
later generally proved [Ou (1988)]. Richter in 1977 and Mandel in 1983
proposed to use nonclassical fields for two-photon int erference and showed
that it is possible to achieve a visibility of 100% in two-photon interference
[Richter (1977); Mandel (1983)]. A quick examination of Eqs. (8.3) and
(8.5) and the related Fig. 8.2 reveals that the reason for 50% visibility is
the contributions of the two cases where two photons come from the same
laser (Figs. 8.2(c) and (d)) and this contribution always exists for classical
fields.
To reach a visibility higher than 50%, we need to reduce this type of
= ¢1(ri)¢2(r2),
contribution and this leads to quantum sources with photon anti-bunching
= ¢i(r 2)¢ 2(ri)- [Ou (1988)]. A single-photon state is the most anti-bunched photon source
(8.6) and gives no two-photon event and thus no contribution from the cases in
Figs. 8.2(c) and (cl). Indeed , for a quantum state of l?,b)= ll)i @11)2, where
each side has only one photon , we can easily check that the visibility of
two-photon interference is 100%:
(8.7)
The first two -photon interference experiment with a two-photon quantum
state was performed by Ghosh and Mandel , who demonstrated a two-
photon interference fringe is consistent with Eq. (8.7) [Ghosh and Mandel
(1987)]. After that, various two-photon interference phenomena were dis-
covered with two-photon sources [Mandel (1999)]. In the following , we will
discuss a number of typical ones.
2
Phase-sensitive two-photon interferenc e may occur if th e two photons together follow
separate paths for int erference. See Section 8.2.2.
222 Quantum Optics For E1:perhnenta lists Applicat'ions: l\Julti-I'hoton Interference and En tanglement 223

8.2.1 Hong-Ou-Mandel Interference function for the inp ut fields is simp ly

P erh aps t he m ost well-kn own tw o-phot on inte rfer ence ph enom enon is t he w;~)= ¢ 10¢20 = A 2ei(cprn+"° 20
)ui(l)u2(2). (8.9)
Hong-Ou-M and el effect [Hon g ct al. (1987)]. It s simpli city in b ot h geom e-
Afte r t he beam sp litte r. eac h wave is sp lit into two:
tr ica l st ru ctur e and ph ysica l pi ct ur e m akes it a favori te exa mpl e in qu ant um
opt ics t ext b ooks. It has b ecom e a st and ard tec hniqu e for test in g p articl e ¢10 -+ tAeicp10 u1 (1) + rAeiSCH
1
1.L1
(2),
indi stin gui shabilit y. Furth ermor e, it is also th e b as is for lin ea r opti ca l qu an- ¢20-+ t' Ae 1'P20 u2(2) + r' Aeicp 20
u2( l ), (8.10)
t um compu t ing [Knill et al. (2001 )] and a numb er of pro toc ols in qu antum
inform ation [Zeilin ger (1999) ]. As we will see in th e followin g, it clearly where t, r, t', r ' are t he compl ex ampli t ud e tran smi ssivity and reflecti vity
illustrates th e importanc e of mod e conc ept in und erst anding qu antum op- for th e two sid es of th e b ea m spli t t er , resp ect ively. Th e tw o-photon wave
ti ca l phenom en a . fun cti on for th e ou t put is t hen
+ r A eicp + r' A eicp
•! a2 W~~t = [tA eicp10 u 1 (1) u 1(2)l[t' Aeicp 20 ,u 2 (2) 20
10 u 2 (1)]
= tr' A 2ei(cpio+'P 20 2
)u1 (1 )u 2(1) + rt' A ei(cp +cp )u1 (2)u 2 (2)
10 20

BS

al

&2 !
-
T, R
bl
=
+tt' A 2ei(cp 10 +'P 2 o) u 1 (1)u2(2) + rr' A 2ei(cp i o+cp 2 o) u1 (2)u2 ( 1)
A 2ei(cp 10 +'P 2 o) [tr'u 1(l)u 2(l) + rt'u1(2 )u 2(2)
+tt'u1(l)u2(2) + rr'u1(2)u2(l)]
Fig. 8.3 Hong-Ou-Mand el interferom et er with two photons . = A 2e i(cp10 +'P 2 o) [tr'u 1(1)u 2 (1) + rt'u1(2)u2(2)
+(tt' + rr')u(l )u(2)] if u 1 = u2 = u. (8.11)
The Hong-Ou-Mandel interferometer involves simply a 50 :50 beam split -
ter and two photons input with one respectively from each side, as shown in Pictorially , the four terms in Eq. (8.11) correspond to the four possibilities
Fig. 8.3. With an input state of Il)i I1)2 to the beam splitter of amplitude shown in Fig. 8 .4. The first two terms (Figs. 8 .4 (a) and (b)) are distinguis-
transmissivity t and reflectivity r, the output state in single-mode case was hable whereas th e last two t erms (Figs. 8.4(c) and (d)) are indistinguishable
derived in Section 6.2.2 as if u 1 = u 2 = u and are summed together in the last line of Eq. ( 8 .11) . No-
tice that the last two terms in Eq. (8.11), which correspond to two photons
2
IW)out = v'2tr(l2)il0)2 - I0)il2)2) + (t 2 - r )ll)il1)2 going to separate ports, can be written as
1 1
= v'2(l2)i I0)2 - I0)i 12)2) for t = r = v'2. (8.8) A 2ei(cpio+'P
20 \tt' + rr')u(l)u(2) = (w~~tc + w~~tJu(l)u(2), (8.12)

wher e w( 2) = tt' A 2ei (cpi o+ 'P2 o) ' w( 2) = rr' A 2ei (cp 10 + 'P2 o) are the two-
The lack of state 11, 1) in the output state IW)out when t = r = l / \1'2 is outc outd
a result of destructive two -pho ton interference. The multi-mode treatment photon wave functions without the indistinguishable mode functions
for a two-photon state from spontaneous parametric down-conversion can u(l)u(2) for the cases (c) and (d) in Fig. 8.4, respectively. The spatial indis-
be found in the monograph by the current author [Ou (2007)]. But here, tinguishability leads to the superposition of the two-photon wave functions.

·-+-· ·~•
we will take a different approach from the point of view of two-photon wave 0 0
0
function discussed earlier in Section 8.1.2. This will illustrate further its
Lo L_o
difference from the traditional single-photon interference.
Taking the same notations in Section 8.1.2, we assume the wave functi-
•- -- • 0
ons for the two input photons are ¢ 1 = Aei'P10 u 1(l), ¢ 2 = Aei'P20 u 2(2) with (a) (b) (c) (d)

the same amplitude of A(> 0). u 1, u2 are their mode functions and the
numbers in the parenthesis denote input ports. So , the two -ph oton wave Fig . 8.4 Four possibilities for the two photons in Hong-Ou-Mandel interferom eter.
 224 Quantum Optics For E:cperimentalists Applications: Multi-Photon Int e1j'erence and Entanglement 225

For a lossless beam splitter, the phases of t , r, t', r' must satisfy 'Pt + 1000 ~-------------,

= 1r due to energy conservation from input to output (see

'Pt' - cp,. - 'Pr'
Section 6.2.1 and Appendix A). Then , for a 50:50 beam sp litter, we have 8 600

u
i;r,(2) (2) ( I
outc + + rr I § 400
Woutd CX: tt )
8
= ei(cp,.+cp,.1)[ei(cpt+cpt,-cp,.-cp,") + l]/V2 0
200

o o c__---~~----~
= ei(cpr-+cp,.,) [ei1r + 1]/ V2 z 260 280 300 320 340
Position of beam splitte r {um)
360

= 0. (8.13)
(a) (b)

Fig. 8 .5 Hong-Ou-Mandel experiment. (a) Experimental layout ; (b) Two-photon coi-

So, the last two superposition terms in Eq. (8.11) add to zero due to de-
structive interfer ence, leading to no appearance for the two photons at se-
parate ports. In this case, the result based on the two-photon wave function
in Eq. (8.11) is the same as that based on the quantum state in Eq. (8.8).
Notice that the interference discussed here is between two-photon wave
functions wi~tc, wi~td so it depends 011 the phase difference of the two
wave functions, which is the phase difference of the phase sum of each pho-
ton, i.e. , the difference between 'Pt + 'Pt' for case (c) and 'Pr + 'Pr' for case
(d) in Fig. 8.4. The initial phase cp10 + cp20 is canceled out in this way.
Notice that, unlike the single-photon interference, the two-photon interfe-
rence does not depend on the phase difference cp10 - <p20 between two input
phases. Another thing is that the complete cancelation depends on the
complete overlap of the mode functions u 1 = u 2 . This means that photon
indistinguishability relies on the mode of the photons (see Section 8.4) .
Experimentally, if we make a coincidence measurement between the two
output fields of the beam splitter, we should expect no coincidence count
except accidental count according to Eq. (8.8) or the discussion above .
However, these arguments on ly apply to the single -mode case. In the expe -
riment, we usually have multi-frequency modes and the two-photon wave
functions become two-photon wave packets. Only when the two wave pac-
kets overlap at the beam splitter, wh ich correspond to u 1 = u 2 = u in
Eq. (8.11), a complete destructive interference can occur, leading to no
coincidence count. Otherwise, we have a non-zero coincidence count de-
pending on how well the overlap is. This is what was observed by Hong
et al. and shown in Fig. 8.5(b) [Hong et al. (1987)]. The experimental
layout is shown in Fig. 8.5(a), where two photons from spontaneous para-
metric down-conversion (SPDC) in a nonlinear crystal of KDP are input
to a 50:50 beam splitter after filtering out scattered light from the pump
field. Two-photon coincidence measurement is performed on the two out -
put fields with the photon counting techniques discussed in Chapter 7. As
ncid ence dip as the position of the beam splitter is scanned. IF: interference filter ; BS:
bea m splitter. Reproduced from [Hong et al. (1987)].
the position of the beam splitter is scanned, which changes the overlap of
the two input photons, a dip occurs in the coincidence count at optimum
overlap for destructive interference.
Hong-Ou-Mandel experiment with two independent photons
In the two -photon wave function discussion above, we take the two -photon
wave function as the product of the single-photon wave functions ¢10, ¢20.
We can do this for two single-mode photons. But if the two input photons
are of multi -mode nature (frequency mode), this is true only when the two
input photons are independent of each other and each photon corresponds
to a single -photon wave packet in a single-temporal mode, as discussed
in Section 4 .3.2. Although the original experiment by Hong et al. was
performed with correlated photons from SPDC, the argument above shows
that the two -photon destructive interference effect should occur even for
two independent photons. Indeed , the Hong-Ou-Mandel effect was observed
with photons from a variety of independent single -photon sources such as
the heralded single photon from parametric down -conversion [Wang and
Rhee (1999); de Riedmatten et al. (2003)], single quantum dots [Santori et
al. (2002)], single atoms [Beugnon et al. (2006)], single ions [Maunz et al.
(2007)] . In the following, we will discuss this situation .
In writing Eq. (8.10), we already assumed the mode functions are diffe-
rent for two input photons but we set them equal in Eq. (8.11) for complete
destructive interference and only qualitatively discussed the situation when
they are different. For a quantitative discussion , we need to specify the
mode functions u 1 , u 2 . For independent photons, we can describe them
with wave packets of different shapes and arriva l times . To treat this case
 226 Qua nt-um Opt ics For Experim ental ists A ppli cation s : Mult i- Phot on Int erf eren ce and Entangl em ent 227

quantum mechanically , we take the input state as /\J!)in = /T1 )i 0 /T2 h with th e visibility of interference defined as
with the two input photon states as

(8.14)
V 12 (T1 - T2 - D /c ) =
J dt 1 dt2 1* (t1 - T 192f
~g ) * ( 2- T 2)

x g2(t 1 - T2 - D /c )g1(t2 - T1 + D /c )
To illustrate the role of mode match in interference , we assume they have
the wave shapes given by g 1 ,2(T) = (1/ ~) J dw¢ 1,2(w)e-iwT besides the =I/ dw¢ ~(w)¢2(w)eiw(T1- T2- D/cf. (8.19)

different arrival time T1.2- Ai_

2(T) = J dw¢1,2(w) eiwTai,2(w) is the creation Here , we used J dtlg 1(t)/ 2 = 21r = J dt/g2(t)/ 2 for single-photon states.
operator for the single temporal mode defined by g 1 , 2 ( T). When the paths are balanced with D / c = T1- T2, the visibility V12depends
Taking the one-dimensional quasi -monochromatic approximation, we on the mode overlap:

=I/
have the output field operators of the beam splitter as 2
V12(0) dw¢~(w)¢2(w)i S 1. (8.20)
Ei \t) = VTE1(t) + VRE2(t - D/ c),
0

0
E~ \t) = VTE2(t) - VRE1(t + D/ c). (8.15) The inequality above comes from the Cauchy-Schwarz inequality and

where T, Rare the transmissivity and reflectivity of the beam splitter, D is

Jdwl¢1(w)l2 = 1 = J dw/¢2(w)l2 for single -pho ton states. The equal sign
stands when ¢ 1 (w) = ¢ 2 (w) or complete mode match between the two input
the displacement of the beam sp litter relative to some symmetric position,
photons. Set P2 ( oo) = P20 and T = R. As the beam splitter position D is
and input field operator E 1,2(t) = (1/~) J dwa1,2(w)e-iwt_ scanned through c(T 1 -T 2 ) , P2 / P20 shows an interference dip from 1 down
To find the coinc idence counting probability in detecting two photons
to 1 - V12(0), i.e., the Hong-Ou-Mandel interference dip.
at two output ports, we first calculate the following quantity:
A (o) A (o)
El (t1)E2 (t2)/T1)i/T2)2 8.2.2 Time-bin Entanglement and Franson Interferometer
= [VTE1(ti) + VRE2(t1 - D/c)] In both Pfleegor-Mandel and Hong-Ou-Mandel interference experiment, the
+ D/c)] /T1)i/T2)2
x [VTE2(t2) - VRE1(t2 interference effects are independent of phases of individual fields. But this is
= [TE1(t1)E2(t2) - RE2(t1 - D/c)E1(t2 + D/c)] /T1)i/T 2)2 not a characteristic of two -photon interference. In this section, we will study
= [Tg1(t1 -T1)g2(t2 - T2) a type of two-photon interference experiment in which the measured two -
photon coincidence counts are a function of the phase difference between
-Rg2(t1 - T2 - D/c)g1(t2 -Ti+ D/c)] IO), (8.16) optical fields involv ed in the interference.
where EjEj/T 1)i/T 2)2 = 0 (j = 1, 2). So, the probability of two-photon
coincidence detection at the two outputs is proportional to :-4-L\T~:
I I
I I
0
(Ti/ (T2IE~o)t(t2)Ei )t (t1)Ei 0 ) (ti)E~o) (t2) /T1)i IT2)2
Coincidence
= /Tg1(t1 - T1)g2(t2 - T2) Laser
Measurement
2 Pulse
-Rg2(t1 -T2 - D/c)g1(t2 -T 1 + D/c)/ . (8.17) Franson -
Interferometer
The total two-photon coincidence detection probability is an int egrat ion of
t1, t2 over the whole photon wave packet:
Fig. 8.6 Schematic for producing a time-bin entang led state and it s detection with a
0 0 0 Franson interferometer.
P2 ex / dt1dt2 (T1/(T2IE~o)t(t2)Ei )t ( t1)Ei \ t1)E~ \t2) /T1)i /T2)2
2TR Consider a spontaneous parametric down-conversion process pumped
ex 1 - T 2 + R 2 V12(T1- T2 - D /c) (8.18) by two coherent puls es with a delay of !J..T, as shown in Fig. 8.6. Each
 AppZ.icat'i.ons: Multi-Photon Interf erenc e and Entangl em ent 229
228 Quanill1n Optics For Exp erim entalists

pump pulse creates a pulsed two-pho ton state of the form To calculate the two -photon coincidence detection probability, we first
evaluate the following
I<]_))=/ dw1dw2{]_)(w1,w2)lw1)ilw2)2, (8.21)
Eoutl (t + T )Eoul2(t) j<]_))tb
,vhi ch was obtained from Eq. (6.32) in Section 6.1.5. The time delay 6.T = [Einl(t + 1)Ei,-t2(t)+ Einl(t + 1 )Ein2(t + 6£2/c)
between the pump pulses can be introd uced as a phase factor of eiwpt:,T in
t he pump profile ap(wp) in <]_)(w1, w2) from Eq. (6.29). Then the quantum +Ein1(t + T + 6L1/c)Ein2(t + 6£2/c)
state for the time-bin ent ang led state is +Einl(t + T + 6L i/c)Ein2(t) ] j<]_))tb/2. (8.26)
l<]_))tb + l<]_)(
= [l<]_)(0)) .6..T))]/V2
J
Here the contributions from the unu sed vacuum input s are zero. The above
+ ei(wi+w2)6 Tlw1)ilw2)2.
= dw1dw2{]_)(w1,w2)[l (8.22)

Here, we ass um e 6.T is mu ch larger than the inverse of the spectral width
can be calc ul ated from
') ( ) \~1'-) _
Einl (t Ein2 t 'I! tb -
f dw1 dw2
27r
m.( ) -i(w1t' +w2t) [l + i(w1 +w2)6T]\0)
'I! W1,W2 e

of <]_)
(w1,w2) so that the two pulses are no t overlapping: = [F(t', t) + F(t' - 6T , t - 6T)]I0 ), (8.27)
(<]_)(0)l<]_)(.6..T) 2ei(wi+w2),0,T 0.
) = / dw1dw2l<]_)(w1,w2)1 (8.23) with a temporal two-photon wave fun ct ion

Since the two pulses are coherent to each other , the two-photon state F(t,, t) = -1 m.(w1,w2) e -i(w1t'+w2t)•
/ dw1 dW2'l! (8.28)
in Eq. (8.22) is a time-entangled state simil ar to that in Eq. (4.53) of 21r
Section 4.4.3. But different from the state in Eq. (4.53) , the state her e Since the two-photon spectral function <]_)(w
1, w2) has a wide spectrum du e
in Eq. (8.22) is a pulsed two-photon state. In fact , if the two-photon to the broad band nature of th e down-conversion and the pump pulse , the
spectral function <]_)(w
1,w2) is factorable: <]_)(w1,w2)
= c/>
(w1)'l/J(w2), which two-photon wave function F (t' , t) has a narrow range around t', t = 0 in
corresponds to a transform-limited two-photon state, th e two-photon state the order of the reciprocal bandwidth of <]_)(w1
,w2). Substituting Eq. (8.27)
in Eq. (8.22) can be rewritten as into Eq. (8.26), we have a total of eight terms:
(8.24) Eoutl(t + 1)E oud t)j<]_)
)tb
=
where lc/>t(T))i J dw1eiwiTcp(w1)lw1)i , l'l/Jt(1))2 =
J dw2eiwT'lj}(w2)lw2)2
2 = [F(t + T , t) + F(t + T + 6Li/c, t + 6£ 2/c) + F(t + T + 6Li/c, t)
are the single-temporal mode single-photon states defined in Section 4.3.2. +F(t + T, t + 6L 2 /c) + F(t + T - 6T + 6Li/c, t - 6T)
The express ion above shows the entang lement of the two photons in time. +F(t + T - 6T , t - 6T) + F(t + T - 6T , t - 6T + 6L2/c)
To test the time entang lement, we resort to Franson two -photon interfe-
+F(t + T - 6T + 6Li/c , t - 6T + 6L2/c)]I0). (8.29)
rometer [Franson (1989)], which consists of two unbalanced Mach-Zehnder
interferometers whose outputs are detected in coincidence (Fig. 8.6). The In Franson interferometer, we set the path difference 6L1 , 6L2 ,...___,
c6T and
output field operators for the two interferometers can be expressed in terms the coincidence window is much shorter than 6T so that t , T « 6T but
of the input as longer than the reciprocal bandwidth of <]_)(w 1, w2)- Notice that the domain
of F(t + 7 , t) in which F(t + T, t)-/- 0 is much sma ller than 6T because of
Eout1(t) = [Ein1(t) + Ein1(t + 6.Li/c) + Eino1(t) - Ein01(t + 6.Li/c)]/2 Eq. (8.23). Then only two terms in Eq. (8.29) are non-zero:
Eoudt) = [Ein2(t) + Ein2(t + 6.L2/c) + Eino2(t) - Eino2(t + 6.L2/c)]/2,
(8.25) Eoutl(t + 1)Eout2(t)j<]_))tb
= [F(t + 7 , t) + F(t + T - 6T + 6Li/c , t - 6T + 6L2/c)]I0). (8.30)
where Ein{i ,2}(t) = (1/ J27r) J dw e-iwt a{l ,2} (w) and 6L 1,2 are the path
difference for the two unbal anced MZ int erferometers . EinOl, Eino2 are the These two surviv ing terms correspond to the two photons from the first
vacuum input s in the unus ed input ports of the two int erferometers. pulse going through the long path and from the second pulse but going
230 Q'Uant'Urn Optics For Experimentalists Applicalions: l\f'Ulti-Photon Int erference and Entanglenicnt 231

t hrou gh the short path , respect ively. If the delay between the pump pulses by two photons. However, because of the involvement of tvvo photons,
matches the path delays, i.e., 6.£ 1, 6.£ 2 c6.T, these two possibilities are
('J there usually is no one-p hoton interference which shows in sing le detector
indistinguishable leadin g to two-photon int erference. counts. This point can be easi ly checked for both phase independent Hong-
The two-photon coincidence probability for the two outputs of the in- Ou-1\Iandel int erference and phase-sensitive Franson interference effects. In
terferometers is then both cases, sing le detector counts are constant, exhibiting no interference
P2 J
CX dtdT tb(<I>IE1
,ut2(t)Elutl(t + T)Eoun(t + T)E un(t)j<I>)tb
0
effect.

=J dtdTIIEoutl(t + T)Eout2(t)j<I>)tbll
2

= j dtdTIF(t + T, t) + F(t + T - 6.T + 6.L i/c, t - 6.T + 6.L 2/c)i2

= 1 + 1V1 cos[w10(6.T - 6.Li/ c) + w20(6.T - 6.L 2/c) - Eo] indistinguishab le
(a) (bl
= 1 + IVIcos (01 + 02 - Eo), (8.31)
wh ere we us ed the normalization: J dw1 dw2I<I>
(w1, w2) 12 = 1 and Fig. 8 .7 Schematics for one -ph oton int erference with two corre lated photons. (a) Di-

V = J + D1,w20 + D2)12
dD 1dD2l<I>(w10
sappearance of one-photon interference due to distinguishability between il and i2. (b)
Restoration of on e-photon interference du e to indistingui shab ilit y between il and i2.
Adapted from [Zou et a l. (19916)] .
X eiD1 (6.T - 6.L1 / c)+i0.-2(6.T-6.L2/ c)

(8.32) To further illustrate this, let us consider a simple scheme shown in

So the two-photon coincidence shows an interference fringe as a function of
the phases 0j = Wjo(6.T - 6.Lj/c) (j = 1, 2) of the two MZ interferom e-
ters. The visibility of the interference fringe depends on how well the path
differences 6.£ 1, 6.£ 2 match the delay 6.T between the pump pulses.
Experimentally, time entanglement of two photons was first observed in
Franson interferometer by Ou et al. [Ou et al. (1990a)] and Kwiat et al.
[Kwiat et al. ( 1990)] ind epen dently. The pulsed time -bin enta ngled state
was realized by Brendel et al. [Brendel et al. ( 1999)].
The Franson interference effect shown with a time-bin entangled two-
photon state is a nonlocal effect in the sense that the two photons from
SPDC can be spatially separated and analyzed locally by two separate MZ
interferometers, respectively. This is similar to a polarization entangled
two -photon state and can be used to demonstrate the violation of Bell's
inequality [Franson (1989)]. Because of this, it has applications in quantum
cryptography [Ekert (1991)] .
8.2.3 Distinguishability in Two-Photon Interference
Quantum Erasers
As demonstrated in both Hong -Ou- Mandel and Franson interferometers ,
two-photon int erference stems from indistinguishability in the paths taken
Fig . 8.7(a), where the pump field for SPDC is split into two for pumping
two SPDC processes. We start with the pump field in a coherent state lap) -
After the first 50:50 beam splitter, th e state becomes lap/ vl2)ilap/ vl2)2.
If we superpose the two output fields with another beam sp litter, interfe-
rence effect will occur. But now let the pump fields be down-converted to
signal and idler fields , resp ect ively, by int eract ing them with two nonlin ear
crystals, as shown in Fig . 8.7(a) . For the two SPDC processes together, the
Hamiltonian of the system is
(8.33)
For simplicity , we use single-mode treatment here. From Chapter 6, we find
the state after the SPDC is
(8.34)
and
where rJ is proportional to X, ap and lrJl2 is related to the down-conv ersion
probability. We now superpose two signal fields, as shown in Fig. 8.7(a). It
is straightforward to find the intensity of the superposed field as = (CLs1+
232 Quantum Optics Fo1· Expe1·imentalists Applications: Jo.folti-Photon Tnte1ference and Entanglement 233

Os2)/./2as obtained from Eq. (8.35). Although Eq. (8.35) is derived based on the
single-mode description of the field, it is straightforward to extend it to
(\hliiliisl\J!2) = ((\J!2lii!1iis1I\J!2) + (\J!2lii!2iis2I\J!2) multi-mode case when the joint spectral function of the down-converted
+ (w21a!1as2I\J!2) + (\J!2la!2as1I\J!2)) /2 1.w 2 ) = </>(w
fields <I>(w 1)'1/;(w2)is factorized and the state can be written
as I\J!2) = lvac) + 171
</>)s
lW)i, same as the cases of independent photons in
= 1111 + l11l2 l(ldli1)I COS<p
2
Section 8.2.1 and the time-bin entangled photons in Section 8.2.2. Then
2
= 1111 ) (8.35) Eq. (8.34) is changed to
which shows no interference effect. The last line is because (li 2 llil) = 0. (8.37)
This phenomenon can be understood with photon distinguishability as
well. When the two photons are produced in SPDC, they are highly cor- and Eq. (8.35) becomes
related in many degrees of freedom such as frequency, momentum, and
(W2IE!Esl\J!2)
= ((\J!2IE!1Es1I\J!2)
+ (W2IE!2EdW2)
polarization. These correlations provide which-path inform ation for one
of the two correlated photons from the other, thus leading to no inter-
ference effect due to the complementarity principle of quantum mecha-
+ (Wl2El1EdW2)
+ (W2IE!2Es1I\J!2)) /2
2 2
nics. This can be seen from the disappearance of the interference term
= 1111+ 11111(w2l?);1)I cos<p, (8.38)
(w21a!2as1IW2) = l111 2ei<p(ldli1) = 0 because (li2lli1) is zero due to dis- where Esl ,s2 = (1/./2) J dwe-iwtas1,s 2 (w) and Es= (Esl + Es2)/~. So,
tinguishability between il, i2. the visibility now becomes
Compared with other interference experiments demonstrating the com-
plementarity principle, where measurement is usually performed on the (8.39)
particle participating in the interference experiment to gain the which-path
which depends on the mode match between 'lj;1(w), w2(w), similar to
information, the scheme here does not make such a measurement and the
Eq. (8.20) for the visibility in Hong-Ou-Mandel interference. Thus, the
which-path information is obtained through quantum correlation. But this
mode functions provide a quantitative description of photon indistinguis-
also tells us how to restore the one-photon interference effect: making il, i2
hability . We will demonstrate more of this in Section 8.4.
photons indistinguishable. This can be achieved by injecting il field into the
There is another way to restore the interference effect through a concept
second SPDC process and aligning it with i2 field, as shown in Fig. 8.7(b).
called "quantum eraser" [Scully and Druhl (1982); Scully et al. (1991)],
If the two modes are complete ly overlapping, there is no way to distinguish
which is a technique that erases the which -p ath information by projection
from which SPDC process the idler photon is generated, i.e., (li 2 lli 1) = 1.
measurement. Consider a variation of Fig. 8.7(a) shown in Fig. 8.8(a),
This leads to coherence and one-photon interference between sl and s2.
Indeed, when (ldlil) = 1, Eq. (8.35) becomes
2
(Wlaliisl\J!)2 = 1771(1 + cos<p), (8.36)
which shows interference effect in the signal detector counting rate. In the
less-than-perfect case of O < l(li2lli1)I < 1, Eq. (8.35) gives an interference
fringe visibility V = l(li2llii)I < 1, i.e. , partial indistinguishability leads to
reduced visibili ty or coherence. This interference effect, dubbed as "induced (a) (b)
coherence without induced emission", was first observed by Zou et al. [Zou
et al. (199lb)]. Fig. 8.8 Quantum erasing of which-path information by projection measurement on
detector B to restore interference effect. (a) BSB to superpose il and i2 seems to prevent
The indistinguishability of photons is closely related to the mode of us from distinguishing between il and i2. (b) Extra elements are added for the distinction
the photons. This is illustrated by the visibility relation V = l(li2lli1)I between il and i2 .
 234 Quantum Optics For Exp erim entalists Applicat ions: Jvfulti-Photon Int erf e1·en ce and Entangl em ent 235

where we superpose two idler fields with a 50:50 beam splitter (BSB) - The
experimental fact is that there is no interference in the superposition of sl
s2 even with BSB inserted to superpose the two idler fields, similar to th~
case of no BS shown in Fig. 8.7(a). It is easy to understand why there
is no interference in Fig. 8. 7( a), as we discussed earlier. But it is not so 60000 ~~--r---~-..--~--,,-----r--,

obvious to explain the disappearance of interference in Fig. 8.8(a) in terms

VWmvVv~
45000

of complementarity principle since, after all, it seems impossible to obtain 30000

which-path information at the output il' of BSB. Therefore, it seems that

the scheme in Fig. 8.8(a) is similar to the scheme in Fig. 8.7(b) where I:::: ..
... ...............
-~ 15000
[/)

6000
_..
indistinguishability between il and i2 should lead to interference between
"
s1 and s2, contradicting the experimental observation. As a matter of fact '
';?140000
0
:§ 3000

there is a way to distinguish between il and i2 without removing BSB '.

It is by using the other output i2' of BSB. As shown in Fig. 8.8(b), we
can add an extra BS to recombine the two outputs il', i2' of BSB. When
the phase difference 0 is right, the outputs of the extra BS are exact ly
j
u
i
vi
120000

100000
-1---~~,-~~-~~-,-J

l000 2000
Position of idler mirror (nm)
3000 rr 2rr 3rr
Phase difference I'.~
v ..
0 = 45°

the same as the inputs to BSB: il" = i l, i2 " = i2 due to interfer ence in
the Mach -Zehnder interf erometer formed with BSB and the newly placed
BS. So, even though il and i2 are mixed after the first BS (BSB), they
are still distinguishable after the second BS. The availability of the second
output i2' makes it possible to distinguish il and i2. On the other hand
the technique of quantum eraser is to use detector B at output il' to mak~
a projection measurement and erase the which -path information for the
restoration of the interference between s1 and s2. To see this more clearly,
we write from Eq. (8.34) the two-photon state after the beam splitters as
I)W2 BS= 1[ i 1 .
/l)s1 + e 'P/l)s2] /l)il' + [/l)s1 - ei'P/l)s2] /l)i2,,
where we used /l)i1 • (/l)i 1, + /l)i2' )/~ and /l)i 2 • (/l)i 1, -/ l)i 2, )/~for
the transformation of the states by BSB. So, the projection measurement
at detector B will transform state /w2)Bs to [/l)s 1 + ei'P/l)s 2] /l)i 1, and
lead to interference between s1 and s2 .
The first quantum eraser experiment was performed by Kwiat et al.
[Kwiat et al. (1992)] and later by Herzog et al. [Herzog et al. (1995)]
who used a variation of the induced coherence experiment by Zou et al.
[Zou et al. ( 1991 b)]. The schemes by Herzog et al. are shown in Fig. 8. 9,
where both the signal and idler fields of the first SPDC process are injected
back together and are aligned with the signal and idler fields of the second
SPDC process in a bow-tie configuration [Herzog et al. (1994)]. When di-
rect reflection is made as in Fig. 8.9(a) , both signal and idler fields will
2rr 3rr
Phase difference I'.~
(a) (b)
Fig. 8.9 (a) Schematic and observation data for enhancement and suppression of two-
photon emission in SPDC by two-photon interference. Adapted from [Herzog et al.
(1994)]. (b) Schematic and observation data for quantum erasing of which-path infor-
mation to restore interference effect . Adapted from [Herzog et al. (1995)].
show one-photon interference effect due to indistinguishability in both sig-
nal fields and idler fields. The interference pattern in the signal and idler
outputs are shown in Fig. 8.9( a). Notice that the interference patterns
of the signal and idler fields are completely in phase, indicating the en-
(8.40) hancement and suppression of the two-photon process by constructive and
destructive interference , respectively. In the quantum eraser scheme shown
in Fig. 8.9(b), the polarizations of both the signal and idler field from the
first SPDC process are rotated 90 degree before being injected back into
the second SPDC process. With the notations in Fig. 8.8, now s1 and s2
as well as il and i2 are distinguishable due to orthogonal polarizations. In
this case, no interference is observed at either signal or idler output field as
we project both s1 and s2 (or il and i2) into 45 degree with a polarization
beam splitter, which is equivalent to BSA (or BSB) in Fig. 8.8(b). The
projection measurement onto i; is equivalent to the detection of a photon
at detector B. So, the detection of the superimposed signal fields, when ga-
ted upon the detection of a photon at detector B, will show an interference
pattern , as observed by Herzog et al. and shown in coincidence count of 45
degree case in Fig. 8.9(b) [Herzog et al. (1995)].
 Applications: l\f'Ulti-Photon Jnt e1jerence and Entanglement 237
236 Q'Uant'Um Optics For Experimentalists

8.2.4 Cavity Enhancement of SP DC by Constructive with

1'1 - ')'2 + 2i0 4(1'1 (8.42)
Multi-pass Two-Photon Interference cD(O)= . , DD(O)= (')' l + '( 2 ')"")2'
')' l + ')'2 - 2'l 0 - ~lH

When the phase is right for constructive two-photon interference , the feed-
2y'1D2 4( fiD2 (8.43)
back scheme shown in Fig. 8.9(a) by Herzog et al. [Herzog et al. (1994)]
G L(O) = 1'1 + 1'2 - 2i0' gL(O) = (1'1 + 1'2 - 2if2)2
leads to an enhancement of two -ph oton conversion rate. Under this condi-
t ion the scheme is equivalent to phase-matched double-pass of the crysta l Here we set ,B-- rv, 1, ,-.,, -- rv, 2 from Eq · (6 ·135) as the coupling constants (a.k.a.
by the pump field and therefore doubles the length of the crysta l. We can decay constants) for ain and Cin, respectively. Cin represe~ts the ~mwanted
extend this idea to multiple passes with a cav ity, as shown in Fig. 8.10, uum mode coupled -in due to losses in the system. ( 1s the smgle-pass
vac 1· d d
which effectively lengthens the crystal of length l by roughly F t imes to parametric amplitude gain and is proportional to the pump amp 1tu e an
2
Fl due to phase-matched multiple-pass (Fi s the finesse of the cav ity), and th e nonlinear coefficient. In Eqs. (8.42) and (8.43) we dropped the 1(1 -
leads to strong en hancement of two-photon production rate. This is similar term in the denominator because the OPO is operated far below thres hold
to the cavity enhancement effect in second harinonic generation obse rved so that \(I« 1'1, 1'2• .
by Wu and Kimble [Wu and Kimble ( 1985); Ou and Kimble ( 1993)]. The Now let us look at the enhancement effect in down-convers1011 due to
cav it y enhan ceme nt effect of spontaneous p aramet ri c down-conversion was resonance. For vacuum inputs, we calcu late from Eq . (8.41) the spectrum
first observed by Ou an d Lu [Ou and Lu (1999); Lu and Ou (2000)] and S (w) of t he field defined by
ha s since be en a robust source of narrow-band two-photon states [Fortsch (a!ut(wo + O)aout(wo + O')) = (A!ut(O)Aout(O ') = S(O)c5(0 - O'). (8.44)
et al. (2013)].
The result is
2
; l ; 2 2 - 16\(1 1'1(')'1 + ')'2) (8.45)
I
I
I
I
S(w) = l9D(O)I + lgL(O)I - [(1'1 + ')'2)2 + 402]2 ·

The rate of down-conversion can be calc ul ated as

p
PDC
Rcavity -- (EC-)(
out t )E(+)(t))
out = 2-rr J dOS(O) ' (8.46)

J where

Fig. 8.10 Enhancement of th e two-photon emission rate of SPDC w ith a cavity by

j};C+\t) = [ft (-)(t W = _1_ Jdf»ii(w) e-iwt
J2n
I
= e-:;;;/dr!A(n)e-'"'.
(8.47)
length ening the crystal effect ive length by F times.

From Eq. (8.45) , we have

To understand the enhancement effect quantitatively, we consider the
theoretical modeling of parametric process in a cavity which is covered 1 Joo 2
161(1 1'1(')'1 + 1'2)
R cavity = 21f -oo dO [(')'l + ')'2)2 + 402]2
in Section 6.3.4. But here we consider the case of far below threshold
where there is only spontaneous process for two-photon generation. From = lr\ F 2 / -rrD.tFo
2
, (8.48)
Eq. (6.135) of Section 6.3 .4 with a0 ut(wo+O) = Aout(O) and ( 2 « (,8+ ')') 2/4 where r (D.t is th e single-pass gain param eter with D.t as the round-
for the far-below threshold case, we find the output operator of a degenerate tr ip time, and F = 2-rr/(1' 1 + ')'2)D.t = 2-rr
/ D.tb.W0 po is th~ ~n esse of t~e
OPO on resonance (D.a = 0, wo is the degenerate frequency of the OPO) is cav ity (which is of the order of the numb er of bounces of light b efore it
related to the input as follows: leaves the cavity) and can be measur ed directly. Here D.wopo= 1'1_+ ')'2
aout(Wo+ 0) = GD(O)ain(wo + 0) + gD(O)aln(wo - 0) corresponds to th e bandwidth of the OPO cavity . Fo = 2-rrh1D.t is the
same quantity without the loss (1'2 = 0). To find the enhancement factor ,
+GL(O)c\n (wo + 0) + 9L(O)cJn(wo - 0) (8.41)
 238 Q'Uantwn Optics For Experimentalists App lieo,t'ions : M'Ulti-Photon Int erference and Entanglement 239

we need the signal rate without the cav ity. In the single pass case, we
simpl y have gD (w) = n7(w) and gL = 0. Her e r;(w) is the gain spect rum
of single-pass spontaneo us down-conversion d term in ed by phase- matching
condit ion with normalization 17(0) = 1. In the exper im ent, we usually
have an interferenc e filt er (IF) in front of the detector. Th e frequency
bandwidth 6.v1 p = 6.w1p/21r of the IF is norm ally smaller than that of
down-conversion so t hat 17(w) 1 for w within 6.w 1p an d is zero for w
T
outside 6-w 1F . Hence, the signal rate withou t t he cav ity is

Rsingl e_pass = lrl2 6-VJ F = lrl2 6-WJ F /2n, (8.49) Fig. 8. 11 Two-photon time corr elation fun ct ion for a mod e-locked two-photon state.

and t he average enhancement factor per mode is Th e state in Eq. (8.51) is the so-ca lled mode-locked (ML) two-photon
B = = F 3/ nFo,
R resonance/ 6-W opo
(8.50)
state first discovered by Lu et al. [Lu et al. (2003)]. Different modes
6-wIF
Rsingl e_pass / of photon pairs are in superpos ition. All the pairs have a common ori-
or roughly the square of the numb er of bounc es of light before it leaves the gin (phase) from the pump field , which provid es the mechanism for phase
cavity. This is consistent with the phas e-match ed multi-pass parametric (mode) locking. It has a comb-like two-photon spectrum, whose Fourier
down-conversion where the conversion rate is proportional to th e square transformation is the two -ph oton time correlation fun ction :
of the crysta l length and the effective crystal length is incr ease d by the f ~L(T) = (EJ(-)(t)ft( -\ t + T)EJ(+)(t + T)EJ(+\t))
number of passes or F. Her e the phas e matching is satisfied by the on
resonance condition. The square law is a result of two-photon constructive = lg(T)F(T)i2 (8.52)
interference. The loss of the system will reduce the effect by a factor of with
F / Fo. = / dn~J,(n) -iDT F( ) = sin[(2N + 1)6.D,T/2]
g (T ) - H, tp H e ' T - sin(6.D,T /2) ·
8.2.5 Mode-locked Two-Photon States As shown in Fig. 8.11, the two-photon time corre lation function has a
shape of regular spikes separated by the round-trip time tr = 2n / 6-0 psR
In the discussion of the cavity enhancement effect in the previous section,
of the cavity. The physical mea ning of Eq. (8.52) is very clear: after a
we only cons idered the single degenerate mode of the cavity (w1 = w 2 =
photon is detected , we need to wait an int eger multiple of the round-trip
wo). For a type-I degenerate parametric down -conversion, the two down -
time tr for the second photon to come out of the cavity and be det ecte d .
converted fields have the same frequency and polarization and thus have a
This behavior is similar to the pulse s coming out of a mode-locked laser .
wide spectral bandwidth [Boyd (2003)]. So, in add ition to the mode with
The two -photon time correlation function shown in Fig . 8.11 is a result of
W1 = w2 = wo, other lon gitudina l modes with w 1 = w 0 + N 6-D,FSR, w 2 =
frequency-entanglement of the two photons generated . Th e spike-like two-
wo - N 6-D,FSR are also on resonance, with 6-0FsR as the free spectra l
photon time correlation function of the mode-locked two-photon state was
range of the cavity and N determ in ed by the dispersion of the nonlinear
observed by Goto et al. [Goto et al. (2003)] and indir ect ly confirmed with
medium. This leads to a two -photon state of many frequency pairs:
a Hong-Ou-Mandel int erferometer where a revival of the Hong-Ou-Mand el
l'l')ML =
m=-N
N

I dO 'lj;(O)at(wo + m6.D,FSR + D,) dip is observed [Lu et al. (2003); Xie et al. (2015)].

8.3 Multi-Photon Interference Effects

x at(wo - m6.0FsR - D,)Jvac), (8.51)
where 't/J(D,)is the spectral function from one cavity mode and is related to Now we can extend two -photon interferenc e effects to multi-photon inter -
GD (D,) in Eq. (8.42). ference effects where mor e than two photons are involved in multi-photon
240 Quantum Optics For Experimentalists Applications: Jvlult'i-Photon Int er-fe'1·ence and En tanglement 241

coincidence detection. The expe rim enta l technique for multi-photon coinci- 25

dence is simi lar to two-photon coincidence. For simpl e coincidence measu- -;:; 20
LPJml E-
rement, we can use an "AND " gate with more than two inpu ts . So, there is 413.lno 15

a coincidence pulse out for countin g only when all the inpu ts have a pulse '"
10
g
in. For time-delayed multi-photon coin cidence measurement, the tec hnique ·5
() 5
Width 21.2µm

is more comp licate d since it involv es multiple time delays. In this sect ion, Plh/1
0
Al•A2°0.8.,,\50,nfmt1/st,II 0 50 100 150
we will concentrate on the phenomena with only simpl e multi-photon coi- 6x(µm}

ncidence measurement.
(a) (b)

8.3.1 Multi-Photon Bunching Effects
We have discussed photon bunching effect before in both the classical wave
and quantum optical language s. We showed that these two are actua lly
equival ent. However, there is a deep er physical principle in photon bun-
ching effect buried under the mathematical calculation. The principle is
quantum interferenc e. In fact, Fano was the first to associate an interfe-
rence effect of two-photon amplitudes to the photon bunching effect [Fano
(1961)]. Glauber later used a similar argument to explain the photon bun-
ching effect [Glauber (1964)]. In the following, we will show this connection
clearly with Hong-Ou-Mandel interferometer and extend it to the case of
more than two photons.
8.3.1.1 Photon Bunching Effects and Two-Photon Constructive
Interj erence
Consider a different detection scheme for the Hong-Ou-Mandel interfero -
meter: instead of detecting the I1, 1) state with two detectors placed at two
different output ports (Fig. 8.5(a)), we place them at the same output port
(Fig. 8.12(a)). This will measure the probability P 2(2, 0) for 12,0) state
instead of the probability of P2(l, 1) in Hong-Ou-Mandel effect.
From Eq. (8.8) , we have
(8.53)
Fig. 8 .12 Ph oton bunching effect in Hong-Ou-Mandel int erferometer: (a) outline of
setup; (b) the resu lt of the ex p er im ent. Reproduced from [Rar it y and Tapster ( 1989 )].
probability is twice the classical probability. This is exactly the same as the
photon bunching effect. Its demonstration was first performed by Rarity
and Tapster , as shown in Fig . 8.12 [Rarity and Tapster (1989)]. To see that
this is the result of two-photon interferenc e, we consider Fig. 8.13, which is
the setup to measure P 2 (2, 0). The two-photon detection measurement for
P 2(2, 0) is accomplished by a beam-splitter (BSd) splitting scheme shown in
Figs. 8.13(a) and (b). This is similar to HBT experiment (Sections 1.2, 1.6).
As seen in Figs. 8.13(a) and (b), there are two possible ways to arrange the
detection of the two photons from one output port of the beam splitter (BS).
If the incoming two photons are well-separated , they behave like classical
particles and we add the probabilities of the two possibilities: P 21 = IAl2 +
IAl2 where we take A as the amplitude for each of the cases. But if they
overlap at the beam splitter, we cannot distinguish the two possibilities,
and we add the amplitudes before taking the absolute value square for the
overall probability: P;r
= IA+ Al 2 = 4IAl 2 = 2P 21. Note that the phases
for the two cases are the same because the overall paths for the two photons
in the two possibilities are the same due to indistinguishability of the two
photons. Therefore, it is constructive interferenc e that is responsible for
the photon bunching effect in the Hong-Ou-Mandel interferometer. This

r
Here, "qu" denotes the prediction from quantum theory. On the other
hand, we may find the classical probability of detecting both photons at
the same side of the beam splitter, i.e., case (a) or (b) in Fig. 8.4. The result
is simply the product of the sing le-photon events due to indep endence:
P{ 1(2, 0) = P{ 1(0, 2) = A (1, 0)P{ (1, 0) = t 2r 2, (8.54)
where A (1, 0), P{ (1, 0) are the probabilities for one input photon from two
BS t
0
T (a) (b)

Fig. 8.13 Two possibilities for two -photon constructive int erference in explaining pho-
sides, respectively. From Eqs. (8.53) and (8.54), we find that the quantum ton bunching effect in Hong-Ou-Mandel interferometer. BSd is for two-photon detection.
2.J2 Quantum Optics For E xperim entalists
was consistent with the original view by Glauber, who first exp lain ed the
photon bunching effect with an equiva lent view of two-photon amp litudes
[Glauber (1964)].
8.3.1.2 Multi-Photon Constructive Interference and Stimulated
Emission
As a genera lization of the two -phot on bunching effect in Hong-Ou-Mande l
interferometer, we cons ider the case of N + 1, i.e., one photon input to
one side of the beam splitter and N photon input at the other side, as
shown in Fig. 8.14. From the output state derived in Eq. (6 .60) , we find
the probability for all N + 1 photons to exit at one output port as
PJ/+1 (N + 1, 0) = (N + 1)/2N+1, (8.55)
which is N + 1 times of the classical probability PJj-tl(N + 1, 0) =
(1/2) x (1/2N) = l/2N-t1 if the input single photon (white circl e) is distin-
guishable from other input N identical photons (black solid circles). This
factor can be understood with the (N + 1)-photon detection scheme shown
in Fig. 8.14, where there are N + 1 different possibilities to arrange the
input single photon (white circle). PJJ+i(N + 1, 0) = (N + l)IAl 2 corre-
sponds to completely distinguishable situation among the possibilities while
Pl,~ 1 (N+l,0) = l(N+l)Al 2 = (N+l)PJf+ 1 (N+l,0) is due to construct ive
quantum interference by adding the amplitudes .
N is
Fig. 8.14 (N-tl) -photon bunching effect in Hong-Ou-Mandel interferometer.
The multi-photon bunching effect discussed above can be used to ex-
plain the phenomena of stimulated emission of atom by incident photons.
As is well-known, st imul ated emission was first proposed by Einstein to
exp lain the spectrum of blackbody radiation [Einstein (1917)]. It is the
foundation for optical amp lification and is thus responsible for lasers . Phe-
nomenologically, when a single photon interacts with an excited atom, it
can stimulate the atom to emit. The atom can, of course, emit a photon
spontaneously. From Einstein's A- and B-coefficients, the rates of the sti -
mulated em ission and sponta neous em ission are the same, and are denoted
Application s: J\,Julti-Phot on Int erf er en cf' and Entaugl f'm ent 243
as R. The overall rate is then 2R. When there are N input photons, each
hoton may stimulate the atom, and the overall rate is then (N + l)R. It
peeins that the description above has nothing to do with the multi-photon
interference effect we discussed in the beginning of this section.
(a) (b)
Fig. 8.15 Photo-emission from two atoms: ( a) spontaneous emission of two independent
atoms and (b) stimulated emission of one atom by the emission of another atom.
To make a connection, let us consider two excited atoms, each emitting
a photon , as shown in Fig. 8.15. According to the description above, there
are two different types in the photo- em ission process. In Fig. 8.15(a), the
atoms in the excited state independently emit photons due to spontaneous
emission, and two -photon detection probability in this case is simply the
product of individual emission probability: p;P =Pl= R. In Fig. 8.15(b),
the detected two photons are from stimulated emission, i.e., the photon
spontaneously emitted from one atom stimulates the emission of another
atom. Since the rate of stimulated emission is the same as that of spontane -
ous emission, we have pr
= R = Pl = P?. Hence , the overall probability
(8.56)
which is exactly the rat io of the photon bunching effect.
Since photon bunching effect is due to a two -ph oton constructive in-
terference effect and the stimu lated emission gives the same result as the
photon bunching effect, we therefore have reason to speculate that stimu-
lated emission is also a result of the two -photon interference effect. In fact ,
we look at the two schemes in Fig. 8.16 where the (N + 1)- photon int erfe-
rence scheme with a beam splitter (Fig. 8.16(a)) is copied from Fig. 8.14.
From Eq. (8.55), we find that the (N + 1)-photon detection probability is
PN+l = (N +l)/2N+l, which is N +l times the probability PJj-tl= 1/2N-tl
when the N + 1 photons were classical particles. The enhancement factor is
 244 Q'Uant'Um Optics For Experimentalists Applications : .!lf'Ulti-Photon Int e1jc1·e11ce and Enta11glcm c·11t 2--15

(a)
- / -~}• N
t
0
•
0

:,,

~ ~ ···~ -D
the laser lincwidth is then
i:),,vlaser = l:),,vc ---
N+l
l
+0 X ---
N+l
N l:),,vc
--
N
. (8.57)

Suppose the laser has an output coupl er of transmissivity T. Then the

output power of the laser is Pout= TNhv/tr. Here, tr= 2L/c = l/ l:),,vFsR
(b) is the round trip time for the intra-cavit y photons with 2L as the round
trip lengt h of a standing wave cavity an d l:),,vFsR as the free spectral range

---- N
Fig. 8 .16 Comparison between ( a) multi-photon interf erenc e and (b) stimulated
sion by N photons. R eproduc ed from [Sun et al. (2007)].
N + l. For the stimulated emission scheme in Fig. 8.16(b), we can likewise
make the same argument as the one for (N + 1)-photon construct ive inter-
ference in Fig. 8.14 since the multi-photon detection scheme cannot make
the distinction between the two scenarios in Fig. 8.16. This also leads to
an enhancement factor of N + l as compared to the spontaneous emission
which is equivalent to the case when the atom-emitted photon is complete l;
distinguishable from the N incident photons, i.e ., the N incident photons
do not interact with the atom at all and are totally independent of its
em ission.
Although Einstein used energy balance to introduce the stimulated emis-
sion but did not present any detail, the quantum theory of light develo-
ped later fully explained it. In essence, it is from the Bosonic relation of
atIN) = + llN + 1), which is consistent with the N + l enhancement
factor given above. But this explanation relies on some complicated opera -
tor algebra. Through our discussion above, we find that there is a simpler
physical principle underlying the phenomenon of stimulated emission that
'
is, stimulated emission is a result of multi-photon constructive interference
in exactly the same way as the photon bunching effect. Experimentally, this
connection between the stimulated emission and multi-photon interference
was demonstrated by Sun et al. [Sun et al. (2007)].
The ratio of stimulated emission to spontaneous emission discussed
above can be used to derive the Schawlow-Townes linewidth of lasers [Scha-
wlow and Townes (1958)]. Suppose that there are N photons in the laser
cavity. Then the next photon emitted by the atoms will have a chance of
N / ( N + l) to be identical to the N photons and 1/ (N + l) chance to be
spontaneously emitted photon. The spontaneously emitted photon has a
linewidth of the laser cavity l:),,vc whi le the stimulated photons have the
same frequency as the laser. So, upon averaging over all N + l photons ,
of the laser cavity. Taking linewidth of the cavity 1:),,1/c = l:),,vFsR/ F and
the finesse F = 2n /T, we change Eq. (8.57) to
2
em is- (l:),,vc)
l:),,vlaser = 2nhv---, (8.58)
Pout
which is the Schawlow-Townes linewidth of a laser. From this simple deri-
vation, we find that the Schawlow-Townes linewidth originates from sponta-
neous emission and is thus of the quantum nature. Therefor e, this linewidth
is a fundamental quantum limit for the linewidth of a laser. The linewidth
of an actual laser is far above this limit due to classical technical issues such
as cavity mechanical vibrations.
8.3.l.3 More Photon Bunching Effects
Now let us generalize the case of N-photon + 1-photon to the case of N-
photon + 1\11-photon with M 2. We start with the 2 + 2 case, i.e., two
photons enter a 50:50 beam sp litt er from one side while other two photons
enter from the other side. This is similar to the Hong-Ou-Mandel effect
for two photons but here, it is for two pairs of photons and we make the
observation at only one output port. Using the technique in Section 6.2.2,
we find the output state is
IW
4) = JI (14,0) + 10,4)) -112, 2) (8.59)
Hence, the probability to have all four photons exit in one output port is
Pr( 4, 0) = 3 / 8. On the other hand, if we treat the photons as classical
particles, the probability would simply be P41( 4, 0) = (1/2) 4 = 1/16. So,
there is an enhancement factor of 6 for the quantum probability.
We can understand this enhancement factor again by counting the pos-
sib le ways to arrange the two pairs of photons with four detectors: there
are Cl = 6 ways, as shown in Fig. 8.17. The phases for all six possibilities
are the same due to the same path for the four photons. Then a complete
construct ive 4-photon interference leads to a factor of 6 enhancement, simi-
lar to the argument presented in two-photon bunching effect. This photon
246 Quantum Optics For Experim entalists Applications: Afulti-Photon hzt c1jere11ce and Entangl em ent 247

e .e )
O 0 0 e O e e O O e the cance lation or a decrease in multi-photon coincidence coun t. In fact .
Hong-Ou-l\Iandel effect is the first such effect for two photons. However.
0 0 e e e O e 0 0 e e 0
- r---7 \ we cannot genera lize it to the case of more photons with a symmetric beam
splitter, as shown in Eq . (8.17) where the 12. 2) state does not disappear
Fig. 8 .17 Six possibilities for detecting two pairs of photons in four detectors. as we would want. On the other hand, we may cons ider its genera lization
with an asymmetric beam splitter with T -=I=R.
pair bunching effect was first demonstrated by Ou et al. with two pairs of
photon from param et ric down-conversion process [Ou et al. (1999a)]. The 8.3.2 .1 Three-Photon Hong-O'U,-Mandel Interferometer
expe rim ent al schematic and the result are shown in Figs . 8.18(a) an d (b),
The genera lizat ion of the Hong-Ou-Mandel interferometer to three-photon
where an ap proximatel y 5-fold enh anceme nt was observed, a littl e smaller
case was achieved and demonstrated by Sanaka et al. in realizing a non-
t han the predicted 6-fold en h anceme nt (see Section 8 .4 for the reason).
linear phase gate [Sanaka et al. (2004)], who cons idered an input state
of 12,1) to a beam splitter with T -=I=R. With the method discussed in
Section 6.2.2, we m ay easi ly express the output state as
+
D3 r
2
50
__ /l_

l fl _ ,
1'113) = Rl3, 0)
+VT(T - 2R)l2, 1)
R 2 IO,
3)
+ VR(R - 2T)ll, 2). (8.60)

D2
il
0 Note that P3(2 , 1) = T(T - 2R) and is equal to zero when T
2
= 2R = 2/3 .
Dl
0 '
Photon Arrival Time Delay (ps)
Under this condition, Eq. (8.60) becomes
(a) (b) 2 v'210,3) - v1311, 2).
1'113)=
3 13, 0) + 3
(8.61)
Fig. 8.18 (a) Schematic for d emonstrating photon pair bunching effect . (b) The result
3
of photon pair bunching. Ad apte d from [Ou et al. (1999a)]. Note that the disapp eara nce of the 12,1) is due to destructive three-
photon int erference. This comp lete canc elat ion of probability am plitud e
For the general case of N-photon + M-photon, it is straightforward to for the output of 12,1) is similar to the two-photon Hong-Ou-Mand el ef-
show from the output state in Eq. (6.56) that the enhancement factor is fect and can be easily und erst ood with the picture in Fig . 8.19, where
CfS+M = (N + M)!/N!M!. A case of 3 x 3 was demonstrated by Niu et al. th ere are thr ee possible ways to obtain an output of 12,1): (a) all thr ee
with the observed enhancement factor of 17, a little short of the theoretical photons are transmitted through the beam splitter with a probability am-
prediction of 20 [Niu et al. (2009)]. The less-th an -p erfect results in both the plitude of fi/3 x fi/3 x fi/3; (b) and (c) one of the two photons
two-pair bunching case in Fig. 8.18 and the three-pair bunching case here from one side is transmitted while the oth er one of the two photons and
are due to photon distinguishability effect that we will discuss in Section 8.4. th e singl e photon from the other side are reflected with each case ha-
This idea can be applied to a weak coherent state to take out two-photon ving a probability amplitude of x-fil3 x /173 /173.
The negative
state for achieving photon anti-bunching. sign is due to an overall 1r phase shift on the two reflected photons, sim i-
lar to that in the Hong-Ou-Mandel effect. Thus the overall amplitude is
8.3.2 Generalized Hong-Ou-Mandel Effects and
( fi/3) 3 - 2 x ( fi/3)( /173)
2
= 0 1 leading to the disappearance of the
12,1) state. Notice that when R = 2T = 2/3, the I1, 2) state will disappear
Destructive Multi-Photon Interference
in the same way.
The multi-photon bunching effects discussed in the previous sect ion are all The generalization of the asymmetr ic beam sp litt er scheme above to
the results of constructive interferenc e for enhancement. We will consider four photons, with two on each side of the BS, is straightforward and will
multi-photon destructive interference effect in this section. The result is be dealt with in Problem 8.1.
2.J8 Quantum Optics Fa,,· Exp erim entalists Applications : l\fult-i-Photon Int e1Jeren cc and Entanglem ent 249

for multi-photon inte rfer ence , distinguishability among different particles

111a.y a.lso lea.d to degra dation of int erference, as we have seen in Hon g-
Ou-J\land el int erfer ence dip. In t his sect ion, we will use the multi-photon
interfere nce effects discussed in t he previous sect ion s to quantitatively cha-
racte rize t he degree of temporal distinguishability of photons.

(a) (b) (c) 8.4.1 N-Photon State from N Single-Photon States and
Photon Indistinguishability
F ig. 8 .19 Three possibilities for the outpu t of 12, 1): (a) all phot ons are transmitted·
(b) and (c) one of the two photons from one sid e is transmitted while the other one and We start with the simplest case of two photons and exam ine how tempo-
t h e sin gle photon from the othe r side are reflected. ral distinguishability of photons ari ses . Consider the output state of the
Hong-Ou-M an del int erferometer for the input of two photons in the single -
8.3.2.2 Number State Filtering Effect
temporal mode states in Eq. (8. 14). The input state of the system is
The dest ructiv e three -ph oton int erferenc e effect in the pr evious section can
l'llin) = IT1)i @ IT2)2
b e easily generalized to an arbitrary input state of IN,1). With a beam
splitter of T , R , we can eas ily find th e probability amp litud e for an output
of IN,1) (Problem 8.4):
= J dw1dw2¢1(w1)eiwiT
1
¢2(w2)eiw2 T2at(w1)al(w2)IO). (8.64)

With the method in Section 6.2.2, we find the state of the system after th e
AN+1(N , 1) = NR), (8.62) 50:50 beam sp lit ter as
which leads to the probability PN+i (N , 1) = rN- 1(T- N R)2. PN+i (N, 1)
equa ls to zero when T = N R = N / (N + l). This effect can be used as a l'llout) =J d0J1dw2~¢ 1(w1)eiwiT1 ¢2(w2)eiw2T2
number state filter when conditioned on the single-photon output at port
X [at (w1) - al(w1)] [at (w2) + al(w2)] IO)
=J
2. This idea was first proposed and demonstrated by Sanaka et al. [Sanaka
et al. (2006)]. Consider an arb itrary state l?/)inh = I:n cnln)i input at port dw1dw2~¢1(w1)eiw1T1¢
2(w2)eiw2T2
1 and a single-photon state at port 2. The singl e photon at port 2 is often
called the anc illa photon. We find from Eq. (8.62) that when cond itioned X [at (w1)at (w2) - al(w1)al(w2)
on the detection of a sing le photon at output port 2, the output state at +at(w1)al(w2) - at(w2)al(wi)] I0)
port 1 is then projected to
= [l<I>2h- l<I>2)2
+ l<I>1,1h2
], (8.65)
ut)1 =
l?/)o NL Cn - nR)ln)1 , (8.63)
n where
where N is a normalization factor. When we choose T, R so that T / R =
no = integer , state lno) disappears from the projected output state l?/)outh-
l<I>2)j= J (j = 1, 2)
dw1dw2<I>2(w1,w2)a}(w1)a}(w2)IO)
Thus, the number state lno) is filtered out at the output.
,1h2 = / dw1dw2 [<I>2(w1
l<I>1 , w2) - <I>2(w2
, wi)] at(w1)al(w2)I0)
8.4 Quantum Interference and Photon lndistinguishability ,w2) = ¢1(wi)eiwiT1 ¢2(w2)eiw2T2.
with <I>2(w1 (8.66)

The complementarity principle of quantum mechanics states that distin - Obviously, l<I>1,1h2 gives rise to Hong-Ou-Mandel int erference effect but our
guishability in evitably degrades the int erference effect. For traditiona l int erest now is on I<I>
2)i, I<I>2)2, which correspond to the cases when both
single-photon int erference, distinguishability may only occur in paths. But incoming photons exit from the same side of the beam sp litt er. Apparently
 y

Applications: Multi-I'hoto11 Inte1J ernn cC'and Entangl em ent 251

250 Quantum Optics For E:i:perimental-ists

they are in a multi-frequency two-photon state. How do they look like in at the beam splitter with well separated time T1, T2. This leads to a two-
time domain? hoton state with two photons well separated as shown in Fig. 8.20(a). The
To find out, we evaluate the single-photon detection rate: ~wo photons in this state are completely distinguishable. ..With notations
in Section 4.3.2, we can write the two-photon state in this case as l<I?2 )=
2)\0) = \l)r 1\l)r 2. The second case corresponds to the situation
(8.67) ;lt(T 1).A_t(T
For this, it is easy to first calculate when the two photons arrive at the beam splitter at the sam~ time T1 =
y = T and the two-photon state in this case becomes \<I?2)= [At (T) ]2\0) =
E,(t)l<l>2)i= J dwe-jwta,(w) J dw1dw2<l>2(w,,w2)iil(w,)iil(w2)IO)
2
/2\2)r. This case is depicted in Fig. 8.20(b), and the two photons in this
state are in one single temporal mode and are completely indistinguishable.
= j dw1dw2<I?2(w1,w2)[e-jwita,!(w2)+e - jw2ta!(wi)]I O). The coefficient /2 in front of \2)r gives rise to the photon bunching effect
(see Section 8.3.1).
Then we have
At A R = 1/2 R = 1/3 R = 1/N

tt
R1(t) = (<I?2IE1
(t)E1(t)l<I?2)1
j dw~dw;dw1dw2<I?;(w~,w;)<I?2(w1, w2)
")I ")I
•
nN
)I

tN
• •
= ~ Tl
2
X [ ej(w~-w1)tb(w2 - w;) + ej(w;-w2)tb(w1 - w~)
• • • •
+ej(w~-w2)tc5(w1- w;) + ej(w;-wi)tc5(w2- w~)]. (8.68) Fig. 8.21 Generation of an N-photon state by superposing N singl e-photon states with
beam sp litters.
We next substitute the explicit form of <I?
2(w1, w2) given in Eq. (8.66) into
the above and arrive at The argument above can be generalized to an arbitrary number of pho-
2 tons. With N single-photon states, we can combine them with beam split -
R1 (t) = 191(t - T1)12+ l92(t - T2)1
ters to form an N-photon state, as shown in Fig . 8.21. This scheme has it s
+c129;(t - T2)91(t -Ti)+
ci 29;(t - T1)92(t -T2) , (8.69) practicality since multi-photon states with N > 2 are not easily available
where we used the normalization J dwl¢j(w)l 2 1 and c12 in lab 3 but single photons on demand have been widely achieved so far.
J dw¢j_(w)¢2(w)ejw(T2-Ti), and 9j(t-Tj) with j = 1, 2 is the single-photon With the method in Section 6.3.2, it is straightforward to show that the
wavepacket function given in Eq. (4.45) of Section 4.3.2. output state of the port shown in Fig. 8.21 has the form of

(a)
\<I?N)out= N;,,
12
.At(T1)At(T2)...At(TN)\0) (8.70)

with
A_t(T) = J dw¢*(w)e-jwTa)(w). (8.71)
As we have shown for the two -ph oton case, the arrival times
T , T2 , ... , TN are critical for the indistinguishability of the N -ph oton state .
1
Fig. 8.20 Two different two-photon states: (a) two photons are well separated in time
When T 1 = T2 = ... = TN = T, we have the N photons completely indis-
a nd distinguishable; (b) two photons are in one temporal mode and are indistinguishable.
tinguishable in one temporal mode:
For the sake of simplicity but without loss of generality, we set 9 1 ( t) = (N![A_t(T)]N {N!
\<I?N)out= Vyjii vfJT. \0) = Vyjii\N)r. (8.72)
92(t) = 9(t). There are two extreme cases: (1) \T1 - T2\ » ~T and (2)
IT1 - T2I « ~T with ~T as the width of the wavepacket function 9(t). 3M ulti-photon states can be produced from two -photon states generated in sp~ntaneous
The first case corresponds to the situation when the two photons arrive parametric processes but with limit ed qua lity of indistinguishability. See Section 8 .4 .2.
252 Quantuni Oph .cs For Experim ental ists Application s : Alnlt i-Photon lnt e1j'en '11ce and Entangl em ent 253

The square of t he coefficient in front of IN)r gives t he probability for the (a)

N-p hoton state generat ion:

(8.73)

On t he other hand, when ITi -Tj I » 6.T , some of the photons in the N-
Fig. 8.22 Two different four-photon states: (a) tvYo pairs are separated and distinguis-
photon state become dist inguish ab le in t ime. If a p arti al gro up of photons
hable; (b) all four p hotons are in one temporal mode and are indistinguishable.
have Ti = Tj among the N photons, we have only partial indistinguis-
hability. According to quantum compl eme ntarit y principle, t his will lead four-p hoton interference scheme, the interference visib ility will be different
to reduced int erfer ence effect in multi -ph oton int erfer ence effect as compa- for the two cases in Fig. 8.22. We cons ider next the photon pair bunching
red to the case in Eq. (8.72) with comp lete indistin guishabilit y. We will effect first discussed at the end of Section 8.3 .1.
invest igate next how partial indi st inguishabilit y affects the multi-photon When the two pairs are well-separated in time, as depicted in
interference effects. Fig. 8.22(a), there is st ill some but less bunching effect . In fact, when
t he two pair s are distinguishable, the input state becomes
8.4.2 Pair Distinguishability and its Characterization
~~))' = I1i : b) ® I1 ; 1; ) ,
I<1> (8. 74)
We have already seen the effect of partial indistinguishability on two-photon = 121, 22). The output state is then
instead of l<I>~~))
interference in Hong-Ou-Mandel effect in Section 8.2 .1. The next examp le is
the case of two pairs of photons in the photon pair bunching effect discussed l<I>i~t)'= (1/2) (\21, 02) - \01, 22)) ® (12~,o;) - 10~, 2;)), (8.75)
at the end of Section 8.3.1. This is one of the first multi-photon interfer ence from which we find the probability P4 (4, 0) = 1/4 and the ratio to the
phenomena with N > 2 that are affected by partial indistinguishability classical probability is then
and reveals the cha llenge encountered in generating higher photon number P~(4, 0)/ Pt 1(4, 0) = 4. (8.76)
states (N > 2) from spontaneous parametric processes for the app lications
This value is reduced from the maximum value of 6 given in Section 8.3.1,
in quantum information such as quantum state teleportation [Bouwmeester
which occurs when the two pairs are indistinguishabl e from each oth er, as
et al. ( 1997)] and quantum state swapping [Pan et al. ( 1998)].
depicted in Fig. 8.22(b). Thus , distinguishability results in degradation of
After successful studies of two-photon interference phenomena in the
the interference effect.
90's of last century, attention started to focus on multi -photon interfer ence
For partial distinguishability of the pairs , Ou et al. considered a multi-
phenomena with N > 2 due to th e discovery of the GHZ states [Greenberger
mode four-photon state made from two pairs of correlated photons from
et al. (1989); Yurke and Stoler (1992); Bouwmeester et al. (1999)]. At the
spontaneous parametric processes [Ou et al. (1999b)]. The photon state
time , spontaneous parametric down-conversion processes with i 2) materi-
has th e form of
als were popular sources of two -photon states, and with stronger pumping
power by pulsed lasers, the higher order processes will produce multiple
pairs of photons for high er photon number state generation. However,
14'4) (; J dw,dw2dw;dw;<P2(w,
,w2)<P2(w;
,w;)

pair production in spontaneous parametric down -conversion is completely X al (w1)al(w~)a~(w2)a~(w;) lvac)' (8. 77)
random so that the pairs of photons produced are likely separated in time whi ch is taken directly from Eq. (6.32) in Sect ion 6.1.5 for a weak parametric
as shown in Fig. 8.22(a). So, ultra-short pump pulses are required [Zukow- process (l~\2 << 1). Here, <I>2(w1,w2)= <l>2(w2,w1)is the two-photon wave
ski et al. (1995); Rarity (1995)] in order to force the two pairs be produced function that is symmetric with respect to w1, w2. The distinguishability
in one single temporal mode as shown in Fig . 8.22(b). In this way, the four between pairs is described in terms of the quantity £/A with
photons be come completely indistinguish ab le. Most likely scenario will be
somewhere in between. If the four photons described here are involved in a
A= J dw1dw2dw~dw;l<1>2(w1
,w2)<P2(w~,w; )l2 (8.78)
 Appl'ications: J,,J11,lti-Pholo11 I nte1fC'1'encc CL'lldEntanglement 255
254 Q-uanfom Opt'ics For Experimen lalis/,s

and the same for the two photons, the difference in their mode fu~1ctions _n_rn:'

E= J (w1,w2)<I>2
dw1dw2dw~dw;<I>2 (w:, w;)<I>;(w1, w;)<I>;(w~,w2). (8. 79)
also lead to partial disLinguishability, as we have demonstrated m the v1s1b1-
lity of the Hong-Ou-l\Iandcl interference given in Eq. (8.20~ in Sec~ion 8.2.1
and of the induced coherence effect given in Eq. (8.39) m Section 8.2.3.
\Vith these two quantities, it can be shown that the four-photon bunching
For a more general two -photon spectral wave function <I>2 (w1;w2), which
effect takes a value of [Ou et al. (1999b)]
is defined through an arbitrary two-photon state in two one-dimensional
1
P4(4, 0)/ Pf (4, 0) = 4 + A! E, (8.80) fields:
l<I>u)~~l)= / dw1dw2<I>2(w1.w2)al(w1)at(w2)lvac), (8.82)
which recovers t he maximum value of 6 for E = A and the value of 4 in
Eq. (8. 76) for E = 0. The value of 6 correspon ds to four-photon bun-
it is stra ightforward to show that the visibi lity of the Hong-Ou-Mandel
ching effect whereas t he value of 4 is on ly due to two -pho ton bunching
for each pair. Thus, the quantity E/ A is a goo d measure of photon in- interference effect is related to <I>2
(w1;w2) as
distinguishability between pairs. The exper im enta lly measured value of
P 4 (4,0)/PJ 1(4,0) = 5.1 ±0.4 from Fig. 8. 18 gives E/ A = 0.4 .
With the input state in Eq. (8. 77) , it is stra ightforward to find the
V2 = I/dw1dw2<1>;(w1,
w2)<1',(w2;w1) I// dw1dw2l<!>2(w1,
2
w2)1 , (8.83)

output state at one output port of the b eam splitter as where we omit the spatial mode in the one-dimensional approximation for

out =
l<!>4) f/ dw1 dw2dw;dw;<!>2 (w;,w;)
(W1, w2)<!>2
2)j with j = 1, 2 for one output
simpli city of argument. The output state l<I>
port of the beam splitt er can be shown to be

xa t (w1)at (w~)at (w2)at (w;) lvac). (8.81) l<I>2)j= / dw1dw2<I>2(w1,w2)a}(w1)a}(w2)lvac)

(j = 1,2), (8.84)
This is a temporal four-photon state in the output port 1. We will study
which is in the same form of Eq. (8.66) but for arbitrary two-photon spectra l
the property of photon indistinguishability according to the value of E/ A
in the next section. wave function <I>
2(w1,w2).
Applying Cauchy-Schwarz inequ ality to Eq. (8.83), we find that V2 = 1,
4
or the maximum interference effect occure if and only if
8.4.3 Characterization of Photon Indistinguishability by
Multi-Photon Bunching Effects (8.85)

Since photon indistinguishability has such an impact on the multi-photon
interference effects, we should be able to use the interference effects to
characterize the properties of photon indistinguishability just like optica l
coherence characterized by the visibility of optical interference. In this way,
we should be able to tell the difference between the two four -photon states
in Figs. 8.22(a) and (b). This becomes especially important if we want
to produce an N -photon state from N single-photon states in the scheme
discussed in Section 8.4.1 .
Let us first revisit the Hong -Ou-Mandel interference effect discussed in
Section 8.2.1. We have shown that it depends on the overlap between the
two incoming photons . If the arrival times for the two photons at the beam
splitter are quite different, temporal dist inguishability between the two pho -
tons will dimin ish the interference effect. Even when the arrival times are
This seems to gives the condition for the complete temporal indistinguis-
hability of the two photons in th e two-photon state I<I>2) j (j = 1, 2) of
Eq. (8.84), for V2 = 1 gives the maximum two-photon bunching. effect
or the complete overlap of two photons at the output, correspondmg to
the situation in Fig. 8.20(b). Indeed, for the factorized <I>2(w1,w2)
¢ (wi)eiw1T1¢(w2 )eiw2T2 in Section 8.4.1, condition in Eq. (8.85) is satis -
fied when T 1 = T2 , and th e output state is a single-mode two-photon state,
as discussed in Section 8.4.1.
However , let us consider a non-factorized <I>;
(w1,w2):
(8.86)
4 This is also obvious from Eq. (8.66) . l<I>1,1h2 = 0, i.e. , maximum Hong-Ou -Mandel
effect when condition in Eq. (8.85) is met.
 256 Q'uantwn Opti cs F01· E:rperim entalists Appl'ications: Multi-Photon Int e1fcren ce and Entangl em ent 257

On the other hand, we have V2 = 0, or no Hong-Ou-1\Iandel interference

effect when
<P 2(u.,•1. t...'2)
-= v(w1 )0(-'-·2) ____ __ (8.89)
2
which gives the crit erion for the comp lete te mporal distinguishability of
the two photons. This is a sufficient cond ition but not a necessary one in a
F ig. 8.23 Generation of n on- factorize d wave function from factor ized wave function.
similar way as the discussion above for the cond itio n in Eq. (8.85). Equa-
tion s (8.85) and (8.89) ca n be generali zed to a state with hi gher number of
with ,I.Ti - _T2I » ~T. _It satisfies the symmetry condition in Eq. (8.85) photons.
and gives nse to V2 = 1 111 Hong-Ou-Mandel int erfer ence. Th e two-photon For the two pairs of photons from param etr ic down-conversion, the four-
state of this form can b e produc ed with unb alanc ed Mach-Zehnder interfe- ph oton state is given by
1,w 2) = ¢(w 1)¢(w 2), as shown
~·om~t ers from the factorized state with <I>2(w
/<I>4)= / dw1dw2dw:dw;<I>2(w1,w2)<I>2(w:,w;)
~n Fi~. 8.23. Notice that the two photons never meet in the beam splitter
111 this case and yet we have V2 = 1 and this demonstrates the peculiar X at (w1)&,!(w2)at (w:)a!(w;) /vac). (8.90)
quantum nonlocal behavior of multi-photon states [Pittman et al. (1996)·
Nasr et al. (2003); Lu et al. (2003)]. -- ' So the four-photon spectral wave function has the form of
Now look at the output state at one output port: (8.91)
/<I>2)j= / dw1dw2¢(w 1)¢(w 2) (eiw1T1 eiw2T2 + eiw 1T 2eiw2T 1) We have just shown that the overlap betw een <I> 2(w1, w2) and <I>
2(w2, wi)
dete rmines the two-photon indistinguishability. So, let us assume we have
X aJ(w1)0,J(w2) /vac) total indistinguishability between the two photons within one pair, i.e .,
= 2/Ti) j IT2)j (j = 1, 2), (8.87) 2(w1,w 2) = <I>2(w2,w
<I> 1). From the discussion in Section 8.4.2, we learned
which, with /T1 - T2 I » ~T , is a two -photon state corresponding to that the indistinguishability bet ween two pairs is determined by E = A.
Fig . 8.20(a) with the two photons well-separated in time T 1 and T2 (~Tis With definition in Eq. (8.91) , the condition for indistinguishable pairs , i.e. ,
the width of the wavepacket ¢(w)). E = A, can be rewritten as
~h~refore, the symmetry condition in Eq. (8.85) is only a necessary <:l>4(w1,w2;w:,w;) = <:l>4(w:,w2;w1,w;) = <:l>4(w1,w;;w:,w2). (8.92)
condit1011 for two photons in the state of Eq . (8.84) to be indistinguishab le
in time but not sufficient condition . This is reflected in the fact that for Note that the permutation is between primed and unprimed variables, in-
any <I>2(w1,w2)that does not sat isfy condition in Eq . (8.85), we can always dicating permutation symmetry between two photons , with one from each
rearrange the order of aj(w1)aj(w2) in Eq. (8.84) so that it has a new form pair. Thus , we have pair exchange symmetry. Again, we need to be cauti -
of ous about the condition in Eq. (8.92) since it is only a necessary condition

j =
/<I>2) j dw1dw21[<I>2(w1, w2) + <I>2(w2, w1)] aJ(wi)aJ(w2) /vac) like the two-photon case.
On the other hand, for the condition for complete distinguishable pairs,

· h q>sym (
J
= dw1dw2<I>~ym(w1,w2)aj(w1)aj(w2)/vac) (j = 1,2) (8.88)
i.e ., E = 0, we have

wit 2 W1,w2) = [<I>2(w1,w2)+ <I>2(w2,w1)]/2, which satisfies the sym -
metry condit ion in Eq. (8.85). This is because the number -state represen -
tat ion of a Bosonic system is already symmetrized with respect to particle
exchange .
(8.93)
Both Eqs. (8.92) and (8.93) are extensions of Eqs. (8.85) and (8.89) to
the four -photon case of two pairs. These can be further generalized to
an arbitrary number of photons. But because of the complexity involved,
 259
Appli .ra.tions: l\Jv,lti-Photon Int erf eren ce and Enta11glernent
258 Quantum Optics For Exp erim entalists

haYing a width of 6T. IP\0 t "' is permutation between to, t1,;and JID{t1,
.... tN}
we 'Nill not discuss it here. Readers who are interested can find mor
discussions in [Ou (2008)] . is any permutation of t 1 , f 2 . ... , t N . Hence , we have
Even though the mathematical description of a tempora l multi-photon PN+l = 21 +1 j dtodti ...dtN L ]IDtotk[
L c *(to:Jl:D{t1,••
•,tN})]
state such as that in Eq. (8.84) cannot clearly and uniquely determine the k,I IP
temporal distinguishability of the photons, we can resort to multi -photon in-
XJl:Dt
t [LG(to;Jl:D{t1,•••,tN})]
0 1
terference expe rim ents for a solution. A simp le scheme is t he multi-photon
IP
bunching effect discussed in Section 8.3.1. We are specifica lly int ereste d in
(8.100)
the N + l scheme in Fig. 8.14 where t he input N -photon state is a tempo- =L +L ·
k=l ki,l
ral multi-photon state such as that in Eq. (8.84) and t he other input is a
single-mode single-photon state who serves as the reference. For simplicity In Appendix B, we find the first sum as
of arg um ent , we ass um e the singl e-photon state ha s a form of
L = (N + l )!N (8.101)
k=l
= ITo) = / dw¢(w)eiwT0 a!(w)lvac)
l<I>1)2 (8.94)
with
and the N-photon state is the one generate d by the scheme in Fig. 8.21 and
has a form of N = _l_
- 2N+l
j dt 0 dt 1...dtNG*(t 0 ; t1, ..., tN) L G(to; JID{t1, ..., tN} ).
IP
(8.102)
N }l = IT1)... ITN}
l<I>
= ! dw1...dwN¢(wi)eiw1T1 ...cp(wN)eiwNTN For th e second sum in Eq. (8 .100), iet us assume To = Ti with i =
1, ... , m and \To -Tj\ » 6T for j = m + 1, ... , N, that is , the sing le photon
al al
X (w1)... (wN) lvac). (8.95) entering at port 2 completely overlaps with m photons in the N-photon
Let us now calculate the N + I-photon coincidence counting probability: state entering port 1 but is well-separated from other N - m photons. Th e

PN+l = ! dtodt1...dtN (et(tN ) ...ftt(to)E(to) ...E(tN))' (8.96)

calc ulation in Appendix B gives
L = m(N + l)!N. (8.103)
where we integrate over all time for overall probability since the input states ki,l

are non -stationary pulses and Hence , the total N + 1-photon coinc idence counting probability is

E(t) = [E1(t) + E2(t)]/h° = j dw [a1(w) + a2(w)] e- iwt_ (8.97) PN+i = (1 + m)(N + l)!N = (1 + m)PJf+i , (8.104)

It is eas ier to first calcu late: where

PN+l = (N + l)!N (8.105)
E(to)E(t1) ...E(tN) l<I>
N )i l<I>1
)2
N
is the probability when To = ±oo or the single photon in port 2 does not
= 2(N!l) /2 L ]IDtot1;;
[E2(to)E1(t1) ...E1(tN)] l<I>N)il<I>1)2 overlap with any photon in the N-photon state in port 1, which gives m = 0.
k= O
With the result in Eq. (8.104), we can proceed as follows to exper imen-
l N
= 2(N+l)/ 2 L
ltot1;; [ I::G(to; JP>{t1, ..., tN})] lvac), (8.98)
tally probe the tempora l structure of the N-photon state : we scan the time
To of the sing le photon in port 2 and as it goes through T1, ... , TN, we will
k= O IP
observe a number of peaks in PN+i, as shown in Fig. 8.24. The height
where G(to; t1, ...tN) = g(to - To)g(t1 - T1)...g(tN - TN) with
of each peak corresponds to the number of photons that overlap with the
g(t) = - 1 / dwcp(w)e-iw
. t (8.99) sing le photon. Note that when all N photons are in one temporal mode
 260 Quantum Optics F01· Experim entalists
Applical ions: ]'.fulti-Photon Inte1fere11ce and Entanglement 261

A A
(a)
IN),
T0 m/11 AT·.IJ =
Wl·tl l
Ll - T l - T·(i
J • J. = 0. l. 2). Here, we take g(t) as real. Note
Tl Tk TV H(O) = 1, H(oo) = 0. When To= ±oo. we have
P 3 = P!f° = 3[1 + H 2 (6T12)]/4. (8.108)
(bl
P11.-
+1

Pi\':1
A ff[~U., With this quantity, Eq. (8.106) can be expressed as

0
1 f
TI Tk 0,.
'
p3
p=3 =l+
2
2H(6T 01)H(6T 02)H(6Tl2) + H 2(6To1) + H 2 (6To2)

2
l+H 2(6T12)

if IT1 - T2I » flT,

= {l + H (6T 01) + H (6To2)
Fig. 8.24 Temporal structure of the N-photon state is scanned by the single photon. (a) (8.109)
Temporal distribution of the photons. (b) Normalized coincidence counting probability 1 + 2H 2(To - T1)
as a function of To.
When To = T 1 = T2, we have P3 / P 3 = 3 and when To = T1 but ~ell
with To = Tn(n = l, ... , N), we have m = N and PN+i/ P'fv+ = l + N, separated from T2, we have P3 / P 3 = 2. These values are consis~ent with
1
exactly the value found in Section 8.3.1 for the (N + 1)-photon bunching the result in Eq. (8.104). In Fig. 8.25, we plot P3/ P 3 as a funct10~ of To
effect. When m < N, photon distinguishability among the N photons will for three values of 6T 12, which correspond to three different scenarios for
reduce the effect of multi-photon constructive interference. By the met- the temporal two-photon state l<I> 2) in Eq. (8.66). Here, we take g(t) as a
hod discussed above, the temporal structure of the N-photon state can be Gaussian function with a width of flT.
revealed experimentally.
On the other hand, Fig. 8.24 only shows an extreme case when pho-
tons in the N-photon state are grouped together with no overlap between
different groups. The more general case of partial overlap is complicated ;,~i .----1,.
, A -'---r-A_A-----:-____
---,-- ------r--1,:,
-~~--~ . ,_,
and was considered in [Ou (2008)]. But to have some idea about how the
situation looks like, we treat next the case of N = 2. In this case, Eq. (8.98) !i ! I !> -(, .j , lj

T/t:,,T
becomes
E( to)E( t1 )E( t2) I<I>2)
1I<I>1
)2 . 8 · 25 p 3 /P
F 1g. 3
00
as a function of To for (a) T1 - T2 = O; (b) T1 - T2 = 71::::.T;(c) T1
-T2 = 21::::.T.
= 23\ { g(to - To)[g(t1 - T1)g(t2 - T2) + g(t2 - T1)g(t 1 - T2)]
+g(t1 - To)[g(to - T1)g(t2 - T2) + g(t2 - T1)g(to - T2)]
8.4.4 Optical Coherence as a Consequence of Photon
+g(t2 -To)[g(t1 -T1)g(to -T2) + g(to -T1)g(t1 -T2)] }ivac). Indistinguishability
Then, the three-photon coincidence counting probability is
2 Classical optical coherence theory, developed in the 195Os [Born and Wolf
P3 = /E(to)E(t1)E(t2)l<I>2)il<I>1)2/ (1999)], was based on the second-order or single-photon int~rference effects .

= t 2
[6 + 6H (6T12) + l2H(flToi)H(flT 02 )H(flT12)
In brief, the intensity distribution shows an interference frmge pattern as

J(x) ex:Ji+ h + 2-/.IJ;hl cos21r(x - xo)/L, (8.110)

2 2
+6H (6T 01) + 6H (6To2)], (8.106)
where where Ji, h are the intensities of the two interfering fields, L is the fringe

H(!iT,1) = J dtg(t - T,)g(t - T1) = J dtg(t)g(t - liT,J) (8.107)

spacing along the x-direction, and

= (E{ E 2) /-/.IJ; (8.111)


262 Quantum Optics For Experim ental ist s
is the degree of coherence between the two fields. Herc, :r 0 in Eq. (8.110)
is related to arg(,).
Quantum col~erence theory was later constructed by Glauber primari ly
along the same lme as the classica l theory, but with quantum forma lism of
operators and quantum states . The physics was hidden beneath t he com-
plicated mathematical formu la . As discussed in previous sect ion s, distin-
guishabi lity of photons leads to degradation of the visibility of interference .
Thus, the two should be related to each othe r. In the following, we will
make an initi al attempt to revea l the connection.
In 1996 , J avanaa inen and Yoo showed that in a sing le realization, an
interference fringe will form in the sup erpo sit ion region of two groups of
photons with the same number N, respectively, i.e., with a state of IN)i IN)2
[Jav anaa inen and Yoo (1996)]. Lat er , t he study was exten ded by Ou and
Su to the superposition of two groups of photon s with differ ent photon
numbers n an d m, resp ect ively, i.e. , with a state of ln)i lm) 2 [Ou and Su
(2003)]. A quantum Monte Carlo simulation in the study by Ou and Su
shows that for the state of ln)i lm)2, there is an int erfere nce fringe forming
with a probability distribution of [Ou and Su (2003)]
P( x ) ex n + m + 2ylnm cos 2n(x - x 0 )/ L , (8.112)
where Xo is arbitrary and L is the fringe spacing. If we compare the above
with Eq. (8.111) , we find that the normalized degr ee of coherence is simply
"Y= l. This is not surprising in the sense that the photons in th e quantum
state In, m) belong to one wave function and are all indistinguishable in the
superposition region . On the other hand, if there is partial indistinguishabi -
lity among the photons, from the discussion in the previous section, we find
that the visibility will drop. Assume that the input state is IN)i IM) 2 but
only n photons among the N photons in mode 1 are indistinguishable from
m photons among the M photons in mode 2. Therefore , only then+ m 2
photons will give rise to an int erference pattern , described by Eq. (8.112).
The rest of the photons , i.e. , N - n photons from mod e 1 and M - m pho -
tons from mode 2, are distinguishable and produce no interference fringe.
Thus the probability distribution in this case is given by
P'(x) ex [N - n] + [M - m] + [n + m + 2ylnmcos2n(x - x 0 )/L]
= N + M + 2ylnm cos 2n(x - x 0 )/ L
= (N + M) [l+!~ cos21r(x - xo)/Ll (8.113)
Comparing with Eq. (8.112) , we have the degree of coherence
"Y' = Jnm/NM. (8.114)
263
Appl ication s: Multi-Photon Int erf eren ce and Entangl e m ent
ote that n/ N and m / M are the percentages of indistinguishable photons
in the two groups, respective ly. Thus the degree of coherence is_re lated
to the percentage of indistinguishable photons among the photons mvolved
in interference. The same conclusion was made by J\landel in a different
argument [Mande l ( 1991)]. . . . . ..
As we demonstrated throughout t his Chapter, the md1stmgrnshab1hty
of photons is closely related to the modes of the photons. \ Vhen photons are
all in one single mode (spec ial or genera l) , they become indistinguish~ble
and give rise to the maximum effects of int erference, which is ch aracte ri zed
as "'I = 1 here for interference fringe. However, photon multi-mode exci-
tations of the optical fields shou ld lead to distinguishability and reduced
interference effect thus a sma ller degree of cohere nce "'f < 1. On the other
hand, if there exist phase corre lation s among different modes, t he optical
field can be described by a genera lized singl e mode, lead in g to indi st in guis-
hab ility and int erference. The ph ase corr elat ion is of course described by
the degree of coherence T Therefore, photon indistinguish ab ilit y should be
related to the coh erence of the field.
8.5 Problems
Problem 8.1 Generalized Hong-Ou-Mand el int erferometer for two pairs of
photons with an asym met ri c beam splitter .
We hav e seen the generalization of the classic Hong-Ou-Mandel two-
photon interferom eter to three photons with an asymmetric beam splitter
(BS) in Section 8.3.2. We will make a generalization to four photons, two
from each side of the BS , i.e., the input state is l<I>in)= l2)i 12)2-
(i) Use the techniqu e in Sect ion 6.2.2 to show that the output state is
2
J<I>out)= TRv16(j4,0 ) + J0,4)) + (T +R -4TR)j2,2)
+J6TR(T - R) (13, 1) - 11,3)), (8.115)
where T , R are the transmissivity and reflectivity of the beam splitter , re-
spectively . Notice that the term of 12, 2) disappears or P4(2, 2) = 0 when
y2+R2-4TR = o or T = (3±vl3)/6, R = (3=t=vl3)/6, leading to generalized
Hong-Ou-Mand el effect for two pairs of photons.
(ii) Suppose the two pair s of photons are in the state of l<I>:n)= 111,b) ®
I1~, 1~) (2 x 2 case) . Use the m ethod in Section 8.4.2 to show
(8.116)
 26.J Q'Uantum Optics For E:rperimentalists
Applications: Multi-Photon Interference a11d Entanglement 265

where P_rl(2, 2) is th e probability for classical particles. Here we use T :::::: that the four-photon coincidence measurement probability in the scheme
(3 ± J:3)
/ 6, R = (3 =i=J:3) / 6. Thus, the visibility is 1/ 3 in this case. shown in Fig. 8.27 has the following dependence on the phase differ ence cp
betw een th e two arms:
Problem 8.2 Three-photon NOON state by projection: Wang-Kobayashi
interferometer. P4 (2c, 2D) ex 1 + cos4cp. (8.118)

The NOON state introduced in Eq. (4.54) of Section 4.5 .1 is of great in- This leads to four-photon de Broglie wavelength . Her e, we take T = (3 ±
terest in meteorology because it is capab le to produce an interfer ence fringe /3)/6, R = (3 =i=/3)/6.
with 21r/N period of phase in JV-photon coincidence measurement , thus en-
hancing the sens itivity of phase measurement by a factor of JV. On the other
hand , pure NOON states of the form in Eq. (4.54) is extreme ly hard to pro-
duce in the lab because states like IN -k, k), k = 1, ... , N -1 are not easy to
elimin ate . For examp le, the three-photon state in Eq. (8.61) has the term
of 11, 2) besides the NOON state for N = 3. However , the contrib ution of
11,2)-term can be made to zero by using the three-photon Hong-Ou-Mande l
effect discussed in Section 8.3.2, as was suggested by Wang and Kobayashi Fig. 8.27 Wang-Kobayashi Int erferometer for four-photon de Broglie wavelength.
[Wang and Kobayashi (2005)].

Problem 8.4 Characterization of JV-photon distinguishability by genera-

a
lized Hong-Ou-Mandel effect of N photons and one photon.

Since multi-photon interference effects depend on the photon indistin-

guishability, we can in principle employ any one of them for the charac-
terization of photon indistinguishability. Here, we will take a look of the
generalized Hong-Ou-Mandel effect for the input of N photons at one side of
Fig. 8.26 Wang-Kobayashi Interferomet er for three -ph oton de Broglie wavelength.
a beam splitter and a single photon at the other discussed in Section 8.3.2.

The schematic of Wang-Kobayashi Interferometer is shown in Fig. 8.26
with T = 2R = 2/3. Show that the three-photon coincidence measure-
ment probability has the following dependence on the phase difference ,6.cp
between the two arms :
P3 (2c, ID) ex 1 + cos 3,6.cp. (8.117)
This dependence on phase ,6.cpis equivalent to the situation when the three
photons are treated as one entity with an equivalent de Broglie wavelength
of Ao/3. Here ,\ 0 is the wavelength of one photon.
Problem 8.3 Four-photon NOON state by projection.
We can generalize Wang-Kobayashi Interferometer to four photons using
the four-photon Hong-Ou-Mandel effect discussed in Problem 8.1. Show
(i) For an input state of IN , 1) to a beam splitter of transmissivity T
and reflectivity R, use the technique in Section 6.2.2 to show the result
in Eq . (8.62). Thus , when T = NR = N/(N +1), we have PN+1(N, 1) = 0,
i.e., the genera lized Hong-Ou-Mandel effect for N + 1 photons.
(ii) For the input of the multi-mode N-photon state in Eq. (8.95) and
single -ph oton state in Eq. (8.94), prove , sim ilar to Eq. (8.104),
PN+i(N, 1) = 1
_ m
(8.119)
PN+l(N, 1) N
when To= Tj(j = 1,2, ... ,m) and ITo-Tkl » ,6.T(k = m+l, ... ,N) . This is
the case when the single photon overlaps with m photons in the JV-ph oton
state but is well separated from the other N - m photons. PN+l (N, 1)
corresponds to the case when !To - Tkl » ,6.T(k = 1, ... , N), i.e., the single
 266 Quantum Optics For Experirnentahsts Applications: Multi-Photon Int e1:ference and Entanglement 267

photon does not overlap with the N-photon state at all. So, in contrast Find the output state l1J!
)01 it and show that the multi-photon detection
to the bumps in Fig. 8.24, PN +l (N, 1) shows dips here as To is scanne d prob ab iliti es are
through Tk(k = l , 2, .. , N).

Problem 8.5 Multi-photon

and a coherent state.
interference effect betw een a squeezed vacuum
( Q3 -
av)2
3µ
As we have shown in Section 8.3.1 , stimulated emission can be exp lained a
4
-6--
Q2V
+3-
V2) ' (8.122)
by multi-photon interference. Some other int eresting quantum phenomena
( µ µ2
can also have an explanation through quantum interference. The photon where a= ,B -JRwith R « l and T--+ 1. The above is similar to the result
number probability distribution given in Eq. (3.94) and shown in Fig. 3.6(a) in Eq. (3.94) with v « l , µ "-' 1. Comparing Eqs. (8.121) and (8.122), we
for the squeezed coherent states is another example . Here , we will use the find that P(2a , Ob) is from the superposition of two contributions: one is
result in Eq. (3.196) for squeezed vacuum states and multi-photon interfe- the two photons from the coherent state and the other is the two photons
rence picture to obtain the result in Eq. (3.94). from the squeezed vacuum. P(3a , Ob) and P(4a , Ob) have similar explanati-
In order to obtain a coherent component for the squeezed coherent state ons. Hence, we demonstrate that the result in Eq. (3.94) can be viewed as
in Eq. (3.88), we mix the squeezed vacuum state /-r) with a coherent state interference between a squeezed vacuum and a coherent state.
/,8)by a beam splitter. We have shown in Section 6.2.4 that when the beam
splitter has T --+ l and R "-' 0 but with ,B-JR= a = constant, the side Problem 8.6 Photon anti-bunching effect in a beam splitter with a sing let
with squeezed vacuum transmitted will have an output state in the form Bell state .
of a squeezed coherent state. To demonstrate the effect of multi -photon
interference , we will assume weak excitation with /a/2 "-' v « l so that we The photon bunching effect in the Hong-Ou-Mandel interferometer
only need to write the squeezed vacuum and the coherent state up to the shows the cooperative behavior of the Bosonic nature of photons. It also
four-photon terms: demonstrates the symmetry between the two photons involved, which is
required for Bosons. On the other hand, Pauli's exclusion principle for
Fermions will lead to separation of particles at the outputs of the beam
splitter or anti-bunching effect [Bocquillon et al. (2013)]. As we have seen
in Chapters 1 and 5, photon anti-bunching is allowed in quantum optics. It
turns out that we can mimic the Fermionic behavior with photons having
some sort of anti-symmetric property.
where we used Eq. (3.196) for the coefficients in the squeezed vacuum state (i) Consider a two -photon polarization state given with a minus sign in
and v = sinh r, µ = cosh r. So, up to the four-photon terms, the input Eq. (4.49) of Section 4.4.2:
state to the BS is
IIJ!-) = (llx)Ally)B - lly)Allx)B)/v'2, (8.123)
which is a Bell singlet state that changes sign when we switch A and B.
Show that when the state in Eq. (8.123) is input to a 50 :50 beam splitter,
the output state is
(8.124)
where the two photons are separated in the outputs, showing the Fermionic
(8.121)
anti -bunch ing behavior.
 268 Qu antu m Opt ics For Exper im entali sts

(ii) Find the output states for other Bell states in Section 4.4 .2.

Tlw properties of the Bell states discussed abov e were used to distinguish
between different Bell states [Braunstein and Mann (1996)] with applicati -
ons in Bell state projection measurement for quantum st ate t eleportation
[Bouwmeester et al. ( 1997)].
(iii) Consider the reverse process of Hong-Ou-Mandel interferometer
(Eq. (8.8)) , that is, the input state is
(8.125)
which is anti-symmetric with respect to A and B. Show that the output
state is l'It)aut = ll)All)s, that is, the two photons are separated from each
other like Fermions [Bocquillon et al. (2013)].
Problem 8. 7 Purity of single-photon and the visibility of HOM effect.
A single-photon wave packet is described by Eq. (4.43) in Section 4.3.2
as
Chapter 9
Experimental Techniques of Quantum
Optics II: Detection of Continuous
Photo-Currents
We mentioned in Chapter 7 that when the intensity is relatively large , the
hotoelectric pulses generated by the detector will overlap and form a con-
p d. l.
tiirnous photoelectric current. This is the situation to be discusse m t 1is
chapter . The methods of measurement and analysis will be totally different
from those in Chapter 7. For a continuous photocurrent , we cannot use a
pulse discriminator to filter out the contribution from the dark cm~rent as

IT)¢= J dwcp(w)eiwTaT(w)IO). (8.126)

in Chapter 7. But because of the randomness of the dark current, its con-
tributions are mostly in the low frequency regime in the form of 1/ f noise.
But if the p eak time T is fluctuating due to uncertainty in emiss ion time, So we can use the method of spectral analysis to filter it out. Therefore,
as often happens for the single photon from a quantum dot [Sun and Wong sp~ctral analysis is the main method in the detection of a strong continuous
(2009)] or the heralded photon from spontaneous parametric process [Ou optical field.
(1997b)], the state is described by a mixed state with density operator
9.1 Photocurrent and Its Relation to Quantum
1
Plp = !1T LT dTIT)¢(TI, (8.127) Measurement Theory

where !1T is the uncertainty in T. When the intensity of the optical field is high, many photoelectrons can be
produced from a detector, resulting in the overlap of the amplifie~ ph~t~e-
Show that if we use the photons described in Eq. (8.127) for the Hong-Ou-
lectric pulses. Thus a continuous photocurrent is formed. For simphc~ty,
Mandel interferometer , the visibility is
we assume all the photoelectric pulses have an identical response funct10n
VHoM = Trp2 = J dwdw'l¢(w)cp(w')l2 sinc 2 [(w - w')!:1T/2], (8.128) of k(t), as shown in Fig. 9.l(a). The pulse generated at tj makes a contri -
bution of k(t - tj) to the photoelectric current at t. The sum o~ the t~tal
which defines the purity of the single-photon state [Du (2015)]. Obviously, contributions of all photoelectric pulses generated at different times gives
2
VHoM = Trp 1 if !1T « Tc with Tc as the width of the single-photon the photoelectric current :
wave packet in Eq. (8.126).
i(t) = L k(t - tj) (9.1)
j

The generation of the photoelectrons is random so tj is a random varia-

ble. Now divide the time axis into small segments of size !1t (Fig. 9.l(b)).

269
 270 Quantum Opt ics For E xperim entalists Rrp crim ental Techniqu es II: D etection of Conhrmous Photo-Cn1'T cnts 271

k(t)~ current operator i(t) and the quantum state \[I of the optical field. The
case of quantum efficiency 11< 1 can be modeled by the optical losses (see
Llr
1 Section 9.4) . In this case. vacuum quantum noise will enter through the
lossy channe l.
(a) (b)

Fi_g. 9. 1 (a) The identical response function k(t) of photoe lectric pulses; (b) Photoelec- 9.2 Spectral Analysis of Photocurrents
tric pulses generated at different t im es.
Although the photoelectric current is generated by light, it is still an elec-
From Glauber's photo-detection theory [Glauber (1963b , 1964)], we have tric current in essence. So, we can study it by using the methods for electric
the probability of detecting a photoelectron within the sma ll segme nt as current ana lysis . Electrical engi neer ing provides us with a unique way for
~Pi (t 1 ) = P1 (t1 )~t with the probability density 1 current ana lysis, that is, the spectral ana lysis of t he current. Because of the
existence of the 1/ f noise in electric currents, the large electronic noise at
P1(tj) = TJ(E(-\tj)E(+\ tj))w = TJ(i(t1 ))w. (9.2)
low frequency forces us to work at high frequency domain whe re t he intrin-
Her e O < TJ< 1 is the quantum efficiency of the detector. \[I is the quantum sic electron ic noise is relatively low. The method of spectra l ana lysis can
state of the optical field. j;( +) ( t) has the form of effectively filter out the low frequency electron ic noise. This is usually done

j;( +)(t) = Jdwa(w) e-iwt = [E(-\t)t (9.3)

with an electron ic spectrum ana lyzer, whose output spectrum is related to
the fluctuations of the input electr ic current by

in the 01;:e-dime_?sionalA
quasi-monochromatic approximation (Sections 2.3.4,
2.3.5). J(t) = E( -) (t)E(+)(t) is the intensity operator. Hence , the average
Ssp(D) = J dT(~i(t)~i(t + T))ej0.T, (9.5)

of the photo-electric current is where (~i(t)~i(t + T)) = (i(t)i(t + T)) - (i(t) )(i(t + T)). The correlation
+ T)) of the currents can be obtained from Eq.
(i(t)) = L k(t - tj)~Pi(tj) = L k(t - t1)P1(tj)~t
function (i(t)i(t (9.1) as
j j
(i (t) i (t + T)) = ( L k (t - ti) L k (t + T - tj))
= J dt'k(t - t')p,(t') = '7 J dt'k(t - t')(i(t'))w. (9.4)

We can see from the expression above that , when the quantum efficiency TJ= i,J
1, we can define a photoelectric current operator: i(t) = J dt'k(t - t')i(t')
and the average of the detecte d photocurrent is simply (i(t)) = (i(t))w,
= \ Lk(t-tj)k(t+T-tj) )
j
that is, the expectation value of the photoelectric current operator i(t)
for the quantum state \[I. We learn from quantum mechanics that the +(L k(t - ti)k(t +T - tj) ), (9.6)
quantum measurement of any physical observable corresponds to a Hermi- i=Jj
~ian operator. Ou a:1d Kimble proved that photoelectric current operator where, due to the discr ete ness of th e photoel ectric pulses , we split the
i (t) = J dt' k (t - t') I (t') is exact ly the Hermitian operator for the photo - double sum in the second line of the expression above into two terms corre-
electric measurem ent process [Ou and Kimble (1995)] . All the stat istical spond ing to the auto- corr elation of each electric pulse and cross -correlation
properties of photocurrent i (t) can be calcu lated from th e photoelectric between two elect ric puls es, respectiv ely. The first t erm of auto -correlation
1
The mod el pr ese nt ed her e for the photocurrent can also be applied to Mandel's sem i- can b e calculated in the same way as we deriv e Eq. (9.4) and we have
classical th eory of photo-d et ect ion for class ical wav es [Mandel et al. (1964)]. But we
need to change the average over th e quantum state to the average over th e fluctuations
of wav es .
\ L k(t - tj)k(t + T - tj)) = / dt'k(t - t')k(t + T - t')p1(t'), (9.7)
J
 272 Qu.a.nt w n Op tics Fo'/· Expcri m Pnta.l is ts
ETperim e'lltal Techniques II : D etection of Continuous Photo-Currents 273

wh ere_p}(t~) .ii, ,obtain : d from E q. (9.2). _ _T he se_concl tc nn in E q . (9.6)

as
conc~1ns c10_ss _-corr elat10n b etween two chfferent t un es and its calcul at ion
reqmr es t he Jomt pro b abilit y !).P2(ti. tJ) = p2(ti, tJ){).f;!).t.i of detect ing two ( i (t )'i (t + T)) = (i (t )i (t + T) h
photo electro ns in the sm all regions !).t;, {).f.i, resp ectiv ely. \i\Tith t his, th e
second te rm in Eq . (9.6) ca n b e expr essed as
= ( / dt'k (t - t' )i (t' ) J dt"k (t + T - t" )i (t" )) \JI

= J dt' dt" k (t - t') k(t + T - t" ) (] (t' )f (t" ))w

\ Lk (t - ti) k (t +T-
ii=,j
tj))
= J dt'dt"k(t - t' )k(t + T - t" )
L k(t - ti)k (t + T - tj )!).A(t i, tj)
= x [(: i(t' )i(t" ) :)w + o(t' - t") (i(t')h ]
ii=}

= L k(t - ti)k(t + tj )P2(ti, tj )!).ti!).tj

T -
= J dt' k (t - t' )k (t + T - t' ) (j (t' )) \JI

=
ii=,j

J dt'dt"k(t -- t')k(t + T - t )p2(t' , t").

11
(9.8)
+ J dt' dt" k (t - t') k (t + T - t") (: f (t') f (t") :) w , (9 .11)

wh ere we used th e commutation relation [i; C+)(t') , i; C-)(t") ] = o(t' - t").

Notice that Eq. (9.11) is exactly the same as Eq. (9.9) with Eqs. (9.2) and
Combining Eqs. (9.7) and (9.8), we hav e the correlation function of the (9.10) when r; = 1. For r; < 1, we can also obtain Eq. (9.9) in this way by
photo -electric current as
introducing losses (see Section 9. 4).
Using Eqs. (9.4) and (9.11) , we have
(i(t)i(t + T)) = (
i ,J
k(t - ti)k(t + T - tj))
(!).i(t)!).i(t + T)) = j dt'k(t - t')k(t + T - t')(i(t'))w
= J dt'k(t - t')k(t + T - t')p 1 (t') + J dt'dt"k(t - t')k(t + T - t") (: !).f(t')!).f(t") :)w,

+ J dt'dt"k(t - t')k(t + T - t )p2(t' , t").

11
(9.9)
where !).f(t) = i(t) - (i(t)) . The first term in the expression above is the
(9.12)

so-called electronic "shot noise" in electrical engineering and is due to the

Here the probability density P2(t' , t") can be obtained
photo-detection theory [Glauber (1963b, 1964)] as 2
P2(t', t") = rJ2(i;C-)(t')EC - \t")i;C+\t")i;C+\t
= rl (:i (t') i (t") :) w.
When 'I]= 1, Eq. (9.9) with Eqs. (9.2) and (9.10) can also be obtained
directly from the current operator i(t) = J dt'k(t-t')i(t') introduced earlier
2 corresponds to the quantum noise of the coherent states (see Section 3. 2 .4).
See footnote 1.
from Glauber's discreteness of photoelectrons. It is proportional to the average intensity
of the optical field. As we mention in footnote 1, the formalism discus-
sed above can also be applied to Mandel's semiclassical theory of photo-
detection with classical wave fluctuations. So, the shot noise does not come
' ))w
from the quantization of light , as most theoreticians believe (see discussion
(9.10) about Eq. (9.33) in Section 9.4), but rather stems from the discreteness of
photoe lectrons. The second term, on the other hand, is from the fluctu-
ations of the intensity and is determined by the properties of the optical
field . For examp le, this term is zero for the multi-mode coherent states.
So, the measurement on coherent states only results in shot noise, which
For a classical field as defined in Section 5.2.4, the second term in Eq. (9.12)
274 Quantum Optics For Experimentalists E:rpe1·imental Techniques II: D etection of Continuous Photo-Cu1Te11.ts 275

is always bigger than zero and thrn, produces some extra noise above the e-ri 1'-t"I (r is the linewiclth of the fluorescence). So. the excess noise from
shot noise level. Some quantum fields, however, can make this term less the thermal field of the fluorescent light of an atomic gas is
than zero so that a sub-shot noise photoelectric current can be produced,
whose detected photoelectric current noise level is lower than those of the (6.i(t)6.i(t + T))e.r
coherent states. The quantum fields in amplitude squeezed states possess
this property (see the end of Section 3.4.3).
= ,l I5 J dt' dt 11k(t - t')k(t +T- t")e-rit'-t" I. (9.18)

For a continuous stationary optical field, the intensity of the field is a Its spectrum is then
constant: p 1(t) = 17(EH(t)E(+)(t)),:r, = 17!0.So. Eq. (9.4) becomes
/(r 2
Sex(D) = (77Io)2lk(D.)122r + D2 )
(i(t)) = 77/ dt'k(t- J
t')I(t') = 17!0 dt'k(t - t') = 171!0. (9.13) = SsN(D.)17Io2I'/(r 2 + D.2 ) > 0. (9.19)

It is always larger than zero and is proportional to the square of the intensity
The first term in Eq. (9.12) for shot noise becomes
of the field. This is the characteristic of the excess noise, which is different
(6.i(t)6.i(t + T))sN = j dt'k(t - t')k(t + T - t )p1(t 1 1
)
from the linear relation for the shot noise. The different dependence on Io
for the shot noise and excess noise is often used in the experiment to check
J
= 17!0 dt 11k(t 11)k(t 11 + T), (9.14)
if the intended shot noise measurement indeed gives the shot noise level or
there is some extra noise. Adding Eq. (9.19) to the shot noise, we have the
where t" = t - t'. Hence, the spectrum of the shot noise is total measured noise level as

j
SsN(D.) = 17!0 dTdt 11k(t 11)k(t 11 + T)ejflT
(9.20)

This is the spectrum of the photocurrent when we measure a thermal field.

= 17Iolk(D.)12, (9.15) Figure 9.2 shows a typical result of the spectral analysis of a photocurrent.
where k(D.) = J dTk(T)ejwT is the response spectrum of the detection sy-
stem, which is determined by the spectral response of the detector and the
subsequent amplificat ion system. Notice that the shot noise level is linearly
proportional to the intensity Io of the field. This is the characteristic of the
shot noise. 0)
-80
For the second term in Eq. (9.12) , we have ,_J
(l)
C/J -85
\: 6.f(t 1 )6.f(t 11
= I5[g(2 )(t 1 - t") - 1]. ·5
): ) (9.16) z
Hence, the second term of Eq. (9.12) becomes

J
-95 t__ __________________ _

\ 6-i (t) 6-i (t + T))ex = 172I5

0.5 1.0 2.0 2.5 3.0 3.5 4.0
dt' dt" k (t - t') k (t +T- t 11
)
0 L5

2
X [g( \ t 1 - t 11)
which is determined by the int ens ity fluctuations of the detected optical
field and is sometimes called the "excess noise". For example, we know
from Section 4.2.4 that thermal fields have g( 2 )(t' - t") = 1 + 1,(t' - t")j 2 .
For the fluor escent light from an atomic gas ensemb le, we have j--y(t'-t 11 )12=
Frequency (MHz)
- l], (9.17) Fig. 9.2 A typical result of the spectral analysis of a photocurrent: (a) the electronic
noise level without the illumination of light field on the detector. This is produced by the
dark currents of the detector and the noise of the electronic amplifiers. Notice the larg e
1/ f noise near zero frequency. (b) The situation after the illumination of light field on
the detector: the flat part is the shot noise leve l (SN). The sharp spikes, especially those
at low frequency are from the excess noise due to intensity fluctuations of the light field.
They may a lso be the input signal modulated at a specific frequency and its harmonics.
The vertical axis is in log-sca le. Courtesy of Jun Jia.
 276 Q,uantwn Optics For E:rpe1·i.mentalists
E:rperime 11tal Techniqu es II: D etrctio11 of Contimwus Photo-Cul'r cnts 277

9.3 Homodyne and Heterodyne Detection Techniques

When the optical field to be detected is very weak, the electronic noise
of the measurement system such as dark currents and amplifier's pick- up
noise ·will overwhelm the photoelectric signal if we make a direct detection
of the optical field, Since the photoelectric currents are continuous, it is
impossible to use the pulse shape for discrimination like we did in the
photon counting technique, On the other hand, we can apply the technique
of homodyne or heterodyne detection to increase the photoelectric signa l
over the background electron ic noise. To see how it works, we first consider
the single-mode case .
Fig. 9.3 Homodyne and heterodyne detection. An in cid ent wea k signal field ain is
mixed with a strong local oscillator field £ with a beam sp litt er of reflectivity near zero
(R « 1) so that all the signal field is detected with out atten u ation.
In homodyne and heterodyne detection schemes shown in Fig. 9.3, a
weak input signal field is first mixed with a loca l oscillator field (LO) be-
fore being measured by the detector. LO is usually a strong field in a
coherent state and thus can be represented by a c-number [ = IEleJ'P_It
is much larger than the signal field: IEl2 » fin = (aJnain)- It is called
homodyne detection when the center frequencies of the two fields are the
same . Oth erwise , it is het erodyne. After the mixing by a beam sp litt er of
nearly zero reflectivity, the field to be measured by the detector is
'.\,1lele. XA.111 (1.
y 0) =
- at e.i'P+ a e- jr..p is the quadrature-phase
111 Ill amplitude of
the input field with its phase determined byAthe phase <p of ~he LO and
XAr_ (0) = at + Gin = X ex i. Xin(1r/2) = (ain - ain)/j = y ex j3 with
171 lll . f 1
.iA
,.pArespectively' rer)resenting the position and momentum operators o t 1e
virtual harmonic oscillator for the single-mode input signal field.
So, apart from a constant term, the intensity measured by th~ detector
is proportional to the amplitude-phase quadrature of the input signal fi~ld
and the proportional constant is just the amplitude of the LO field, which
can be controlled in the expe rim ent . Making it large can lift the overall
detected signal far above the electronic noise background. In this _case. even
though the signal from the direct detection of the original weak mput field
is small and overwhelmed by the electronic noise background, homodyne
detection with the aid of a strong LO field can still detect it. Here , because
the output from the homodyne detection is the weak input field multiplied
by the amplitude IEI of the LO, the original signal is _amplified ~y the
involvement of the LO field. Of course, this amplification effect is only
on the optical part but the electronic noise background is not amplified.
Obviously, we cannot make the amplitude of the LO infinitely large, for the
detector will eventually be saturated by the dominating LO field.
The first constant term in Eq. (9.22) can be filtered out by spectra l
analysis since it is at zero frequency. In the spectral analysis ~f t~e pho-
tocurrent, the frequency of our interest in the input signal field 1s diffe~ent
from that of the LO field. So, to treat this case, we need to consider
multi -mod e situation. In this case, the LO field is still a single-mode field
in the coherent state, set at frequency w 0 . The multi-mode treatment ~s
on the input signal field, which is described as a one-dim ensional quasi-
monochromatic field E}:\t). Similar to the single -mod e case, we assume

a= ain + [,
IEl2>> (ft;;;,)ft}:)). The field seen by the detector is then
(9 .21)
where we only consider the homodyne detection scheme and absorbed the
(9.23)
reflect ivity coefficient R into [ which denotes the LO part after the beam
splitter. Since R rv 0, the input signal will almost completely transmit
through the beam splitter without loss: the detected field contains all the where E}:\t) is in the form of Eq. (9.3) but with ii(w) replaced by Clin(w)
information of the input field. So, after dropping the small higher order and the LO field is in a coherent state. 3 Here, because of the normal
term, we have ordering encountered in the photo-detection theory described in Chapter
IEl2+ [ ain 5, we replace the field operator of the LO field with a constant . Then, the
aAtA
a= At
+ [* ain
A At A
+ ain ain
2
IEl + EaJn + [ * ain 3Here, we assume the signal field has the same sp~tia l m~de as th~ LO_ field so that

= IEl2+ IEIXin(cp),
we can drop the spatial mode. In Section 9.6 , we will conside r the situation when the
(9 .22) spatial modes are different.
 278
Qirnntu ,rn Optics Fo , · ExpC'rimenta lis ts
E:rpC'ri.m.ental Techniques Il : Dcif'ction of Contirmous I'holo-Cu.nents 279
intensity m eas ured by the detector is
Io -= (i;(-)j;(+)) Because of the normal ordering, the second term is zero for vacuum. So,
2 th e shot noise from the first term is also regarded as the vacuum quantum
= /E/ + E*eJwot(Ei\~)) + Ee-Jwot(i;(+)) + (E(-) j;(+)) noise although it is ind epend ent of the iuput fielcl. Theoreticians usuall y
;::::;
/E/2. zn rn in
prefer this association and this viewpoint can be furth er confirm ed by the
Henc e, from Eq (9 14) 1 ,1 . (9.24) following arrangement.
. . ' t le s 1ot nois e of the homodyne detection is

with its spectrum as

(6.i( t )6.i (t + ') )SN ;::::;17/E/
2 I dt" k( t")k( t" + 7 )
(9.25) 9.4 Vacuum Noise and Beam Splitter Model of Losses

2 When the quantum efficiency 17= 1, using [Et) (t'), E}:\t 11 )] = 5(t' - t") ,
. . . SsN(w) = 77/E//k(w)/2. (9_
26 we have (6.X t 1 (t')6.X fn(t 11 ) ) = (: 6.X frJt')6.X t 1 (t 11 ) :) + b(t' - t"). With
This is mdependent of the input signal field. The part tl . t . . l )
the input signal fi ld f 1a is ie ated to this, Eq. (9.32) is changed to
. e comes rom the second term of Eq. (g .12) U .
Eq. (9.23) and keeping only terms up to /E/2' we have
t' t")
P2( ,
')I 3 [/E/+(Xi~(t'))+
;::::;77-E/
A
. smg
(Xi~,,(t"))]
(6.i(t) 6.i(t + ,) ) = 1£1
2
J
dt 1dt k (t - t')k(t - t" + ,)(6.X'fn(t')6.X'fn(t
11 11
))

==(6.Z(t)6.Z(t + ,) ), (9.33)
I ~772l;l2[(: xi~ (t')Xi~ (t") :) + (fin (t') ) + (fin (t"))], (9.27)
II

P1(t )P1(t ) ;::::;

77/El [IE/+ (Xfn(t ')) + (Xi~(t"))] where Z(t) ==/ElJ dt'k(t-t')X'fn(t'). If the input signal field is in vacuum ,
we have (6.X'fn(t')6.X'fn(t 11 ))vac = b(t' - t"). This is the result of quantum
+772/E/2[(Xi~(t'))(Xi~(t")) + (iin(t')) + (i (t"))] (9.28) fluctuations of the vacuum fields. Substituting it into Eq. (9.33), we then
Here, in ·
obtain the first term of shot noise in Eq. (9.32). That is why theoreticians
xi~ (t) ==Et)(t)ej(cp-wot) + Ett\t)e-j(cp-wat) like to regard the shot noise as the contributions from vacuum quantum
Notice that in th · · (9 -29) fluctuations although photo-detection theory based on classical waves can
A'P e express10n above, the frequency is shifted by w th t
Xin(t) is a slowly-varying part of the field Th' b o so a also give rise to the shot noise (see discussion about the origin of the shot
E (9 3) t is can e seen by using
o write out Xi~(t) explicitly as
A •

q. . noise in Section 9.2). From Eq. (9.33) , we find that similar to Eqs. (9.4)

X,';,(t)= I dw[ii],.(w)ej(,p+wl-wot) + a,n(w)e-j(,p+wt-wot)l

and (9.11), homodyne detection realizes a quantum measurement of opera-
tor Z(t), which is the Hermitian operator corresponding to this quantum
measurement process.
v27r/dn[ ain
= _1_ At(Wo + ")ejc.p+ J-ru + ain(w
H
A
0
. .
+ D)e-J'P-JDt], (9.30) When the quantum efficiency 77< 1, we can rewrite Eq. (9.32) as
where we shift the frequency to n = w - . tl .
. - wo m 1e mtegral So th · (6.i(t)6.i(t + ,) )
no f ast oscillating terms like eJwot in X'P (t) C b. .
(9.28), we have
. ' ere is
in · om mmg Eqs. (9.27) and = ?] I£I2 J dt' dt 11 k (t - t') k (t - t" + T)
P2(t', t") - P1(t')P1(t");::::;
772/E/2(:6.X 'P(t')6.X'P (t") ·) (9 )
A

with 6.X 'P(t) = X'P ( ) 'P in A in · .31 X [(1-?J) (6.X::)(t')6.X::)(t"))vac+ 77(6.X'fn(t')6.X'fn(t

11
))]. (9.34)
h t ~n - in. t - (Xin (t)). Hence, for homodyne detection the
p o ocunent correlat10n function in Eq · (9 •12) is
. ch ange d to ' Here, Xv is some ind epen dent mode in vacuum. So, there are two contri -
butions in the express ion above: (1 - 77)(vacuum noise) + 77(input field).
( D,.i (t) D,.i (t 77/E /2 / dt' k (t - t') k (t - t' + 7 )
+ ')) ;::::; The first part is the contribution from vacuum of some independent mode
2 2
+77 /E/ I dt' dt"k(t - t')k(t _ t" + ,)
Xv whereas the second part is the contribution from the input field. The
coefficients in front of them indicate that they are coupled through a beam
splitter with transmissivity 77as shown in Fig . 9.4: one input side of the
X (: 6.X in'P(t')6.X in'P(t") ·)
.. (9.32) beam splitter is the vacuum field while the other is the input field. This
 281
E:rpe?"imC'ntal Techniiques JI: Dr.tec/i,on of Continuous Photo-CurrC'11ts
280 Quantum Optics For Expcrimt>ntalists

is determined by the input field and is independent of the LO field. The

, Non-ideal Detector 17< I LO field only determines the size of the shot noise level, making it much
~--------------------------7

I\
a
I\
: t I -rJ
I I
Ideal :
I
higher than the electronic noise. Equation (9 .J 8) indicates that the :::;pectral
function of the photocunent, after being normalized to the shot noise, is
[[.
111 : Detector :
I only related to \ 'P(D) of the input field, whose ratio to the vacuum noise
:T=rJ<l 17=1 :
'------------------------- I
(the term of 1 in Eq. (9.38)) is the signal-to-noise ratio of the homodyne
detection. Therefore, we also regard homodyne detection as a quantum-
Fig. 9.4 The beam splitter model for non-ideal detector wi .
due to losses. Vacuum noise is coupled in through tl 1e unusetdh port.
quantum efficiency 'r/ < 1 limited measurement since its noise is purely of quantum nature (vacuum
quantum fluctuation). So, it can be used to measure the quantum corre la-
beam ~plitt er model shows that the less-than -p erfect quantum efficienc
1;s caused by the loss, that is, t he mixed field of the input and
e s oes not enter the detector complete ly. (1 - ?1)p t O f 1·t .
16 tion of optical fields.
To find a specific form for x:'P(D), we make a change of D -t -D in the
first term in the expression for Xt 1 (t) in Eq. (9.30) and rewrite it as
out and tl 1 A ·' ar is coupled
ms ost. nd vacuum noise is introduced through loss.
On the other hand , the beam splitter model above raises a que f
abo utb the
Xi,,(t) = J d[! [ at(wo - n)eN e-j<lt + ii(wo+ n)e-j~ e-,rn l
. E quantum noise from LO: when we replace it w1'tl 1 a non-operator
s 1011
nun~l :1 '
.
we did not consider the contribution of quantum noise from LO
at _a m Eqs. (9.33) and (9.34). Is it legitimate to not consider the quant
= J dnX~(n)e-j!lt, (9.40)

·th Rfrom LO? In fact : since (6 .X) co h -- 1 an d we use a beam splitter

nmse 2 um where _xcp (D) = at(w0 - D)eJ'P + a(w0 + D)e-jcp is the multi-mode

R(6 2 X) « 1- to couple m the LO field ' 1·t s quantum noise contribution is

wi quadrature-phase amp litude given in Eq. (4.74) of Section 4.6.3. With
coh - R-+ 0. So, no need for quantum noise of LO. Eq. (9.40), we can calculate the correlation function

9.5 Spectral Analysis of Homodyne Detection (6.X fn(t')6.Xt-Jt "))

= 2_ ;· dD'dD"(6X'P(D')6X'P(D"))e-j(D't'+D"t")
fMaking
· a Fourier transformation of Eq · (9 ·32) , we o bt am
J
. t h e output 21r
unction of current spectral analysis for the homodyne detection: = 2_ dD' dD" (6X'P(D ')6X'Pt (D"))e-j(D't'-D"t"). (9.41)
SHD(D) = j dT(6i(t)6i(t + 7 ))ejDT_ (9.35)
21r
The second lin e is due to (X'P(D)}t = _K'P(-D). Since (6Xfn(t')6Xt-Jt"))
For a stationary field, we may define only depends on T = t' - t" for continuous waves, we must have
(: 6 Xfn(t') 6 Xfn(t") :) = x'P(t' - t"). (9.36) (6X'P(D')6X'Pt (D")) = Scp(D')c5(D'- D"). Substituting this into Eq.
Equation (9.35) is then changed to (9.41), we have
2
SHD(D) = TJIEllk(D)l 2 [1+ TJX'P(D)]
= SsN (D) [1 + rJX'P(D)] (9.37)
(6Xt-,,(t')6X fn(t")) =
2
1
7!" J
dD' Scp(D')e-jD'(t'-t"). (9.42)

or To find the quantity in Eq. (9.36), we use the relation

(9.38)
(6Xfn(t')6Xfn(t")) = (: 6Xfn(t')6X frJt ") :) + c5(t'- t"). (9.43)

With this and Eqs. (9.36), (9.39) and (9.42), it is straightforward to find
where
(9.44)
(9.39)
 283
. . l T I . es [[· D etection of Contin'Uous Photo-CuTT cnt s
Expcrimenta ec iniqu .
282 Q'Uantum Optics For Exp e1·im cnlal'ists

-50 .-----------------7
Substituting this into Eq. (9.38), we find the spectral function for homody n
detection
-t-0
(9.45) (i)

So. we just need to calcu late (6..X'P(S1')6..X'Pt (O")) for the input field in
order to find the spectral function in homodyne detection.
As an example, we consider the homodyne detection of the multi -mode -75 L------- 3-_o
-- :,_,
-:-- -1
:-;:
_o
-- -1
:--;
_,--; s.o
1.0 t .5 2.0 ~-S
squeezed state discussed in Section 4.6.3. From Eq. (4.76), we have Frequency (MHz)

( ) f homod ne detection of a multi-mode squee-

(.X (S1).Xt(St')) = Sx (S1)5(S1- O'), Fig. 9.5 Spectra of photocurrent SHD o .. y (w) for = 0o; (iii) SHv(w) for
zed state: (i) spectrum of the shot noise; (u) SH D <p
(Y(D)Yt (O')) = Sy(S1)5(S1 - St'), (9.46)
<p= 00 + 1r /2. Courtesy of Wei Du.

where Sx(D) = (IG(D)I - jg(S1)1)2 < 1 for noise reduction, but Sy(S1) =
Mode Match and Local Oscillator Noise in Homo dyne
(IG(D)I + fg(S1)1)2 > 1 for noise amplification . For the multi-mode squeeze d 9.6
state, we have (.X'fn(t)) = 0. Hence , Scp(D) = Sx(D) for <p = 00 and Detection
Sy (0) for cp = 00 + 1r /2. Then the observed spectral function of homodyne . . 1 d ren detection, we did not consider
detection for the multi-mode squeezed state is In all our previous discussions on 1.omo J . t . 1 field
. . b e assumed that the mpu signa

!;:~~;
the spat ial n1odes. This is ecause w_ H we know that it
= (1 - r1) + '7(IG(l1)1± lg(l1)1) .
2
(9.4 7) an:!nth:
~~}it~~:;:~e!~~y
s:::c~i:::a:p:::t:od:;:;:~
3
optical
fields
in
is p . t vVe will consider this non-ideal case m the followmg.
Here "-" is for cp = 00 and "+" is for cp = 0o + 1r /2. When Tl = 1, the e~penm:~~~l modes are included, we cannot use ~he l_-dim a~pr~xima-
the spectrum of the detected photocurrent, after normalization to the shot
noise, is the spectrum of squeezing Scp(w). When Tl < 1, the first term in
Eq. (9.47) is obviously the contribution of vacuum so the vacuum noise is
Smce sp .
tion. Assume the optical .e
pendicul~r t~ the cross -section
tl:
fi ld ro agates along z-direction which is per-
detector and the field has a transverse
° .
ft(+) = ft(+) (x, y , t) (we ignore the z
coup led in and reduces the amount of noise reduction effect due to squee - distribution m. x - y plane, that is\nterested in the cross -section of the de-
zed state. Figure 9.5 shows experimentally observed spectra SHD(Sl) of coord inat e here because we are only t . e the transverse distribution
photocurrent at cp = 00 , 0o + 1r / 2 and the shot noise spectrum S s N ( S1) for tector). The spatial modes of t~e field de er~ift(+) (x t) = ft(+) (t)u(x, y)
. 1 e and we can wnte the field as ' y'
. d J d d \ ( . y) 1~
?
the squeezed state detection . Notice that log-sca le is used in Fig. 9.5 so m x - y p an
. th node function and is normalize : x y u x'
= 1.
that the ratio of SHD (S1) to Ss N (S1) is simp ly the difference between the where u (x, Y) is er - E (5 26) d (5 28)
. th Glauber photodetection fonnulas m qs. . an .
two curves. For cp = 00 , SHD(Sl) is lower than the shot noise level SsN(Sl) In this case, e
by about 3 dB at around 2.5 MHz, realizing a quantum noise reduction are changed to
by a factor of 3 dB = 50%. Notice from Eq . (9.47) that when Tl = 1,
Sx(S1)Sy(S1) = (IG(S1)12 - jg(S1)12)2 = 1, which means noise reduction
J
P1(t) = Tl da(ftH(x,y,t)ft(+)(x,y,t))
(9.48)

and noise amplification should be the same amount in log-sca le. However,
Fig. 9.5 shows different amounts for reduction and amplification. This is
due to Tl < 1. From Fig. 9.5, we find the amount for reduction is -3 dB and
p2(t, t') = T72 JJdada' (: i(x, y, t)i(x'' y'' t') :)
(9.49)

that for amplification is 8 dB. From Eq. (9.47), we can deduce the efficiency
where da = dxdy, da' = dx' dy' ·
coefficient Tl = 0.55.
 28--! Q·llantmn Optics For Expe1 ·im,ental-ists
E:rperimental Techniqlle.s II: Detection of C'onti.n.1wus Photo-Cu.rrents 285

In homodyne detection, if the input field and the LO field have different
the LO is proportional to the square of the LO intensity (IE/2 ) whereas
spatial modes, denoted by u1 ,2 ( .r, y), the superposition field is then
the shot noise is only proportional to the intensity of LO. These intensity
jj;(+)(;c,y,t) = E 1(,,~)(t)u1(x,y) +Ee-Jwotl 12(.r.y). (9.50) dependent properties of noise can be used to check if the shot noise level
where J daluu(x,y)l 2 = l. From this, we can easily find obta ined in the experiment. i.e .. the vacuum noise level, is truly of the shot
noise nature or has some contribution from the intensity fluctuations of the
j daft(-)(x,y,t)ft(+\x,y,t) /E/2 +,8/E/X 1~(t), (9.51) LO field.
The excess noise of the LO field originates from the laser noise. Diffe-
where ,8 = I J daut(x, y)u2(x, y)/ and we drop (E;:;\t)Ei(,~\ t)) because its rent lasers have different noise spectra. Solid state laser s like Ti:sapphire
contribution is much smaller than the other two terms. Using Cauchy- and YAG lasers have noise usually in the range within 1 l\IHz. So. their
Schwarz inequality, we have ,B s; 1 (the equal sign stands for u 1 (x, y) = noise is not a big problem when our working frequency is usually at a few
u2(x, y)) . Substituting the above into Eqs. (9.48) and (9 .49), we can prove 1\IHz. However, the commonly used semi-conductor lasers have a wide noise
that 77is changed to 77/ ,8/2 , that is, the mismatch of the spatial modes leads spectrum ranging from 10 to 100 MHz. Fortunately, we can use the balan-
to the reduction of quantum efficiency. (Readers who are interested in this ced homodyne detection technique discussed in the following to cancel the
can prove it as an exercise .) excess noise from the LO [Yuen and Chan (1983)].
Although the argument above is for spatial modes, it applies to the mis-
match of other modes, such as temporal modes in pulsed fields. In fact, the 9. 7 Balanced Homodyne Detection
underlying physical principle of homodyne is the interference between the
input field and the LO field and the mode match parameter ,Bis equivalent As shown in Fig. 9.6, balanced homodyne measurement is achieved with
to the visibility of the interference fringe. two detectors. We use a 50:50 beam splitter to combine the input field
In homodyne detection, there is another issue we need to address that and the LO field and evenly illuminate on the two detectors. We measure
is, the effect of the intensity fluctuations of the LO field. We can s~e its the photocurrent difference from the two detectors. A simple single-mode
effect from the simple single -mod e model. Assume the intensity fluctuation model gives
of the LO field is from the amplitude of the LO field: E -+ E + 13.E.From a1= (ain + a2 = (E - (9.54)
Eq. (9.22), we have So, the difference from the two detector outputs is
2
ata ~IE+ 13.E/ +IE+ 13.E/Xin(<p) a!a 1 - a!a2 = IEIXin(<p). (9.55)
2 We find from the above that the LO intensity part is cancelled. Hence, its
/E/ + /E/[Xin(<p)+ 213.E]. (9.52) fluctuations will not contribute to the output of the balanced homodyne
Here, we only take the first term of 13.E. Since E is really large, a small detection.
ion of 13.E
relative fluctuat A / E can make 13.Every lar ge, which is added to the
signal term of Xin ( <p)to become extra noise.
In the exper im ent, the intensity fluctuations of the LO field will show
up in the shot noise spectrum, i.e., the photocurrent spectrum without the
input signa l. This is equiva lent to the case of direct illumination of the
detector by only the LO field. From Eq. (9.17), we obtain the excess noise
as
2 2
(D) = (77(1El ) )2/k(D) / h(D) = SsN (D)77(/E/
Be.1: 2
)h(D), (9.53)
where h(D) = J dT[gl,6(T) - lJe-JS-h gives the spectrum of the excess noise
2

after being normalized to the shot noise. Notice that the excess noise from Fig. 9.6 Balanced homodyne detection scheme.
 E:rperim('ntal Techniques II: Dclect-io11 of Ccmti1111ons Photo-C'llrrents
287
286 Quantum Optics For Experimentalists

For multi-mode case, we have ·s completelv, I unbalanced and we ,vill find a modulation signal at Jo, as
IL
shown in Fig. 9.7(a). Now gradually increase ,\ to see the reduction of the
(9.56) modulation signal at J0 . as shown in Fig. 9.7(b). Fine tune,\ to minimize
We first assume the detectors are identical, i.e., 171= 172 = 17,k 1 (t) = k 2 (t) == the modulation signal at J0 . The difference between the case of ,\ = 0
k(t) and obtain the same result as the single-mode case: and the minimized modulation tells us how well the scheme is balanced. A

L = il -12 = JdTk(t-T)17[Ei-\T)Ei+\T) - E~-)(T)E?\T)]

good balancing scheme usually leads to 30 dB reduction of the modulation
signa l.

= JdTk(t - T)77IEIXi~\T). (9.57) modulation-..::---

....
ignal i
-- -- -7
If the two detectors have different quantum efficiency: 77 1 > 17
2 but have the . • •. > ',,~ "' • ~, ,' 11
I

same time response 4 : k 1 ( t) = k 2 ( t) = k (t), we can balance them by slightly

§ -~5
misaligning the detector of higher 77 1 to reduce it to match the lower one: z _,,, \
77~= 172. We can also achieve balance by selecting a non-50:50 beam splitter _.,L._·~
_; ___________ _
with 172T = 771 R:
n
Frequency (MHz)
Frequency (MH7)

(b)
(a)
(9.58)
Fig. 9.7 Spectra of homodyne detection for (a) unbalanced case and (b) nearly balanced
We also obtain similar result to Eq. (9.57)
case. The modulation signal disappears for the completely balanced case .

i_ = i1 - i2 = J dTk(t -T)[771Ei-)(T)Ei+\T) - 772E~-\T)E?\T)]

= J dTk(t - T)~IEIXi\~)(T). (9.59) 9.8 Intensity Fluctuations and Self-Homodyne Detection

But the quantum efficiency is changed to 77eff =~'which is better than
the misalignment technique.
For the most general case of different time response and quantum ef-
ficiency for the two detectors, we cannot achieve balancing in the whole
spectral range . But we can balance them at one specific frequency. This
requires balancing the amplification gains of the two photocurrents, that is,
i_ = >-1i1- >-2i2ex:i1 - Ai2, where,\= >-2/,\ 1 is the relative amplification
gain.
At one specific frequency, the photocurrent fluctuation of each detec -
tor is proportional to the intensity fluctuation of the optical field at that
frequency. Hence, we can adjust ,\ to cancel the contribution to the photo-
current difference i_ from the LO intensity fluctuations at that frequency.
The following are the specific steps.
First, we apply some intensity modulation at some specific frequency
Joon the LO field (usually by an electro-optic modulator) to simulate the
intensity fluctuations. Then, we set ,\ = 0 so that the detection system
4
This is usually the case when we use detectors made in the same manufacturing
Measurement of intensity fluctuations is actually more common than ho-
modyne detection in the experiment. A direct detection of an optical field
will make a measurement of the intensity of the optical field. But as we
will see in this section, when the average intensity of the measured opti-
cal field is high, the measurement of intensity fluctuations is equivalent to
the homodyne detection of one particular quadrature-phase amplitude with
the local oscillator being the coherent component of the optical field itself.
Thus this case is also known as self-homodyne detection.
As' a matter of fact, we already encountered this in Section 3.4 .3 when
we discussed about coherent squeezed state: when we choose the squeezing
angle 0 so that it is related to the phase angle <fa of the coherent component
by 0 = n+2<pa, we obtain the amplitude squeezed state. When we measure
intensity (photon number) of the field in this state, we obtain a reduced
quantum fluctuation in intensity (see Eq. (9. 79) of Problem 9.2). Amplitude
squeezed states is a special kind of quadrature-phase amplitude squeezed
state . In the measurement of intensity, the coherent component of the field
run.
acts as a local oscillator for one particular quadrature-phase amplitude.
 288 Quantum Optics F01· Experimentalists Experimental Techniques II: Detection of Continuous Photo-CurTcnts 289

To see mor e clearly the role of coherent compo nent in self-homodyne de-
tection , let us corn,ider a single-m ode field with a large coherent component:
(a) = rei'-Powith r » l. Define a field fluctuation operator !:ia = a - (a)
and we h ave
a,ta, = f(a)f2 + (a)!:iat + (a)*!:ia + !:iat !:ia
r2 + r!:iX(<.po),
where we dropped the sma ll quadratic field fluctuation term because it is
usu ally of the order of one , which is much small er than r 2 . !:iX (<po)
!:iae-i'-Po+ !:iate i '-Po. Comparing the above equation with Eq . (9.22), we
see that the coherent p art (a) acts exact ly as [, the am plitud e of the local
oscill ator for homodyne detect ion and the quadratur e-ph ase angle is fixed
at <po,the ph ase of the coh erent compon ent (a) = rei'-Po
. In particul ar , if the
coherent component (a) is real, intensity measurement always corresponds
to homodyn e detection of X = a + at, which is the ampl itud e of the field.
The argument above is for the single-mode case. With the multi-mode
case, let us go back to Eqs. (9.9) , (9.14) and (9.16) for the full treatment
of photo-detection:
(£ii (t) £i i (t + T)) = (£ii (t) £i i (t + T))SN + (£ii (t) £i i (t + T))ex
Comparing to Eq. (9 .32), we find that the equation above has exact ly the
same form. So, when the measur ed field has a larg e cohere nt component,
the dir ect int ens ity fluctuation measurement is equivalent to a homodyne
meas urement of one particular quadrature-phase amp litud e of the measur ed
field. Intensity fluctuation measurement is t hu s somet imes known as self-
homodyne.
(9.60) vVe will come back to th is in Sect ion 10.1.4 when we discuss abo ut
quant um corre latio n in int ensit y for twin beams and its relatio n to Einstein-
= Podolsky-Rosen ent angl ement .
9.9 Photo-detection for Ultra-Fast Pulses
9.9.1 General Consideration
In qu antum optics , generation of nonclassical states of light oft en reli es on
nonlinear optical int eract ion , which requir es high power to reach strong
interaction. This leads to pulse lasers , which usually have extremely high
peak power. Commonly used pulse lasers are mode-locked lasers with a few
100s of femto -second pulse width and a few tens of MHz repetition rate.
Th e quantum fields produced with these lasers also have simi lar temporal

= J dt'dt"k(t - t')k(t + T - t")rJ2(: !:ii(t')!:ii(t"):) profiles. For example, we treated in Section 6.1.5 the case of pulse-pumped
para m et ric processes, wher e the modes of the fields are temporal modes.

+ j dt'k(t - t')k(t + T - t')77I0 (9.61)

In this section, we will discuss photo-detection of this type of fields.
Compared to the continuous cas e, th e main difference is the detector
Now similar to the single -mode case, we write !:iE = E - (E) with the response bandwidth versus the bandwidth of the optical fields. In the con-
coherent component (E) = reii_po -i wot (w0 is the center frequency of the tinuous case, optical bandwidth is well within the response bandwidth of
2 2
measured field) and r » (ti E), i.e ., coherent component is much larger the detector and the method of spectral ana lysis provides the spectrum of
than the fluctuation of the field. Then E = reii_po-iwot + !:iE and we can the optical field. In the pulsed case, however, the ultra-short optical pul-
make the approximation for the intensity operator: ses are too fast for the detector to respond and the photocurrent is then
a tim e average of optical pulses. Furthermore , due to the characteristic
f(t) = jj;t (t)E(t) = r 2 + r(!:iEt eii_po-iwot
+ !:iEe-i'-Po+ iw 0 t) + !:iEt !:iE
of mode-locked lasers, the ultra-short pulses rep eat with a repetition rate
2
+ r!:iX'-Po(t).
r (9.62) of a few tens of MHz (80 MHz for a typical Ti: sap hir e laser) and form a
Here !:iX'-P(t) = !:iEtei i_po-iwot
0
+ !:iEe-i '-Po
+iwot_With Io= (i(t)) = r(r + quasi -cont inuou s field. So, the photocurrent has a form of
r 2 , we have lif(t)
CX)
(!:iX'-Po
(t))) r!:iX'-Po
(t) and Eq. (9.61) becomes
i(t) = I: k(t - nTrep)In, (9.64)
(Cii(t)Cii(t + T)) n=- =

= 77r
2 2
J dt' dt" k(t - t')k(t + T - t") \ : !:iX'-Po
(t')!:iX'-Po(t") : )
where Tr ep is the time between pulses, k(t) is the response function of the
detector and In is the total photon number of then-th pulse:

+77r J
2
dt'k(t - t')k(t + T - t'). (9.63) In = J dT(Et (T)E( T)n- (9.65)
 290 Quantum Opti cs For E:tperim e:ntalists
Experimental Techniqu es II: D et ection of Continuous I'hoto-Cnrr ents 291

In may change from pulse to pulse due to intensity fluctuation or mo- IEI> > 1. The output current difference operator is similar to the con-
dulation. This fluctuation may have classical or quantum origin and th
tinuous case in Eq. (9.57) but the time integral is over the pulse due to
modulation is usually the signal encoded in the pulses. If the detector':
Eq. (9.65). So, the current difference operator in this case takes the form
response time TR is much long er than Trep, Eq. (9.64) can be approximated
as

i(t) J d,k(t - ,)I(,), (9.66)

of

L (t) = k(t) I:[ E£o(T)E\n(T) + h.c] dT

which is in the same form as the continuous case in Eq. (9.4). Equation j
= k(t) IEI dw [A~ 0 (w)ain (w)e- i'P+ h.c.], (9.68)
(9.64) can be viewed as the discrete version of Eq. (9.4). Then, we can use
where the input field is a quantum field described by operator
spectral analysis method to measure the fluctuations of the optical pulses
just like the continuous case.
E in =
A 1 / A
ain
(
w )
e -iwtd w. (9.69)
On the other hand, if TR is comparable to Tr e p, spectral analysis of
the photocurrent will show strong frequency components at the repetition Here, we assume the spatial modes are matched and only concentrate on
frequency and its harmonics [Slusher et al. (1987)]. These frequency compo - the temporal part of the fields by using the quasi-monochromatic and one-
nents can be very large to overwhelm the subsequent electronic amplifiers . dimensional approximation for the input field. Equation (9.68) only descri-
So, they need to be handled right after the photo-detectors. Two methods bes a single pulse contribution and shows that the spectral property of the
can be used. The first one is to directly use a good low pass filter(> 100 dB) photo current is determined mainly by the detector response function k (t).
to block out these strong frequency components. The other is to use two The field fluctuation is then multiplied on all frequency components in the
nearly identical detectors and perform a balanced detection of the current spectral distribution and is exhibited as pulse-to-pulse variation per expla -
difference of the two detectors. The latter method is commonly employed nation of the general expression in Eq. (9.64). Hence , quantum average
in intensity difference measurement for the detection of twin beams and in should also be understood as pulse-to-pulse average.
balanced homodyne detection where strong modulation in the LO field is The pulsed input field is usually described by a set of orthonormal tem-
canceled in the difference of the photo-currents. We will address balanced poral mode functions characterized by {</>j (w)} (e.g., the fields from pulsed
homodyne detection of pulsed fields here. -pumped parametric processes in Section 6.1.5) and the LO field amplitude
A LO can be decomposed as
9.9.2 Homodyne Detection of Pulsed Fields
Aw(w) = L ~j</>j(w), (9.70)
Although the spectral analysis of the pulsed case is the same as the conti- j

nuous case after blocking out the repetition frequency and its harmonics, where the coefficient ~j is given by
there is one more important issue that does not occur in the continuo us
case, that is, the temporal mode match. We have seen in Section 9.6 how (7 = l~jleieJ = J Aw(w)<f>;(w)dw (9.71)
the spatial mode mismatch can lead to poor detection efficiency. The effect
with L j l~j 2 = 1. Substituting Eq. (9. 70) into Eq. (9.68), the output of
1
is the same for temporal mode match, which arises only in the pulsed case
because of the time integral in Eq. (9.65). homodyne detection can be rewritten as
Suppose that the LO field is a transform -limit ed pulse in the form of i_ (t) = k(t) IEI L ,~jIXj (0j + <p) (9.72)
j
. 1 / Aw(w)e -iwt dw,
Ew(t) = IElei'P (9.67)
where Xj (0) is the quadrature-phase amplitude operator for mode j and is
with ALo(w) satisfying the normalization condition J IALo(w)l2dw = l. <p defined as
is the phase of the LO field and the amp litud e of the LO field is strong: (9.73)
 292 Q'Uantv.m Optirs Fm· E:rpe·1"imentalists E:rperimental Teclmiques U: Drtection of Continuous Photo-Currents 293

with
Herc. ·we used Lj /(J /2 = 1. So, the consequence of mode mismatch /(1 /2 <
(9.74) 1 it:>equivalent to a loss of 1 - 16/2 or a drop of the quantum efficiency
by a factor of /( 1 /2 . This is exactly the same as the case of spatial mode
as the annihilation operator for temporal mode j [¢J (w)] first defined in mismatch in Section 9.6.
Eq. (6.38) in Section 6.1.5.
On the other hand, if there arc multiple modes excited in the input field
A Notice that the phase of the measured quadrature-phase amplitude as in the case of pulsed parametric processes, the situation becomes com-
Xj (0j + <p) depends not only on the overall phase <pbut also on the phase 0. plicated. Suppose we are looking for quantum noise reduction. Assuming
J
of quantity (j, which may be different for different mode j. Furthermore . there is no correlation between different modes, we obtain from Eq. (9. 72)
2
/(j / can be viewed as the mode -matching efficiency for each mode, whi le 0 '. 2 2
is equivalent to the homodyne detection phase for different temporal mod~
S_(w) = SsN(w) L.)(j/ (ll Xj(0j
A

Bj of the input field.
9.9.3 Temporal Mode Match
First, let us consider the case when the temporal mode of the LO field
matches with one of the temporal modes of the input field , that is,
ALo(w) = </>j
0 (w). Then because of the orthonormal property of </>j(
w),
we simply have (j = bj,jo from Eq . (9.71), and Eq. (9.72) becomes
L(t) = k(t)/E/Xjo(ejo + <p).
The noise spectrum is then
(9.76)
where SsN(w) = /k(w)E/ 2 is the shot noise spectrum . So, the homodyne
detection measures only the quadrature-phase amplitude of temporal mode
Jo of the input field. Other modes have no contribution. In this case, the
LO field acts as a temporal mode filter to pick up only mode j 0 of the field
and filter out all others.
Next, for the more general case when the temporal mode of the LO does
not match to any particular temporal mode of the input field, instead of
responding to one particular temporal mode , the homodyne detection will
be the sum of the contributions from all the temporal modes with different
homodyne detection phase 0j and different mode-matching efficiency /(j /2
for each mode.
If the input field is in a single mode, i.e., on ly one temporal mode,
say j = 1, is excited and the rest is in vacuum , we then obtain the noise
spectrum from Eq. (9.72) as
+ cp)). (9.78)
j
If the phase angles {0j} are all same, we can adjust the global phase <p
of the LO to achieve minimum values for all modes. In this case, the
gains of all the modes are synchronized and quantum noise is squeezed
together for all the modes. But if the phase angles {0j} are not all same,
we have the worst scenario: one mode has optimum squeezing while others
may not have or sometimes even have noise increase. In this case, the
effect of noise squeezing will be reduced and this reduction effect cannot be
accounted for by the simple model of losses, for the unsqueezed modes have
(9.75)
noise increasing with the gain or the pump power, eventually leading to
no squeezing as a whole . This behavior was observed in some experiments
in measuring pulsed squeezing [Guo et al. (2012, 2016b )] . So, temporal
mode matching issue is more complicated than spatial mode match issue
and needs to be dealt with carefully.
9.10 Problems
Problem 9.1 Measurement of the shot noise and intensity noise of LO.
In the case of ideal balanced homodyne detection, when the input field
is in vacuum, prove that the difference between the photocurrents of the
two detectors always gives rise to the shot noise of the LO field no matter
what quantum state the LO field is in. On the other hand, the sum of the
two photocurrents always leads to the intensity noise of the LO field.
Problem 9.2 Self-homodyne of the amplitude -squeezed state.
2, •
The amplitude squeezed state /a , -re 21 '-P°')= D(a)S(-re 1 '-Pa)/O)with
• , A A

2
S_(w) = SsN(w)[/6/
2
(ll Xj 0 (0j0 +cp)) + Ll (j/ 2
]
ei'-P°'= a//a/ is depicted in Fig. 3.5(a), showing noise is squeezed in the am-
#1
2 2
plitude of the field. This can be confirmed by direct intensity measurement
/6/ (ll Xj 0 (0j0 +<p))+l-/6/ 2
A

=SsN(w) [ ]. (9.77) via self-h omodyne discussed in Section 9.8.

29-1 Q'Uanlum Opt ics Fo, Expe1·im entali sts

(i) Assuming InI » r, calculate (n). (ii 2 ) by using Eq. (3.82).

(ii) Show that direct intensity measurement gives (6 2 i ) with i = fl = 0,ta,
in the following form Chapter 10
(6 2 i ) = (6 2 ft)
= lal2 (1 - 2e- r sinh r ) Applications of Homodyne Detection
= (i ) (l-2 e- rsinhr) , (9. 79) Technique: Quantum Measurement of
which shows intensity fluctuation is smaller than the shot noise level of Continuous Variables
(6 2 i) s N = (i ) . Here we dropped higher order terms due to Ia:I2 » r.
Note that Eq. (9. 79) can be rewritten as
(6 2 n) = (n) (1 - 2e- r sinh r) < (n), (9.80) Physics is a science of measurement. In the measurement of physical quanti-
ties , many are in the form of continuous variables , which have a continuous
showing sub-Poisson photon statistics (Section 5.4).
spectrum of measurement outcomes. The amplitude and phase of an opti-
cal field are continuous at least in classical sense. Classically, the physical
quantities are well defined and can in principle be measured with arbitrary
precision. This is true even for continuous variables. In quantum mecha-
nics, however, energy is quantized. Intensity, for example, is directly related
to the photon number after quantization of the fields and has discrete values
when we measure it. This discreteness will ultimately lead to measurement
uncertainties of continuous variables. This is somewhat similar to the error
generated in digitizing an analog signal in electrical engineering. Further-
more, measurement of an optical field is usually done with photo-detector
by converting light into electrical signal. As we discussed in the previous
chapter that the origin of shot noise in photocurrents is due to the discre-
teness of electrons generated by the illumination of light. This has nothing
to do with quantization of the optical field: shot noise exists even in the
semi-classical theory of photo-detection.
In addition to the experimental difficulties mentioned above, we learn
from the quantum measurement theory, that quantum mechanics has the
Heisenberg uncertainty principle, which seems to give rise to measurement
imprecision in principle. Indeed, in many cases, quantum mechanics is the
culprit for measurement inaccuracy and sets the so-called standard quan-
tum limit. On the other hand , as we will see in this chapter , quantum
mechanics also provides a powerful tool of entanglement to tackle measure-
ment uncertainty problem and reduces the inherent quantum noise to some
extent that is not allowed by the classical theory. This is because quantum
entanglement can lead to higher correlation than what classical physics can

295
 Appli cations: Qwmturn M easur em ent of Cont inuous \la,iabl es 297
296 Quantum Opti cs Fo, Expcl'im entalists

have, as demonstrated in the violat ion of the classical Bell's inequalities
and Cauchy-Schwarz inequalities. Then the magic of quantum intcrfe ._
1· 1 · rence
0 ~ rn 111. It plays a cruc ial role in cance ling out the correlated quant
d · . um
noise an give_sn_se to noise reduction. This even works for the reduction
of the shot noise 111 photocurrents mentioned ab ove.
In ~his ch ap_ter , as an appli cat ion of the homo dyne measurement tec hni-
ques discussed m the previous chapter, we will find how to reduce quar 1t
· · h um
noi ~e 111 t e measurement of physical quantities of continuo us nature. The
mam focuses are the quantum noise and it s reduction in optical int erfe-
rometers and in quantum am plific at ion . For an appli cat ion in quantum
me~surei1:ent, we will find how to use homod yne tech niqu e to comp lete ly
char acte n ze a quantum state of a single- mode optical field by quantum
state tomography and then transport it by quantum state teleportation.
10.1 Squeezing and Correlation of Quantum Noise
10.1.1 Quantum Noise in Intensity
Fr~m the discussion of photo-detection th eory in Chapter 5 and direct calcu -
lation of t~e photo-current in Chapter 9, we learned that the photocurrent
~roduced 111 photo-detection process is directly related to intensity of an op-
t ical field. As a matter of fact, it can be shown in [Ou and Kimble (1995)]
that th e process of photo-detection is a quantum measurement process for
the photocurrent op erator:
which is related to intensity correlation function
l 2) (t' - t") = (: i(t')i(t") :) ; 15. (10.4)
with i(t) = i;t(t)E(t).
From the discussion in Section 9.2, the first term in Eq. (10.3) is the
shot noise contribut ion which exists for any state of the detected field and
only depends on the average intensity. The second term, on the other hand,
depends on the state of t he field and is related to the int ens it y fluct uat ion s
of the field. It can become negative and leads to a photocurrent noise
2
below the shot noise level. For this to happen. we need gC) (t' - t") < 1,
and ant i-bunch ed light sat isfies this.
Int ensit y noi se squeezin g (first called amplitude squeezing) was amo ng
the first observed effects of quantum noise reduction [Machida et al. (1987):
Machida and Yam amoto (1988)]. It is closely related to photon numb er
squeezing or sub-Poisson light [Teich and Saleh (1988)] and is a special
case of quadratur e-ph ase amplitude squeezing through the self-h omodyne
detect ion discussed in Section 9.8. We present ed an example in Problem
9.2.
10.1.2 Quantum Noise of Quadrature-phase Amplitudes
and its Reduction: Squeezed States
We introduced the squeezed state first for a single-mode field in Section 3.4
and then for a multi-frequency mode field in Section 4.6 .3 and a multi-
temporal mode field in Section 6.1.5. Experimentally, since homodyne de-

j
i(t) = 77 drk(t - r)Et(r)E(r) , (10.1)
tection measures directly the quadrature-phase amp litud e X(t) of an optical
field, it is used to measure the quantum noise reduction of squeezed state.
with 7as the quantum efficiency of the detector and k( r) as the response
Slusher et al. first observed a noise level at about 0.3 dB below the shot
noise level with a squeezed state generated in a near degenerate four-wave
function of the detector and

E(r)
A
=- 1 J dwa(w) e-J WT. (10.2)
mixing process in atomic sodium [Slusher et al. (1985)]. Soon after, Wu et
al. obtained a relative ly large amount of squeez ing from an optical para-
metric oscillator (OPO) below threshold [Wu et al. (1986)], which has since
for
. a. one -dimensional quasi- monochromatic field . So , the phot ocurren t become a common source of squeezed state of light for a variety of applica-
is directly related to the int ensity i(t) = i;t(t)E(t) of the field. From tions. Nowadays, more than 10 dB of squeezing was routinely achi eved in
Eqs . (9.14) and (9.17), we find the fluctuation of the photocurrent is the lab [Vahlbruch et al. (2016)] with more or less the same OPO scheme

\ i (t) i (t + T)) = 77lo J dt' k (t') k (t' + T)

as the original [Wu et al. ( 1986)].
The schematic of a typical set up for generating the squeezed state
2
+11 J
15 dt' dt" k(t - t')k(t + T - t")
of light is shown in Fig . 10.1. The light source is usually a frequency-
stab ilized la ser operating at a single frequency of w0. Then it is frequency-
doubled to 2wo by eith er intra-laser or externa l cavity enhancement
X [gC
2
\ t' - t") - l], (10.3)
 298 Quantum Optics For Experimentalists
Spectrum
Laser(%) Spatial Mode LOQ7 Analyzer
Cleaning Cavity
Squeezt - V
State ' Auxiliary beam
l - ··-..----.·-)'
OPO - /
DBS Cavi ty
Fig . 10.1 Experimental set- up for generat ing squeezed states of light SHG·
l1 · . · . seco n
armomc ge~erat1on; PC: pump mode m atc hin g cavity ; DBS: dichroic b eam splitter ·
LO: lo ca l oscillator.
technique (Section 6.3.2). With the availability of the second harmonic
field at 2wo, w~ use it to pump the optical parametric oscillator (OPO) .
!h: OPO consists of an optical cavity with an efficient nonlinear crysta l
ms1de. The ~avity is transparent to the pump beam so that the pump pas-
~es the non~1~ear crystal inside the OPO only once. 1 The aux iliary beam
from the ongmal laser serves multiple purposes. The first is to match the
~patial mode of the pump to that of the OPO cavity . Since the pump beam
1s n~t resonant to the O~O cavity , we have to use the pump mode matching
cavity (PC) to accomp lish the mode match indirectly. To match PC to the
OPO cavity , we use the OPO reversely for generating a beam at 2w from
the auxiliary beam. This beam has the same spatial mode as theo OPO
cavity and can be mode matched to PC. The second purpose is to match
the _OPOcavity with the spatial mode of the LO field for homodyne de-
te_ct10n. Tra~smitted beam through the OPO cavity by the auxiliary beam
will ser~e -this ~urpose. The third is to use it for locking the OPO cav ity to
wa• This 1s ach ieved by shifting the frequency of the locking auxiliary beam
to one nearest hig_her spatial mode of the cavity at an off-set frequency
(~q. (2.61) of Sect10n 2.2.1) and locking the shifted auxiliary beam to the
hi~her sp~tial mode. Since the higher spatial mode has a larg e frequ ency
shift (typically 100 MHz) from wa, it will not be detected by homodyne
measurement.
1
Some designs use double resonance scheme , which h as the OPO cavity lock ed to the
pump beam at 2wo and then tunes the temperature of the crysta l to h ave the cavity a lso
resonant to wo [Wu et _al. _(1986~; Vahlbruch et al. (2016)]. In this case , mode matching
the pu~p to ?PO cavity 1s straightforward . This design is usually used when the pump
po':er is r elatively low but double resonance is a lso less stab le than th e single resonance
design.
Applicahons: Q·uantwn l\leasurernent of Continuous Variables 299
To achi eve a good squeezing, we must choose a proper transm issivity T
for the output coupler. If the intra-cavit y loss (mostly due to crystal ab-
sorpt ion and surface reflection) is L. then from Eq. (6.141) in Section 6.3.4,
the best squeezing out of the OPO is about L /(T+ L) or lOlog[L /( T+L)]
in dB sca le of the vacuum noise or shot nois e level. So, we wish to have
a large T but th is will increase the threshold of the OPO and reduce the
amount of squeezing according to Fig. 6.8 in Sect ion 6.3.4. Thus a trade-off
must be made depending on the ava ilab le pump power.
d
The detection of t he squeezed state is clone with the balanced homody ne
' detection scheme and the outpu t photocurrent is ana lyzed by an electron ic
spect rum analyzer to obtain a noise level. The local oscillator (LO) for the
homodyne detection is from the or iginal laser at w 0 after the spatia l mode
cleaning .
A typical result is shown in Fig. 10.2 where Fig. 10.2(a) shows the
detected photocurrent noise level in log-sca le as the phase of the LO is
scan ned. Th e phas e ins ensitiv e trace \J!01 is the vacuum nois e level obtained
by simply blocking the light from the OPO. In this diagram , the center
frequency of the spectral analyzer is set at a frequency that gives rise to
the most squeezing (typically 1-2 MHz) with the frequ ency span set to zero.
In Fig. 10.2(b) , we plot the maximum nois e W+ as well as the minimum
noise \JI_ obtained from Fig. 10.2( a) as a function of th e quantum noise
gain Gq, which is obtain ed exper imentally by averaging the maximum and
the minimum of Fig. 10.2(a): Gq = ('ll+ + w _ )/2. Th e theory of the OPO
is covered in Sections 6.3. 4 and 9. 5. Th eoreticall y, G q = [(GD + gD) 2 +
(GL + gL) 2 + (GD - gD) 2 + (GL - gL)2]/2 = G'IJ + gb + G'i + gt, where
GD,gD,GL , gL are from Eq. (6.137) of Section 6.3.4. The solid line in
'l'ol
0o+n 3
Gain , Gq
LO Phase, 00
Fig. 10.2 Observed nois e level for the photo-current from homodyne d etection .
(a) Noise level W1 as the phas e of the LO is scanned. Wo1 is the shot noise leve l.
(b) Noise as a function of the gain of OPO. Reproduced from [Ou et a l. (1992a)].
 300 Q uant 'urn Optics For Experimentalists
Fig, 10.2(b) is a theoretical curve of '11± = (GD ± gD) 2 + (GL ± gL) 2 versus
Gq. The good agreement between the theory and experiment supports our
understanding of the physics in the process.
In the theory of Sections 4.6.3, 6.3.4, and 9.5. the calculations are stra ig-
htforward but seem to lack physical intuition about how the noise reduct ion
is achieved. To find th is, let us look a little further into Eq. (4. 75), whose
correlat ion function leads to the spectrum of squeezing in Eqs. (4. 76), (9.45)
and (9.46). We rewrite it as
Y(D) = [a(wo+ D) - a,t(w0 - D)]/i. (10.5)
Here we are only interested in Y(D) since it gives noise reduction in Sy( D),
Let us rewrite the inpu t-output relation obtained in Eq . (6.135) of
Section 6.3.4 for degenerate OPO below threshold:
aout(Wo+ D) = GD(D)ain(wo + D) + gD(D)a!n(wo - D)
alut(wo - D) = GD(D)a!n(wo - D) + gD(D)ain(wo + D), (10.6)
where we used G'o(-D) = GD(D),g'o(-D) = gD(D). From Eq. (6.136) of
Section 6.3.4 for GD, gD, we find that when the threshold is approac hed ,
GD, gD become very large and GD (D) gD(D). Then from Eq. (10.6),
we find aout(Wo+ n) and alut (Wo- n) become almost the same when th is
happens~t Since Eq. (10.6) ~s an equat ion fo_rqua~tum operators, aout(Wo+
D) and aout(wo - D) are highly corre lated m their quantum fluctuations.
Now come back to Eq. (10.5) for quantity Y(D), which is the difference
between two highly corre lated quantities aout(Wo+ D) and alut(wo - D).
Then q:1antum fluctuations of aout(Wo+ D) and alut(wo - D) are cance -
led in Y(D). So, the quantum noise reduction in Y(D) is a consequence
of quantum destructive interference through quantum correlation between
frequency components of wo + D and w0 -D generated in the OPO. Interes -
ting ly as we will see later in Sections 10.3.4 and 11.3, quantum correla tion
and destructive interferenc e are the underlying physics for quantum noise
cancelation in quantum amplifiers and SU(l,l) interferom ete r.
10.1.3 Quantum Correlation of Quadrature-phase
Amplitudes: EPR Entangled States
In studying quantum nois e reduction in the squeezed states of light, our
concern is the quantum fluctuations in the quadrature-phase amplitude of
one single-mode field. For measuring this, we employ one set of homod yne
detection device, and as we have just showed in the previous section, two
Applications: Quantmn .Meas1tremr11t of Continuous Variablf's :301
correlated frequency components work together for destructive quantum
interference bet"\veen the two correlated amplitudes of the two frequency
components, Then next question naturally arises: cau we direcLly observe
this type of quantum correlation between the two modes of the field? For
this. we need two sets of homodyne detection devices to measure separate ly
the quadrature-phase amplitudes of the two modes and compare the res-._1lts.
Historically, the first observation of quantum corr lation between two
continuous variables of quadrature-phase amplitudes was done in the con-
text of demonstration of Einstein-Poldosky-Rosen paradox, lead ing to the
demonstration of non-locality of quantum mechanics . In 1935, Einstein,
Podolsky and Rosen (EPR) [Einstein et al. (1935)] proposed a gedanken ex-
periment involving a system of two particles in the following wavefunction:
(10. 7)
which can also be written in the momentum space as
(10.8)
Here C, C' are normalization constants. 2 So the two particles are spatially
separated by x 0 but perfectly correlated in both positions and momenta.
From their view of local realism, EPR concluded that quantum mechanics
is incomplete since canonically conjugate variables of position and momen-
tum for one of the particles could be assigned some definite values without
disturbing it through the perfect correlations from the measurements of the
other particle, in apparent conflict with the Heisenberg uncertainty princi-
ple. Later, J. S. Bell extended the wavefunctions in Eqs. (10.7) and (10 .8)
to the Wigner function in x - p phase space [Bell (1987)]:
W(x1, x2; Pi, P2) = JJdy1dy2e-i(PiYi +P Y )'ljJ( x 1 + ~1 , x 2 + y;)
2 2
X '1/J*
( X1 - Y2l'X2 - y;)
= C" 5(x1 - x2 + xa)5(p1 + P2)- (10.9)
This demonstrates more directly the perfect correlations between positions
and momenta than the two wavefunctions in Eqs. (10.7) and (10.8).
In 1989, Reid showed that the outputs of a nondegenerate optical pa -
rametric amplifier (NOPA) possess the same correlation properties of the
EPR state except that the two particles are replaced by the virtual harmo -
nic oscillators of two output optical modes of the NOPA and the positions
and momenta of the particles by the quadrature-phase amplitudes of the
2
The o-function in the wavefunction makes it un -normalizable so we use C, C ' (rv 0).
 302
Quantam Optics For Experimentalists
Apphcat'ions: Qua11t'Um J.Jeas'U1·cment of Cont .imwns Variables 303

optical modes [Reid (1989)]. Consider the two-mode squeezed state discus-
ow, we can still make the same argument as EPR did. that is, t~si:1g
sed in Section 4.6. The quadrature-phase amplitudes X ., y,, and XA yA 1
B, B /c l ) fl
X to infer a value for XA or -YB for YA, but inferences are not defimte
of the two NOPA output modes A, B are given in Section 4.6.2 and are / E )?. 6.?. y -
but with errors of 6; 111.XA = (6 2 (XA -XE))= 2/(G+g - and inf A=
A A

related to the input modes a, b by

(t:.2 (Y4 +Ys)) = 2/(G+g) 2 . As long as 6.~11fXA~Tnf1A -± f(G+g) 4 < 1.
XA - XB = (Xa - Xb)/(G + g), Y4+YB= (Ya+ Yb)/(G + g). (10.10) the inferred values violate the Heisenberg uncertamt y relation and the EPR
paradox still stands. . .
Here, g is taken to be positive for the simplicity of argument. Problem 10.3
The condition 4/(C+g) 4 < 1 corresponds to a n01se squeezmg of 1/(C+
will deal with the case of arbitrary phase for g. So, when the a-ain
6 ,g
c g)2 < 1/2 or 3 dB of noise squeezing. However, Reid suggested a better_way
bcc01:1e very }arge, both XA - XB and YA + YB go to zero or XA = XB
A A A A

of inference of XA and YA by using quantities AXE or -AYE, respectively

and Y4 = -YB. Since these are operator identities, if we make measure -
[Reid (1989)]. Then the inferenc e error is given by
ment of XA, XB or YA, -YB, their values will be perfectly correlated or
anti-corr~la!ed. From Chapter 3, we know that the quadrature-phase am- 2
t:.inf X A (A)= I\ t:.
2
(XA - AXE) )= (G2 + g2)(1 + A2 ) - 4AGg. (10.14)
plitude X, Y are proportional to the position and momentum operators of
which reaches a minimum value of 1/(C 2 + g2) when~= 2Cg/(C 2 + g 2 ).
the harmonic oscillator describing a single-mode of optical field and they are
Similarly, the inference error 67111YA(A) = (6 2(YA +AYE)) has a minimum
conjugate to each other satisfy ing commutation relation [X, Y] = i . The-
value of 1/(C 2 + g 2 ) for the same A value. So, the condition for an EPR
refore, the two output modes of NOPA exhibit the same type of quantum
correlation as the two particles described by EPR. paradox is C 2 + g 2 > 1 or g 2 > 0. .
In practice, each field consists of multiple frequency modes. Smee spa-
To compare with Eq. (10 .9), we can find the Wigner function for the two
tial modes out of the cavity are well-defined, we can use one-dimensional
output modes of NOPA. If the input states to the NOPA are the vacuum
description with quasi-monochromatic approximation for each field, in a
states, i.e., the input Wigner function is [Eq. (3.173) in Section 3.7.2]

c~r
way similar to the multi-frequency mode squeezed state in Section 4.6._3
Wa,b(x,, y,; X2, Y2) = exp [ - xl ; Yilexp [ - x§;y~l,(10.1!) but for two fields now. The correlation is between quadrature-phase ampli-
tudes defined in frequency domains of the two fields, as in Eq. (4.74). We
we may derive the output Wigner function from the transformation given can proceed with the same argument above.
in Eq. (10.10) as (or see alternative method in Problem 10.1) The first experimental demonstration of EPR paradox in continuous va-
riables was carried out by Ou et al. [Ou et al. ( 1992b)]. The experimental
WA,B (X1, Y1;X2, Y2) = 1 exp { - l[ (x1 + X2)2 + (y -y ) 2 Je-r
71" layout is shown in Fig. 10.3. The EPR correlated fields are g~nerated fr~m a
4 2 4 1 2
non-degenerate OPO below threshold. This device is theoretically described
- 41 [(x1 -x2) 2
+ (Y1 +y2) 2 ]er} as a non-degenerate parametric amplifier by Eq. (6.128) in Section 6.3 .4.
The non -degeneracy of the OPO is achieved via a type -II i 2 )_nonlinear pro-
• C'" 6(x1 - x2)6(y1 + Y2) as r • oo . (10.12)
cess where the two correlated fields have the same frequency but orthogonal
Here r is related to the amplitude gain by G = cosh r and C'" is some polarizations. So, the two fields can be separated by a polarization be~m
normalization constant .
splitter (P in Fig. 10.3) and directed to two separate homodyne detection
So at infinit e gain the system can mimic the EPR correlated two-particle schemes for the measurements of XA(01), XE(02), respectively. 01, 02 are
system. But we know that infinte gain is impractical, then what happens determined by the phases of L0 1 and L02, respectively. The A parame -
for the finite gain? Can we st ill demonstrate the EPR paradox? ter from Eq . ( 10 .14) is optimized by controlling the electronic gain g in
At finite gain, the correlation will not be perfect. From Eq. (10.10), we Fig. 10.3. The exper imenta l results are shown in Fig. 10.4, where measure-
find for a, bin vacuum states
ments of the inferred errors of 67111X and 67 111Y are shown in a log-scale.
Both values are below the value of 1, the limit set by the Heisenberg un-
certainty relation, thus demonstrating the EPR paradox.
 305
App licat io11s: Quantwn J\.Jeasurement of Continuou.s Variab les
304 Qua n tu m Optics For Experimenta lis ts

Du an et al. and Simon each deriv ed ind ep end entl y a necessa ry and suffi-
cient cou<lition for insep arabilit y of a qu antmn stat e [Du an ct al. (2000) ;
Simon (2000 )], whi ch , wh en appli ed to t he case here, h as t he form of
(10.16)

2
Th e result s in [Ou et al. (1992b )] give u~~nf x + 6 TnJY )/ 2(1 + >..) = -2. 8
dB ;::::::
0.52, clearly sati sfying th e ins ep arabili ty condi t ion in Eq. (10.16) and
demonstrating th e quantum entangl em ent b etwe en th e two fields from th e
non-degenerat e OPO.

10.1.4 Quantum Correlation between Intensities:

Twin Beams
In Section 4.6.1 , we find that ther e exists a perfect photon number cor-
relation betw een the two modes in th e two-mode squeezed states which ,
from what we learned in Section 6.3.4, can be generat ed from a parametric
<l>(Q. 81, 82)
amplifier with no input (spontaneous mode). This perfect photon number
correlation persists even at large gain for the parametric amplifier. In this
dFig. 10.3t OP
Experim ental sch ematic for demonst1·at1·ng EPR paradox with a non-
egen era e 0. Reproduced from [Ou et al. (1992b)]. section, we will explore more of this type of correlation.
Historically , nonclassical intensity correlations were first demonstrated
by Heidmann et al. with an optical parametric oscillator (OPO) above
1.0 threshold [Heidmann et al. (1987)]. They measured the noise of the inten-
sity difference between the signal and idler fields of the OPO and observed
a noise level 30% below the shot noise level of the two intensities, indicating
a quantum correlation in the intensity fluctuations of the two fields.
For the twin beams out of a parametric amplifier, the early experiments
usually did not have enough gain so that the output intensities are too low in
the spontaneous mode to produce a significant photocurrent to overcome
the dark currents of the photo-detectors for a direct confirmation of the
0 0.2 0.4 0 0.2 0.4 perfect photon number correlation exhibited in Section 4.6.1. So, in order
Time (s)
to boost the intensity, Aytur and Kumar injected a coherent state as input
Fig. ~0.4 Experimental results that demonstrate EPR parad f . . to the parametric amplifier [Aytur and Kumar (1990)] and obtained two
Wos is the vacuum noise for each beam. Reproduced from [Oo: e;ra~~(:~~~~)]. vanables. bright outputs from the amplifier that have the similar photon number
correlation as the twin beams discussed in Section 4.6.1.
The EPR correlated fields also provide an example of two t Consider an optical parametric amplifier which has the same input-
t 1 d fi 1 . q uan um output relation as Eq. (4.58) in Section 4.6 and has a coherent state \a)
en ang e e ds, which cannot be expressed in the form of a separable state:
input at mode a:
(10.15) (10.17)
 306
Quantum Optics For Experimentalists
Applications: Quantwn J..1easurement of Continuous Variables 307

For simplicity of argument, we assume g > 0. It is straightforward to ·h

2 So. the noise reduction factor is
2
tl:at, with G - g = 1, A_t A - iJtiJ = a,ta,- Mb.Then the photon nu:b 0;
difference of the two outputs has a fluctuation of (,6.2(NA- ANB)) - (G2 + g2)(G2 + g2>,.2)- 4)..G2g2
(10.23)
2 2
(,6. (N.4 - NB))= (,6. (Na - Nb))ja) = /a/ 2 . (10.18) (,6.2 (NA - ANB))sn - G2 + )..
2 2
g

Now let us compare this noise level with that of two coherent fields whose When ).. = G / g, the noise reduction factor reaches a minimum value of
ph~t~n nm~bers are at the levels of the two outputs or the shot noise level. 2
(,6. (NA - )..NB)) _ (G _ )2 = l (10.24)
This 1s a fair comparison because two fields in coherent states with noise (,6.2(NA - ANB) )sn - g (G + g) 2 '
levels at shot noise level are completely uncorrelated quantum mechanicall
which is about 3 dB better than the value in Eq. (10.20) at large gain.
In fact, it can be shown (see Problem 10.2) that for two classical fields y.
Experimentally, because the scheme by Aytiir and Kumar is relatively
always have (,6.2(JA-fB))c1 2::(,0.2JA)cs+ (,0.2JB)cs = (].4) + (iB). He:e
easy to implement, it has become a standard technique for testing the twin
the subscript "cs" denotes coherent states. '
beams property for a number of parametric amplifiers achieved in different
The two output fields from the parametric amplifier with a coherent
nonlinear media including the regular x( 2 ) materials [Aytur and Kumar
state
2 2input2 have 2photon numbers of (NA)= G 2/a/ 2 + g2 c2/a/2 , (NB)=
2 (1990)], four -wave mixing in atomic vapor [McCormick et al. (2007)], and
g /a/ + g g /a/ , respectively , for /a/ 2 » 1. So, the shot noise level
four-wave mixing in optical fiber [Guo et al. (2012)] . The typical experimen-
for the number difference of the two fields at the intensity levels of the two
outputs is simply tal arrangement is shown in Fig. 10.5, where a coherent signal is injected
2
in the input port of an optical parametric amplifier (OPA). In addition
( ,0. (NA - NB)) sn = (,0.2 NA) sn + (,0.2 NB) sn to the amplified signal, an idler beam is generated conjugate to the signal
= (NA)+ (NB) = (G2 + g 2 )/a/ 2.
(10 .19) beam . Both the signal and the idler beams are respectively directed to two
Note here that the shot noise level is obtained by taking the number statis- identical detectors where the difference of the photo-currents are measured
tics as the Poissonian distribution: (,6.2NA,B)sn = (NA,B)- Hence, compa- with an electronic spectrum analyzer. With the help of a removable mirror
red to the shot noise level, the noise level of the photon number difference (M) and a beam splitter (BS), an auxiliary coherent beam is used to me-
of the two output fields has a reduction factor of asure the shot noise level equivalent to the total shot noise level produced
by the two twin beams together. To ensure they are equivalent, the DC
R = (,6.2(NA - NB)) _ 1 output levels of the two detectors are used as references and are matched
(,6.2(NA _NB)) sn - c2 + g2 · (10.20)
between the twin beams and the auxiliary beams. The results are shown
When the gain is infinity, we achieve 100% reduction. This is somewhat in Fig. 10.6 where trace (i) is the equivalent total shot noise level from the
similar to the EP R correlation discussed ear lier al though (,6.2(NA _ NB)) = two twin beams and trace (ii) is the noise level for the intensity difference
2
/a/ -/=0. Here we consider relative noise level with reference to the sum of of the twin beams, showing more than 5 dB noise reduction. Trace (iii) is
the shot noise levels of the two beams. the electronic noise when no light is on.
At finite gain, the nois e reduction is less than 100%. Similar to the cal- It is int eresting to note that intensity corre lations discussed above is a
f the inferenc e error ,6.fnfXA , !).InfYA ear lier, we should consider
c;1lation Ao special case of the EPR correlat ion discussed in the previous section . As we
NA - )..NB, whose noise variance, under the approximation of /a/2 » 1, is
calc ulated as
2 2
(,6. (NA - )..NB))= [(G + g 2)(G 2 + g 2A2 ) 4)..G2g 2]/a/2. (10.21)
s1crnal
0 tnJect,on
~----, ~~~
~~-----
l M
identical
detecto1s
-
) J
7'-----~
OPA I k-~
The corresponding shot noise level is Pump "'~-~
signal
·i• BS :
1
2 auxiliary coherent beam
(,6. (NA - )..NB))sn = (,6.2 NA)sn + A2 ( /). 2 NB)sn
2
= (NA)+ A (NB) = (G 2 + A2g 2)/a/2. (10.22)
Fig. 10.5 Experimental arrangement for the observation of intensity corre lat ion s bet-
ween two fields of twin beams . M: mirror; BS: beam sp litt er; SA: spectrum analyzer.
 308
Quantum . Optics For Expe1'imen t,alists
Applications: Q·uantum J\IcasurPm ent of Continuous \/m·iables

- 65
10.2.1 Quantum Nois e Analysis for a Mach-Zehnder
- 70 Interferometer
(i)
53' - 75 The quantum noise performance of a linear interferometer was first ana lyzed
:s ...,,,.....,.,..__- by Carl Caves in a semina l paper that st udi es the sensitivity of using a
Q)
> -80
C) Iichelson int erferom eter as a tool to detect gravitationa l waves [Caves
....i
Q)
Cf) - 85 (1981)] . In that st ud y, he cons idered t he effect of rad iation pressure on
·o suspended mirrors. But for simpli city, we will fix the mirror s at t he moment
z - 90
so as to ignor e the effect of radiation pressure and only concentrate on the
-95
effect of quantum noise of light on the phase measurement sensitivity.
0.5

r, &in
1.0 1.5 2.0 2.5 3.0 3.5 4.0
A
Frequency (MHz) A A
a.111 Homodyne
Fig. 10.6 Spectr~m o~ the int ensit y differ enc e . (i) Total shot noise leve l for the two
Detection
b ea ~ s to?eth er ; (u) Noise leve l for th e int ensity difference of the twin beams· (iii ) El
trornc noise. Courtesy of Jun Jia. ' ec-

discussed in Section 9.8, measurement of intensity fluctuations is basically

self-ho~odyne measurement of the X-quadrature-phase amplitude for the Fig. 10 .7 Layout of a Mach -Z ehnd er int erferomet er for th e measur ement of a small
field~ wit~ a strong real coherent component. 3 Therefore , the variance of pha se shift fJ with a homodyne d et ect ion. fp s is the phase sensing field photon number.
the mtens1ty difference is equivalent to the variance of _x _ XA Th
1·t is
· . . A B· US
not surpnsmg to see noise reduction in the int ensity difference. But Consider a Mach-Zehnder interferometer depicted in Fig. 10.7. Assume
measurement of Y corresponds to the phase measurement which we did the two beam splitters are 50:50 and lossless. They are described by the
not measure in the study of twin beams. So, we cannot demonstrate the input-output relation
EPR paradox or EPR entang lement through the inseparability criterion of A = (Clin+bin)/ \!'2,
B = (bin - Clin)/ \!'2;
Eq. (10.16) here.
Clout= bout= (Beicp+ A)/\1'2
(A- B ei'P)/\1'2, , (10.25)
where we place the overall phase difference cpof the interferometer on the B
arm only. It is straightforward to express the outputs of the interferometer
10.2 Quantum Noise and Its Reduction in Linear in terms of the inputs:
Interferometers
Clout=t(cp)ain + r(cp)bin; bout= t(cp)bin+ r(cp)ain, (10.26)
where t(cp) = eicp/ coscp/2; r(cp) = -iei'P / sincp/2. So, the interferometer
2 2
~s a ~racti~al application of squeezed states, we will study quantum noise acts as another beam splitter with phase sensitive amplitude transmissivity
m a lm ear mterferometer and see how we can improve the sensitivity in t( cp) and reflectivity r( cp).
phase. measurement in this device · In Chapter 11 , we w·11 d"lSCUSS Ill
· more
1 For a coherent state let)input at CLinand vacuum at bin, the output
~et~1ls the precision phase measurement in a general platform that is not photon number 4 at output port b of the int erferometer is simply
lnmted to linear interferom eters.
h -- (boutbout)
At A -
- let!2 (1 - COS i.p)/2 -- fps (l - COS i.p), (10.27)
3 4 Forsing le-mode description of the field, the intensity of the field is different from th e
For a complex coherent component , refer to Problem 10.3 (Eq. (10.ll0)).
photon number by a constant. So, we will use intensity symbol for photon nurn.ber.
 310
Quantum Optics For E:rpcri111e11talists Appli cations: Q11antum Ji.feasw· em cnl of Continnous Variabl es 311

whe~·e IP~ = (iJt B) = /~/ / 2 is the photon number of the field subject to the
2
output is dark, but with a small shift, the output becomes what Eq. (10 .31)
phase shift (pha se sensmg field). As we will prove later in Chapter 11 tl gives. On the other hand, the detector cannot detect anything less than
the pha , · . fi Id · I . 1at one photon. So. the minimum detectable phase shift is c5m= 1/ -jl;;/2
. se sens mg e 1s w iat matters m phase measurement accuracy .
Fiom Eq. (10.27), a sma ll ch ange of 5 in <..pwill lead to a change in t he when 5h = 1. This is similar to SQL.
output photon number: However. for low light level , the dark currents of t he detectors will
overwhelm the signa l (one photon). So, we cannot use direct detection.
5h = fps5 sin <p. (10.28)
Fortunately. we know a detection technique that can overcome the dark
This is ~he sig1~al due to t he phase change 5. The signal size due to the phase current problem. This is t he hom odyne detection method for measuring
change
.
1s maxmmm at <p = 1r/2 ·· 51bl\l -- I ps 5 . Bor• tl 1e noise
· per1ormance
.c
of t he quadrature-phase amp litud es that we discussed in the previous chapter
the mterferometer,_ we calc ul ate the variance of the output at <..p = 1r/ 2 for (Sect ion 9.3).
t he coherent state mput: With a sma ll phase shift 5 for <..pand o. = i /a I. we can eas ily find
2f
I:::. - (bt b At A At A 2
b - out out boutbout)7r j 2 - (boutbout)7r / 2 =
?
/a/-/2 = fps· (10.29) (XL) = (xt 2
(5/2)) cos (J/2) + (Y}in(o/2)) 2
sin (J/2)
So, the signal-to-noise ratio of the phase measurement is 1 + /a/ 6
2 2
for 5 < < 1, /a/2 > > 1. (10.32)

SN R = (5h)2 = (fps5)2 - 2 Here Xb(o/2) = beib/ 2 + bte-ib / 2 and Ya(b/2) = (aeib/ 2 - a,tcib / 2)/i are
b 1:::,.2
f I - fp s5 (10.30)
b ps the quadrature-phase amp litude s of correspon ding fields. Obviously , /a/ 2 62
Th e minimum detectable phas e shift is then 5m = 1/ y if1 ps · Th.lS lS· th e in (Xb2out ) corresponds to the phase signal while the nois e of the phase
so-called standard quantum limit (SQL) of phase measurem ent , which has m eas urement is simply 1, th e vacuum nois e, in Eq. (10.32). Hence , the
a 1/-J]v;: -dependence on the phase sensing photon number N = J signal-to-noise ratio of the linear interferom eter for measuring X is
F h d . . ps ps ·
rom t e envat1011 above, it is hard to see where the nois e in Eq. (10.29) (10.33)
~omes from. It seems to come from the photon number fluctuations in the
mput coherent state. It turns out that Eq. (10.29) stands even for the which lea ds to the standard quantum limit 5sqL = 1/ffi with N = 2Ip s ·
number state (/N)) input with no photon number fluctuation. The noise A close exa mination of the origin of the nois e term in Eq. (10.32) reveals
stems
.
from the interferometer itself · at 1nr -- "Tr/2
" ,
· t er £erome t er 1s
the 111 · that it is from the vacuum input at bin, the unused input port . We can
basically a 50:50 beam splitter with /t(1r/2)/ = /r(1r/ 2)/ = 1/ h and this confirm this by noting that at <p = 0, we hav e bout = bin, which is in vacuum
~eads to randomness in one of the output even with no randomness at the state.
mput. A_nother way to think about it is that the culprit of the noise is the
vacuum mput of the unused port as we will soon see in the following. 10.2.2 Sub-shot Noise Interferometry with Squeezed States
. From Eq. (10.30), we find that the larger fp s is, the better the sensitivity
is. But the average detected photon number at the output 1·s J h. l I Knowing the origin of th e noise in the interferometers, we can use quan-
b ps, w 1c1 a so
ecomes ~arge. Sooner or later , it will saturate the detector and it is thus tum noise reduction technique to reduce it . Caves was the first to suggest
not practi~al to work at (fJ = 1r/2. The way to circumvent this problem is to to inject the squeezed vacuum into the unused port of the interferometer
make the mterferometer work at <p = 0 or at the dark fringe. In this case, [Caves (1981)]. From Eq. (10.32), we can easily see that with a squeezed
Eq • ( 10. 28) does not stand and we need to go to higher order and O btain state at the unused input port bin instead of vacuum, Eq. (10.32) becomes

(10.34)
(10.31)
Since_ the _interferom~ter ~orks at dark fringes now , Eq. (10.29) will lead Here we choose the phase of the squeezed state so that (XL (J/2)) is at
to zero noise and an mfimte SNR. This is not true. At no phase shift, the the minimum value of e-r with a squeezing parameter r > 0.
 312 Quantum Opti cs For E:.cper im entali sts Ap pl ica t io ns : Q u.a11tu m J\.leasu re lll ent of Con ti 1wons Vm·ia bles 313

The first demonstration of sub-shot noise interferometry was performed for a small phase chan ge c5« 1 and the inpnt port a in (E t in Fi g. 10.8) in
soon after the generation of the squeezed states [Xiao et al. ( 1987); Grang ier a coherent state (Ia)), we have th e phase signal:
et al. (1987)]. vVe will discuss next only the experiment performed by Xiao
et al. [Xiao et al. (1987)]. c5L = JaJ2 c5sin cp= 2I psc5sin cp. (10.36)
Different from th e detection schemes we discussed in the previo us So, we have the largest signal when <p = 1r / 2. Notice that the phase signal
section, Xiao et al. employed a self-homodyne scheme shown in Fig. 10.8. is twice as large as what Eq. (10.28) gives b ecaus e we have tv.ro detectors
P1,2 are two electro-optic modulators (EOM) with one for adjusting over- here.
3;.llphase <p and the other for introducing phase signal c5. The input port For the nois e, we find that for <p = 1r / 2, we have [t(1r / 2) I = [r(n / 2) I =
E1 is in a coherent state Jo ) but the other input port Es can be in va- 1/ J2 and the interferometer is simply a 50:50 b eam splitter. Hence , the
cuum or a squeezed state from OPO. Both outputs are detected and t he outcome is the same as a balanced homodyne detection scheme:
photocurrents are subtracted before being analyzed by a spectral ana lyzer.
(10.37)
The combination of the interferometer and detection system is similar to
the balanced homedyne detection discussed in Section 9. 7 except that the where bin= Es in Fig . 10.8 , YEJ <fJa
) = (bine-i(f)o.
- bJnei(f)a
)/ i and ei (f)n =
50:50 beam splitter is replaced by the interferometer or a phase dependent a/Ja[. When Es is in vacuum , (YlJ<p J) = 1. So, we obtain SNR 0 =
beams splitter , as we found in the previous section. From the input-output (c5L)2 / (6.2j_ ) = 2Ipsc52 and c5m= 1 ~ ' which is the SQL. Note that
relation in Eq. (10.26), we find in this case, (6. L)v = [a[ = 2lps or the total shot noise from the two
2 2

A - At A At A detectors. On the other hand , when Es is in a squeezed state and with

J_ = aoutaout - boutbout
- 2 2 At A At A the right choice of <fJa,we have (6.2 L)sq = JaJ2
e-r = (6.2 L)ve-r and the
- [Jt(<p)J- Jr(<p)J](ainain - binbin) noise is reduced below the shot noise level.
+[t(<p)r*(<p)- t*(<p)r(<p)](ainbln- a!nbin)- (10.35) Figure 10.9 shows the phase signal level and the noise level with vacu1-:m
(Fig. 10.9( a)) or a squeezed state (Fig. 10.9(b)) in the unused port Es .
Notice that the noise level is about 3 dB below the shot noise level (dashed

~+
line at O dB) for the squeezed state input and the SNR is improved also
about 3 dB, thus achieving sub-shot noise interferometry.
~-=- ---·l<"--m-1
__ _
(b) 4.0 -
(a) 4 .0-
ON
3.0
20 ON
2.0
1.0
10
¢(dB)
¢ (dB) 0
0 --
·1.0
·I0

·2 0

3 .0 Of F

- 4.0
0 0 .1 02
T ,me (s ec )

Fig. 10.9 Phase signa l level and the noise l~vel for phase measurem~nt with (a) vacuum
or (b) a squeezed state in the unused port Es - Reproduced from [Xiao et al. (1987)].

Fig. 10.8 Layout of a Mach-Zehnder int erferomete r for phase measurement. Reprodu-
ced from [Xiao et al. (1987)].
 ""' 315
Applications: Quantum .Measurem .en t of Continuous Va,riablrs
31--l Qua11,tum Optics For Expe1"imental'ists

10.2.3 Quantum Noise Analysis for LIGO E~1

(t) = _ l _c-zwol
12n
I
'
dni i(D)e-int.

LIGO is an acronym for Light Interferometer Gravitational-wave Obser-

vatory, It employs a long baseline l\Iichelson interferometer to detect a
jj_;b(t) =
171
e-iwot [Eo
+ _l_
12n J dDb(D)e-irtt], (10.38)

minute phase change due to space distortion when a gravitational wave ·where we treat the strong coherent laser as a c-number and shift the Eo
passes by, On September 14. 2015, after over four decades of persistent frequency zero to the center frequency w 0 of the strong laser. The fields
endeavor, LIGO detected the first signal generated by the merger of two riffht after the 50:50 beam splitter are
black holes [Abbott et al (2016)]. Since the operation principle of LIGO is b E1 = [Efn(t)+ Eti(t)]/J?., E2 = [EiJt) - Efn(t)]/J?.. (10.39)
to measure phase shift, it is subject to the same quantum noise discussed
After they are bounced back by the mirrors and recombined in the beam
in the previous sections, In addition, there is also the radiat ion pressure
noise due to randomness of the photon number impinging on the free ly sus- splitter, the output fields are A A A,
pended mirrors. 5 This noise is also of quantum nature. Carl Caves, in his Eiut = [E~(t) + E;(t)]/J2, E~11 t = [E;(t) - E 1 (t)]/J2, (10 .40)
seminal paper on quantum noise in interferometers [Caves (1981)], analyze d where E'-(t) = i; (t- 2x 1 (t- L/c)/c) with x 1 (t- L/_c) as the posit_ion ope-
1
the effects of these two sources of noise based on a single -mode model. In rators of the two freely suspended mirrors. Propagation delays ~re mcluded
this section, we will make an analysis with a multi-frequency mode model in the expression above and L is the length of each arm assummg they ~re
to calcu late the noise spectrum of the interferometer. balanced. Substituting Eqs. (10 .38) and (10.39) into the above, we ~btam
Ad - -iwo[t-2x1(t-L/c)/c]E /2 - e-iwo[t-2x2(t-L/c)/c]Eo/2 + D.E(t)
E t - e o
OU e-iwo(t-2L/c\iwo/c)D.x(t - L/c)Eo + D.E(t), (10.4l)
A A = c5
A _ c5x with c5x. = x - L(j 1, 2) is the relative displa-
wh ere uX - X1 2 J J . . -i2wofJx 1 /c ~
cement for the two mirrors and we made the approx1mat10n e ~
L

\ J .(
1 - i2woc5x
1 / c, and
e -iw 0 (t-2L/c)
D.E(t) = ----
A d"~iaA(") L/ )
~i e -ir2 t-2 c . (10.42)
i, v12K
- 2 When a gravitational wave passes by, the nature of th_e ~ave is such that
it stretches the space in one direction and compresses 1t m the ortho~onal
E2 -L---
•----~ -- direction. So, a disturbance of hcw(t)/2 or -hcw(t)/2 is addedAto X1 or
2 +x 2(t)-L.
i;:1
I 1E:llf x respectively: c5x1 = hcw(t)/2+x1(t)-L,c5x2 = -hcw(t)/
T;~e freely suspended mirrors are illuminated by the light fields and thus
Fig . 10.10 Michelson int erferom eter with two freely suspended mirrors for gravitatio nal their motions are subject to radiation pressure forces:
wave detection.
M d2xj ( t) = p(1) (t) (10.43)
dt2 ph

Consider a Michelson interferometer with two freely suspended mirrors with

as shown in Fig. 10.10. A strong coherent laser enters the interferom eter p(j)(t) = ~E\t- L/c)E 1 (t - L/c)
J
ph
from the bright port Efntogether with a vacuum input at the dark port
J

Efn. Assuming that the spatial modes are all matched throughout the {E * + _l_
= '5:_ dD[bt (D) ± at (D))i 0
(t-L/c)}
2 0 v12K
interferometer, we can describe the fields with a one -di mensional quasi-
monochromatic approximation (Section 2. 3. 5): X {Eo+ _l_
v12K
J dD'[b(D') ± a(D'))e-iO'(t-L/c)}

5
2 e-iO(t-L/c) }
Free suspension of the mirrors is for seismic isol ation cons id erat ion . ~ ~\Eo\ +'5:_{E*fdn[b(D) ± a(D))--- +h.c. · (10.44)
~ 2 2 o y12K
 316 Quantum Optics Fo1· E:rpcrimentalists Applications: Q'llantum Afeasuf'cment of Continuous \lm'iablcs 317

Here for one -dim ens ion al a.pproximation , we know from Section 2.3.5 that where we used hc n-( - D) = hc n ·(D) and bhew = hc n-/L as the strain.
2
/Eo/ = Rin is the input photon flux. The radiation force on mirrors is t hen and
F = 2nkoR;n and h. = 2nwo/c. Solving Eq. (10.43) in frequency domain
with (10.53)

i:1(t) = j dfli:1([1)e -m', hcw(t) j dflh c w (fl)e-;'", /( is the opto-mechanica l coupli ng constant whi ch couples the light fields
with the motion of the mirrors through radiation pressur e. Taking the
P;fi)
(t) = j dDF;~\n)e-mt. (J = 1. 2) (10.45) power of the input laser as Po = nwoR in = nwo/Eo/2. we can rewrite
we h ave Eq . (10 .53) as

4Powo M = 2L = AK
where
(10.46) JC= l\ID 2 c2 ' v hsQL
(10.54)

wit h
P;fi)(D) = {Eo[bt(-n) ± a,t(-D)) + E; [b(n) ± a(D)]} eiDL/c (10.47)
with j = 1 -t + and j = 2 -t - . Hence , hsQL = V = (10 .55)

(0±1 - c5±2)(D)= - M~ 2 [P;~\n)- fr;~\n)J+ hew where Alµ, = 111/2 is the reduc ed mass for two free masses. This is the stan -
dard quantum limit for position measurement of a free mass 111µ,[Braginsky
= hew - M~ 2 [Eoat(-D) + E6a(D)]eiDL/c _ (10.48)
et al. (1992)]. With these notations , Eq. (10.52) takes the new form of
Substituting
A

E~ut(t) = ----
J .
the above into Eq. (10.41) and using Eq. (10.45), we have
e-iw 0 (t-2L/c) KE W
dfl e-i nt{-. _o_o [Eoat(-D) +E*a(D)]e 2iDL/c
Xout(fl) = Xin(fl)
;-:;-;:;;:-Ohew(D)
-iDL/c
Yout(D) = ~n (D) - JCXin (D) + v 4JC h
A A A

iNJf12c o e · (10.56)
SQ L
+a(D) e2iOL/c + (iwo/c)Eohew(D) eiOL/c }- (10.49) From Section 9.5, we find the measured spectrum in homodyn e detection
is given by
Writing Eiut(t) in th e 1-dim form as in Eq. (10.38), we have
A

E~ut(t) =
e-iwo(t-2L/c) J A

dDA(D)e-iD(t -2 L/c) (10.50)

(10.57)

with spectral density S(D) defined through

with
KEowo (Yout(D)Y}ut(D')) = S(D)o(D - D'). (10.58)
= a(D) + iMf12c
A

A(D) [Eoa t ( -n) + E6 a(D)]

For vacuum input , we have (~n(D)X:~(D')) = (Xin(D)Xin(D )) = c5(D-
A A O A A t '

+( iwo/ c)Eohew (D)e-iDL/c_

(10.51) D'). Single -sided spectral density is used for characterizing the gravitational
Using the input field Ea = /Ea /ei'Po as LO for the homodyne detection of wave strain spectrum so that 6
the output, we define Xout (fl)
A • A
=
A(D)e-i'Po + A_t(-D) ei'Poand yout(D) = (c5hew(D)bh0w(D')) = ~Scw(D)o(D - D'). (10 .59)
[A(D)e-i'Po - AT(-D)ei 'Po]/i, which are the operators measured in homo-
dyn e detection from Chapter 9. Then we have the input-output relation Combining all of the above, we have the measured spectrum as
for the interferometer:
S(D) = 1 + JC2 + 2/CScw(D)/h}QL· (10.60)
Xout(fl) = Xin(fl)
6E quation (10.58) defin es a two-sid ed spectral density so we tak e half of th at in defini-
Yout(D) = ~n(D) - JCXin(D) + Mc5hew(D)e-iDL /c (10.52) tion.
 Applications: Quantum J\Ieasrn·em cnt of Continuous Vori.ables 319
318 Q1rnntum Opt'ics For Experimentalists

I (I ' 1
Obviously 1 the last term is the signal while the firnt two terms are quantum mirror~
c•
noise contribution. So, t he minimum detectab le GW signa l is -N
I ll''

2 1 + K?
hiQL ( 1 )
Scw( D) = hsQL - Ye,+K , = ~ (10.61) 10 ''
2 mirror
mirror mirror
v.rhich h as a minimum of hiQL when K = 1 or Pa5QL= Ji.JD2 c2 /4w 0 . 10'"

To better und erst an d the physical meaning of the quantum noise ) 10 100 1000

spectrum in Eq. (10.61 ), we may rearrange the second exp ression in Q /2rr. (Hz)

Eq. (10.56) as (b)

(a)
Youl(fl) = + hne-i<JL/ c, (10.62)
SQL Fig. 10.11 ( a) Miche lson interferometer with power enhancement cavities. (b) Fre-
quency dependence of (Trace (i)) for Po= PtQL and hsQL(O) (Trace (ii)).
where we define a quantum noise operator

h n-
A = hsQL iSlL/c tn(D) - KX in(D)
v12e vl2fc
1 A A
ex ID ~n (D) - P,= Xin (D)
vPo o 0
whos e single -sided spectral density is the quantum limit for gravitationa l
wave detection. Ind eed, we find Shn (D) = Scw(D). The expression in
the second lin e of Eq. (10.63) helps us id entify the sources of the quantum
noise. The first term is inverse ly proportional to -,/N with N as the photon
number. This is just the standard quantum limit for phase measurement
due to shot noise in photon detection. Th e second term is proportional
to -JN or ( 6. 2 N) for coherent states and is thus caused by the radiat ion
pressure noise due to the fluctuations of the photon number.
To hav e a sense of what LIGO is dealing with, let us take some number
from LIGO design: Mµ = 30kg/4 (there are four free masses), L = 4000m,
n = 21r X lO0rad/s (GW signal frequency rv lO0H z ), Wo = 1.8 X 1015 /s ,
we obtain, hsqL rv 2 X 10- 24 / ./Hz and Pa5QLrv 1.5 X 108 W. This power
is impossibl e to achieve in the lab. So, the shot noise dominates in the
conventiona l Michelson interferometer. To reduce Pa5QL, LIGO resorts to
a design with power enhancement cavities as shown in Fig. 10.ll(a).
With the enhancement cav ity , it can be shown [Kimble et al. (2001)]
that the output field operators take the same form 7 of Eq. (10.56) but with
a different frequency response in the opto-mechanical coupling constant K
which is now changed to
K = 2,-y4
(P 0 / Pa5QL)
[t2('y2 + [12)
--------------
7
Because our definition of the quadrature-phase amp litud es in Eq. (10.56) is off by a
factor of .J2 from the cit ed reference, the la st term is off by the same factor.
Here
1
= cT / 4L with T as the transmissivity of the coupling mirrors and
(10.63) p_SQL= M £ 2 1 4 /4wo is the optimum input power to re ach hsqL at D = T
. -sQL ,.._,_, 4W
Take the numbers given before and T = 0.033 , we h ave P 0 1.0 x 10
,.._,_, •
This power is more manag eabl e in the lab, espe cially with a power recycling
mirror right before th e beam splitter. With Kin Eq. (10 .64), the minimum
detectable GW spectrum Sew takes the form of
(10.65)
where hiqL(D) = 8n/MD 2L2 . Figur e 10.ll(b) shows the frequency dep en-
dence of JScw(D), which has a minimum at D = 1 . It is dominated by the
shot noise term at high frequency: Scw(D » 1) ex D / Po, whereas at low
2
4
frequency, the radiation pressure effect kicks in: Scw(D « 1) ex Po/D .
The current noise performances of two advanced LIGO interferometers at
Hanford , WA (Hl) and Livingston , LA (Ll) are shown in Fig. 10.12 [Mar-
tynov et al. (2016)]. Narrow spikes are well-characterized suspension vi-
brations and calibration lines and 60 Hz electr ic power grid harmonics.
The noise are mostly from photon shot noise at frequencies higher than
150 Hz, and from other sources at low frequency. On September 14, 2015
at 09:50:45 UTC , the two LIGO detectors at Hanford, WA and Livingston ,
(10.64) LA observatories both detected in coincidence with the signa l GW150914
from the merger of two black holes of masses 36 and 29 solar mass [Abbott
et al. (2016)]. The signals from the two detectors are shown in Fig. 10.13.
 321
Applications: Quantum . l\1easu1·em ent of Continuous Va1·iables
320 Quantum Optics For Experimentalists

It thus becomes a fundamental problem related to quantum measurement.

So, the question is this: what role do es quantmn noise play in the noise
l· performance of an amplifier?
The study of quantum noise in amplifier started as ea rly as in 1960s
[Haus and Mullen (1962); Heffner (1962)}. It was generally believed that
quantum noise must be added through the internal degrees of the amplifier
in the amplification process to make quantum cloning or simple duplication
of microscopic observable impossible [Hong et al. (1985)1, for otherwise the
microscopic quantum superposition could be amplified to macroscopic scale
and Schrodinger cat state could be easily produced [Glauber (1986)}. The
added noise is the key in the decoherence process from microscopic quantum
- Advanced LIGO, L1 (2015) Enhanced LIGO (2010)
- Advanced LIGO, Hl (2015) - Advanced LIGO design world to macroscopic classical world.
10-24 Quantun1 theory of amplification was established around 1980s [Caves
102 103
Frequency, H z (1982); Ley and Loudon (1984)} and experimenta lly tested [Levenson et al.
(1993); Ou et al. (1993)}. A phase insensitive amplifier can be described
Fig. 10.12 LIGO noise performance data in 2016. R eprod uced from [Martynov et al. quantum mechanically by [Caves (1982); Ley and Loudon (1984)}
(2016)].
(10.66)
aout = Gain + F,
r-1
N
1.0 where G is the amplitude gain of the amplifier. In Eq. (10.66), the commu-
I
0.5 tation relation for Cloutrequires the existence of F, which is related to the
0
r-1 gain medium of the amplifier and thus is also dubbed the term "interna l
0.0
C degrees" of the amplifier. These degrees are usually unattended and left
(0 -0.5 in vacuum state, as shown in the conceptual diagram in Fig. 10.14. F is
I....
..µ
V') -1.0 responsible for spontaneous emission, which is a source of extra noise ad-
Hl observed (shifted, inverted)
ded to the signal from the amplifier. Even though F may be related to
0.30 0.35 0.40 0.45 8
many modes of the internal degrees of the amplifier, we can define a new
Time (s) operator a = frt/g with g = - 1. From Eq. (10.66), we can eas ily
0
show that [a , ab} = 1 so that a0 corresponds to a single mode annihilation
Fig. 10.13 The gravitational signals of GW150914. Adapted from [Abbott et al. 0
(2016)]. operator. When F( or a0 ) is in vacuum, the photon numbers of the input
and output are related by
10.3 Quantum Noise in Amplifiers (10.67)

10.3.1 Quantum Amplifiers in General Obviously , g 2 is the spontaneous emission of the amplifier and acts as the
Nois: pe_rformance of_a~ amplifier is always of great concern in the practical noise from the vacuum contribution of the internal degrees.
apphcat10ns of ma~mfym~ a small signal. Many techniques were invented If signal information is encoded in the quadrature -phase amplitude X =
~o sup~ress the n01se and mcrease the signal -to -n oise ratio. However even a+at, Eq. (10.66) becomes
m_a~ ideal ampli~er ~hen all the technical noise such as thermal n~ise is (10.68)
Xout = GXin + gXo,
~hmmat_ed, there 1s still quantum noise due to uncertainties generically em-
edded m quantum mechanics, as we have seen in the previous discussions. 8 Any phase of g is absorbed in operator ao.
 322
Quanturn Optirs For E1:pe1·imentahsts
Applicat-ions: Quantum li1casurement of Contimwus Variables 323

'A A
&::!. : a.Ill .-------. a OIi[,' ,----, A ~----~
:a. aAOll f ,:,
t
'-:::,y, : 0.
i In Amplifier Out :t :
•\;S/, Ill

Amplifier 0 ut

vacuum •
Bo internal
degrees
: :
:- - - - :
sq ueezed
1n ~----~
A
ao
t internal
degree s
vacuum
Fig. 10.14
Conceptual diagram for the quantum nois e in a n amp lifier.
Fig. 10.15 Concept u a l diagram for the quantum noi se reduction iu an amp lifier by
co uplin g the in te rnal modes of the amplifier to a squ eezed state.
where XJ =
-
aJ. + aj.t (J. --
A . ) .
out, z.n , 0 . Then the noise of the signal field at the
output of the amplifier as specified by the variance of
' C
Xout is. , given
_. b
y the extra noise in one quadrature-phas e amplitude where signal is encoded
x )-
(6 2 out - c2 (Ll" 2xA
in ) + g 2(6-? Xo). A

(10.69) can be suppressed whil e most of the extra noise is dumped into the unus ed
conjugate quadrature , as already implied in Eq. (10.69). This idea will
Hfere we as~ume the input signal field is independent of the internal mod work provided the internal modes a0 of the amplifier can b e accessed from
o the amplifier Norm 11 th · · • e
. . a ~' e mput is m a coherent state and the internal outside, as shown in Fig. 10.15. A parametric amplifier is such an amplifier
mode is unattended and is in vacuum. So (62 X· ) = ( " 2xA) _ .
· b · ' in L\ o - 1. This where the internal mode is simply the idler mode: a0 = bidl and Eq. (10.72)
case is est i_llustrated in Fi~. l~.14 by the noise circles in the input and becomes
output. Them are two contnb ut10ns in the output: the inside shad d . -
cle_r e~re~~nts the directly amplified noise from the input and the o:ts~:e (10. 72)
solid _nng is the ~xtra vacuum noise added from the internal degrees of th
amplifier. If the mput signal is denoted by (X· ) tl th t • _e where (6 2 xidl(0)) is the phase dependent noise from squeezed vacuum and
. l (X A in ' 1en e ou put signal is has a form of (6 2xidl (0)) = e- 2r cos 2 0 + e2 r sin 2 0 (Eq. (3.83)) with a
simp Y out) = G (Xin) and an output signal-to-noise ratio (SNR) is
squeezing parameter r . When 0 = 0, maximum suppression of the extra
(Xout) 2 G 2(X~n)2
SN Rout==---,-- 0 noise is achieved.
2
(6 Xout) c 2(6 2xin) + /g/2(62Xo) This strategy was first proposed by Milburn , Steyn-Ross, and \i\Talls
_ G 2 (Xin) 2 [Milburn et al. (1987)] and later demonstrated by Ou , Pereira, and Kimble
- c2 + g2 · (10.10) [Ou et al. (1993)], whose results are shown in Fig. 10.16. Here, a homodyn e
detection scheme is used to measure the quadrature-phase amplitude of
We_cai~ find the noi~e figure of the amplification, which is defined as the
the output field of a non-degenerate optical parametric amplifier (NOPA)
rat10 of the output signal-to -noise ratio (SNR) to that of the input:
whose idler mode is inject ed with a squeezed vacuum state. The NOPA was
NF== SN Rout = c2 c2 realized with a type-II optical parametric down-conversion process insid e
SNRin c2+g2 2c2_ 1· (10.71) a cavity which is resonant to both polarization components of the down-
T~is leads_ to the famous 3 dB degradation (1/2 = -3 dB) in SNR for hi l converted fields for gain enhancement. The theoretical model of such a
gam amplifier (G » 1) Th · -· f h. g1 device is presented in Section 6.3.4. In Fig. 10.16, trace iii ('1i0 ) is the
. . e ongm o t is degradation is the extra added
vacuum nmse from the internal modes of the amplifier. vacuum noise level; trace i ( <I>c)is the output noise level of the amplifier
when the idl er field (a0 ) is in vacuum. This, according to Eq. (10.69), is
10.3.2 Quantum Noise Reduction in Amplifiers G 2 + g 2 and gives the noise gain from vacuum noise of the amplifier. Trace
0)) is the output noise level when the idler mode is coup led to the
ii ( <I>(
Although the extra added noise cannot be avoided it can b d squeezed vacuum state. For the right phase 0, <I>(0)is below <I>crealizing
With th -1 b·i· , e rearrange
e avai a i ity of the squeezed states for quantum noise reduction. suppression of the extra noise of the amplifier.
'
 325
Applicalions: Q'Uant'Um J\leasurc rncnt of Conti111wus , 'ariabl es
32J Q'Uant'Um Optics For Experimentalists

·9····, I\ I\ : /0___
\
~----.Amplifier f--------:,____
,
t
:, : [/ill {/(}I/( : :, }
10 ii
<t>(0) correlated : : In ._____ -----"
Out ;:
noise
•
:____
o
: /\
00
internal
degrees :_________
:

Fig. 10.17 Conceptual diagram for the quantum noise reduction in an amplifier by
ent ang ling the int ernal modes of the amp lifier and the input signa l field. The output
noise leve l depends on the ph ase diff ere nce between the input and the int er nal rnode.
__ ______ __ _____ _______ __i~--
noise reduction here is applied not only to the excess noise of the internal
degrees but also to the input nois e. So, the amplified output may have a
ll1o better SNR than that at the input port.
From Eq. (10.68), we find the output noise as
-2 ---~~----..J--__J (10. 73)
8- 8-+rt
LO PHASE, 8 If the internal mode a0 and the input field CLin are correlated and the phase
Fig. 10 ·16 Measured noise levels at the out t f
is adjust ed correctly, there will be cancelat ion of the quantum noise between
fl.er whose idler mode is coupled t p~ o a non-degenerat e parametric ampli-
the correlated quantum fluctuations resulting in noise reduction, as we have
homodyne detection Reproduced fo a sq[Oueeze lsta(te. LO is the local oscillator in the
· rom u et a. 1993)]. shown in Section 10.1.3 with the EPR entangled state. In this case, the
input cannot be arbitrary but needs to be prepared in the entangled state
. Notice that
. a. dashed. 1me
· (t race
. iv· ) is
• drawn, corresponding to the di- with the internal mode of the amplifier. Since a non-degenerate optical
rectly amplified mput signal noise or the noise-free case with infi ·t parametric amplifier (NOPA) can output an EPR entangled state with
of squeezin g. so, th. is
· sch eme only suppresses the extra noise m ande amount
thus at correlated quantum noise [Ou et al. (1992b)], we use one beam of the EPR
best preserves the signal-to-noise ratio (SNR) during the am 1·fi t· entangled state to combine with a coherent state by a beam splitter with
cess (NF - l) h. Pi ca 10n pro-
. ' w ich corresponds to noiseless amplification. Since the t « 1 to generate a coherent signal S = ta for the amplifier, as shown in
ext~a n01se is added to the other quadrature, the quantum state is altered Fig. 10.18. The other correlated beam is coupled to the internal mode a0
durmg the amplification process and thus is not truly noise-free. of the amplifier, which in this case is also an NOPA with its idler mode as
the int erna l mode.
10.3.3 <:Juantu~ Correlation for Quantum Noise Reduction Amplifier
in Amplifiers Correlated fields r------
I
Coherent
I
1 aI\ 111
.
, ~--- '"
I Clo ut
Up to now, in the discussion of quantum noise in an amprfi . 1 : NOPA(G)i----+----
that internal m d (F) f . i er' we assume 1 State ex 1
. o es o the amplifier are independent of the signal (a. )
becau~e the mput field is usually arbitrary or the internal modes a. int
accessible
1 . . O n th e ot h er hand, quantum mechanics also allows the re no
corre- NOPA
at~on of quantum fluctuations, which gives rise to quantum entanglement __________
I
f.,l, V
]

an can be used to subtract out quantum noise, as we have seen in Sections

Fig. 10 .18 Schematic for an amp lification scheme by using an EPR corre lated state
10.1.3 and 1?.1.4. Therefore , if we make the internal mode of the amprfi .
entangled with th e mpu
· t mo d e, as shown in Fig. 10.17 the cor 1 t' i er generated from an NOPA with gain parametersµ,, v.
. .
their quantum · d , re a ion in
n01se may 1ea to the nois e cancelation in the output. The
 """'
326 QU,ant'llm Opt ic s For E xpe1·imenta.lis ts A pp lica ti ons : Quan tum .M easure m en t of Con t in uous Varia bles 327

Assume the NOPA for generating the EPR state has an amplitude gain (a) (b l
.72 ~------------,
of f-l while the NOPA acting as an ;:tmplifier is describ ed by an amplitude
gain G. With t « 1, th e output signal and nois e ar e calculated from
Eqs. (10.6 8) and (10.73) to be
(Xout) = GSGfo =
(6 X aut) = (µ + v 2)(G 2 + g2) - 41-wGg,
2 2 (10.74)
-86 L---------- - ~
where v = J
µ 2 - 1, g = JG2 - 1 and the relative phase is adjusted so 100 200 300

Time ( ms)
400 100 200
Time ( ms)
300 400

that the noise is minimum. Here , because t « 1, the contribution to the

noise performance from th e coher ent state is negligible. Then the output
signal-to- noise ratio is
A 2 2 A
SN Rout = (Xout) / (6 Xout)
0252
(10.75)
- (µ2 + v2)(G2 + g2) - 4µvGg'
which has a maximum value of G 2 S2 when we adjust µ of the EPR state
so thatµ= G. This is certainly better than the SNR of G2S 2/(G 2 + g2)
when the signal and the idler of the amplifier are uncorrelated and are in
coherent state and vacuum, respectively. 9 Notice that when µ = G, the
output noise of the amplifier is simply 1, which not only excludes the extra
internal noise of the amplifier (the solid part in Figs. 10.14 and 10.15) but
is even smaller than the input signal contribution as well (the shaded part
in Figs. 10.14 and 10.15 and the first term in Eq. (10.72)). This is because
of the destructive interference between the signal and idler input fields, as
expressed in the form of the addition of two field amplitudes in Eq. (10.68).
This only occurs when the input field is correlated with the idler field.
The proof-of-principle experiment was performed by Kong et al. [Kong
et al. (2013a)] with two optical parametric amplifiers achieved in atomic
4-wave mixing processes. The application to a fiber amplifier was done by
Guo et al. [Guo et al. (2016a)] . The experimental schematic is shown in
Fig. 10.18. The noise levels and signal levels are shown in Figs. 10.19(a)
and (b), respectively. "On" and "Off'' in Fig. 10.19(b) denote the "on"
and "off'' states of the input signal. Trace (i) is obtained when both the
amplifier and the entangled source are turned off. This corresponds to the
input of the coherent signal. Trace (ii) is obtained when the amplifier is
turned on but the entangled source is turned off. This corresponds to the
case of uncorrelated input to the amplifier. Trace (iii) is obtained when
both the amplifier and the entangled source are turned on and the phase of
9
This SNR can be obtained by settingµ= 1, v = 0 in Eq. (10.75).
Fi g. 10.19 (a) Noise levels for a n a mplifi er with an EPR ent a ngled stat e cou~l ed b etwe en
th e input signal and th e int ern al mod e . (b) Encod ed signal levels at th e mput (Tra ce
(i)) and th e output of th e amplifier with its intern a l mod e coupl ed to vacuum (Trac e
(ii)) or ent a ngled stat e (Tr ac e (iii)). Adapted from [Kong et al. (2013 a )].
the entangled source is such that the noise is minimum. This corr esponds to
the case of correlated input. Trace (iv) in Fig. 10.19(a) is the sam e as Trace
(iii) but with the phase of the entangled source scanned. The sinusoidal
change of the noise level indicates an interference effect between the input
signal and the idler fields as discussed before. The data shown in Fig. 10.19
clearly demonstrates the noise cancelation effect due to correlated input to
the amplifier. The output noise is reduced about 2 dB between Trace (ii)
and Trace (iii) in Fig. 10.19(a) with SNR enhanced about 2 dB as well in
Fig. 10.19(b).
Since correlation from entanglement is an extra resource here , it is not
fair to find the noise figure of the amplifier by comparing the SNR of the
output to that of the input directly. But we can treat the entangled source
and the NOPA as one amplification device and see how it performs. This
time, the input signal is the coherent state. It is straightforward to find
(Xout) = Gfo
2
(6 2 Xout) = G2t 2 + [µG~ - vg]
+[vG~ - µg]2 . (10.76)
This device has an effective amplitude gain of G' = Gt. The first term in
the noise expression is the contribution from the coherent state. The output
2 2 2
noise has a minimum value of 2G 2 t 2 - 1 when µ 2 = G 2 (1 - t )/(G t - 1).
This gives the output SNR~ut = G 2 t 2 \a\2 /(2G 2 t 2 - 1) which is the same
as Eq. (10.70) or the case shown in Fig. 10.14 when the amplifier's internal
modes are vacuum. So, we cannot use this scheme to amplify an arbitrary
signal, which is usually uncorrelated to anything in the amplifier. But we
can use the combination of the correlated source and the coherent state as
 329
Applications: Qua.ntuni J\Ieasmement of C'onfrnuous \lariables
328 Quant ,um Optics For Experimentalists

~· probe to some sa mpl es for signal encoding . If t he signal after the sam 1 a ____ a bout
ts too.w eak to be detected, we can us e the cur rent scheme to amp lify it ~0 ; 0
better measurement , as shown in Fi g. 10,20, NOPA
r-------
Correlated Source
7
Sample r - - A - rfi 7 b
1 jn/t ) _ _.__ ___I'-__ I M2 mp• ier I
I BSI In) I
I .----- Signal I I Fig. 10.21 Phas e-sensitive amplifier for noiseless amplificat ion.

I FOPAl Idler I FOPA2

wh ich means SN Rout = SN Rin or noiseless amplification . The drawback
I t-
L _!l - - - - .J L - - _!! 2~ 12_ .J here is that only half of the input is amp lified because bin is in vacuum. But
we can solve this by inj ect in g bin with sign al, as shown in Fig. 10.21 where
Fig, 10 .20 Amplification of a weak sign al from a samp le probed by correlated s .
Reproduced from [Guo et a l. ( 201 6 a)]. om ces .
the input signal is first equally split into two before send ing to both signal
and idler input port of an NOPA and the two outputs are comb ined with
another 50:50 b eam sp litt er . Using the notation in Fig. 10.21 , we rewrite
. The effect _of quantum noise cancelation in amplification by qu antum
· t er £erometer
mterference will be applied in Section 11 ·3 ·2 to tl1e SU(l , 1) m Eq. (10.77) as
for the enhancement of phase measurement sensit ivit y. a= Gao - geicpbb,
b = Gbo - geicpab, (10.80)
10.3.4 Phase-sensitive Amplifiers
where, for the sake of general argument, we include a phas e - eicp in g,
In the genera l discussion of the amplifier's noise in Section 10 ..3 1, we consi·-
d which is related to the phase of the pump field to the NOPA. For the beam
er~d the output port (aout) that is directly coup led to the input port (ain)- splitters, we hav e
In fact, there are other output ports related to the internal modes. These
ao = (ain + bin)/v12, bo = (bin - ain)/v12,
ot~er output ports also carry the amplified information of the input signa l
(10.81)
which we can make use of provided that we can get access to these modes. aout = (a - b)/v12 , bout= (b + a)/v12.
In the case of a non-deg~nerate optical parametric amp lifier (NOPA), we Combining the above with Eq. (10.80) , we have
have access to both the signal and idler (internal) modes in both the input A - G A. icpAt
and output sides. Then , Eq. (10.66) for a general amplifier becomes aout - am + ge ain
A A i At (10.82)
A - A At bout = Gbin - ge cpbin•
aout - Gain + gbin
Both Cloutand bout are decoupled from each other and are described re-
bout = Gbin + gaJn, (10.77)
spectively by a degenerate parametric amp lifier of the form of Eq. (6.135)
whe:e w_erepJace operator F for the internal modes with idl er mode opera- studied in Section 6.3.4 but with their pump fields 180° out of phase. Since
tor b. Smee bout also contains the inform ation about the input , let us add they are decoupled, we only need to consider one, say aout. With the cohe-
the two outputs: rent state input , the output photon number is
2
Xaout+ Xbout -(
- G + g)(Xa in + xb iJ.
A A
(10. 78) (alutClout) = 2+ i(la.12+ 1) + Ggeicpa.* + Gge-i 'Pa.
c 21a.1
2

Notice that this is an operator equation so that = IC+ gei(cp-2cpa) 12

la.12
+ g2. (10.83)

The last term is from spontaneous em ission. But the equ ival ent gain G' =
(Xaout + xb o?Lt)= (G + g)(Xa in + Xb,;J
2 2 2 2 = G2 + g2 + 2Gg cos( cp - <pa) is phase dependent. So,
IG + gei(cp - 2cp 1
0 )

(~ (Xaout + xb ouJ) = (G + g) (~ (Xain + xbiJ), (10. 79)

- Q

Applications: Quantnm l\feasnrcment of Cont-imWLLS Variables 331

330 Quant 'um Optics For Experimentalists

this amplifier's gain depends on the phase of the pump field relative to where the Wigner function W(,r. y) is defined in Eq. (3.167) for a single-
the input coherent state phase. This type of phase-sensitive amplifier was mode field.
The integral in Eq. (10.86) is known as the Radon transformation and
studied thoroughly by Caves [Caves ( 1982)].
For the quadrature phase amplitudes, we have especia lly can be inverted to obtain the ..Wigner function [Herman (1980): Leonhardt
(1997)]:
Xout = GXa; + gXa;n (<p),
171"
1=Pe(::r)I<(Xcos0+Ysin0-x)dxd0,
11

Yout = GYain - gXa;n (<p+ 1r/2), (10.84) W(X,Y) = - 12 (10.87)

21r O -(X)

where Xa;n(<p) Cline-i'-P+ aJnei'-P,Xain = Xa;n(O), Yain where the integration kernel is
Take <.p= 0, Eq. (10.84) changes to

Xout = (G + g)Xain'
K(x) =
2
1=\~\ej~xdr
-(X)
(10.88)

Yout = (G - g)Ya;n·
Notice that the above are operator equations so SN Rout = SN Rin or
noiseless amplification for X but attenuation for Y. Therefore, dege-
nerate parametric amplifiers can achieve noiseless amplification for one
quadrature-phase amplitude but attenuation for the other orthogonal one.
The drawback is that we need to phase lock the input phase relative to
the pump phase and the phase noise (classical) of the input can transfer to
intensity noise in the output. Experimental demonstration of the noiseless
amplification was performed by Choi et al. [Choi et al. (1999)].
(10.85) The function above diverges in the regular form so it can only exist as
a generalized function inside integrals similar to the Dirac's b-function.
Leonhardt presented a full characterization of this function and its use in
the inverse Radon transformation [Leonhardt ( 1997)].
Therefore, we can measure Pe (x) for a number of fixed 0 ranging from
0 to 1r and perform the inverse Radon transformation in Eq. (10.87). This
technique is similar to the CT scan in X-ray tomography [Herman (1980)],
and is thus dubbed "quantum state tomography".
Alternatively, we can rewrite Eq. (10.87) as

W(X, Y) = : 2 (K(X cos0 + Ysin0 - x))(x,e)• (10.89)

10.4 Complete Measurement of Quantum States: Quantum 2
State Tomography Strictly from the mathematical expression in Eq. (10.89), the Wigner
function W(X, Y) is just an ensemble average of the kernel K(X cos 0 +
In Chapter 9, we discussed homodyne detection and found that the homo-
Y sin 0 - x) over a pair of variables (x, 0), which can be measured expe -
dyne detection corresponds directly to the quantum measurement of the
rimentally by homodyne detection: the LO phase gives 0 and the photo-
quadrature-phase amplitude Xe = ae-je + ateje where 0 is the phase of
current output of the HD measurement gives x. Thus, we can make N
t~e local oscillator (LO) of the homodyne detection. Then, we can measure
measurements of (x, 0) by monitoring x while changing 0 in a controlled
Xe many times and record each outcome to build up the statistics. In terms
way such as phase scanning of the LO field. From these data, we may
of the language of statistics, we make an ensemb le measurement of Xe on
reconstruct the Wigner function as
the quantum state of the field, i.e., we measure Xe for a large number of the
identical states. An examp le of the ensemb le is a series of pulses and each 1 N
= + Ysin0m
pulse corresponds to a sample in the ensemble. In this way, we can obtain
the probability distribution Pe (x) for the quantum state of the optical field .
W(X, Y) lim - -
2
N-+= 21r N °
m=l
K(X cos0m - Xm)- (10.90)

In practice, there are a number of more efficient and accurate approaches

On the other hand, we found from Eq. (3.185) in Section 3.7.2 that the
to measure experimenta lly the Wigner function [Leonhardt (1997)]. The
marginal probability distribution of the quadrature-phase amplitude Xe is
technique of quantum tomography was thoroughly reviewed by Lvovsky and
related to the Wigner function of the quantum state of the optical field:
Raymer [Lvovsky and Raymer (2009)]. The first experimental measurement
Pe(x) = j dy W(xcos0 -ysin0,ycos0 + xs in0) , (10.86) of the Wigner function of a quantum state was performed by Smithey et al.
 Applications: QLLantum I\Ieasurernent of Continuous Variables 333
332 Quantum Optics For Experimentalists

on the squeezed state generated from a pulsed parametric down-conversion Classical lnformation
[Smithey et al. ( 1993)]. The Wigner function of a single-photon state was
measured by Lvovsky and J\Ilnyek , which showed for the first time the
nonclassical negativen ess of the Wigner function [Lvovsky et al. (2001)] .

10.5 Complete Quantum State Teleportation

w(}/If

EPR Entangled Bob

Quantum state teleportation is a transfer of the quantum state of a system Alice
Source,, NOPA (G)
''--- - - - - - - - --- - - -- - - -- - -'
to another different (likely remote) system without physically transporting
the original system. Sounds like some magic tricks? Yes, it is the magic
play of quantum entanglement. It is possible because of the existence of Fig. l0.22 Schematic of the quantum state tel eportation of an unknown sta te wit~
.
con t 1nuous vana· bles by EPR entangled sources from . a non-degenerat e OPA (NOPA) o
nonlocal correlations due to quantum entanglement [Bell ( 198 7)]. The idea
am.plitude gain G = cosh r.
was first proposed by Bennett et al. for teleporting an unknown spin state
of a particle to another [Bennett et al. (1993)] and demonstrated experi -
mentally by Bouwmeester et al. using photons in a polarization entangled
states [Bouwmeester et al. (1997)]. However, the teleported quantum states
of this scheme are in a discrete Hilbert space with finite dimension. Tele- is the output field to which the unknown state will be teleported. The EPR
portation of quantum states with continuous variables was first proposed entang led fields provide the quantum channel that transmits the quant~m
by Vaidman [Vaidman (1994)] and analyzed for a more practical system entang leme nt. Homodyne detections (HDA x and HDy) are mad~ by Alice
by Braustein and Kimble [Braunstein and Kimble (1998)]. The first ex- to measure Xf of the mixed field 1' and Y~ of the mixed field 2 after _the
perimental teleportation of quantum states with continuous variables was beam splitter. The results of the measurement are the ph~toc~rrents ix~,
performed by Furusawa et al. [Furusawa et al. (1998)]. iy_,, which are transmitted to Bob through classical c~mmumcat10n channel.
Both quantum state teleportation schemes by Bennett et al. and by B~b uses this classical information to modify field 3 m order to recover the
Vaidman require the sharing of quantum entanglement between the sen-
input unknown state Win. .
der and receiver. The discrete scheme is aided by the Bell polarization From Eq. (10.12), we have the Wigner function for the EPR sources as
entangled states (Section 4.4.2) while the scheme by Vaidmen is via the
EPR entangled states (Section 4.6.2). The implementation and confirma -
tion of a state teleportation in the discrete scheme by Bennett et al. can
be performed with the photon counting techniques discussed in Chapter 7 WEPR(x2, y2; X3, y3)
[Braunstein and Mann ( 1996); Bouwmeester et al. ( 1997)]. The scheme of = _l_ exp {- ! [(x3 + x2) 2 + (y3 -y2)2]er
continuous variables by Vaidman can be realized [Furusawa et al. (1998)] 41r2 4
with the homodyne detection technique discussed in Chapter 9, which we _! [(x3 - x2) 2 + (y3 + Y2/]e-r }, (10 .91)
4
will consider here.
To begin with, Alice and Bob share two fields (fields 2 and 3) of EPR
entang led sources from a non-degenerate OPA (NOPA) of amplitude gain
G = cosh rand Alice wants to te leport an unknown state in field 1 described where we changed the variables x1, Y1;x2, Y2 to X2,Y2;X3, Y3 to SUit th e
by its Wigner function Win to Bob. As shown in Fig. 10.22, Alice combines notation in Fig . 10.22 and r to -r for convenience later. .
From Eq. (6.72) in Section 6.2.3 for the transformation o~ the Wigner
one EPR entang led field (field 2) that she owns with the unknown state by
function for a 50:50 beam splitter, we have the Wigner function after the
a 50 :50 beam splitter. The other EPR field (field 3) that is owned by Bob
 Q'Uant'Um Optics For E1:pe ·l'irne11talists Applications: Q-uanfllm fi.Ieasw-e111c11t of Continnous , ·a,.iobl es 335

50 :50 beam sp litt er as which is dominated by field 2 and thus contains little information abont the
input state. This is because the field from the EPR entangled sources is in a
W B s (;1:~, Y~; .i:;,y;; .r·3, Y3)
therma l state of average photon number sinl? r if the other field is discarded
_
--exp1 2 .:r~+
{ -- 1 [ ( ::r3+---
, :r;)2+ (y3---- y~+ y;)2]e r
[Yurke and Potasek (1987)] and the superposed fields are dominated by the
41r 4 /2 /2
-i[( X3 - x\ -:n
x; ) 2+ (Y3 + y(-:ny;
)2]e-, }
large thermal state when r ---+oo .
However, when measurements are made on X{, Y1with resu lts ix~, i'r~,
the system is projected to a state of
I J I I
. ( X1 - X2 Y1- Y2)
X W in fri ' fri . (10.92) Pproj = Tr1 1
21 (Iix~, i1'j) (ix~, iy; IPsys)• (10.96)
v2 v2
If we discard the field going to Bob (field 3) and only care abo ut the The Wigner function of the projected state is then
fields for homod yne detection, we trace out field 3 by int egrat ing over x 3 ,
y3 in Eq. (10.92) an d the result is
Wproj (x3. y3) = J dx;dy~ W Bs(x~, Y~;x; , y;; X3, y3) Ix~
=ix~ ,y;=iy;

= ~
41r2
exp (- X§ \yj)
2 cos 1 r
j
dxdyVVin(x, y)

cosh r 1n 2
x exp { - - - [(x - ix~ v 2 - x3 tanh r)
2
(10 .93) +(y + iy; v'2- y3 tanh r ) 2 ]}

When Alice makes the homodyne measurement of X{ and Y1,she obtains 1 ( X§ + yj) Id , d , [ cosh r ( ,2 , 2 )]
= 41r2 exp - 2 cosh r x y exp - -2- x +y
photocurrents of ix~ and iy; which have a joint marginal probability of
x Win (x ' + ix; v'2+ X3 tanh r, y' - iy; v'2+ y3 tanh r)
P(ix~ , iy;) = j dy~dx;WBs(ix~ , y~;x; ,iy;)
1 ( X§ + yj ) J
d ,d ,c5( , , )
21rcosh r exp - 2 cosh r x y r x 'Y
l Id 'd' [ (ix; +x;)2+(y~ +iyc:)2]
21rcosh r Yi x 2 exp - 4 cosh r - x Win (x' + ix; v'2+ X3 tanh r, y' - iy; v'2+ y3 tanh r)
· I
ix~ - X2 Y1- iy;)
I · (10.97)
X Win ( y'2 , y'2
where c5r(x',y') = (cosh r/21r) exp [ - co~ l1r (x'2 + y' 2 )] ---+c5(x')c5(y') as
2 2
= ---- l ; ·
dxdy exp [ - (ix 1 v'2-x) +(iy,\1'2-y) ] r ---+oo. Th en, we hav e for large r
1

2
21rcosh r 2 cosh r
1 ( X§ + yj) . .
xWin(x, y) , (10.94) Wproj(X3,y3)~ h exp - h Win(x3+ix~v'2 , y3-iy;v'2)
27T cos r 2 cos r
where we made changes of x = (ix~ - x;)/\1'2and y (y~ - iy;)/\1'2. = P(x 3, y3)Win (X3 + ix~ v'2,Y3 - iy; v'2), (10.98)
This is a convolution between Win and a Gaussian of width cosh r. When
which is ju st the shifted input Wign er function weighted on the detection
r ---+oo, the Gaussian is much broader than vVin and can be pulled out of
probability in Eq. (10.95).
th e integral and Eq. (10 .94) becomes

P(ix~ , iy;)
1
h exp
27TCOS r
(
-
·2
i X' 1
cos r
·2
+h iy,2 ) I dxdyWin(x, y)
If Bob do esn 't do anyth ing on field 3, the output state is an integration
over ix,,1 iy,2 and is simply the thermal state described by P(x3, y 3), which
is just one side of th e EPR entangled state with the other side discarded.
·2 ·2
1 ixf + iy; ) However , if Bob, upon receiving the information of ix~, iy; from Alice after
exp ( - (10 .95) her homodyn e det ection measurem ent , makes displacements of X3 ---+X3 +
21rcosh r cosh r '
336 Quanhl '1n Ophr,s Fm· Expcrim ,enl,alists Appl-i.ccdions: Quantum Ji.leasu .rcm cnt . of Conh.n'llous Va'l"iables

ix~ J2 and y3 • y3 - i 1 ;; J2 on the transmitted field 3. he will obta in that the output \i\Tigner function is
lVproj(X3,y3) ex: lVinCt3,y3) in field 3, thus recover the input state and 1 { 1[ )') )2] -
1Vovt(x 1 , y1; X2.y2) = 7r exp -
4 (.r1 + J:2 - + ( Y1 - Y2 e
r
completely telcport the state from field 1 to field 3. 4 2
The displacement in Wigner phase space can be implemented by a co- - ~ [(x1 - ,r2/+ (y1 + Y2/]er}
herent state through a low-coupling beam splitter, in a similar way to that
discussed in Section 6.2.4 where a coherent squeezed state is produced by • Cc5(x1 - x2)c5(y1+ Y2) as r • oo . (10.100)
combining a squeezed vacuum state with a coherent state with a beam
sp litt er. The coupled -in coherent state is modulated in both amplitude an d Problem 10.2 Limit for the fluctuations of intensity difference of two
phase by an amplitude modulation for the displacement .of x 3 • x 3 +ix, 1 )2 classical fields.
and a phase modulation for the displacement of y3 • y 3 -iv. 2 J2. This stra -
1
Consider the intensities of two single -m ode fields: f A = T]AtA, f B =
tegy was used by Furusawa et al. in the first continuou s-vari ab le quantum
r]BtB with 7] being some constant.
state teleportation experiment [Furusawa et al. ( 1998)]. Since then, tele-
portations of a number of non-trivial quantum states have been achieved (i) Prove the following result of normal ordering operation:
in the lab, including squeezed vacuum states [Yonezawa et al. (2007)], a
(iA - iB) 2 = : (iA - iB) 2 : +T](L + iB), (10.101)
single-photon state [Lee et al. (2011)], and a "Schrodinger kitten" state
[Takeda et al. (2013)]. which leads to
2 (iA -
(!:::.. 2(L
iB)) = (: !:::.. - iB) :) + T]((iA) + (iB)). (10.102)

10.6 Problems (ii) Use the Glauber-Sudarshan P -representation to show

2(lA - fB)
(: !:::.. :)cl 2: 0 (10.103)
Problem 10.1 Wigner function of the state from a non-degenerate para -
metric amplifier with vacuum input. for any classical states of fields A, B with equal sign for coherent states.
From Section 6.1.2, we found that the output state from a non- This gives the classical limit for the fluctuations of intensity difference
degenerate parametric amplifier with vacuum input is simply a two-mode of two fields:
squeezed state, which can also be produced from two single-mode squee - 2(j A
\ l::.. -
2(j A
j B)) cl 2: \ l::.. - j B)) cs = T]( \j A) + \j B)) · (10.104)
zed vacuum states by a 50:50 beam splitter with 180° phase shift and r /2
squeezing parameter (Section 6.2.4) . So, we can start by writing down the Here "cs" stands for the coherent state . The right -hand side is also the shot
Wigner functions for the two single -mode squeezed states as (Eq. (3.176) noise level of intensity fluctuations.
in Section 3. 7.2)
Problem 10.3 Quantum entanglement and correlation in a non -degenerate
parametric amp lifier with an arbitrary phase of the amplitude gain g.
W1(x,y) = ~ exp [-
2
~(xer+ye-r)],
2 2
Consider a non-degenerate parametric amplifier whose input and output

W2(x,y) = ~ exp [ -
2
~(xe-r+y2er)].
2
(10.99) relations are given by
A= Ga+ g[}; i3 = cb + gat (10.105)

Here, we changed the notations x 1 , x 2 to x, y. Using the input-output with g = \g\ei'Pg. Here, the phase cp9 is set to be arbitrary and is usually
relation in Eq. (6.72) of Section 6.2.3 for a 50:50 beam splitter here, show determined by the phase of the pump fields in parametric processes.
 App/ica t ions: Quantum J.Jeasurnne1d of Continvous Vm'iablrs ;339
338 Qnantum Opt ics For E:r:perimentalists

(i) Construct the quadrature-phase amplitudes: XA (0A) = A.e-i 0,1 + A_tei0A (1980)]. In contrast, quantum information splitting by a beam splitter will
and XB(0B) = i3e-i0B +f3tei 0s. If inputs a,bare in vacuum , show that have vacuum noised added from the unused vacuum porL and have output
SNR degrad ed by 3 dB for shot-noise limited input [Shapiro (1980)].
(10.106) Consider a variation of the phase-sensitiv e amplifier shown in Fig. 10.21
and where the output beam splitter is taken out. VVestudy the two outputs a,
b of the NOPA as a function of the inputs a;n,bin.
([XA(0A) - XB(0B)] 2)
= 2(G 2 + lgl2) - 4Glgl cos(0A + 0B - cp9 ), (10 .107) (i) Find the relationship between a, b and CLin,bin·
which has a minimum of 2(G-lgl) 2 when 0A +0B = 2nK+cp9 (n = integer). (ii) For the coherent state \a) input Aat CLinand vacuum at bin, find the
photon numbers at the two outputs a,b. Show that they are phase -sensitive.
(ii) We now define
(iii) Show the noise at the outputs is phase-insensitive.
XA = XA(0A - <pg/2), XE= XB(0B - <pg/2);
(iv) At the phase with optimum outputs , show that the ratios of the output
YA= XA(0A - <pg/2+ K/2),
SNRs to the input SNR are given by [Levenson et al. (1993)]
YB= XB(0B - <pg/2+ K/2) = -XB(0B - (f)g/2 - K/2). (10.108)
2 A A A A SN Ra SN Rb 1( 2Gg )
~alculate yi (~A - XE)) and (L~2 (YA + YB)) and show (.6.2(XA - 1
SN Rin = SN Rin = 2 + G 2 + g 2 .
(10.111)
XB))(.6. 2(YA + YB)) = 4(G - lgl) 4 < 1 if (G - lgl) 2 < 1, exhibit ing the
EPR paradox. Thus, we have
SNRa SNRb 2Gg (10.112)
(i~i) F?r an injection of coherent state la ) with lal 2 » lgl2 rv 1, calculate SNRin
+---=l+c2
SNRin + g2'
(A), (B) and show the self-homodyne relation (Section 9.8)
which is larger than the value 1 of the classical limit for quantum infor-
2 + l(.A.)l[XA(cpa) - (XA(cpa))],
_At.A_~l(.A.)1 mation splitting [Shapiro (1980)] and approaches the maximum value of 2
i3ti3 l(B)l 2 + l(B)l[XB(((Jg - ((Ja) - (XB((f)g - ((Ja))]. (10.109) when g • oo.

(iv) From Eqs. (10.109) and (10.107), show the intensity correlation:
2
(.6.2(f _ >..i)) = (IA)+ >.. (iB) (10.llO)
A B (G+lgl)2 '

where fA = A.tA., fB = iJtiJ, and>..= G/lgl. This is exactly the same

as Eq . (10.24) . So, as we discussed at the end of Section 10.1.4, intensity
correlation in twin beams corresponds to one special EPR-type quadrature -
phase amplitude correlation (it is XA - XE for real injection and positive
amplitude gain of g > 0).

Prob lem 10.4 Phase -sensitive amplifier for quantum information tapping .

The phase-sensitive amplifier discussed in Section 10.3.4 can be used

to split quantum information without the degradation of signal-to -noise
ratio . This type of device was dubbed "quantum information tap" [Shapiro
 Chapter 11

Quantum Noise in Phase

Measurement

We have already discussed phase measurement sensitivity and the standard

quantum limit to it in Chapter 10 with a linear interferometer. But phase
measurement is not limited to linear interferometers. There are many sche-
mes for phase measurement. On the one hand, phase of an optical field is
introduced classically to describe the wave oscillation states of the field in
a continuous way. On the other hand, we find in quantum mechanics that
phase is not associated with any physical observable that can be represen-
ted by a Hermitian operator. So, we can only treat phase as a parameter
in a similar way as time in quantum mechanics. Furthermore, the concept
of phase is applied to any wave to describe its state of oscillation, which
can be sensitive to a variety of physical quantities. For example, matter
waves for atoms or electrons can form a matter wave interferometer whose
phase difference is sensitive to gravity. Therefore, measurement of phase
has a broad significance and is intimately related to precision measurement
of many physical quantities.
Since homodyne measurement is one type of phase measurement in
which the input field and the LO field int erfere to produce a phase sen-
sitive photocurrent, phase measurement is an application of this quantum
optical technique. Indeed, we studied phase measurement in Mach-Zehnder
and Michelson interferometers in Chapter 10. But because of its broad im-
pact, we shall devote this whole chapter to the discussion of general phase
measurement. We will present a general argument leading to the ultimate
quantum limit, i.e., the Heisenberg limit , for phase measurement and make
some general discussions. A new type of int erferometers based on nonlinear
interaction will be introduced. Finally, the technique for phase measure-
ment can also be applied to amplitude measurement of optical fields, which
will lead to the topic of joint measurement of both phase and amplitude.

341
 •
3"12 Quantum , Optics For E:rpc1·i111cntalists
11.1 Phase Measurement in General
It is well-known that quantum nature of electromagnetic fields leads to
limitations on how precise a physical quantity of an optical field can be
measured. It is generally believed that given an arbitrary state of light,
Heisenberg uncertainty relation sets the lower bound on the sensitivity of
the measurement. On the other hand, if we arc allowed to prepare the
system in some specific states, according to quantum theory of measure-
ment, a physical quantity can be measured to arbitrary precision, provided
that the states are the eigen-states of the operator representing the physical
quantity in quantum mechanics. For the phase of an optical field, however,
the answer is not so straightforward, mainly because of the fact that there
does not exist a Hermitian operator for phase in an infinite-dimensional
state space for a quantized optical mode [Dirac (1927); Reitler (1954); Sus-
skind and Glogower (1964); Caruthers and Nieto (1968)]. A more modern
approach [Pegg and Barnett (1988, 1989); Barnett and Pegg (1990)] in iden-
tifying a quantum mechanical operator for the phase in a finite-dimensional
state space leads to the following limiting state as the eigen -st ate of a phase
operator (which is also defined by a limiting process):
s of a variety of physical quantities.
10) = lim (s + 1)-1/2 eim01m), (11.1)
s-+oo
m=O
which resembles the eigen-state of the position operator (defined through a
limiting process). One problem with the phase state in Eq. (11.1), however,
is that the average photon number of this phase state is infinite in the
limiting process. Thus it is not a physical state, reflect ing the difficulty
encountered in the search for a physical phase operator. It becomes a
common consensus that with unlimited resource of energy, it is possible to
measure a phase shift to arbitrary precision but for a real system of finite
energy, the precision of phase measurement is limited no matter how the
phase is defined.
In addition to the theoretical approach in solving the difficulty in defi-
ning a quantum mechanical phase, another approach from the experimenta l
point of view was adapted in defining an operational phase as related to a
measurement proc ess [Noh et al. (1991, 1992, 1993)]. However, it encoun -
ters some problems at low average photon numbers. It turns out that the
problem lies in the discrete output in the photon-counting technique used
for the measurement [Ou and Su (2003)], which is an intrinsic property of
quantized fields (we will show lat er that this also lead s to the Heisenberg
Quantum Nois e in Phas e Ji.leasurem ent
limit in precision phase measurement). Thus. this proYicles another per-
spective 011 the quantum mechanical phase problem: phase is a classical
concept based on the continuous wave picture for the description of the
state of wave oscillation. This continuous v-r1:wepicture of phase is incom-
patible with the discrete energy of quantized fields in quantum description
unless the total energy tends to infinity so that the discreteness is trans-
formed to continuum, as we demonstrated in Section 2.3.3 and quantum is
transformed to classical by correspondence principle .
On the other hand, we do see a phase-like quantity in the quantum
superposition state description of the system:
(11.2)
where c c2 > O and quantity 6.cp has all the properties associated with the
1, - '
classical phase we are familiar with in optics because its change can lead to
a shift in the interference fringe in the measurement for the superposition
state in Eq. (11.2). But this phase depends on other physical quantities such
as the propagation distance for optical fields and magnetic field for atomic
states . Since interferometer output is very sensitive to the change of this
phase quantity, its determination is the basis for the precision measurern.ent
Usually, the phase quantity discussed above comes into play as a part of
the mode function such as eikz for optical fields or time evolution function
eiwt with w depending on the magnetic field for the atomic states. In these
cases the phase quantity is a parameter that depends on other physical
quan~ities. Therefore, everything comes down to the problem of est im ating
the phase parameter 6.cp through its action on the state of the system, as
shown in Fig. 11.1.
Fieldo •
j4>)
•
j4>')
Phase
LM_ea_s_ur_e_1n_e1_1t~
Phase Shift S
Fig. 11.1 Phase shift 8 on an optical field ii to change its state from I<I>)to I<I>')·
Particularly in the quantum theory of light, if a phase shift 6 is induced
by a linear optical element on a single mode optical field, it can be described
by a state evolution unitary operator:
(11.3)
U5= exp(inb),
 4
Quantllm Nois e i11 Phas e .Measur em ent 345
344 Quantum Optics Fol' Exp erim entalists

Heisenberg limit:
where 11= ata is the number operator with a as the annih ilat ion ope rator
1 (11.6)
for the optical mode (see Section 3.2.3 for the evolution of a sing le-mo de 6</>HL (N)"
field) . If the optical field is in state IcI>),the state after the phase shift is
then lcI>')= U0 lcI>)(Fig. 11.1). We can then make a quantum measurement that Heisenberg limit should be understood as an asymptotic limit
Not e . t 1 es
on lcI>')to estimate the phase shift c5. In fact, the measurement for the at large mean photon number, that is, the ph~se u:1certam y approac 1
est im ation of c5 plays an essential role in precision measurement and ha the order of (N)-1 for large (N). We treat it m tlus way throughout the
been widely used in practical applications as well as in fundamental studie . chapter. .
The most common way to measure 5 is the interferometric method by However, Shapiro et al. proposed the followmg state
comparing lcI>')with some reference field. Homodyne measurement is a kind
of interference between the input field and a strong loca l oscillator (LO) \cI>)ssw=A0
l \ )
rn+l m
(l\I » 1, A~ J6F(i) (11 .7)
with the LO acting as the reference. In Section 11.3, we will introduce some m=O
genera l schemes of phase measurement which are different from traditional s an optimum state in precision phase measurement [Shapiro et_al. (1989)].
2
interferometric method such as a Mach-Zehnder int erferomete r. ~urprisingly, this state has a photon number fluctuation (6 N) m th~ o:·der
of exp((N)/A2) for large l\f. Hence, Eq. (11.4) would le_ad to a l~m~t of
2), which is much better than the Heisenberg lnmt of
11.1.1 Ultimate Quantum Limit in Precision Phase 6 </> exp(-(N)/A
Measurement Eq. (11.6) for larg e (N). _
As a matter of fact, the validity of Eq. (11.4) is not general [Susskmd ~nd
Since we only have finite amount of energy in a rea listic physical world, Glogower (1964)]. For exam ple, for the state of v~cuum, _the left -han .d side
we will limit our discussion under the finite energy constraint throughout of Eq. (11.4) is obviously zero thus violating the meq~ali~y. Therefore, the
this chapter. But can we still measure phase to an arbitrary precision with arguments based on the Heisenberg uncertainty rela~10n m_E~- _(11.~) _c~n-
finite energy? In some sense, the failure to find an eigen-state of phase not hold in general and the question remains: What is the lnmt m preci~10n
with finite photon number implies a negative answer to the question. The phase measurement given the available total mean nu~b~r _ofphotons_• _In
traditional and straightforward argument for the limit on the precision in the following, we will show that the ultimate quant~m- lm~it m_the precision
phase measurement comes from the Heisenberg uncertainty principle for of phase measurement is precisely the Heisenberg limit given m Eq. (11.6).
phase and photon number [Dirac (1927); Reitler (1954)] :

(11.4)

where 6</>and 6N are the fluctuations for phase and photon number,
respectively. Therefore, shot noise (6N = j (6 2N) rv '1{fi)) due to par-
ticle nature of light will place the so-called shot-noise limit or the standard
quantum limit (SQL) on the sensitivity of phase measurement:
Fig. 11.2 Phase m easu rement by a Mach -Z ehnder int erferomet er.
6 'f/SQLrvM"
~ >-1- (11.5)
A simple argument
On the other hand, quantum mechanics does not set any restriction on
Classically, phase is just the argument of the complex field amplitude used
the fluctuation 6N of photon number. Intuitively, one would argue that
to describe an optical field. Many factors may change the_ value of phas~-
because of energy constraint, 6N should be bound by the mean number
The traditional method of measuring phase shift is interferometry. Th_is
of photons, that is, (6 2N) rv 0( (N)2). Thus given total mean number of
method relies on optical interference effect for the comparison of phases m
photons, the limit on precision phase measurement should be the so-called
3cJ.6 Qnant.um Optics For E:i;p erim entalists
two paths. If we fix the phase delay of one path, any detected change in the
output intensity of the interferometer will indicate a phase shift experience d
in the other path leading to a measurement of the phase shift. To be
more specific, as shown in Fig. 11.2. a coherent optical field is split by a
beam splitter into two fields which are later recombined to form interference
fringes. If the interferometer is properly balanced, the output intensity has
the form of
lout= Iin(l - cos¢)/2, (11.8)
where Iin is the intensity of the input field and ¢ is the relative phase shift
between the two interfering paths. If we have a well-defined amplitude in
the input field, any change 6Iaut in the output intensity must come from
the change 6¢ in the relative phase. The sensitivity is highest when we set
¢ = n/2:
(11.9)
Classically, there is no limit on how small the change 6Iout in intensity
can be. Therefore, in principle, there is no limit on how small a phase shift
6 ¢ can be measured. In quantum theory, however, the particle nature of
light does not allow infinite division of energy thus setting a lower limit on
6Iout· We can rewrite Eq. (11.9) in terms of photon number as
(11.10)
where Nin is the total input photon number and 6Nout is the change in
output photon number. The minimum 6Nout that is allowed by quantum
theory is simply one, corresponding to the change of one quanta. Therefore,
the quantum limit for phase measurement is
1 interference effect [Mandel (1991); Englert (1996)]. The visibility will be
6A->- (with N = Nin/2) (11.11)
'Y - N'
which is the Heisenberg limit. N = Nin/2 is the total number of photons
in the arm of the interferometer that experiences the phase shift.
Furthermore, if classical states of light are used as the input to the
interferometer, photon statistics at the output is at best the Poisson dis-
tribution, i.e., 6Nout 2: ~- But for optimum sensitivity at ¢ = 1r /2,
Nout = Nin/2 = N. Hence, from Eq. (11.10), we arrive at the standard
quantum limit in a conventional interferometer [Caves ( 1981)]:
6 </JsQL2".1/VN. (11.12)
Thus nonclassical state of light must be employed in order to surpass the
standard quantum limit to reach the Heisenberg limit .
Qnantum . Nois e in Phas f' l\Icasur em f'nt 3.!7
The semiclassical argument c1.bovcfor the Heisenberg limit is indepen-
dent of the input field and nms equally well if we describe the optical field
quantum mechanically. However, it is limited to the specific scheme of in-
terferometry for phase comparison and to the detection scheme of intensity
measurement. Nevertheless, the semiclassical argument above clearly de-
monstrates that it is the quantization of optical fields that is responsible for
the Heisenberg limit in precision phase measnrement. Next, we will present
another more general argument for the Heisenberg limit that is independent
of phase measurement scheme .
A general argument by the complementarity principle of quantum
mechanics
The complementarity principle of quantum mechanics [Bohr (1983)] con-
cerns the particle and wave duality of light. Although light exhibits both
wave-like and particle-like behavior, it is impossible to observe both of them
simultaneous ly. When we apply the complementarity principle to the phe-
nomena of interference, we find that it is impossible to obtain the complete
which-path information for the two possible interfering paths of a photon
and to observe in the meantime the interference effect in a single expe-
riment. In other words, the interference effect will disappear if we know
exactly from which one in the two possible interfering paths the photon is
coming to the detector, whereas the appearance of interference is always a
manifestation of intrinsic indistinguishability of the path of the photon. In
a more quantitative language, the fringe visibility of interference and the
which -way information are related through an inequality which sets an up-
per bound on the amount of which-way information and the visibility of the
zero if we know exactly which path the photon goes through whereas 100%
visibility in the int erfere nce pattern leads to no knowledge of the which-
path information at all. If we have some partial information about which
path the photon goes through, the visibility of int erfere nce will lie between
0 and 1. Furthermore, if, without disturbing the interference system, there
exists a possibility, even in principle, for the distinction of two interfering
paths, all interference is wiped out. Notice that it is not necessary to actu-
ally carry out an experiment for the distinction in order for the interference
to disappear. The mere possibility that it can be performed is sufficient
to suppress the interference effect [Zajonc et al. (1991)]. This supplement
of the complementarity principle is the key in our next argument for the
 Qu.ant1lrn Noise in Phase J.Icasuremc11t 3.J.9
:348 Qnant1c1n Optics Fm· Expc"l'imcntalists

fundamental quantum limit in precision phase meas urement.
Consider a single-photon interferometer shown in Fig. 11.3. In one of
the interfering paths (Field A), we add a device which couples to another
field labeled as ii and allows the measurement of the photon number in .A
without destroying the photons. This is a kind of quantum nondemolition
measurement ( QND) of photon number [Braginsky et al. ( 1992)]. In this
way, it is possible to obtain which-path information for the single photon
without destroying it (no disturbance to the photon number of the interfe -
rence system). It is known [Milburn and Walls (1983); Imoto et al. (1985);
Kartner and Haus (1993)] that the optical Kerr effect can be used to imple-
ment a QND measurement of the photon number. In this case, the field to
be measured (A in Fig. 11.3) imposes a phase shift on another beam called
the probe beam (ii in Fig. 11.3). Measurement of the phase shift on the
probe beam provides the information about the photon number in A and
will influence the interference pattern. Thus, this interference system provi -
des us a platform for the discussion of the precision of phase measurement
in connection with the complementarity principle. Sanders and Milburn
were the first to make such a connection [Sanders and Milburn (1989)] . We
will use it to derive the limit for precision phase measurement.
I1)
B Let us now examine the visibility of the single-photon interferometer
A with the QND device in path A. Assume that a single-photon state is fed
Observation
of Interference
I<I>) Uqnd
l;f,.I) Phase
Field a ':L'
genera l state l<I>)similcu to Fig. 11.1 for the probe fielcl a..Then the output
state for the two fields is
Uqndll)Al<I>)a= ll)Aei"atal<I>)o = ll)Ai<l?')a. (11.H)
Thus, according to Eq. (11.3), the probe field ii is subject to~ phase shift~-
imposed by the input of a single photon in the signal field A. 1\Iore_over,if
the input state to field A is an N-photon state IN) A, it is straightforward
to see that the phase shift on field ii will be N 1-1,. So; l't is simply the phase
shift on field ii induced by a single photon in field A.
Next, we perform some measurement on the probe field ii to es:im~te
the phase shift (Fig. 11.1). If we can detect the phase shift in field a with
precision better than K, by whatever means, we will be able to tell whether
a photon is in path A or not. Hence, if we use this device i~ one arn: of
the single-photon interferometer. we will know the which-path mformatio~,
and according to the complementarity principle, the interference effect will
disappear. On the other hand, if we can observe a 100% visibility _in the
single -photon interferometer in Fig. 11.3, it implies that it is impossible to
detect the phase shift"' in field ii no matter what kind of method or strategy
we use for the extraction of the phase shift. Therefore, the visibility of the
interferometer is directly related to our ability to resolve the phase shift K,
due to a single photon.
into one of the input ports of the interferometer, and to be more general,
we assign a mixed state described by the density operator Pa to the pro_be
field a. Thus the input state for the total system is described by the density

Measurement operator
Ptot = 11)(11@ Pa· (11.15)
Fig. 11.3 Single-photon Mach -Zehnder interferometer with a QND measurement device
for photon number in one of the arms to obtain which -path information.
After the first beam splitter, the state for the system becomes
P~ot = 11/J)(1/JI@Pa, (11.16)
In the QND measurement of photon number by optical Kerr interaction,
with
two fields A and ii (one is called the signal wh ile the other the pro be) are
11/J)= (11)AIO)B + IO)All)B). (11.17)
coup led through a Kerr medium and the state evolution is determined by
the unitary operator [Kartner and Haus (1993)] After passing the QND device, the state of the system has the form of
t - iK"-0.
UA 't"A A/ e-iK"-a
ta.A . ,t,'t;i
aA
a.;itA
iK"-a.t A// A A/
Ptot = UqndPtot qnd - e
Uqnd e , (11.13)
A -

- Ptot

where K, is a parameter that depends on the strength of the interaction and = 1(1lA, OB)(lA, OBlei"'a.ta.Pae-i"'a.ta.+ IOA,lB) (OA, lB I
is adjustab le. To see further the physica l meaning of K,, let the input state
+IOA, lB)(lA,OBIPae-i"'a.ta. + llA,OB)(OA, lBlei"'a.ta.Pa)· (11.18)
to the QND device be a single -photon state for the signal field A and a
 Quantum No'i.se i11 Phase }.frasuTem ,c11t 351
350 Quantum Optics For Experinienlalists

From this state, we can calculate the probability of detecting

a photon the total me an photon nmnb er in field a must satisfy (N) 1/ f-t., which
at one of the output ports of the interferometer: P± = (.A.~.A.±)with provides a lower bound on the photon numb er requir ed in field ii in order
A.±= (.A± ei1>B)/ /2. It has the form of to resolv e a phase shift i-i,. 1\Ioreover. Eq. (11.24) can also be written as

1 1-v (11.25)
p± =
2 [1 ± V COS ( <J)- E)], (11.19) h. > (N) ,
with the visibility which sets a lower limit on the minimum detectable phase shift, given the
total mean number of photons available in field a. If a phase shift K can
(11.20) be resolved by whatever means , as the previous argument shows, this will
result in the disappearance of the interference pattern, or v rv 0. From
and E as the phase of T:r(ein;atar\) [Sanders and l\,1ilburn (1989)]. For a
Eq. (11.25), we have K 1/(N). So the minimum detectable phase shift in
single mode field, Pahas a general form of
field a is of the order of 1/ (N) or the Heisenberg limit.
(11.21) The argument above is for the case when there is only a single-mode
m,n in field a. With a multi-mode field probing the phase shift, the problem is
in the number state base. Therefore, the visibility of the interference pat- equivalent to multiple phase measurement schemes [Shapiro et al. (1989);
tern is Braunstein et al. (1992); Lane et al. (1993); Braunstein (1992)]. The mul-
tiple measurement schemes divide the available energy into multiple fields
(11.22) with each sensing the same phase shift. Optimization of the measurement
strategy can be performed based on quantum information theory for the
with Pm = Pmm· For the purpose of comparison with the unit visibility, let estimation of the phase shift. For this measurement scheme, we can make
us evaluate the quantity 1 - v as follows: a similar argument as the single- mode case but a modification of the uni-

1- 1-1~
V cc Pmeim•I <; 11- ~Pmeim,1
tary operator for QND measurement is needed to include the coupling to
multiple fields. The modified unitary operator has the following form:
A At A~ t ]
Uqnd = exp [iKA A L...tajaj ' (11.26)
= 21~Pmcim•/ 2
sin~" I <; 2 I
Pm sin~" I, (11.23) j

where a multi-mode field with modes characterized by the annihilation

where we used Lm Pm= 1. By using the inequality sinx < x in Eq. (11.23), operators { aj} is coupled to a single-mode of A.that is one arm of a single-
we end up with the following inequality photon interferometer. Note that the unitary operator in Eq. (11.26) in-
1-v volves a hypothetical interaction that is used here purely for the s~ke of
1 - v < (N)K or (N) > -- (11.24) argument. It can be easily checked that a single photon in field A will
K",

with (N) = Lm mPm as the average photon number in field a. This
inequality sets a lower limit on the total mean number of photon required
in field a in order to resolve the phase shift of K in the phase measurement
of field a. The argument runs as follows: When it is possible, by whatever
means, to resolve the phase shift Kin field a,we can tell whether the photon
entering the interferometer passes through path A.or B in the single-photon
interferometer. Since we know the which -path information, according to
complementarity principle, the interference effect in the interferometer will
disappear or equivalently, v rv 0. Thus from Eq. (11.24), we find that
induce a phase shift K in all the modes {CLj}. Joint measurements on all the
modes can be performed to estimate the phase shift. By following the same
line of argument as the single mode case above, we can easily show that the
1
precision in the joint phase measurement cannot be better than (Ntat)-
with (Ntat) = Lj (a}aJ),the total mean photon number in all the modes.
Therefore, we have generalized the proof for the fundamental limit to the
multi-mode case. Furthermore, we did not specify here how the energy is
distributed among different modes. Thus the argument applies to the case
of uneven distribution of energy as well as to the case of equal partition of
 4
352
Quantum Optics For Experimentalists
Quanh,m Noise i11 Phase Ji.Jeasuremenl 353

energy in all fields involved [Shapiro et al (1989). B. .

L t 1 ( __ c_. , raunstem et al (19g 2) which reflects the fact that it is possible to make a precise measurement of
ai~ 1993); Braunstein (1992)] . --· ;
phase in this state no matter how small the phase shift is. With a finite
otice that although the loss of int erference (v rv 0) 1. l' h
to resolve the h l ·f . e ies on t e ability s. on the other hand, we have v = lsinc[(s + l)r,;/2]/s inc(h-/2)1 and v is
p ase s 11 t K, , it does not require actuall ·£ . .
measurement of the phase shift. Actuall th . per ~rmmg the different from 1 only when r;_ 2:,2/ s = l / (N)e.
field aA . ffi . . Y, e mere passmg of the prob
. is su cient to wipe out tl1e smg · le-p l1oton interference effect '] e (iv) For a number state lj\I), v = l , and it is impossible to resolve a phase
11o_wit works , note that while the single photon in field A . o ee shift no matter how large (N) = j\J is. This reflects the fact that no phase
shift of K, in field a, the effect of field a on field A . tl can cat~se phase fluctuation is induced by the photon number state and that the number
induce l h'f is 1e same: it will also
. a p Jase s i t depending on the photon numbers in field If he A state is not good state to measure the phase shift K, due to its random
is a large fluct t · · h a. t ere
ua 10n m t e photon number of field ·t ·11· phase nature.
phase of field A. d ·f
A

a, i wi impose on the
tl . an i the phase fluctuation is large enough due to ·th
(v) From the examples in (i) and (ii), we find that photon number fluctua-
. 1e size_of K, or the p~oton number fluctuation in a, the interference ;;ir er
m ~he smgle-photon mterferometer will be averaged out l d' 1ge tions play some role for a vanishing visibility in the single -photon interfero-
This effect b b , ea mg to v rv 0 meter. But there is no direct connection. For examp le for the phase state
form of . e/een y the following examp les where we find the exp lici~
of Eq. (11.7), which has a large photon number fluctuation, v l - 6K,/7r
. . ~i~i i i y o~ some known states. It is straightforward to calcu la
the visibility for various known states by using Eq. (11.22): te when (N) > > 1 and r., < < 1. Therefore, for r;, < < 1, v l, which means
that the state in Eq. (11.7) is not suit ab le for the probe field a to sense
(i) For the coherent state /a), v = e-lal 2(i-cos"') e-(N)"'2; 2 . a small phase shift. Indeed, there are a number of studies [Schleich et al.
[Sanders and Milburn (1989)]. v rv O when (N) >> 1/ 2 h' ~I- K, <<_ 1 (1991); Braunstein et al. (1992); Lane et al. (1993)] showing that the state
tent with the shot-noise limit of 1/ !Tfi') . h K, ' w ic is consis-
£ h V \1 v J mp ase measurement sensitivity in Eq. (11. 7) has a phase distribution that is not appropriate for measuring
or co erent state interferometry. In this case the ph 0 t b the phase shift in high precision.
ation is (62 N) _ ; N) h' ' on num er fluctu-
. - \ w ich leads to a phase fluctuation of Li¢ = K, !Tfi') It is puzzling to notice from examp le (ii) above that for the thermal
for the smgle-photon interferometer. When Li¢= K, !Tfi') rv . V \lV J
state, the disappearance of interference pattern (v rv 0) is not necessarily
1
rence fringe will diminish. V \ vJ 7r, the mterfe-
related to the existence of a scheme of measurement on the state to resolve
(ii) For th e thermal state described by l the phase shift r.,, for we have from examp le (ii) v rv O when (N)r., >> 1
t 1e density matrix but the phase-shifted thermal state p' = UpUt = p does not contain any
L.,n Pn/n)(n/ with Pn = (N)n/((N) + lr+l ,
A

Pth
information about the phase shift. This fact seems to contradict the com-
1 plementarity principle , which states that interference should always occur
V=r:,--~~~-:--:-:-~-~--
[l + 4(N) ( (N) + 1) sin 2 r.,/2]1/2 whenever there does not exist in principle a method to find the which-
1 path information. To solve this puzzle, we point to the fact that mixed
[l + (N)2 r.,2p;2 for (N) > > 1 and K, < < 1. (11.27) states are obtained after tracing out some other correlated states as a re-
sult of our lack of interest or ability to know these states (e.g. reservoir
No~ice_that is significantly different from unity only when K, >
which is consistent with the Heisenberg limit Likew· th l rv 1/(N), fields for thermal state). Once we enlarge the state space to bring in these
flu t t' f ·
c ua 1011 o a thermal state is (Li2N) = ;N)2 ised he P. 1oton number correlated states to make a pure state, the whole system will carry the in-
fluctuation is L'.i,1-,_ ; ) . . . \ an t e mduced phase formation of the phase shift. The question is then: Does there always exist
'+' - r.,,N • This is consistent with Eq. (11. 27 ).
a phase measurement scheme that can resolve the phase shift r., whenever
(iii) For the phase state in Eq. (11.1), this modified state is utilized in field a for sensing the phase shift and causes
v = 0 in the single-photon interferometer? We will address this question in
v = lim _l_
s---+oo s +l
Isin(ssin+r.,/2l)r.,/21 _- O for any K, =J.0, (11.28) Section 11.1.3 when we discuss the general scheme of phase measurement.
 .'35l Quantvm Optics For E:rpcrimentalists Quantum Noise in Phase .Aleasure111e11t

11.1.2 A Necessary Condition for the Heisenberg Limit ,·isibility, indicc1ting that it is impossible to resolve the phase shift of size
1-i. rv 1/ (N) no matter what we do on field c1. Thus in order to obtain the
I_t i~ known that squeezed state interferometry can achieve the Heisenberg
sensitivity at Heisenberg limit in pha se measurement, we mus t uLilize sLaLes
h1111t[Bondurant and Shapiro (1984)]. Som e other schemes [Ynrke et al.
sa tisfyin g
(1986); Holland and Burnett (1993); Jacobson et al. (1995)] were discovered
that have the same sensitivity. But it is not common for a phase measure- (~ 2N) (N)2 (11.33)
ment scheme to achieve the Heisenberg limit. For example. coherent state
for sensing the phase shift. Notice that the condition in Eq. (11.33) is only
interferometry only reaches 1/ ,/(N) sensitivity, or the standard quantum
a necessary condition, because we have a counter-example in the state in
limit (SQL). As we proved in Eq. (11.12), if classical sources are used in
Eq. (11.7), which has a large photon number fluctuation but gives v rv 1
a conventional interferometer shown in Fig. 11.2, the sensitivity is always
as we showed in example (v) above. It can be easily checked that the
limited by SQL or 1/ ,/(N). To achieve the Heisenberg limit , nonclassical
phase measur ement schemes that have been discovered so far to achieve
sources must be used. So what are the general requirements for the optical
Heisenberg limit all utilize states satisfying the condition in Eq. (11.33).
fields that can achieve the Heisenberg limit when they are employed in a
phase measurement scheme?
Let us now consider those states which have relatively small photon 11.1.3 General Consideration in the Search for Schemes
number fluctuations so that Reaching the Fundamental Limit

for large (N). From the necessary condition derived in the previous section, we find that
(11.29)
in the search for phase measurement schemes that have the sensitivity rea -
We will use these states in the probe field a in the single-photon interfero-
ching the Heisenberg limit, we must first look for those states that satisfy
meter with a QND measurement device having a coupling constant
the necessary conditions in Eq. (11.33) . Then we need to construct a scheme
1 in which we employ these states in field a for sensing the phase shift. So far ,
,.,, rv (N), (11.30)
there have been a number of schemes that reach the ultimate limit in phase
which is also the phase shift in field a induced by a single photon in field A. measurement [Bondurant and Shapiro (1984); Yurke et al. (1986); Holland
Assume further that photon distribution Pm for these states is smooth so and Burnett (1993); Jacobson et al. (1995)]. Among them, some are the
that Pm rv O for those m with lm-(N)I > J(~ 2 N). Then the contribution conventional interferometers with different detection methods [Bondurant
to the sum in the visibility formula in Eq. (11.22) only comes from those and Shapiro (1984); Holland and Burnett (1993)] while others utilize un -
terms with Im - (N) I ;S J (~2
N) and we can approximate Eq . (11.22) as conventional interferometers which do not use beam splitters as their wave
divider [Yurke et al. (1986); Jacobson et al. (1995)]. In the following, we
V"" I L Pmeim,I· (11.31)
will make a general discussion and thus will not limit it to a particular type
of interferometer.
In order to detect the phase shift, we will take measurement on field a.
But for Im - (N)I ;S J(~2 N), because ,.,,J(~ 2 N) << 1 as a result
However, direct photo-detection does not revea l any information about the
of Eqs. (11.29) and (11.30), we can approximate eimK, with ei(N)K, and phase of the field . Therefore, we need to first transform the state of the
Eq . (11.31) becomes field a into some other state for which photo -detection is sensitive to the
phase (e.g. homodyne). Let the state of field a be l<I>') with or l<I>)without
v"' I L Pmei(N),1 L Pml = 1.
lei(N)K, (11.32) the phase sh ift . Consider a unitary operator U which corresponds to some
m phase measurement scheme. It operates on the state l<I>) or l<I>')
and resu lts
in the state:
Therefore, with those states satisfying Eq. (11.29) in field a and a phase
shift of size ,.,, rv 1/ (N), we can observe interference pattern with 100% lw) = Ul<I>),or lw') = Ul<I>'), (11.34)
:35G Quanl :u,m, Optics Fo , · Expe1·imenl alists Qv,anturn Noise in Phase l\tfeasurenie11/ 357

which will be phase sensitive, that is, detection on I'11')will result in signifi- Let us now go back to the question raised at the end of Section 11.1.1
cantly different outcome from 1'11).Our goal now is to detect the difference about whether there always exist a phase measurement scheme that can
between I w) and I w'). This can be easily achieved if we select those states resolve the phase shift 1-1.whenever 1 = 0 in the single-photon interferometer.
1

for 1'11)such that detection on it will yield null result whereas detection on Vie will look for a scheme for phase measurement. Consider first the case
1'11')gives nonzero result. One such state that can serve as 1'11)is simply with a pure state in the general form of
the vacuum state. Thus any detection of a photon in the state I w') will be (11.36)
an indication of a phase shift.
Furthermore, phase is a relative quantity. We often need a reference in
in the number state base for field a. For any state I<I>)with nonzero norm
order to find the change in phase. Therefore, we will bring in another field
in a Hilbert space, it is possible to find a unitary transformation Uwso that
called b as the reference (e.g. the field in the other arm of the conventiona l
1
optical interferometers with beam splitters or the local oscillator in homo- Uwl<I>)
=IO)= 1'11) or l<I>)= u; IO), (11.37)
dyne detection). After we find the state (or mixed state) I<I>)satisfying the
where IO)is the vacuum state and contains no photon. So it can serve as the
condition in Eq. (11.33), it is useful to enlarge the state space to include
state 1'11)with special feature for distinction. Thus the interferometer has
the field b so that fields a, bare correlated. Therefore, the state for the total
the form shown in Fig. ll.4(b). Notice that only single-mode field is used
system has the general form:
in this interferometer which is different from a conventional interferometer
(11.35) (see more discussion on unconventional interferometer in Section 11. 3.).
m,n
Obviously, we have from Eqs. (11.36) and (11.37)
in the number state base. The unitary operator Uab acts on the enlarged
state space of two modes. Normally, the output will also consist of two Cm= (m1u; 1 10)= (mlUJIO) = (OIUwlm)*. (11.38)
modes. Detection can then be performed on the two output modes and With a phase shift on the state l<I>),we find the output state becomes
comparison is made between the two modes for the extraction of the phase
(11.39)
shift.
where Eq. (11.37) is used. With no phase shift, the output field is simply
!w) jw')
=>0-1
ab =>
in IO)and has no photon, but with a nonzero phase shift, the output state
is no longer the vacuum state and will contain photons. Thus detection of
E
(a) (b)
any photon in the output field is an indication of nonzero phase shift. A
better measure for this will be the probability P of detecting any photon in
Fig . 11.4 General scheme of phase measurement with some unitary operation Uab, u;;/ the output. Obviously, we have P = 1 - Po with Po being the probability
as the generalized beam splitters. ( a) The scheme with a reference field b. (b) The
scheme with a vacuum input state I '1!) = lvac) and no reference field. of no photon. With the output state in Eq. (11.39), we find

If the state Iw) is easily available such as the vacuum state, we can
generate the special state I<I>)as the phase sensing state by the inverse pro-
Po= l(OIIJ.i')l
2
= l(Ollf1>e""lf;'IO)l
=
2
I~
lc,,,le'='1
2
2

, (ll.40)

cess of Uab . Then, we form a general type of interferometers as shown in where we used the closure relation I:m lm)(ml = 1 and Eq. (11.38) in
Fig. ll.4(a). Notice that Uab is a general type of unitary operator that sa- the last equality. Therefore, we have for the probability of detecting any
tisfies our requirement for producing a unique state IW) from I<I>).Thus we photon:
have generalized our discussion to a broad class of unconventional interfero-
meters. Specifically for a conventional interferometer with beam splitters, (11.41)
Uab = exp[0(atb - abt)J given in Appendix A.
 358 Qv,aritv,rn Opt'ics For E:rperinientalists
where we used Eq. (11.22) with h" = c5 for the visibility v of the single-
photon interferometer. Therefore. if we use this scheme for detecting the
phase shift on the probe field 8, induced by a photon in field A of the single-
photon interferometer in Fig. 11.3 in Section 11.1.1, we find that when ever
u = 0, P = l indicating that we are able to detect the phase shift of c5.
Thus we have shown that whenever the interference disappears (v = 0), we
will have at least in principle a method to know whether the single photon
passes the path A or not with 100% probability. A good example for the
pure state is the phase state in Eq. (11.1) with finites . As a matter of fact
such a scheme achieves the Heisenberg limit. From Eq. (11.41). we see tha;
the quantity v as expressed in Eq. (11.22) is a good measure in the search
for optimum phase measurement schemes.
Next, let us consider a more general case with a mixed state of
PA= LPmnlm)(nl . (11.42)
mn
As we have discussed at the end of Section 11.1.1, let us assume that, when
we enlarge the state space, we are able to obtain a pure state of the form:
(11.43)
m,>-.
which, after tracing over the fields b, will reproduce the mixed state in
Eq. (11.42). The states {l,\)b} characterize the other states in fields b that
are correlated with field a. It is always possible to make the states {l,\)b}
a set of orthonormal states with (,\'IA'') = 6N>.."• Therefore, after tracing
over the fields b and comparing with Eq. (11.42), we have
Pmn = L Cm(A)c~(A). (11.44)
>..
Consider now the vacuum state IO)alO)bfor all the relevant modes in
fields a, b. As before, it is possible to find a unitary operator Uab so that
ab = IO)a IO)b. We can then run through the same argument as t he
(; ab I<I>)
case of a pure state for field a. The only thing different here is that the
cr iterion for finding the phase sh ift is the detect ion of any photon in any
mode of the fields a, b. Therefore, we have proved that if we can write the
state of the system in the form of a pure state after enlarging the state
space, it is always possible to find a measurement scheme to reso lve the
phase shift due to a sing le photon whenever the visibility of the single-
photon interferometer is zero .
Q'llantmn Nois e in Phase lifcasun 'rnent 359
11.2 Phase Measurement Schemes Reaching Heisenberg
Limit
Among the states that are available in the lab, only the thermal states
and the single- and two-mode squeezed states have the variance of photon
number in the order of the square of the mean, thus satisfying the condition
in Eq. (11.33). tiore specifically, the variance is (~ 2 N) = 2(N)(l + (N))
for the squeezed states of total mean photon number of (N) and (~ 2 N) =
(N)(l + (N)) for the thermal states or the two-mode squeezed states before
tracing out one mode (Section 4.6.1). In this section, we will discuss a
number of schemes that utilize the squeezed states and two-mode squeezed
states and other nonclassical states to achieve Heisenberg limit in precision
phase measurement.
11.2.1 Schemes Reaching Heisenberg Limit with
Conventional Interferometers
Let us first consider squeezed states. There are a number of ways to utilize
squeezed states to form interferometer. As a matter of fact, the first interfe-
rometer that beats the shot noise limit for the sensitivity in precision phase
measurement employs a squeezed vacuum state in the unused input port
of a conventional interferometer [Xiao et al. (1987); Grangier et al. (1987)].
We have discussed this scheme in detail in Section 10.2.2. In this section,
we will consider a number of phase measurement schemes based on the
conventional interferometers such as Mach-Zehnder (MZ) interferometers.
However, using only squeezed vacuum in the unused port cannot re-
ach the sensitivity set by the Heisenberg limit. Bondurant and Shapiro
proposed to use a squeezed coherent state instead of the coherent state to-
gether with a squeezed vacuum in the unused port and proved that such
a scheme can reach the Heisenberg limit [Bondurant and Shapiro (1984)].
The ir argument runs as follows.
Consider a Mach -Zehnder interferometer in Fig . 11.5, where a squeezed
vacuum state of I - r) and a squeezed coherent state Ir, -ia) (Section 3.4)
are injected into the two input ports. Referring to Fig. 11.5, for identical
50:50 beam splitters, we have
A= (ain + bin)/V2, ' B = (bin - Clin)/V2,
A' = A.ei<p,B' = Bei0 ,
Clout=(A' - B')/V2, bout= (B' + A')/V2. (11.45)
 361
Quantum Noi.se in Phase J\Ieasu:rement
360 QuantV,m , Optics FoT E:1pe1"iniental-ists

\N,N) is input to a 50:50 beam splitter, which is now the state to probe
I\ C/ill 1-r) a phase shift 5. From Eq. (6.58), we ha.ve the photon number distribution
b.Ill
Ir, - i a)
for the output state of the first BS:
(2k) l(2N - 2k)l
P(k) = 22N[k!(N - k)!]2. (11.51)
This photon number distribution leads to a photon number fluctuation
2
(6. 2N) rv N 2/2, as shown in Fig. 11.6 where we plot (6. N) as a function
2
F ig . 11.input.
states 5 Mac h-Z eh nd er interferometer with s q u e e ze d vac uum and squeezed coherent of N 2 in log-log sca le an d it approaches a lin ear dependence of (6. N) =
N 2/2 (straight lin e). In fact, for larg e N, the distribution in Eq. (11.51)
For a dark fring e output at aout, we set 0 - 7r, cp _
- uK <<1 and approac hes
1 (11.52)
Clout= ieio 12(Clinsin 5/2 - ibin cos 5/2) P(x) = --=====
1rJx(l - x)
2 2 2
(i5/2)ain + (1 + i5/ 2)bin· (11.46) with x = k/N, from which we find (6. x) = 1/2 or (6. N) = N /2 rv
We measure Xa = aout + aA out t b Y homod yne detection and have N 2 . This satisfies the necessary condition in Eq. (ll.33) for reaching the
Heisenb erg limit when this state is us ed to probe a phase shift.
A A A =A-(5 / 2)(Ya,n + Y,b in ) + XAb inl
Xa (11.47)
10000 .-----------------
where
- ) xb = b·m + bti:1-'y,b in -- (b- btin )/ i.. For th e mput
rnA • in -
. states of Clin in
I r and bin m Ir, -ia), we have 1000

(Xa) = 5(µ + v)a ,

(6. 2 Xa ) = (µ - v) 2 (ll.48) 100

with µ = cosh r , v = sinh r. So, th e signal-to-noise ratio is

10 !000 10000
10 100
SN R = (Xa)2 - 52(µ + v)2a2
(11.49)
(6.2 Xa) - (µ - v)2 2
Fig. 11.6 Photon number fluctuation (b. 2 N) as a function of N in log -log scale. The
We now define the phase sensing photon number as N = (AAtAA) _ 2 straight lin e is (b.2 N) = 2
N / 2.
a2( + v)2/2 If ps - - - vv +
µ . we keep Nps constant and assume v » 1 so that -
l/(µ + v) l/2v, Eq . (11.49) becomes µ - Now we need to find a phase measurement scheme and check if the
phase measurement precision reaches Heisenberg limit. For this, we set the
SNR = 852(Nps - v2)v2 '.S252N;s· (11.50) working point of the Mach-Zehnder interferometer such that if the phase
The maximum SNR of the right -hand side is reached when v2 = N /2 shift 5 is zero, the output state is the same as the input state and the photon
When
. SNR ,.___,
1 ' we obt am
· th e mimmum
· . measurable 5 ,.___,
1/N ps h · number difference between the two outputs is zero. This is achieved when
Heisenberg lim it . m ps, or t e we set 0 = 0, cp = 5 « 1 in the MZ interferometer in Fig. 11.5. From
H . An~ther scheme using a Mach -Zehnder interferometer for reaching the Eq. (11.45), the input -output relation then has the form of
e1sen erg limit employs the photon number states at the inputs . Photon Clout= ClinCOS 5/2 + 'ibinsin 5/2,
number states can be c~nsidered as amplitude squeezed states. It has no (11.53)
bout = bin COS 5/2 + iCLinsin 5/2.
photon number fluctuat10ns. In 1993 Holland and B tt
inject the state IN N) . ' urne proposed to When there is a phase shift, the output photon difference will not be zero.
' mto a Mach -Zehnder interferometer [Holland and
This can be confirmed by calculating the variance of the fluctuation in
Burnett (1993)] . In Section 6.2 .2, we have found the output state when
 363
Quantnm No 'isc in Phas e M ca.surement
362 Quantum Opti cs For Experimentalists

where we take t = cos b/ 2, r = sin b/ 2 for the 1\IZ interferometer with a

photon . number difference (~ 2N(out) ( 5)) with f.J(out)
- = atouta,rm/ - bAout
t bA
out• phase di.ffercnce of b. The probability to have the output r:.;tateIN - m , N +
For tlus, we first calculate
m) for the interferometer is then
2
Am(b) = \(N - m. N + m\U(5)\N, N)\
(11.54) = \(N - m, N + m\e8(/;at-a &t)/ 2\N. N) \2 . (11.57)

For the input state of IN, N), we can easily find (N~out))(5) = 0. which me-
ans that the photon number distribution is symmetric for the two outputs.
2
Next we calculate in the same way (N~out) (5)):
2
(N~out) (5)) = -((aJn&in - &Jnain)2 ) sin 2 5
2 any non-zero m. But J 0 starts to drop for appreciable N 6. Especially, the
= (aJ)in&Jnain + &JnainaJn&in)
sin 5
= 2N(N + 1) si1i 5,
where we used the twin photon state property: N~in)IN, N) = 0.
In this case, since (N~out)) = 0 for any value of 5, we can't use it as
signal for the detection of the phase shift 5. However, (N~out) (5)) -=/-o
2 put is not easily performed in the experiment. To circumvent this shor-
for nonzero 5, so it can be used as the signal. But the noise is not simply
}(~ 2N~out)(5)). In fact, when 5 = 0, we have ((N_)m) = 0 for any non-
zero integer m so that any detection of nonzero N _ is an indication of 5 -=I-0.
But N_ is quantized and only takes 0, ±1, ±2, .... Thus the noise of N_ is
2N 252 for (N) » 1 and 5 « 1,
simply 1. Therefore, SN R = (~ 2N~out)) ,:::::
and the minimum detectable phase shift is 6min rv 1/N = 1/(N), which is
the Heisenberg limit.
The above situation can also be understood in terms of the distribution
of photon number difference. For 5 = 0, the output has a state same as the
in_put IN, N). When 5-=/-0, we will have other states like IN - m, N + m)
with m-=/- 0. Whenever the variance (~ 2N~out\5)) = (N~out)2(5)) 2:, 1,
there will be some significant chance to have N~out) = 2m -=/-0, indicating
a non-zero phase shift 5. This requires 5 2:, 1/N. Thus, the minimum
detectable 5 is at the Heisenberg limit.
To further demonstrate the validity of the argument above, let us consi -
der the probability of detecting an imbalance in the photon number at the
two outputs, that is, the probability to have IN -m,N +m)(m-=/- O) in the
output . For this, we need to find the output state of the interferometer.
Since the interferometer is equivalent to a beam splitter, we can find its
evolution operator in Eq. (A.6) from Appendix A:
U(5) = exp[5(bat - a&t)/2],
When N is large and m « N, Holland and Burnett showed that P2m (b)
J;,, (Nb) with Jm as the mth-order Bessel function [Holland and Burnet~
(1993)]. We are particularly interested in Po(b) because the probability P
of detecting any non-zero m is simply P(b) = 1 - Po(b) = 1 - ff;(Nb).
When N fJ « 1, J O rv 1 and P (b) rv 0, indicating we are unable to detect
first root of J 0 (x) = 0 is x 1 = 2.405, that is, when Nb~ 2.4, we have
(11.55)
P( b) = 1. This means that we will be able to detect the phase shift of
b 2.4/ N and this is the Heisenberg limit for phase measurement.
However, measurement of the photon number distribution in the out-
tcoming, Campos et al. proposed to employ parity measurement at only
one of the outputs of the interferometer [Campos et al. (2003)]. The pa-
rity operator is defined as 6 = ( -1 )at a = exp( ina ta). It was shown that
(O)N = PN[cos(2b)] with PN(x) as the Legendre polynomial and this me-
asurement scheme also leads Heisenberg limit in phase measurement sensi-
tivity [Campos et al. (2003)].
In practice, state \N,N) is hard to generate in the lab. But the twin-
beam state \'17ab) discussed in Section 4.6.1 possesses the same twin photon
property N_\'17ab) = O that is required here. Let us now evaluate it for this
scheme.
Suppose the two input ports ain, bin are injected with a twin-beam state
discussed in Section 4.6.1 and given by
(11.58)
with
(11 .59)
Here, we changed the notation a, b --t ain, bin· From Section 4.6.1, we know
that (aJnain- &Jnbin)m\'17ab) = 0 for any non-zero integer m.
Similar to the case of \N,N) state input, we can find from the input-
output relation in Eq. (11.53) that the average photon number diffe-
. (out)) h • • ( A 2N(out) ( s:)) _
rence at the output 1s (N _ = 0 but t e vanance 1s u _ u -
(11.56)
 365
Quantum Nois e in Phas e lvfeasur em ent
364 Q'Uant'Um Optics For Experim entalists

2 2
4~L2v 2 sin 2 5 = 4((N) + 1) (N) sin 2 5 where µ 2 = cosh !171, v 2 = sinh 1771,
and (N) = lvl2 is the total mean number of photons in one arm of the
interferometer. When (N) » 1 and 5 « l, (!~2 N~aut\5)) '::::4(N)252_
Using an argument similar to the case of IN, N) injection, we find that
for the twin -b eam state l'Tlab),when (!~2 N~out)(5)) '::::4(N)25 2 rv 1, there
is a significant chance to obtain N _ -/=-0. Then, the minimum detectable
phase sh ift is 5min rv 1/2(N), which is the Heisenberg limit. This can be
c_onfirmed by considering the probability P of detecting N _ -/=-O. Obviously,
P = l - Po with Po as the probability of detecting N _ = 0. We can find
Po by first calcu lating the characteristic function for N _:
C(r) = (eir6.N)
1/(1 +4(N) 252 sin 2 r)½ ((N) » 1,5 « 1) (11.60)
transformation in Fig. 11.4 can reach the Heisenberg limit. In this section ,
we will replace the beam splitters with some more general types of unitary
transformation.
The first scheme is to use the squeezing operator
S(O = exp[~(a? - h.c.)] (~=real) (11.62)
to form a single-mode interferometer. As is known, when acting on vacuum
state, the squeezing operator produces a squeezed vacuum state that has the
2 2
variance of photon number fluctuations as (6. N) rv (N) (Problem 3.5),
thus satisfying the condition in Eq. (11.33) for reaching the Heisenberg
limit. So, we can choose I<!?)= S(O\vac) for probing the phase shift 5.
Then we have U<t> = s-
1 (~) and 1'11)= \vac) in the general scheme in

Fig. ll.4(b). The singl e-mod e squeezed vacuum state interferometer takes
and then making a finite Fourier transformation:
the form shown in Fig. 11.8.
p _ (~ d cos2mr
5
m - 7r Jo r (l + 4(N) 252 sin 2 r) ½·
s(~)\<P)
A
(ll. l) \vac) a
Thus Po = 2K(-4(N)25 2)/1r with K(x) as the complete elliptic function
of the_ first kind. Figure 11. 7 shows P as a function of (N)5. It is obvious \w) \w')
that P '::::0 when (N)5 « 1 and starts to rise when (N)5 rv 1, which is an
Fig. 11.8 Single-mode squeezed state interferometer.
indication of nonzero phase shift 5. So we find in this figure that the mini-
mum detectable phase shift is of the order of 1/ (N) and the interferometer
Without the phase shift, the output state of the system is ~imply
is operated at the Heisenberg limit.
\w)aut = 5- 1 s\vac) = \vac). With a phase shift 5 on the state\<!?)= S\vac),
the output state becomes

p
0.8 \w')out = 5-lei&) aoSlvac), (11.63)
0.6
and the average photon number at the output of the interferometer is then
0.4

0.2 (w'\ata\w')aut = 4(N)(l + (N)) sin 2 5, (11.64)

4 6 10 where (N) = sinh 2 is the average photon number in the state \<!?). Since
(N)8 the output of the interferometer is vacuum when there is no phase shift,
Fig. 11.7 Probability P of detecting any non-zero N_ as a function of (N)8 for a twin-
whenever we detect a photon at the output , we may conclude that there is
beam interferometer. Reproduced from [Ou (1997a)]. a phase shift. So, as before, we take the noise in the output simply as one
photon. Thus we have the signal-to-noise ratio as
2
SN R = 4(N) (1 + (N)) sin 5
11.2.2 Schemes Reaching Heisenberg Limit with
4(N) 252 ( (N) >> 1, 5 << 1). (11.65)
Unconventional Interferometers
Therefore, the SNR is significant only when 5 rv 1/ (N). Since the detection
Let us go back to the general scheme of phase measurement discussed in
of a single photon will indicate a nonzero phase shift, a better quantitl to
Section 11.1.3. We have demonstrated in the previous section that a num -
characterize the sensitivity of the interferometer is the probability P of
ber of phase measurement schemes using beam splitters as the unitary
 366 Quanfllm Opt.ics For Experimentalists Quantum Noise in Phase Aleasu.1·ement 367

finding any photon in the output. Obviously P = 1 - P0 with Po being the so that
probability of finding no photon in the output. Po can b e calculated as
UMl<I> )M = I0)all\I )b = IW)J..J. (11.71)
n
ro = (,T''
'±'
I : e -at a : I,T,'
'±'
)out With a phas e shift of 5 in field a. the stat e l<I>'
) b ecomes
1
) = ~(IO)all\I)b + eiJ..Jc5ll\I)al
!<I>')=eic5atal<I> O)b)- (11.72)
J1 + 4(N)(l + (N)) sin 2 5
and the output state then has th e form of
1
~==== 2 2 for (N) » 1, 5 « 1. (11.66) lw' )M = UMl<I>')
J1 + 4(N) 5 '
This expression can also be derived from Eq. (11.41) through the visibility = eiJ..Jc5/2 (cos l\I()IO)alM)b +isin l\I()ll\I )alO)b)·
(11.73)
2 2
v for squeezed states [Sanders and l'viilburn ( 1989)). Thus P is significantly Thus, if we measure photon number at output port a, any detection of
different from zero when 5 ?:,1/ (N) and the interferometer is operated at photon will indicate a phase shift. The probability of detecting any photon
the Heisenberg limit. This scheme was also discussed by Yurke et al. who in output port a is
used signal-to-noise ratio for the sensitivity [Yurke et al. (1986)).
P- = sin~') l\I 5/ 2
1 - cos l\I u~)
= - (1 (11.74)
As the second example, we consider the state 2
which is significantly different from zero only when 5 r-v 1/( l\I / 2) = 1/( N)a-
(11.67) Therefore, such a scheme reaches the Heisenberg limit.
The criterion here for the detection of a phase shift is different from
which is a photon number entangled state of two modes. This is the the genera l discussion in Section 11.1.3 because of the form of the unitary
so-called NOON state (Section 4.5.1) [Kok et al. (2002)). Obviously, operator in Eq. (11.68). It does not annihilate all the photons to produce
2 2
(l::!..N)a = M /4 = (N);, thus it also satisfies the necessary condition a vacuum state as required for the general scheme but rather preserves the
in Eq. (11.33). However, it is not so easy to find an evolution process total photon number.
to produce a distinctive state lw) as discussed in Section 11.1.3. But the Because of the unusual form of the unitary operator (which depends on
M -photon entangled state above looks similar to the one-photon entangled the total input photon number M), this scheme is not likely to be practical
state ll)ab = (ll)al0)b + I0)all)b)/v0 discussed in Section 4.3.1, which can as compared to the scheme discussed earlier with single-mode squeezed
be produced by a 50:50 beam splitter from a single-photon state. From the states . It is used here as another examp le of unconventional interferometers
form of the evolution -operator of a beam splitter (JBS = exp[0(ba,t - a,bt)J which can achieve the Heisenberg limit . But a unitary operator in a slight ly
(see Appendix A), we find that it is a one -photon creation and annihilation different form from Eq . (11.68) can be simulated in a trapped ion system
process. For M -photon creation and annihilation as one entity, we consider [Leibfried et al. (2002)] and is given as
the evolution operator
l;(aY11 b-h. c.)
U 1-~~
A -

- e . (11.75)
A _ ( '7f [ AAt M )
UM - exp M! (ab ) - h.c.] , (11.68) This is a one-to -M photon beam splitter. We have encountered this type
4
which comes from the Hamiltonian given by of interaction in Section 6.1.4 where the evolution operator in Eq. (6.19)
for t he photon number doubler is of the same form as Eq. (11.75) but with
(11.69) NI= 2. So, the unitary operator in Eq. (11.75) describes a one -to-M photon
It is easy to check that converter. Using this type of interaction but with only partial conversion
((v'M! = 1r /4), we can transform a photon in b-field into a superposition
UMIO)alM)b= /2(I0)alM)b - IM)al0)b), state of the form

UMIM)alO)b = 12(I0)alM)b + IM)al0)b), (11.70) (11.76)

 369
Quantum Noise in Phase /i.Ieasurement
368 Quantum Opti cs For Experimenta l-ists

.
0 bv10us 1y, tl 1c st a t,e- I'T'') lrns a nonzero average photon numb er. By mea-
If we use field a to sense t he phase shift, t he pha se sh ifted state is '±' L • •

suring the photon number in one of the outputs of the mterferom ete r' we
)al O)b + I0)all)b),
lcI>'h-AI = ~( eil\Jc511\I (11.77) can detect the phase shift 5.
The outp ut state is then \'l!' )
lw' h-J\J = OJcI>
') •
= ei.l\Jo/ 2 ( cos J\.15I0)al l )b + is in l\I5 11\l)al 0)b) , (11. 78)
2 2
which is similar to Eq , (11.73), The probability of detecting photons in . 11 .9
F 1g. SU(l,l) inter ferometer with vacuum inpu t.
field a is then
Mb . · th tput of the int erfero -
1 To find out the average photon num b er 111 e ou
Pa(5) = sin 2 - = -( 1 - cos Mb), (11.79)
2 2 meter' we find it eas ier to cons ider the evolution of the operators tha~ that
Notice that Pa shows 21r/ Jil per iod and is lvl times more sensitive to 5 . t F'g 11 9 fields a b are related to the 111put
of the states. R efernng O 1 · · ' '
than a traditional MZ int erferometer, Th en we can use the same argument vacuum fields ao, bo as
before to claim that the interferometer bas ed on the unitar y transformation a= µao + vbl , b = µbo + VCL6, (11.82)
in Eq, (11.75) can reach the Heisenberg limit. Experiment was perform ed
- sinh jnJ Assume field a experiences a phase shift of
by Leibfried et al. to simu lat e the unitary operator in Eq. (11.75) with where µ, = cos h I'f/I' v - ., . 1 t d t th
b of the interferometer are re a e o e
M = 2, 3 and confirm the phase dependence in Eq. (11.79) [Leibfried et al. 5. Then the outpu t field s aout, out
(2002)]. fields a, b as A A A At -ic5 (11.83)
Next, let us cons ider the thermal state because its photon number fluc- fiout = µa eio - vbt ' bout = µb - va e .
tuation satisfies the necessary condit ion in Eq. (11.33) for reaching the Hen ce,
5 At
Heisenberg limit. But it is impossible to impl ement an interferometer ba- . -u) ao + 2ie
5-: •

= (1 + 2iei·c5;2µ,2 s111 . ic5/2µv s1·11 -2 bo

A

sed on the thermal state alone due to its lack of phase coherence . However, aout
A

2
as we discussed before, we can always enlarge the state space to include A At (11,84)
==Gao+ g bo,
field b to form a pure state I<I>) ab so that we may find coherence between . ·c5;2 2 · J.:/2 d g = 2i ei 0 / 2 µ,v sin 5/2. The photon
with G ==1 + 2iei µ, s111u an -
fields a and b. We hav e shown in Section 4.6 that a two-mode squeezed
number operator at output field aout is then A

state from the nondegenerate parametric down-conversion process has two

A = \G\2atfio + \g\2 bob6+ Cg* fiobo+ G*g albl. (ll.8 5)
correlated fields with each having the thermal statistics, or in other words, na out o ) . h
12 4(N) 2 sin 2 5/2 for ( (N) > > 1 wit
(6 2N)a ,b = (N)((N) + 1). Such a state can be .generated from vacuum by • (

It is easi ly found that na out - g -

A ) _ \ r--.,
s· .1 t the
applying the operator Sab(TJ)(see Section 4.6) (N) -
= \v\2 being the average photon number in field a. 11111 ar _o
· h · 6 1s one
s ueezed state interferometers discussed ear lier, t e n01se na oul
(11.80)
~oton so that SNR = 4(N)2 sin2 5/2. Therefore, SN_R 1 only when r--.,

Now let us reverse the process by assigning the two- mode squeezed state 1/ (N). Furthermore, sim ilar to the sque~zed state 111terferomete:~ we
r--.,

fields as the fields a, b in Fig 11.4(a) and set J<I>)ab= Sab(TJ)Jvac)aband find the probability of detecting any photon 111the output field aout
Jw) = lvac)ab· Then the unitary operator required in Eq. (11.34) is Uab = p = I'- Po = l _ (: e-naaut :)
s-;;/
( 77). The int erferomet er for the whole system then has the same form
4(N)((N) + 1) sin 5/2
2
as the two -mode interferom eter shown in Fig . 11.4(a). In this case, Jw) = 2
= Jvac)abwhen the phase shift is zero. But when we have a
s~i/(TJ)l<I>)ab = 1 + 4(N)( (N) + 1) sin 5/2
2
nonzero phase shift 5, the output state becomes (N)25 for (N) >> 1, 6 < < 1. (11.86)
1 + (N)252'
Jw') = s-;;/(17)eioa)asab(TJ)lvac)ab· (11.81)
 :370 Quantum Optics For Experim entalists
Therefore, the probability of detecting a phase shift of size b is significantly
different from zero only if b 1/ (N). Thus the minimum detectable phase
shift is 1/ (N), or the Heisenberg limit. Such a scheme was first discussed by
Yurke et al. along the line of SU ( 1,1) interferometer [Yurke et al. (1986)].
The word ··SU(l,1)'" was due to the symmetry group of the unitary opera-
tors used in single- and two-mode squeezed interferometers in Eqs. (11.62)
and (11.80), respectively. By comparison, the unitary operator for a lossless
beam splitter in Eq. (A.6) falls in the SU(2) symmetry group.
Notice that the criterion here for detecting a phase shift is simply the
detection of any photon in field aout alone without considering field bout·
This is quite different from the criterion discussed earlier in Section 11.1.3
for the general case of mixed state. The general criterion is the detection
of any photon in any relevant modes which will include both modes aout
and bout· The reason for the difference is that the output fields aout, bout
are actua lly in a twin photon state as Eq. (11.84) indicates and the photon
numbers of modes aout, bout are perfectly correlated in such a way that
the two modes have exactly the same photon number at all time. Thus
detection of photon in mode aout alon e is equivalent to the detection of
photon in any of the two fields.
11.3 Non-conventional Interferometers
We have already discussed in Section 11.2.2 a number of phase measure-
ment schemes using some devices other than the traditional beam splitters
in interferometers. They can reach a phase measurement sensitivity at Hei-
senberg limit. In this section, we will make a general discussion about phase
measurement with non-conventional interferometers and concentrate on a
specific type, i.e ., SU(l,1) int erferometers with coherent state boost, which
is more practical to work with in the lab.
11.3.1 General Consideration
As we found before, in order to increas e the sensitivity of phase measure-
ment in a traditional int erferometer such as the MZ interferom eter, squee -
zed states are employed to reduce the quantum noise in the measurement.
Some other nonclassical states with quantum correlation such as the photon
number correlated twin -beam state may also be used. So, the attention is
focused on the quantum states that are used in the interferom eter but with
the hardware structure of the interferomet er unchanged. This is straight -
371
Quantum Nois e in Phas e J\Jcas1tTem e11t
forward thinking after Caves pointed out, in a seminal paper [Caves (1981)].
that quantum noise is the limiting factor in a traditional interferometer. So.
most research effort is on finding the correct quantum states with various
quantum correlations by reducing the noise for the increase of the. ~h~se
measurement sensitivity. On the other hand, the measurement sens1trv1ty.
which is usually characterized by th signal-to-noise ratio (SNR). depends
not only on the noise but also on the signal, that is,
Signal
SNR=-.~. (11.87)
Nmse
Signal increase will also improve the the measurement sensitivity even with
the same noise. As a matter of fact, we already encountered an example
of signal increase in Eqs. (11.74) and (11.79) in Section 11.2.2 where the
change in the measured signal of bN due to a phase change b is bN =
NobP = MNob (No is total number of photons). This is llI times as large
as the case of a traditional interferometer. This increase in phase signal
is because the interference fringe depends on sin(l\H) which has a slope
at b - O A1 times as large as sin b for a traditional interferometer. The
ach ie~err:ent of th is is by the employment of a non-traditional beam splitter
described by Eqs. (11.68) and (11.75).
The increase in the signal thus relies on the hardware or structt:re cl~ange
in an interferometer, which is on the unitary transformation Uab,Ui:f>in
Fig. 11.4. On the other hand, the quantum state change can be ~iewed ~s
software programming. Therefore, both strategy should be considered 111
order to find the optimum phase measurement scheme. In the following, we
will revisit the SU(l,1) interferometer which shows the advantages in both
aspects .
11.3.2 Coherent State Boosted SU{1,1) Interferometers
Although both sing le- and two- mode squeezed states used in SU ( 1, 1) in-
terferometers can be produced in the lab, there is a serious problem that
makes their practical application difficult. For input in vacuum, we can find
the photon number of the squeezed states as (N) = lvl2 = sinh r with r as
2
the squeezing parameter. However, it is usually hard to achieve very larg e
squeezing parameter in the lab so that the fields used to sense ph~se
have a quite low photon number to be compet itive with the trad1t10nal 111-
terferometry involving las ers which usually have a larg e number of photons.
To incr ease the photon number for phase measurement, we can use a cohe-
rent state from a laser to inj ect into the SU(l,1) interferometer. This idea
 •
Quantum Noise 'in Phase Afmsuremc11t 373
372 Q-uant :1L
'm Optics FO'I' Expffim('_ntalists

was first proposed by Plick et al, and dubbed as ·'coherent-state boosted where t(c5) = ci 8 / 2 cosc5/ 2; r(c5) = -iei 8 / 2 sinc5/ 2. Herc c5 is the phase shift
SU(l,1) interferometer" [Plick et al. (2010)]. Furthermore, the detection and we assume the beam splitters are identical with 50:50 transmissivity
scheme in the discussion of previous sect ions is mostly the photon number and reflectivity. A A
counting, whose sens itivit y is limited by the dark counts of photo-detectors. For a coherent state Ia) input at a111 and vacuum at b,;n. bout is at the
As we have shown in Chapter 9, homodyne detection can overcome the dark dark fringe at c5= 0 and the output of this port is simply
count problem and achieve quantum-limited measurement. We will analyze (blutbout} = faf 2 (1 - cosc5)/2 = Ips(l - cosc5), (11.90)
the performance of the coherent-state boosted SU ( 1.1) interferometers and where fps = (bTb)= fn\ 2 /2
is the photon number of the field subject to
compare it with the traditional IZ interferometers in this section. the phase shift (phase sensing field). Note that the phase sensing field is
what matters in phase measurement accuracy as shown in Secti?n 11.1.l.
b 111,'Y
We usually make homodyne measurement at the dark port, i.e., bout of the
interferometer with c5= 0. With a small phase shift c5« 1 and a = i\a\,
la )
we can easily find
(XLJ = (XL (c5/2)) cos 2 (c5/2) + (YL(c5/2)) sin 2 (c5/2)
2
1 + \a\ 2 c52 for c5« 1, \a\ » 1. (11.91)
A bA + bAt X (c5/2) = bei8/2 + ht e-i8/2 and Ya(c5/2) =
(a) (b) H ere X bo-ut out out, b '
(iiei8/2 _ a,te-i8/ 2)/i are the quadrature-phase amplitudes of corresponding
Fig, 11,10 (a) An SU(l,1) interferometer with parametric amplifiers (PAl, PA2) as the fields. Obviously , \a\ 2 c52 in (Xtut) corresponds to the phase signal while the
equivalent beam sp litters , (b) A traditional Mach-Zehnder int erferom eter .
noise of the phase measurement is simply 1 from vacuum quantum noise.
Hence, the signal-to-noise ratio of the Mach Zehnder interferometer is
So, referring to the general scheme in Fig. ll.4(a), we have Iw) = 2
SN RJ\.Iz = \a\ 2 c52 / 1 = 2Ivsc5 , (11.92)
fa)alvac)b as the input state and Uab is the parametric amplifier from which
a two-mode squeezed state can be produced (Section 4.6). With these, we which leads to the standard quantum limit 6sQL = 1/ffi with N = 2Ips·
redraw the configuration in Fig. 11.lO(a) together with a traditional MZ For the SU(l,1) interferometer in Fig. 11.lO(a), a parametric amplifier
interferometer in Fig. 11.lO(b) for comparison . Thus, this scheme is similar (PAI) now acts as a beam splitter to split the input signal beam (ain)
to a traditional interferom eter except that the wave splitting and combi- into the amplified signal beam (a) and the accompanying beam (b). Anot-
nation elements are not regular beam splitters but parametric amplifiers . her parametric amplifier (PA2) acts as a beam combiner to complete the
For general discussion , we assume the two parametric amplifiers (PA1,PA2) interferometer. Even though there is no injection at mode bin for the am-
have different gains denoted by G 1 , g 1 and G 2 , g2 , respectively. A coherent plifier, vacuum still contributes with quantum noise. The full input-output
state is input at both interferometers in field ain while field bin is in vacuum. relation for the amplifiers is given by (Section 4.6)
At A A At
For the MZ interferometer in Fig. 11.lO(b) , with input-output relation a= G1ain + g1bin' b = G1bin + g1ain;
for a lossless beam splitter given by aout = G 2a - g2bAt e - i8, bAout = G 2 bAei8 - g2 aA
A A t, (11.93)
2
a = (CLin+ bin) I h, I h;
b = (bin - CLin) where we assume the amplifiers have amplitude gains G1, G2 with \G-i\ -
8 8 \gi\2 = l(i = 1, 2). We choose the sign of g2 soAthat PA2 acts as a de-
aout = (a - bei )/ h, bout = (bei + a)/h, (11.88)
amplifier. We introduce a phase shift of c5on field b. Therefore, the output -
it is straightforward to express the outputs of the interferometer in terms input relation of the interferometer is
8
of the inputs: aout = [Gr(c5)ain + gr(c5)blnle-i ,
A A At (11.94)
bout = Gr (c5)bin + gr (c5)ain,
(11.89)
 •
374
Quantum Optics For E:r:pe1'irnentalists
Quantum Noise i'll Phase l\Jeasurc111e11t 375

16
with Gr(5) = G1G2c> -g1g2, gr(5) = G2g1ei6 -G1g2. The interferometer
we have b0111 bin.v,,hich is in ,,acm1m. so the noise is simply vacuum
works best at dark fringe when 5 = 0 with Gr(0) = G 1 G 2 - g g and
12 noise, just as in the 1\IZ interferometer. Thus the improvement in sensiti-
gr(O) = G 1g2- G2g1, which has a unit overall gain Gr(0) = 1 for G = G
[Yurke ct al. ( 1986)]. 1 2 vity is achieved not by reducing the ,-acmnn quantum noise at the unused
in1;ut port. which is usually done with a 1\IZ interferometer but rather by
\i\Tith a coherent state input at ain and no input at bin as shown in
enhancing the signal level via amplification in tbc S1:!(l,l) interferometer.
Fig. 11.lO(a), similar to Eq. (11.90), we can easily fincl the dark output
port (bout) intensity as Since the vacuum noise at the unused input port (bi11)is the noise source
of the SU(l,l) interferometer discussed above, we can reduce it with a
2
(blutbo1it)= 1Gr(5)1 lal 2+ lgr(5)12(1al2 + 1) squeezed state input at this port, i.e ... (Xt) = e-r . Then Eq. (11.96) is
2 2
2G g ja/ (1 - cos 5) for /a/ 2 » 1 and G 1 = G 2 = G
2 modified to
2
= 2c 1;/ (1 - cos 5), (11.95) (11.99)

where 1// = (btb) = gf(lal + 1) gfla/ 2 (lal 2 » 1) is the photon number Then the sensitivity of phase measurement can be further increased from
2

of the phase sensing field b of the SU ( 1, 1) interferometer (SI). Com paring the standard quantum limit by a factor of
the above with Eq. (11.90) for the MZ interferometer, we find that the SNR~j/SNRMz = 2G2 er. (11.100)
fringe size is increased by a factor of 2G 2 for the same 1 and 1s 1
ps ps . This enhancement factor combines the signal increas e by amplification and
For homodyne detection around the dark fringe with 5 « 1 and a= i/a/,
we have from Eq. (11.94) the noise reduction by squeezed states at the input.
Another interesting case is when there is no injection of the coherent
(XLJ = /Gr(5)12(XL (({Jc))+ /gr(5)12(X~in(-(()9 )) state at all. Setting lal 2 = 0 in Eq. (11.97), we have the signal-to-noise
2 2
1 + G g (4/al
2
+ 2)5 2 for 5 « 1, G1 = G2 = G, (11.96) ratio without coherent injection

where eiipc = Gr(5)/1Gr(5)1, ei1P9 =

gr(5)//gr(5)/. Hence, the signal-to- SNRsr = 2G 2 g25 2 = 21;/U;/+ 1)52 , (11.101)
noise ratio for the SU(l ,l) interferometer is 81
where ]ps = g 2
. This leads to the Heisenberg limit [Yurke et al. (1986)}:
2 2
SN Rs1 = G g ( 4lal 2 + 2)5 2 /1 £ 1/j21s1(1s1 + 1) 1/N ' (11.102)
Um -_ ps ps rv
2 2
4G 1;/ 5 for /a/
2
» 1. (11.97)
with N = lf/ » 1 as the number of photons probing the phase shift.
Comparing this with Eq. (11.92), we find that under the condition of the
Squeezed vacuum injection at the idler port can further increase the SNR
same nu~ber of photons for the phase sensing field, i.e., 1ps = 1//, the but the photon number also increases because squeezed vacuum contains
SU(l,l) mterferometer has a better signa l-to-noise ratio than the MZ in-
photons.
terferometer with an enhancement factor of
For the SU(l,1) interferometer, it is not surprising that signal is in-
SNRsI/SNRMz 2G 2 for /al 2 » 1. (11.98)
creased since the field containing the phase signa l is amplified by PA2.
However, as we have shown in Section 10.3.1, amplification of the signal is
The sens itivit y of phase measurement for the MZ int erferometer with co-
also accompanied by the amp lification of noise and extra noise is usually
herent state and vacuum inputs is at the so-called standard quantum limit.
added leading to a worse SNR. So, amplification alone cannot explain the
Then the SU(l,l) interferometer improves upon the standard quantum li-
2 enhancement of the SNR in the SU(l,1) interferometer. In fact, the role
mit by 2G fold. The physical picture of this enhancement in sens itivit y is
of the first parametric amplifier (PAl) is crucial here. It produces a pair
stra ightforw ard if we compare the output fringe intensities in Eqs. (11.90)
of entang led fields (a, b) that have their noise correlated. We have demon-
and (11.95) for the two interferometers: the fringe size is increased by a
2 strated in Section 10.3.3 that SNR can be improved with correlated fields
factor of 2G . But in the meantime, at 5 = 0 (dark output) with G = G ,
1 2 input to the amp lifier. Therefore, it is noise cancelation due to quantum
 •
376 Quantum Opt'i cs F01· E:rperim ental-is ls
Quantum Noise in Ph ase l\! eas w ·cm en t 377

interference that leads to signal amplification without noise amplification (c)

and thus improvement in SNR. cor re lated tii) Des tru c ti\'e
amplifi cati on ---------_ int erfer ence
To understand better the nois e cancclation du e to quantum interfer ence ---...____ r
we calculate the output noise at dark pmt at an arbitrary phase J. It i~
(a) (i)
.,,.
-~ -
----PA2 (\)

(i i )
straightforward to obtain for coherent stat e and vacuum stat e input
2 uncorrelated
(t1 Xbn11J
= /Gr(J)/ 2 + /gr(J)/ 2 amplificati on
\11)
_
.
No lnt er1erence
(b )
= (GI+ gf) (G~ + g~) - 4cG1G2g1g2cos J. (11.103)
( -~
•
1)
T
UAI-----.....
- --------
A'- ...
( 111)
0.1 0.2 0.3 11.-l
(ii)
Scannin g Tim e (s)
The first term corresponds to the case when there is no correlation between
the inputs to PA2 and the second one is the interference term from the Fig. 11.12 (a) Sch ematic for th e case of qu a ntum-correlat ed amplifi ers forming the
quantum correlation betwe en the inputs to PA2. SU (l,1 ) interferomet er. (b) Sch em a tic for th e case of uncorrelat ed amplifi er~ wi t h am-
plifi ed noise. Th e circles repres ent sizes of nois e levels with lab els correspond111g to part
- -- ----- ( c). ( c) Quantum noise measur ed by homodyne det ection for the b-output port of the
g I
,
;-

06 j o.c, SU(l ,1) interferomet er in Fig . 11.lO(a). Trace (i): shot noise ; Trac e (ii): nois e level_sfor
-si o.5 ! (a)
:s
0
04 .
05
:'.'.; OA
ea ch parametric amplifi er alon e ; Trac e (iii): noise level for two uncorrelated amplifiers
in seri es ; Traces (iv): noise level for correlated p a rametric amplifiers in SU(l ,1) interfe-
-~ 0.3 o (U rom et er with phase scanned in time ; Trac e (v): same as (iv) but with phase locked at
-~
] 0.2 ] 0.2 dark fring e . Adapted from [Hudelist et al. (2014)].
0. 1 0. 1
bl)
ci3 ()
0
From Eq. (11.103), we find that (!12XbuJ = l for J = 0 when G1 = G2,
200 400 600 800 200 400 600 800
Scanning Time (ms) Scanning Time (ms) 0

Fig. 11.11 Int erference fringes observed at two outputs
Reproduced from [Jing et al. (2011)] .
Experimental realization of an SU ( 1, 1) interferometer was first achieved
by Jing et al. with atomic four-wave mixing processes as the two parametric
amplifiers (Jing et al. (2011)]. Figure 11.11 shows the interference fringes at
the two output ports. Notice that the two interference fringes are in phase
in contrast to the situat ion of 180° out of phase for a MZ interferometer.
This is typical of SU(l,l) interferometers [Chen et al. (2015)].
The schematic with noise behavior of an SU(l,l) int~rferometer is de-
picted in Fig. ll.12(b) as compared to an uncorrelated case in Fig. ll.12(a).
The circles at each stage represent the noise sizes with labels corresponding
to the noise levels labeled in Fig. 11.12 (c). The quantum noise performance
of an SU(l,l) int erferometer was first measured by Hudelist et al. whose
results are shown in Fig. ll.12(c) [Hudelist et al. (2014)]. Trace (iv) re-
presents the noise level as the phase of the interferometer is scanned in
time. It shows the interference pattern predicted in Eq. (11.103). The SNR
improvement of 4 dB by the SU(l,l) interferometer as compared to a MZ
interferometer was also demonstrated by Hudelist et al.
of an SU(l ,1) int erferom eter.
i.e. , it is at the shot noise level. However , the lowest noise level shown in
Trace (v) in Fig. ll.12(c) is about 3 dB above the shot noise level (Trace (i)).
This is due to the existence of losses inside the interferometer. The losses
inside the SU(l,l) interferometer will destroy the quantum correlation with
the introduction of uncorrelated vacuum noise. These uncorrelated vacuum
noise will not be canceled but amplified above the shot noise level. We will
make a quantitative analysis of the effect of losses next.
11.3.3 Loss Analysis for the SU(1,1} Interferometer
It is well-kn own that losses are the limiting factor that hinders the applica -
tion of squeezed states in precision measurement. We will examine next the
effect of losses on the sensitivity of the SU(l,l) interferometer. There are
two types of losses: inside and outside the interferometer. We start with
loss outside the interferometer first.
Effect of loss outside the interferometer
Loss outside the interferometer may come from propagation loss, less-
than-perfect homodyne mode match, and most likely the finite quantum
efficiency of the detectors. We can place all these losses into an over-
 378
Quantum Optics For Ex:pe1·irnrntalists
Quantum Noise in Phase !ifeasnrcmcnt 379

~1,
11lo~s Aof L and mo?el it by bea1~1splitter of transmissivity (1 _ L): So with loss, the fringe size is only reduced by l -L2 . However, the qu ant um
bout - bout /1=-L + bo\lL. It lS stnughtforward to find that Eq. (11.9G) noise is not so. With J « 1, we find from Eq. (11.105)
becomes

(X~'") = [lg1g2 - G1G2eir5 /1 - L2/ 2

(11 .104)
So the signal-to -noi se ratio is reduced by a factor of 1 - L. Such a reduction + lg2GLv .1 - L11 - g1G2ei'8 v 1 - L2I 2]4ial 2
can be compensated by the increas e in gains (G, g). The sensi ti vi ty enhan-
cement is unlimited. +G~L 2 + g~L 1. (11.109)
. _R~call that for squeezed state-bas ed schemes, sensitivity enhancement With condition in Eq. (11.107) and J « 1, Eq. (11.109) becomes
is l11111
ted by the loss even for large amount of squeezing. In this sense
the scheme with the SU ( 1, 1) interferometer is less prone to detection los~
(XL,J = 1 + 2g~L1 + (4ia/ 2 + 2)giG~(l - L2)r52 . (11.110)
~han s_q~~ezedstate-based interferometers. The underlying physics for this Henc e, the signal-to-noise ratio is
m~ens1t1v1tyto loss is that the SU(l ,1) interferometer is operated at vacuum
no~se level (G1 = G2) or above (G1 f. G2). So the introduction of vacuum SNR¼L = (4/al 2 + 2)giG~(l - L2)r52/ (1 + 2g~L1)
noise through losses will not change it s noise performance too much. On 4I;/G~(l - L2)r52/( 1 + 2g~L1). (11.111)
t~e other hand,_ if we inject squeezed vacuum into the unused input port
Compared to Eq. (11.97) for the case without losses, the SNR is reduced by
(bin) to further mcrease SNR as in Eq. (11.100) , loss will limit the effect of
a factor of ( 1- L 2 ) / ( 1 + 2g~L 1) and compared to the MZ interferometer , the
squeezing just like squeezed state -bas ed schemes.
enhancement in SNR is 2G~(l - L 2 )/(1 + 2g~L1) 2(1 - L2)/ L1 for large
Effect of loss inside the interferometer g~. Thus, like the MZ int erferometer with squeezed state, the enhancement
is limited by the loss L1.
For losses insid e the interferometer , the situation is not as good. We consi - Furthermore, for the case of no coherent state injection, the SNR beco-
der t~o situ~_tions: (i) losses in the propagation between the two parametric mes
am~hfiers; (11)losses inside the parametric amplifiers. For the first case, we
rr A bA' - = 2giG~(l - L2)r52/(1 + 2g~L1)
1

agam model the losses by beam splitters: a' = ~a+ SN Rrfh (11.112)
~ A A v.1-.u1 V.L..11ao, -
v 1 - L2b + ~bo. We also assume the amplifiers have different gains of
G1, G2, respectively. Then the idler output port is
and if we set gf = g~ =If/ and L1 = L2 =L ,
2
bout= G'(J)bin + g'(J)afn SNRr;h' =2I;/U;/ +1)(1-L)J /(1+2I;/L). (11.113)

+g2$ial + G2~bo, (11.105) So the minimum measurable phase is at the Heisenberg limit for small
with photon number with I;/ « 1/ L but for large photon number of If/ » 1/ L,
Eq. (11.113) becomes
G'(r5) = G1G2ei15~ - g1g2~
U;/ +
1

g'(r5) = g1G2ei 15
~ - g2G1~- SNRr;;?L = l)(l -L)r5 2 /L, (11.114)
(11.106)
For strong coherent state input with /a/2 » 1 and which only improves upon the standard quantum limit by (1-L) / L, similar
to the case of strong coherent injection. This shows that loss is the limiting
g2G1~ =g1G2~ =gGfi-=-I
(11.107) factor for reaching Heisenberg limit in precision phase measurement.
for 100% visibility, we find the output intensity as The second type of losses of the interferometer is the losses inside the
2
(blutbout) 2G g 2(1 - L) /a/2(1 - cos J) parametric amplifiers. This type of losses cannot be modeled as beam
splitters but was considered as coupling to the outside vacuum of a non-
= 2G~I;/(l - L2)(l - cosJ). (11.108) degenerate OPO in Eq. (6.153) in Problem 6.6. In Problem 11.1, we treat
380 Q'Uant'Um Optics For Experim ,entalists Q'Uantnm Noisf' in Phase l'vfeas'Urement 381

the SU ( 1, 1) interferometer with this type of parametric amplifier and find transmissivity T and measure each quadrature-phase amplitude respecti-
the SNR for phase measurement is given in Eq. (11.138) as vely, as shown in Fig. 11.13. Suppose the beam for signal encoding is in
a coherent state la) with o_ = lalejtpo. A phase modulation of 5 « 1
SN R~ 1 = 41;/ 0 2 52 /(1 + 4g'2 )
and an amplitude modulation E « 1 are applied to the input beam si-
;::::::
1;/5(f3h)
2
for c» 1
multaneously, as expressed in terms of the phase of eJ0 ;::::;:1 + j5 and
2
= 21;/5 (T/2L), (11.115) the amplitude transmission of e-E ;::::::1 - E, the modulated field 1 is then
where we used Eq. (6.154) for C, g, G', g'. This SNR is en hanced from A.= ClineJ0 e-E;:::::: Clin(l + j5 - E).
SN LMz in Eq. (11.92) by a factor of T /2L . otice that L/T;::::;: S is the The outputs of the 50:50 beam splitter are given by
maximum squeezing from one of the two NOPAs. So the enhancement is li- a= (A+ B)/v'2, h= (B - A.)/v'2, (11.116)
mited by the overall vacuum noise leaked into the two parametric amplifiers
where Bis in vacuum. Now we measure Xa = ae-jtpo +at eJtpoat one output
through intra-cavity losses.
and i\ = (be- j tpo-b t eJtpo)/ j at the other by homodyne measurement. Since
the input is in a coherent state, the noise is simply the vacuum noise for
11.4 Joint Measurement of Conjugate Observables
both Xa and 1\:
Heisenberg uncertainty relation in quantum mechanics sets the limit on the (11.117)
measurement precision of two conjugate observables. With squeezed states, The measured signals can be calculated as
one can measure one observable more precisely than the standard quantum
(Yb) = -[lalejtpo (1 + i5 - E)e-jtpo/v'2 - c.c.] /j = -v'2lal5
limit at the expense of worse precision in the conjugate observable. In some
applications, however, we need to obtain the information embedded in two (Xa) = lalejtpo(l + i5 - E)e-jtpo/v'2 + c.c. = v'2lal(l - E). (11.118)
conjugate observables. For example, the real and imaginary parts of the Hence, the signals due to modulations are Sxa = 2lal 2 E2 , Syb = 2lal 2 52
linear susceptibility of an optical medium correspond respectively to the and the signal -to-noise ratios are
phase and amp litude modulation of an optical field passing through the 2 (11.119)
SN Rx = 2IpsE2 , SN Ry = 2Ip8 6 ,
medium. In this section, we will discuss the problem of joint measurement
of two conjugate variables. where fps = lal 2 is the photon number of the probe sensing beam.

\a)
Pump

\0) -~----------------
B

Fig. 11.13 Joint measurement of two orthogonal observables by using a beam sp litter. Fig. 11.14 Joint measurement of two orthogonal observables by using a non-degenerate
parametric amp lifier for splitting the incomin g field for simultaneous measurement.

However, the actually measured SNRs can be sma ller than those given
11.4.1 Classical Measurement Schemes
above due to detection losses . On e way to circumvent this is to use a
For the joint measurement of two quadrature-phase amplitudes, one simple 1 Because E <<1, we dropped the vacuum term that resulted from transmission loss of
method is to split the modulated beam into two parts by a beam splitter of e-E.
 382 Q1Lanf1lm Optics For Ea.:p c1·im c11talists
parametric amplifi er as the beam sp litt er before t he homodyue detection
~s shown in Fig. 11.14. A parametric amplifier is described by the followin~
mput-output relation:
where A is the modulated signa l beam in the coherent state and
vacuum. It is stra ightforward to shmv that the output signals are
2
(Xa) = 4G 2 f p 5 E2 , (Yb)2 = 4g2fp 5 62
and t he output noise are
The output sign al to noise ratios for t he amp lificatio n scheme are t hen
SN R amp= 4G2fpsE2 SN R amp= 4g2fps62
x c2 +
g2 , y 02 + g2 · (11.123)
The superscript "amp " is meant for th e amplification scheme. At large gain
2
of g » 1, the results in Eq. (11.123) are the same as that in Eq. (11.119).
The above schemes are classical schemes for joint measur ement in the
sense that their performance is at shot noise limit and joint measurem ent
leads to 3 dB = 1/2 reduction in SNR (for G » 1) due to vacuum noise
introduc ed (from field B) in equal information splitting.
11.4.2 Joint Measurement with EPR Correlated
With the availability of quantum resources such as squeezed states, we have
d~monstrated in Section 10.2 .1 that it is possible to enhance the sensitivity
of phase measurement beyond classical limit. However , the noise decrease
in the X-component is always accompanied by noise increase in the Y-
component in order to preserve th e Heisenberg uncertainty relation. Thus
injecting a squeezed state in the unus ed port in the schemes in Figs. 11.13
and 11.14 will not work for th e joint measurement of X and Y. It seems
that the fundamental law of quantum mechanics prevents us from improving
beyond classical limit the sensitivity of joint measurement of the modula -
tions encoded in two orthogonal conjugate observables .
On the other hand, quantum mechanics allows quantum correlations
that are stronger than classical correlations through the quantum magic
of entanglement, as beautifully presented in the Einsein -Poldosky-Rosen
(EPR) paradox. In Section 10.1.3 , we demonstrated that we can infer
through EPR correlations two conjugate observables with precision better
Quantum Nois e in Phas e .l\Ieasvr em cnt :3s3
@ - x,tJ
IO)
NPA B
G. g A
(11.120) lu)
B is in }'
- "
(11.121) Fig. 11.15 Joint measurement of two orthogona l observab les with entangled fields ge-
nerated from a non-degenerate parametric am plifi er.
than what is allowed by Heisenberg unc erta int y relation. Next, we exp loit
(11.122 )
this quantum cor rel at ion for the improv ement of the sensitivit y of joint
measurement of t he modulations encoded in two orthogonal conjugate ob-
servabl es .
On e way to make us e of this is to enco de information on one of the
EPR correlated beams and combine the two EPR corre lat ed beams with a
50:50 beam splitter. We hav e shown in Sect ion 6.2.4 that when two squee-
zed states (Eq. (6.86)) are combined in a 50:50 beam splitter, a two-mode
squeezed state or an EPR entangled state can be produced (Eq. (6.90)).
If we reverse the process with two EPR entangled beams combined in the
beam splitter, as shown in the beam splitter part of Fig. 11.15 , the two
output beams will be in squeezed states with one squeezed in X while the
other in Y (see also Problem 10.1). Thus, if we can encod e signal in both
States
X and Y on th e input beam and measure X on one output and Y on the
other, the encoded signal is only split half for two sides but the noise can
be squeezed on both X and Y, leading to joint measurem ent of X and Y
with improved SNR.
However , the parametric amplifier in the spontaneous mode has small
photon number outputs for signal encoding . So, we resort to a coherent
state input to the PA for photon number boosting, similar to the case in
SU(l ,l) int erferometer discussed in Section 11.3.2 . This scheme is shown
in Fig. 11.15. The parametric amplifier is again described by
(11.124)
wh ere ain is in a coherent state la) and binis in vacuum. We changed
the sign of g so that the corresponding quadrature -phase amplitudes are
squeezed in noise. We use the output field B to probe the modulations:
B' = B(l + j5 - E). The outputs of the BS are related to the inputs by
a= (A+ B')/v2, b= (B' - A)/v7i. (11.125)
 •
Quantum Noise in Phase l\IeasiLrement 385
384 Q'U,antum Optics For E:L:pe1·imentalists

through a pha se modulator (Pl\I) ancl an amplitude modulator (Al\I) for

The noise of Xa = ae- )'-Po+ a.t eJ'-Poand 1\ = (bC'-)'-{Jo- l)te]'-PO)
/j ('Po _
signal encoding at orthogonal observables. The second parametric amplifier
a/ IcvI) arc stra ightforw ard to calc ul ate as
(NPA2) acts as a bea m combiner to complete the interferometer, ~nd the
(11.126) output fields of signal mode and idler mode are denoted by a and b. Take
the amplitude gains of the two param et ric amplifiers as G1,91 and G2.92,
which is ~elow !he shot noise level, as expected for EPR entang led state
between A and B. Vie dropped both 5 and Edependent terms because 5, E « respectively. Then the operators are related by
A AO t A A A

l. The signals due to modulations are also straightforward to calculate: A= G1a in + 91bJn, B = G1bin + 91ain (11.129)
A A / A A A t

a = G2A - 92B t, b = G2B' - 92A .

(Xa) 2 = 2fpsE2, (Yb)2 = 2lp 8
2
6 , (11.127)
B' is themodulated field and B' = B(l+j5-E). The signs of the arn.plitude
where fps = 9 ial .
2 2
The SNRs are then gains are chosen so that the interferometer works at a dark fringe. Under
SNREPR = 2fpsE2 SNRE_PR = 2fps62 this condition , correlated noise generated by NP Al is canceled and noise
(11.128)
X (G-9)2" 1 (G-9)2' variance is reduced because of the phase sens itiv e de-amplification of NPA2.
To evaluate the noise of the outputs , we can ignor e the modulation terms
This result improves upon the classical SNRs in Eq. (11.119) by a factor of
by setting 5 = O = E since 5, E « 1. Then, the outputs are related to the
1/(G - 9) 2 = (G + 9 )2 .
inputs by
- At A ~A t ( )
11.4.3 Joint Measurement with SU(1,1} Interferometer a= Gain+ gbin' b = Gbin + gain' 11.130

The scheme in the previous section involves reduction of the shot noise in
where G = G 2G 1 - 9291, g = G 291 - 92G 1 are the phase sensitive gain at
photo-detection and is thus sensitive to losses at detection. To avoid this the dark fringe.
For arbitrary quadrature angles 01, 02, the noise levels of the outputs
problem, we can use the amplification scheme for the replacement of the
BS and combine the schemes in Fig. 11.15 and Fig . 11.14. The new scheme of the interferometer can be calculated as
2
is shown in Fig . 11.16, which is just the SU(l,l) interferometer that we \ 6 2Xa(0 1)) = (G2G1 - 9192)2 + (G192 - G291)
(11.131)
discussed in Section 11.3.2. We will next analyze the performance of the
SU(l,l) interferometer on the joint measurement of two quadrature-phase
\6 2.Xb(02)) = (G2G1 -9192)2 + (G192 -G291)2.
amplitudes. Note that the noise variances of different quadrature-phase amplitudes of
the output fields are determined by the gain 91 and 92 of the parametric
amplifiers only and are independent of the angles 01, 02, which is different
from the squeezed state interferometry [Xiao et al. (1987)]. When 91 =
9 , the output noise variances are minimized to
2
Xa,b(0)) = 1, which \6 2

means that the output noise levels of the two-amplifier comb in ation can
be reduced to shot noise level on both output sides of the interferometer
at the same time. This is quite different from the single amplifier case
where output noise level is amp lified from the shot noise level. This is
Fig. 11.16 Joint measurement of two orthogonal observables with an SU(l,1) interfero- because of a destructive quantum interference effect for noise cance lation
meter.
due to quantum corre lation [Ou (1993); Kong et al. (2013a)]. Thus, the
The first parametric amp lifier (NP Al) amplifies the input signal field SU ( 1, 1) int erferometer can be applied to the joint measurement at two
ain at coherent state la) to field A and produces a corre lated idler field different outputs of two arbitrary quadrature -phase amplitudes which are
B. The idler input field binis in vacuum. The idler field B then passes not necessarily orthogona l to each other.
 Quantum Noi se 'in Phas e J\1easurem ent 387
386 Quantum Optics For Exp erim entalists

For the output signals at two output ports , we find 11.5 Problems
2 2 2
(Xa) = 4g~fps6 , (Yb)2= 4G~fpsE , (11.132)
Problem 11.1 Effect of losses inside the parametric amplifiers on the per-
where fps=\ iJ+B) = grlal2(lal2 » 1) is the photon number of the probe formance of SU ( 1, 1) interferometer.
sens ing field .B. Losses inside the parametric amplifiers were considered in Eq. (6.153)
Comb~ning Eqs. (11. 131) and (11.132) , we obtain the SNRs of outputs
in Problem 6.6. Use that result and assume the two parametric amp lifier
at a and bas
in the SU(l,l) interferometer in Fig. 11.lO(a) are identica l.
SN R (X ) = 4g~ I ps52
SI a 2 2
(G2G1 - 9192) + (G192 - G291) (i) With a phase sh ift 5 on b-field, prove that the output fields of the
2
(11.133 )
int er ferometer are
SN R (y;) = 4G~fp 8 E
SI b 2 2
(G2G1 - 9192) + (G192 - G291)
where subscript SI represents the m eas urem ent scheme of SU(l,1) inter- (11.135)
ferometer. With a fixed amplitude gain 9 1, SN Rs/s of quadrature -phase
amplitudes have the m aximum value when 92 • oo. In this situ at ion ,
where a01, 'b01, a02, 'b02 are the vacuum modes coup led in through the losses
SN RsI(Xa) = 2 (G1 + 91)2 Ip s 6 2 , SN RsI(Yb) = 2 (G1 + 91)2 I psE 2 . No- inside the two NOP As and
8
tice that the optimum condit ion here is differ ent from the minimum output Gr(5) = g2 - G 2ei 8 , gr(5) = gG(l - ei ),
noise condition (91 = 92) discussed in Eq. (11.131). It is because the mo- 8 (11.136)
G'y(5) = gg' - CG' ei8 , g'y(5) = G'g - Cg'ei .
dulated signal is amplified more than the noise with the increase of gain
92· G, g, G', g' are given in Eq . (6.154).
Comparing to the results in Eq. (11.119) for the classical joint measu-
(ii) For 5 « 1 and a strong coherent state at Clin, prove that
rement schemes in Figs. 11.13 and 11.14, we have an SNR improvement
factor of (.x2bout
) = (-2 _ 0 2)2 + 41 sz0 262 + (gg' _ cc')2
9 P
SNRsI(X(0)) 2
(0 = 0, n / 2). (11.134) +(gG' - Gg')2 + c;' 2 + g'2
SNR c(X(0)) = (G 1 + 91) 2 12
= 4I;; G 252 + 1 + 2(gG' - Gg') + 29 (11.137)
This improvement factor is the same as the one for the scheme using the
EPR states. However, the scheme with EPR states can only be applied where Iffs1= ('btb) = g 2 lain 12. Hence, the SNR is
to joint measurement of orthogonal variables of Xa and Yb because of the 2 2 2 (11.138)
SN R~I = 4I;; G 5 /(1 + 4g' ),
property of squeezed states. The other quadrature -phase amp litud es will
not have minimum noise. The situation is different for the scheme with the where we used Eq. (6.154).
SU(l,1) interferomet er, which can be applied to the joint measurement of
two arbitrary quadrature-phas e amplitudes. This is so because the noise Problem 11.2 An unconventional interferometer with a parametric am -
outputs of the interferomet er given in Eq. (11.131) do not depend on the plifier and a beam splitt er [Kong et al. (2013b)].
phase angle 0, i.e., the noise is canceled for all quadrature-phase amplitudes
due to a destructive quantum interference in the int erferometer. The joint measurement scheme depicted in Fig. 11.15 is actually an
Th e experimental demonstration of joint measurement of two orthogo- unconventional interferometer in which the beam splitting is done with a
nal observables with precision beating the classical limit was performed by parametric amp lifier but the beam recomb in at ion is achieved with a regular
Li et al. [Li et al. (2002)] with a scheme proposed by Zhang and Peng (see beam splitter. This int erferometer scheme works as long as the two outputs
Problem 11.3) [Zhang and Peng (2000)]. A, B from the parametric amplifier (NPA) are degenerate in frequency. For
 389
Q 11antwn Noise 'in Phase /lfeaswnn ent
388 Quant'Um Optics For Experim entalists

A bA. arc in vacuum or coherent state, the relations above give rise to
the genera lity, ass um e t he b ea m sp litt er has a transmissivity of T. Next If a,l'Tl 1 171 C d t"
. d t· · XA + XA ·1,ndyA _ YB in contrast to the noise re uc 1011
A A •

!GI» 1.
C '

we will ass ume a stro ng coherent state inj ect ion at airi with noise re u c ,1011 111 A B c . . . )
.11l xr4. - XABan. d yAA + yAB (Eq . (10.10)) for the case of pos itive gam
.
(g > 0
(i) Replace t he modulators in Fig. 11.15 with a phase sh ifte r of phase cp. • ' d · S t· 10 1 3 At the end of Sect ion 10.1.4. we discussed the
discusse 111 ec 1011 . . . . .
Show that t he int ensit ies at the two outputs of the BS exh ibi t int erfere nce . · ·ect · at on ly one inpu t of the parametnc amplifier and found
sc11cme of mJ 1011 , A A f •

frin ges as <p changes. Find the visibility of t he int erfer en ce as a fun ction of it always leads to noise reduction in XA - XB- Things are different or
g, a, an d T and show that the visibility b ecomes 100% on one sid e when
inj ect ion at two inputs.
T = g 2 /(G 2 + g 2 ) or on the other side when T = G 2 /(G 2 + g 2 ).
------------------------ + \
(ii) Let us concentrate on the homodyn e detection on output a only. Show
~:PM
that the SNR ha s a maximum of la) :
SNR111ax = 4Iµs6 (G 2 2
+ g2) (11.139) B

when T = 4G g /(8G g + 1) ---+1/2 for

2 2 2 2
G » l. Given that the pha se A
meas ur ement SNR with a classic al interferometer is SN Rc1 = 2Iµs52 , the 10) :
improv ement over classical scheme is then :
I
-------- -
SN Rl\lax = 2(G2 + 2) (11.140)
SNRc1 g . Joint measurement of two orthogonal observables by direct detection.
Fig. 11.17
This is a small improvement over Eq. (11.128) because we optimized T.
(i) Consid er now the injection scheme shown on the le!t part of Fig . 11.l~
Problem 11.3 A scheme of direct detection for joint measurement with (dashed box). Show that for real a, the outputs ~ave_(A)~ (G A-g)a / J'i -
an EPR source and a beam splitter [Zhang and Peng (2000)]. (f3) = real. So , direct detect ion of fields A, B will yield X A, X B ·
The joint measurement scheme depicted in Fig: 11.15 involves homodyne (ii) Consider the det ect ion scheme on the right part of Fig. 11.17 (outside
detections which require extra beams for local oscillators (LO) and phase dashed box). Let us take out the PM and AM modulators first. Show that
lock to observe Xa and 1\. This makes the experimental impl ementation for large (A), (B) or a= (G - g)a/J'i » 1, we have
2
complicated. In Section 9.8, we discussed the scheme of self-homodyne ata+&th= A,tA.+ f3tf3 a(XA + XB) + 2a ,
detection by direct detection without local oscillators. We will use it here
ata - &th~a(YA - YB)- (11 -143 )
to simplify the detection scheme.
In Section 9.8, we showed that direct detection of a real bright field Th the sum and differenc~ currents make measurements of XA + XB ,
( (a) = real) measures X and in the discussion part at the end of y u~, y respectively and we achieve self-homod yne detection of these
Section 10.1.4 , we find that the twin beams generated by the injection q~antit~~ together. Show that these currents have fluctuations below the
2
of a real coherent state (la) with a = real) in a parametric amplifier have shot nois e levels by a factor of 1/ (G + g) · . . .

intensity correlation that is equivalent to noise reduction of XA -XB. This Contrary to the scheme of bright twin-beams generation 111 Section
occurs for a parametric amplifier with all positive gain param eters : g > 0. 10.1.4, where only XA - XB is measured via ~elf-ho~odyne, Athe s;hen~e
Now consider the case of negative gain, that is , presented here achieves measurement of both X A X B an~ YA - B ~ia
self-homodyne and they have the EPR-type corre lations , which demonstra-
(11.141)
tes the EPR paradox with self-homod yne detection.
which leads to
(iii) Now let us move the PM and AM modulators in place as shown. Cal-
XA + XB = (G - g)(Xa in + xb in),
culate the signa l sizes for the sum and difference currents and show that
YA - YB= (G - g)(Yain - Ybi,J. (11.142)
390 Quantum Optics Fm· Experimentalists

the sum current measures amplitude modulation E (Al\I) and the difference
current gives the phase modulation 5 (P:M).

(iv) Show the signal-to-noise ratios at the two output currents are Appendix A
SN R = 2Ipsf2 . SN R_ = 2Ips52
(G-g)2' (G-g)2 (11.144)
+
Derivation of the Explicit Expression
with fps = a Therefore, we can achieve joint measurement of the infor-
2 A

mation encoded in two orthogonal quantities at a sensitivity better than

for U of a Lossless Beam Splitter
the standard quantum limit simultaneously. Experimental demonstration
of this scheme was performed by Li et al. [Li et al. (2002)].

This derivation of U bears some resemblance to angular momentum opera-

tors in rotation in quantum mechanics.
For a lossless beam splitter, since we have t 2 + r 2 = 1. we can assign t
as cos 0 and r as sin 0. With this. we can rewrite the operator relations in
Eq. (6.46) as:
~l = ~ta1~ = cos0a1 + sin0a2. (A.l)
{ b2 = uta 2U = cos0a2 - sin0a 1,
which is similar to the transformation of the two -dimensional rotation of
angle 0. As in any transformation, we consider an infinitesimal transfor-
mation of 50 < < l and make a linear approximation of
U(50) 1 + i50i, (A.2)
where i is a function of a 1, a2 to be determined. Because r)r)t = r)tr) = 1,
we have i = ft or i is a Hermitian operator. For the transformation of
finite 0, we have:
U(0) = U(0/N)N
= lim (1 + if0/N)N
N-+oo

= exp( i0i). (A .3)

Substituting
Eq. (A.2) into Eq . (A .1), we obtain:
[a1, i] = -ia2, [a2,i] = ia1. (A.4)
From the commutators: [a1,ai] = [a2,a!] = 1, we have:
i = -ia2aI + ia1a! + J(a1, a2). (A.5)
But i is an Hermitian operator, hence f(a 1 , a2 ) = 0 and the final expression
for U is
(A.6)
with t = cos 0, r = sin 0. The above expression is the same as that in
[Campos et al. (1989)], which is based on angu lar momentum theory .

391
 Appendix B

Evaluation of the Two Sums in

Eq. (8.100)

It is straightforward to find the first sum in Eq. (8.100) as

L
k=l
= 2!+1 / dtodti ...dtN L JPtot1,,[IL G(to; JP{t1,... , tN})
k IP'
n
= ~N:/ j dtodti ...dtN L G*(to;JP'{ti, ... , tN}) L G(to; JP{ti, ... , tN})
IP'' IP'

= ~N:/ I:=/JP'{dtodti ...dtN}G*(to ;ti , ...,tN)

IP''

x L G(t 0; {ti, ... , tN}})

JP{JP'
IP'

= ~N:/ N! j dtodt1...dtNG*(to; ti, ... , tN) L G(to; JP{ti, ... , tN})

IP'
= (N + l)!N (B.l)

with

N =
2
}+1/ dtodti ...dtNG* (to;ti, ..., tN) L G(to; JP{ti, ... , tN} ).
IP'

Here, we made a change of variables: JP'{ti, ... , tN}-+ {ti, ... , tN} in evalu-
ating the integral.
To evalu ate the second sum in Eq. (8.100), let us assume T 0 = Tn
with n = 1, ... , m and ITo - Tnl » !:lT for n = m + 1, ... , N, that is , the
single photon enter ing at port 2 complete ly overlaps with m photons in
the N-photon state ente ring port 1 but is well-separated with other N - m

393
394 Quant-um Optics For Experimentalists Evaluatio11 of the Two Stl'rns ·i11 Eq. (8.100)

photons. \i\Tecalculate one arbitrary term as follov,·s: becomes

J\.1 =
2 iv+l
1 dtodt1 ...dtN
IP'~
G*(tk; JP~{t1,...to---,tN})

Pkl = ;+1
2
j dtodt1...dtN'ffDtt,. [LG*(to;JP'{t1, ..., tN})]
0
x L G(t 0: JP{ti, ...t, .... t N})
IP" IP'

x'f¥t0 t 1 [ L G(to; JP,{t1, .... tN})] = _!!2_

2N+1
j dt dt 0 1 ... dtN G*(to; JP~{t1- ...t1,:... , t N})
IP' IP'~.

= ;+1 / dtodt1 ...dtN

2
L G*(tk; JP'{ti, ...t tN}) 0 .... x LG(to; JP{t1,---k --,tN})
• IP'' IP'

x L G(tz; JP{t1,...t ,tN} ). 0 ... (B.2) = j dt dt1...dtNG*(to; t1, ...t1;;..., tN)

0
IP' 2N+l
IP'~.

x L G(to; JP{t1,...tz..., tN})

IP'
Since k -1-l , we can make a change of variable: t 0 +-+tz without the change = rn(N - l)lN. (B.4)
of the integral. This is because to, tz are both inside the permutation JP'
of G*(tk; JP'{t1, ...t 0 ... , tN}) and permutation inside permutation does not Again, we made a change of variables JP~{t1,...tk , ... , tN}-+ {t1, ... , tN} in
change the result. Hence, we have evaluating the integral in the last line of Eq . (B.4) and JP~has (N - l)l
terms. So, we have the final sum
L = rn(N + l)N(N - l)lN = m(N + l)lN. (B.5)

Pk1 =
2
;+j dtodt1...dtN LG
1
IP''
*(tk;JP'{t 1 , ... t 0 ... ,tN})
k#l
Here, the sum has a total of N(N + 1) terms.
x LG(to;JP{t1 , ---k-- ,tN }). (B.3)
IP'

To eva lu ate the integral, we note that G(to; t 1 , ... tN) = g(to - To)g(t 1 -
T1) ...g(tN -TN) with g(t) having a width of ~T. Now since t 0 is not in the
permutation of G(to; JP{t1 , . . . tz... , tN} ), this function always has a term of
g(to -To), which will determine the result of the integral depending on the
location of to in G*(tk; JP'{t 1 , .. . to ... , tN}) in the first sun1 in Eq. (B.3). If t 0
is outside the first m terms, e.g ., {t 1 , ... , trn, ... ,to, ... , tN }, the integral will
be zero because J dt 0 g*(t 0 -Tj)g(to -To)= 0 for ITo -Til » ~T(j = m+
1, ... , N) . But if to is inside the first m terms, e.g., {t1, ... ,to, ..., trn, ... , tN },
we have G* (tk; JP~{t1, ...to ... , t N}) = G*(to; JP~{t1, ...tk ... , t N}) because T 0 =
Ti for j = 1, ... , m. Here JP~,JP~ are the permutations exclud ing t 0 and
tk, respectively. There are totally m such non-zero terms and Eq. (B.3)
 Bibliography

Abbott, B. P. et al. (2016). Observation of gravitational waves from a binary

black hole merger, Phys. Rev. Lett. 116, p. 061102.
Agarwal, G. S. (2013). Quantum Optics (Cambridge University Press, New York ,
NY).
Andrews , D. L. and Babiker, M. (2013). The Angular Momentum of Light (Cam-
bridge University Press, New York, NY).
Ashkin, A., Boyd, G. D., and Dziedzic, J. M. (1966). Resonant optical second
harmonic generation and mixing, IEEE J. Quantum Electron. 2, p. 109.
Aytur, 0. and Kumar, P. (1990). Pulsed twin beams of light, Phys. Rev. Lett.
65 , p. 1551.
Bachor, H. A. and Ralph, T. C. (2004). A Guide to Experiments in Quantum
Optics, 2nd edn. (Wiley-Vch, New York).
Barnett, S. M. and Pegg, D. T. (1990). Quantum theory of optical phase corre-
lations, Phys. Rev. A 42 , p. 6713.
Bell, J. S. (1964). On the Einstein-Podolsky-Rosen paradox, Physics 1, p. 195.
Bell, J. S. (1987). Speakable and unspeakable in quantum mechanics (Cambridge
University Press, Cambridge).
Bennett, C. H., Brassard , G., Crepeau, C., Jozsa, R., Peres , A., and Wootters,
W. K. (1993). Teleporting an unknown quantum state via dual classical
and Einstein-Podolsky-Rosen channels, Phys. Rev. Lett. 70 , p. 1895.
Bennink , R. S., Bentley , S. J., and Boyd , R. W. (2002). Two-photon coincidence
imaging with a classical source, Phys. Rev. Lett. 89, p. 113601.
Beugnon, J., Jones, M. P. A., Dingjan , J., Darquie, B., Messin, G., Browaeys, A.,
and Grangier, P. (2006). Quantum interference between two single photons
emitted by independently trapped atoms, Nature 440, p. 779.
Bocquillon, E., Freulon, V., Berroir, J.-M., Degiovanni, P., Placais , B. , Cavanna,
A., Jin , Y., and Feve, G. (2013) . Coherence and indistinguishability of single
electrons emitted by independent sources, Science 339 , p. 1054.
Bohr, N. (1983). in J. A. Wheeler and W. H. Zurek (eds.), Quantum Theory and
Measurement (Princeton University Press, Princeton).
Bondurant, R. S. and Shapiro, J. H. (1984). Squeezed states in phase-sensing
interferometers, Phys. Rev. D 30 , p. 2548.

397

398 Quantum Optics FO'I· E:rpc1·ime11tali.sls
Born. l\I. and Wolf, E. (1999). Principle of Optics. 7th cdn. (Pergamon. Oxford).
Boto, A. N., Kok, P., Abrams, D.S., Braunstein, S. L., \i\Tilliams, C. P .. and Dmv-
lin g, J. P. (2000). Quantum interferometric opt ica l lith ogr ap hy: Exp loitin g
entang lement to beat the diffract ion limit , Phys. Rev . Lett. 85. p. 2733.
Bomvmeester, D., Pan, J . \ iV., Daniell, 1\1., Weinfurter, H., and Zeilinger, A.
( 1999). 0 bserva tion of three-photon Greenberger -H orne -Zeilinger entang le-
ment, Phys. Rev. Lett. 82 , p. 1345.
Bouwmeestcr, D., Pan, J . Vv.,Mattel, K .. E ibl , 1\1., Weinfurter, H., and Zeilinger,
A . (1997) . Exper im enta l quantum te leportat ion , Nat ur e 390 . p . 575 .
Boyd. R. W. (2003). Nonlinear Optics, 2nd edn. (Academ ic Press, San Diego,
CA). · photon emiss ion s from a sin gle atom, Phys. Rev. A 18 , p. 2217 .
Braginsky. V. B .. Khalili, F. Y.. and Thorne, K. S. (1992). Quantum l\Ieasnrement
(Cambr idg e Univers ity Press. Cambridge, Eng land ) .
Braunstein, S. L . (1992). Quantum limit s on precision measurements of phase,
Phys. Rev. Lett. 69 , p. 3598.
Braun ste in , S. L. and Kimbl e, H. J. (1998) . Teleportation of contin uou s quantum
variables, Phy s. Rev. Let t. 80 , p. 869.
Braunstein. S. L., Lane, A. S., and Caves. C. 1\1. (1992) . Maximum-likelihood
analy sis of multiple quantum phas e measurements, Phys. R ev . Lett. 69 ,
p. 2153.
Braun ste in , S. L. and Mann, A. (1996). Measurement of the Bell operator and
quantum teleportation, Phys. Rev. A 51 , p. Rl 727.
Brendel , J. , Gisin , N ., Titt el, W. , and Zbinden , H. (1999). Puls ed energy -time
entangl ed twin-photon source for quantum communication, Phys. R ev . Lett.
82 , p. 2594.
Brun e, M. , Hagley , E., Dr eye r, J. , Maitre, X., Maa li, A., Wund erlich , C., Rai-
mond, J.M., an d Haroche, S. (1996). Observing th e progressiv e de coherenc e
of the "meter" in a quantum measur ement , Phys. Rev. Lett. 77 , p . 4887.
Buck , J. A. (1995). Fundamentals of Optical Fib ers (Wil ey, New York).
Burnham , D. C. and Weinberg , D . L. (1970). Phys. Rev. Lett. 25 , p. 84.
Campos , R. A. , Gerry , C. C., and Benmoussa, A. (2003). Optical interferometry
at th e Heisenberg limit with twin Fock states and parity measur ements ,
Phys. Rev. A 68, p. 023810.
Campos , R. A. , Saleh, B. E. A., and Teich , M. C. (1989). Quantum-m echanical
lossless beam sp litt er: Su(2) symmetry and photon stat istics, Phys. Rev. A
40 , p. 1371.
Caruthers, P. and Nieto, M. M. (1968). Rev. Mod. Phys. 40, p. 411.
Casimir, H. B. G. (1948). On the attraction betw een two perfectly conducting
plates, Proc. Kon. Nederland. Akad. Wetensch. B51 , pp. 793- 795.
Caves, C. M. (1981). Quantum-mechanical noise in an int erferometer, Phys. Rev.
Q 23 , p. 1693.
Caves, C. M. (1982). Phys. Rev. D 26 , p. 1817.
Caves, C. M. and Schumaker, B. L. (1985). New formalism for two-photon quan-
tum optics. I. Quadrature phases and squeezed states , Phys. Rev. A 31 ,
p. 3068.
Bibliography 399
Chen . B., Qiu, C .. Chen. S .. Guo. J .. Chen. L. Q .. Ou. Z. Y .. and Zhang. \ V.
(2015). Atom-light hybrid interferometer. Phys. Rev. Lett. 115. p. 043602.
Cho i. S.-K., Vasi lyev, 1\1., and Kumar. P . (1999). Noise less opt ica l amp lification
of images. Phys. Rev . Lett. 83. p. 1938.
Clauser, J. F. (1974). Experimental distinction between the quantum and classi-
cal field-theoretic predictions for the photoelectric effect. Phys. Rev. D 9 ,
p. 853 .
Co llett, l\'I. J. and WRlls, D. F. (1985). Squeezing spectra for nonlinear optical
systems, Phys. Rev. A 32 . p. 2887.
Dagenais. l\I. and l\Iandel. L. (1978) . Investigation of tvm-t.ime correlations in
Dayan, B., Pe·er, A .. Friesem. A. A., and Silberberg, Y. (2005). Non lin ear inte-
. ractio n s with an ul trahigh flux of broadband entang led p ho tons, Phys. Rev.
Lett. 94 , p. 043602.
de Riedmatten. H., l\Iarcikic, I., Tittel. W .. Zbinden, H., and Gisin, N. (2003).
Quantum interference with ph oton pairs created in spatially separated sour -
ces, Ph ys . R ev. A 67. p. 022301.
de Riedmatten. H. , Scarani, V .. l\Iarcikic, I.. Acin. A., Tittel, W .. Zbinden, H ..
and Gisin , N. (2004). J. Mod. Opt . 51 , p. 1637.
Deleglise, S., Dotsenko , I. , Sayrin, C., Bernu, J., Brun e, M., Raimond, J.M. , and
Haroch e, S. (2008). Nature 455 , p. 510.
Diedrich , F. and Walth er , H. (1987). Nonclassical radiation of a single stor ed ion ,
Phys. R ev. Lett. 58 , p. 203.
Ding , Y. and Ou , Z. Y. (2010). Frequency down convers ion for a quantum network,
Opt. Lett. 35 , p. 2591.
Dira c, P. A. M. (1927). Pro c. R. Soc. London Ser. A 114 , p. 243.
Dirac, P. A. M . (1930). The Principles of Quantum Mechanics, 1st ed n. (Claren-
don , Oxford).
Du , S. (2015). Quantum-st ate purit y of h era lded single photons produced from
frequency-anticorrelated biphotons, Phys. R ev . A 92 , p. 043836.
Duan , L.-M., Giedke, G ., Cirac, J. I. , and Zoller, P. (2000). Inseparability criterion
for continuous variable systems, Phys. Rev. Lett. 84 , p. 2722.
Dur , W. (2001). Multipartite entanglement that is robust against disposal of
particles , Phys. Rev. A 63 , p. 020303(R).
Einst ein , A. (1905) . On a heuristic point of view about the creation and conversion
of light , Ann. der Phys. 18, pp. 132- 148.
Einstein , A. (1909). Phys. Z. 10 , pp. 185- 193.
Einstein , A. (1917). Zur quantentheorie der strahlung, Phys. Z. 18 , p. 121.
Einstein , A., Podolsky, B., and Rosen, N. (1935). Can quantum-mechanical des-
cr iption of physical reality be cons id ered complete? Phys. Rev. 47 , p. 777.
Ekert, A. K. (1991). Quantum cryptography based on Bell 's theorem , Phys. Rev.
Lett. 67 , p. 661.
Englert, B. G. (1996). Fringe visibility and which-way information: An inequality ,
Phys. Rev. Lett. 77 , p. 2154.
Fan, J. , Dogariu, A. , and Wang, L. J. (2005). Opt. Lett. 30 , p. 1530.
 400
Quantum Optics For Experirnentalists
Bibliography 401

Fano , U. (1961). Quantum theory of interferen ce effects in the mixing of light

Greenberger, D. l\I., Horn e, 1\1. A., and Zeilin ger , A. (1989). in l\I. Katafo s (ed.),
from phase ind ependent sources, Am . J. Phys. 29 , p. 539.
Bell 's Theorem. Quantum Th eory, and Conceptions of the Universe (K luwer
Fearn, H. ;:i,nrl Loudon, R. (1987). Quantum theory of the lossl ess beam splitter,
Opt. Comm. 64 , p. 485. Academic. Dordr echt , Th e Net h erlands).
Guo, X., Li, X ., Liu, N., and Ou , Z. Y. (2016a). Quantum information tapping
Fearn. H. and Loudon , R. (1989). Th eory of two-photon int erfer enc e, J. Opt. Soc.
Am. B 6 , p. 917. using a fiber optical param etr ic amplifier with noise figur e improved by
correlated inputs , Sci. Rep. 6, p. 30214.
Fortsch , M., Forst , J. U., Wittmann, C., Strekalov, D., Aiello , A. , Chekhova,
Guo, X. , Li , X. , Liu , N., Yang, L. , and Ou , Z. Y. (2012). An all-fib er source
l\f. V. , Silberhorn , C., Leuchs, G .. and l\Iarquardt, C . (2013). A versatile
of puls ed twin b ea ms for quantum communication, Appl. Phys. Lett . 101.
source of sin gle photons for quantum information processing , Nat. Comm.
4 , p. 1818. p. 261111.
Guo, X. , Liu, N. , Liu, Y. , Li , X. , and Ou, Z. Y. (2016b) . Generation of continuous
Foster , G. T., Mielke, S. L. , and Orozco, L. A. (2000). Intensity correlations in
variable quantum entanglement using a fiber optical param etric amp lifier,
cav it y QED , Phys. Rev. A 61 , p. 053821.
Opt. Lett. 41 , p. 653.
Franson , J. D. (1989). Bell inequality for position and time , Phys. Rev. Lett. 62 ,
p. 2205. Hanbury Brown , R. and Twiss , R. Q. (1956a). Correlation between photons in
two coherent beams of light , Nature 177 , pp. 27- 29.
Friberg, S. R., Hong, C. K. , and Mandel , L. (1985a). Phys. Rev. Lett. 54 , p. 2011.
Hanbury Brown, R. and Twiss, R~956b). A test of a new type of stellar
Friberg, S. R., Hong, C. K. , and Mandel , L. (1985b). Intensity dependence of the
interferometer on sirius , Nature 178 , pp. 1046- 1048.
normalized intensit y correlation function in parametric down-conversion
Opt. Comm. 54 , p. 311. ' Hanbury Brown, R. and Twiss , R. Q. (1958). Interferometry of intensit y fluctu-
ations in light II. An exper im enta l test of the theory for partially coherent
Furusawa, A., Sorensen , J. L. , Braunstein , S. L., Fuchs, C. A., Kimble, H. J ., and
light , Proc. Royal Soc. London A 243 , pp. 291- 319.
Polzik, E. S. (1998). Unconditional quantum teleportation Science 282
p . 706. ' --- ' Haroche, S. and Kleppner, D. (1989). Cavity quantum electrodynamics, Phys.
Today 31(1) , p. 24.
Gardiner, C. W . and Collett, M. J. (1985). Input and output in damped quan-
Haus, H. A . and Mullen, J. A . (1962). Phys. Rev . 128 , p . A2407.
tum systems: Quantum stochastic differential equations and the master
Heffner, H. (1962). Proc. IRE 50 , p. 1604.
equation, Phys. Rev. A 31 , p. 3761.
Heidmann, A., Borowicz, R. J. , Re yna ud , S. , Giacobino, E., Fabre , C. , and Carny ,
Gatti, A ., Brambilla, E., Bache, M., and Lugiato, L. A. (2004). Correlated ima-
G. (1987) . Observation of quantum noise reduction on twin laser beams,
ging, quantum and classical, Phys. Rev. A 70 , p. 013802.
Phys. Rev. Lett. 59 , p. 2555.
Gentle, J. E. (1998) . Numerical Linear Algebra for Applications in Statistics
(Springer-Verlag). Beitler, W. (1954). The Quantum Theory of Radiation, 3rd edn. (Oxford Univer-
sity Press, London).
Ghosh, R. and Mandel, L. (1987). Observation of nonclassical effects in the in-
Herman, G. T. (1980). Image Reconstruction from Projections: The
terference of two photons , Phys. Rev. Lett. 59, p. 1903.
Fundamentals of Computerized Tomography (Academic , New York, NY).
Glauber, R. J. (1963a). Quantum theory of optical coherence, Phys. Rev. 130,
p. 2529 . Herzog, T. J. , Kwiat, P. G. , Weinfurter, H., and Zeilinger , A . (1995) . Comple-
mentarity and the quantum eraser, Phys. Rev. Lett. 75 , p. 3034.
Glauber, R. J. (1963b). Coherent and incoherent states of the radiation field
Phys. Rev. 131 , p. 2766. ' Herzog , T. J., Rar ity, J. G., Weinfurter, H., and Zeilinger, A. (1994) . Frustrated
two -ph oton creation via interference , Phys. Rev. Lett. 72 , p. 629.
G lauber, R. J. (1963c). Photon corre lations, Phys. Rev. Lett. 1, p. 84.
Holland, M. J. and Burnett, K. (1993). Interferometric detection of optical phase
Glauber, R. J. (1964) . Quantum coherence, in C. deWitt , A. Blandin , and
shifts at the Heisenberg limit, Phys. Rev. Lett. 71 , p. 1355.
C . Cohen -Tannoudj i (eds.) , Quantum Optics and Electronics (Les Houches
Hong, C. K., Fr ib erg, S. R., and Mande l, L. (1985). J. Opt. Soc. Am. B 2 , p. 494.
Lectures) (Gordon and Breach, New York), p. 63.
Hong, C. K. and Mande l, L. (1986). Phys. Rev. Lett. 56 , p. 58.
Glauber, R. J. (1986). in E. R. Pike and S. Sarker (eds.), Frontiers in Quantum
Hong, C . K., Ou, Z. Y., and Mande l, L. (1987). Measurement of subpicosecond
Optics (Institute of Physics, Br isto l).
t ime intervals between two photons by interferenc e, Phys. Rev. Lett. 59 ,
Goodman, J. W. (2015). Statistical Optics , 2nd edn. (W iley, New York, NY).
p. 2044.
Goto, H. , Yanagihara, Y., Haibo Wang, T. H., and Kobayashi, T. (2003). Obser-
Huang, J. and Kumar, P. (1992). Observation of quantum frequency conversion,
vation of an oscillatory corre lation function of multimode two -ph oton pairs,
Phys. Rev. A 68 , p. 015803. Phys. Rev. Lett. 68 , p. 2153.
Hudelist , F. , Kong , J. , Liu , C ., Jing , J. , Ou, Z., and Zhang, W. (2014). Quantum
Grang ier, P. , Slusher, R. E., Yurke , B., and La.Porta , A. (1987) . Squeezed -light -
metrology with parametric amplifier -based photon corre lation interferome-
enhanced polarization interferometer, Phys. Rev. Lett. 59 , p. 2153.
ters , Nat . Comm. 5, p. 3049.
 402 Qu,antwn Optirs F01· Experimentalists
ImoLo, N., Haus , H. A., and Yamamoto, Y. (1985). Quantum nond emolition
measurem ent of the photon number via the optical ken effect, Phys. Rev.
32, p. 2287.
Ja cobson , J., Bjork, G., Chuang, I. , and YamamoLo , Y. (1995) . Photonic d e
Broglie waves, Phys . Rev. Lett. 74 , p. 48355.
Javauaain en , J. and Yoo , S. l\1. (1996). Quantum phase of a Bose-Einst ein con-
densate with an arbitrary number of atoms, Phys. Rev. Lett. •76 , p. 161.
Jin g, J., Liu, C., Zhou, Z. , Ou , Z. Y., and Zhang , W. (2011). Reali zat ion of a
nonlin ea r int erferom ete r with para m etri c amplifiers, Appl. Phys. Lett . 99 ,
p. 011110.
Kartner, F. X. and Haus , H. A. (1993). Quantum-nondemolition m eas urem ents
and the "collapse of the wave fun ctio n '· , Ph ys . Rev. A 47 , p. 4585.
Kell ey, P. L. and Klein er , W. H. (1964). Theory of electromagnetic field meas u-
rement and photoelectron counting, Phys. Rev. 136. p. A316.
Kimble, H. J .. Dagenais, l\I., and l\Iandcl , L. (1977). Photon antibunching in
resonance fluorescence , Phys. Rev. Lett. 39 , pp. 691- 695.
Kimble, H. J., Levin , Y., Matsko, A. B., Thorne, K. S. , and Vyatchanin , S. P.
(2001). Conversion of conventiona l gravitational-wave interferometers into
quantum nondemolition interferometers by modifying their input and/ or
output optics , Phys. Rev. D 65 , p. 022002.
Knill , E. , Laflamme, R., and Milburn, G. J. (2001). A scheme for efficient quantum
computation with lin ear optics, Nature 409, p. 46.
Kok , P., Lee, H., and Dowling , J . P. (2002). Creation of larg e-photon-numb er path
entang lement conditioned on photodetection , Phys. Rev. A 65 , p. 052104.
Kong, J., Hudelist, F., Ou, Z. , and Zhang, W. (2013a). Cance llation of int ernal
quantum noise of an amplifier by quantum correlat ion , Phys. Rev. Lett.
111, p. 033608.
Kong , J., Ou, Z., and Zhang , W. (2013b). Phase-measurement sensitivity beyond
the standard quantum limit in an interferometer consisting of a parametric
amplifier and a beam sp litter , Phys. Rev. A 87, p. 023825.
Kozlovsky, W . J. , Nabors, C. D ., and Byer, R. L. (1988). Efficient second harmo -
nic generation of a diode-laser-pumped CW ND : YAG laser using monoli -
thic l\1GO : LiNb03 external resonant cavities, IEEE J. Quantum Electron.
24 , p . 913.
Kwiat, P., Vareka , W., Hong, C., Nathel, H., and Chiao, R. (1990). Correlated
two -photon int erference in a dual-beam Michelson int erferometer, Phys.
Rev. A 41 , p. 2910.
Kwiat, P. G. , Steinberg, A. M., and Chiao, R. Y. (1992). Observation of a "quan -
tum eraser": A revival of coherence in a two -photon interfer ence exper i-
ment , Phys. R ev . A 45, p. 7729 .
Lamb, W . E . (1995). Anti -p hoton, Appl. Phys. B 60 , pp. 77- 84.
Lamb , W. E. and Retherford, R. C. (1947) . Fine structure of the hydrogen atom
by a microwav e method, Phys . Rev. 72 , pp. 241- 243.
Lane, A. S., Braunst ein , S. L., and Caves, C. M . (1993). Maximum -likelihood sta -
t istics of multiple quantum phase measurements, Phys . Rev. A 4 7, p . 1667.
Bibliography
Lee, N., Benichi, H. , Takeno, Y .. Takeda, S., Webb, J., Huntin gton, E., and
Furusawa, A. (2011). Teleportation of non-cl ass ica l wave packets of light,
Science 332. p. 330.
Legero, T., Wilk, T., Hennrich, M., R emp e, G ., and Kulm. A. (2004 ). Quantum
beat of two single photons, Phys . Rev . Lett. 93. p. 070503.
Leibfried, D. , Del\Iarco , B., l\1eyer , V. , Row e, M .. Ben-Kish, A., Britton. J.,
Itano , W. M. , Jel enkovic, B. , Lang er, C., Roscnband , T., and vViueland.
D. J. (2002). Trapped-ion quantum simul ator: Experimental application to
nonlin ea r int erferom ete rs , Ph ys . Rev. Lett. 89 , p. 247901.
Leonhardt , U. (1997). l\Ieasuring the Quantum State of Light (Cambridge Uni-
versity Pr ess, Cambridge) .
Levenson, J. A., Abram, I., Riv era, T., Fayolle, P. , Garreau, J. C., and Grangi er.
P. (1993) . Quantum optical cloning amplifier, Phys. Rev . Lett . 70 , p. 267.
Ley. M. and Loudon , R. (1984). Opt. Acta 33 , p. 371.
Li, X ., Chen, J. , Voss , P ., Sharping, J. , and Kumar , P. (2004) . Opt. Express 12 ,
p. 3737.
Li, X. , Pan , Q .. Jing , J. , Zhang, J. , Xie, C., and P eng , K. (2002). Quantum dense
coding exploiting a bright Einstein-Podolsky-Rosen beam, Phys. Rev. Lett.
88, p. 047904.
Li, X ., Yang , L. , Ma , X. , Cui, L. , Ou, Z. Y., and Yu , D. (2009). All-fiber source
of frequency-entangled photon pairs, Phys. Rev. A 79 , p . 033817 .
Liu, B. H ., Sun, F. W., Gong, Y. X., Huang, Y. F., Guo , G. C., and Ou,
Z. Y. (2007). Four-photon interferen ce with asymmetric beam splitters,
Opt. Lett. 32, p. 1320.
Liu, N., Liu, Y., Guo, X., Yang , L., Li, X., and Ou, Z. Y. (2016). Approaching
single temporal mode operation in twin beams generated by pulse pumped
high gain spontaneous four wave mixing , Opt. Expr ess 24 , p. 001096.
Loudon, R. (2000). The Quantum Theory of Light, 3rd edn. (Oxford University
Press , Oxford).
Lu, Y. J., Campbell , R. L. , and Ou, Z. Y. (2003). Mode-locked two-photon states ,
Phys. Rev. Lett. 91 , p. 163602 .
Lu, Y. J. and Ou, Z. Y. (2002). Observation of nonclassical photon statistics due
to quantum int erference, Phys. Rev . Lett. 88 , p . 023601.
Lu, Y. J. and Ou , Z. Y. (2000). Optical parametric oscillator far below threshold:
Experiment versus theory, Phys. Rev. A 62, p. 033804.
Lukens, J. M ., Dezfooliyan, A., Langrock, C., Fejer, M. M ., Leaird , D. E ., and
Weiner , A . M. (2013). Phys. Rev . Lett. 111 , p. 193603.
Lvovsky, A. I., Hansen , H ., Aiche le, T., Benson, 0 ., Mlynek, J., and Schiller,
S. (2001) . Quantum state reconstruction of the single -photon Fock state,
Phys. Rev. Lett. 87, p . 050402.
Lvovsky, A. I. and Raymer, M . G . (2009). Continuous -variable optical quantum -
state tomography, Rev. Mod . Phys. 81, p. 299.
Machida, S. and Yamamoto, Y. (1988). Ultrabroadband amplitude squeezing in
a semiconductor laser , Phys. Rev. Lett. 60 , p . 792 .

40--l Qua:ntwn Optics For Expe1 "imcntalists
l\Iachida, S .. Yamamoto, Y., and Itaya. Y. (1987). Observation of amplitude
squeezing in a constant-current-driven semiconductor laser. Phys. Rev.
Lett. 58 , p. 1000.
1\Iandel, L. ( 1983). Pho ton interfer enc e and correlation effects produced by inde-
pendent quantum sources, Phys. Rev. A 28, p. 929.
l\Iandel, L. (1991). Coherence and indistinguishability. Opt. Lett. 16 , p. 1882.
l\1andel, L . (1999). Quantum effects in one -photon and two-photon interference,
Rev . Mod. Phys. 71 , p. S274.
l\1andel, L., Sudarshan, E. C. G .. and Wolf, E. (1964) . On a heuristic point of
view about the creation and conversion of light , Proc. Phys. Soc. 84, p. 435.
Mandel, L. and Wolf, E. (1997). Optical Coherence and Quantum Optics (Cam-
bridge University Press, New York , NY).
Martynov, D. V. et al. (2016). Sensitivity of the advanced ligo detectors at the
beginning of gravitational wave astronomy, Phys. Rev. D 93 , p. 112004.
Maunz, P., Moehring, D. L. , Olmschenk, S., Younge, K. C., Matsukevich, D. N.,
and Monroe, C. (2007). Quantum interference of photon pairs from two
remote trapped atomic ions, Nat. Phys. 3 , p. 538.
McCormick, C. F., Boyer, V., Arimondo, E., and Lett , P. D. (2007) . Strong
relative intensity squeezing by four-wave mixing in rubidium vapor, Opt.
Lett. 32, p. 178. --
Mehta, C. L. and Sudarshan, E. C. G. (1965). Relation between quantum and
sem iclassica l description of optical coherence, Phys. Rev. 138, p. B274.
Michelson, A. A. (1890). Phil. Mag. 30 , p. 1.
Michelson, A. A. (1920). Astrophys. J. 51 , p. 257.
Michelson, A. A. and Pease, F. G. (1921). Astrophys. J. 53, p. 249 .
Milburn , G. J., Steyn-Ross, M . L., and Walls, D. F. (1987). Phys. Rev. A 35,
p. 4443.
Milburn, G. J. and Walls, D. F. (1983). Quantum nondemolition measurements
via quantum count in g, Phys. Rev. A 28, p. 2646.
Monroe, C., Meekhof, D. M. , King, B. E., and Wineland, D. J. (1996) . Science
272 , p. 1131. ---
Myatt, C. J., King, B. E., Turchette, Q. A., Sackett, C. A., Kielpinski, D., Itano,
W. H., Monroe, C., and Wineland, D. J. (2000). Nature 403, p. 269.
Nasr, M. B., Saleh, B. E. A., Sergienko, A. V., and Teich , M. C. (2003). De-
monstration of dispersion-canceled quantum-optical coherence tomography,
Phys. Rev. Lett. 91 , p. 083601.
Niu, X.-L., Gong, Y.-X., Liu , B.-H., Huang , Y.-F. , Guo, G.-C., and Ou, z. Y.
(2009). Observation of a generalized bunching effect of six photons, Opt.
Lett. 34 , p. 1297. --
Noh, J., Fougeres, A., and Mandel, L. (1991). Measurement of the quantum phase
by photon counting, Phys. Rev. Lett. 67 , p. 1426.
Noh, J., Fougeres, A., and Mandel, L. (1992). Operational approach to the phase
of a quantum field, Phys. Rev. A 45 , p. 424.
Noh, J., Fougeres, A., and Mandel, L. (1993). Measurements of the probability
distribution of the operationally defined quantum phase difference, Phys.
Rev. Lett. 71 , p. 2579. --
Bibliography --105
Ou, z. Y. (1988). Quantum theory of fourth-order interference, Phys. Re,·. A 37.
p. 1607.
Ou , z. Y. (1993). Quantum amplificaLion with correlated quantum fields. Phys.
Rev. A 48, p. Rl 761.
Ou, Z~996). Quantum multi-particle interference due to a single -photon
state, Quantum and Semicl. Opt. 8, p. 315.
Ou, z. Y. (1997a). Fundamental quantum limit in precision phase measurement,
Phys. Rev. A 55, p. 2598. .
Ou, z. Y. (1997b). Parametric down-conversion with coherent pulse pumpi~1g and
quantum int erference between ind ependent fields, Quantum and Semi-class.
Opt. 9, p . 599.
Ou, z~(2008). Characterizing temporal distinguishability_ of an n-~hoton state
by a generalized photon bunching effect with multiphoton mterference,
Phys. Rev. A 77 , p. 043829.
Ou, z. Y., Hong , C. K., and Mandel, L. (1987). Relation between input and
output states for a beam splitter, Opt. Comm. 63 , p. 118. .
Ou, z. Y. and Kimble, H.J. (1993). Enhanced conversion efficiency for harmomc
generation with double resonance, Opt. Lett. 18 , p. 1053. . .
Ou , z. Y. and Kimble, H. J. (1995). Probability distribution of photoelectric
currents in photodetection processes and its connection to the measurement
of a quantum state, Phys. Rev. A 52, p. 3126. .
Ou, z. Y. and Lu, Y. J. (1999) . Cavity enhanced spontaneous param~tnc down -
conversion for the prolongation of corre lation time between conJugate pho-
tons, Phys. Rev. Lett. 83, p. 2556. . .
Ou, z. Y. and Mandel, L. (1988). Observation of spatial quantum beatmg with
separated photodetectors, Phys. Rev. Lett. 61 , p. 54.
Ou, z. Y. and Mandel, L. (1989). Derivation of reciprocity relations for a beam
splitter from energy balance, Am. J. Phys. 57 , p. 66. . .
Ou, z. Y., Pereira, S. F., and Kimble, H. J. (1992a). Realization of the Emstem-
Podolsky-Rosen paradox for continuous variables in nondegenerate para-
metric amplification, Appl. Phys. B 55, p. 265. . .
Ou, z. Y., Pereira, S. F., and Kimble , H. J. (1993). Quantum noise reduction m
optica l amplification, Phys. Rev. Lett. 70, p. 3239 . . .
Ou , z. Y., Pereira, S. F. , Kimble , H.J., and Peng, K. C. (1992b). Realization of
the Einstein-Podolsky-Rosen paradox for continuous variables, Phys. Rev.
Lett. 68, p. 3663.
Ou, z. Y., Rhee, J. K., and Wang , L. J. (1999a). Observation of four-pho_ton
interference with a beam splitter by pulsed parametric down-conversion,
Phys. Rev. Lett. 83, p . 959. .
Ou, z. Y., Rhee, J. K., and Wang, L. J. (1999b). Photon bunching and multipho-
ton interference in parametric down-conversion, Phys. Rev. A 60 , p. 593.
Ou , z. Y. and Su, Q. (2003). Uncertainty in determining the phase for an optical
field due to the particle nature of light , Laser Phys. 13 , p. 1175.
Ou, z. Y., Zou, X. Y., Wang, L. J., and Mandel, L. (1990a). Observation of
nonlocal interference in separated photon channe ls, Phys. Rev. Lett. 65 ,
p. 321.
406 Quantwn Optics For E1:per'irncntalists
Ou. Z. Y., Zou. X. Y.. Wang, L. J .. and l\Iandcl. L. (1990b). Experiment on
nonclassical fomth-order interference, Phys. Rev A 42, p. 2957.
Ou, Z. Y . J. (2007). l\1ulti-Photon Quantum Interference (Springer. New York,
NY).
Pan, J. VI/., Bouw1neester. D., Weinfurter, H., and Zeilinger. A. (1998). Experi-
mental entanglement sv:apping: Entangling photons that never -interacted,
Phys. Rev. Lett. 80, p. 3891.
Pegg, D. T. and Barnett, S. l\1. (1988). Europhys. Lett. 6. p. 483.
Pegg, D. T. and Barnett. S. M. (1989). Phys. Rev. A 39, p. 1665.
Pfleegor, R. L. and Mandel, L. (1967a). Phys. Lett. 24A. p. 766.
Pfleegor, R. L. and Mandel, L. (1967b). Phys. Rev. 159, p. 1084.
Pittman. T. B.. Shih, Y. H., Strekalov, D. V .. and Sergienko. A. V. (1995). Optical
imaging by means of two-photon quantum entanglement, Phys. Rev. A 52,
p. R3429.
Pittman, T. B., Strekalov, D. V., Migdall, A., Rubin, M. H., Sergienko, A. V.,
and Shih. Y. H. (1996) . Can two-photon interference be considered the
interference of two photons, Phys. Rev. Lett. 77, p. 1917.
Planck, M. (1900). On an improvement of wien's equation for the spectrum,
Verhandlungen der Deutschen Physikalischen Gesellschaft 2, pp. 202- 204.
Plick, W . N ., Dowling, J. P., and Agarwal, G. 8. (2010) . Coherent-light -boosted.
sub-shot noise, quantum interferometry, New J. Phys . 12, p. 083014.
Polzik, E. 8. and Kimble, H. J. (1991). Frequency doubling with K Nb03 in an
externa l cavity, Opt. Lett. 16, p . 731.
Prasad, 8 ., Scully, M. 0 ., and Martienssen, W. (1987). A quantum description of
the beam splitter, Opt. Comm. 62, p. 139 .
Rarity, J. G. (1995). in D. M. Greenberger and A. Zeilinger (eds.), Fundamental
Problems in Quantum Theory (Ann. NY Acad. Sci. vol. 755), p. 624 .
Rarity, J. G. and Tapster, P. R. (1989). Fourth -order interference in parametric
downconversion, J . Opt. Soc. Am. B 6, p. 1221.
Reid, M. D. (1989). Demonstration of the Einstein -Podolsky -Rosen paradox using
nondegenerate parametric amp lification, Phys. Rev. A 40, p . 913.
Rice, P.R. and Carmichael, H . J. (1988). Sing le-atom cavity -enhanced absorption.
i. photon statistics in the bad-cavity limit, IEEE J. Quantum E lectron. 24,
p. 1351.
Richter, G. (1977). Abh. Acad. Wiss. DDR 7 N , p . 245.
Sanaka, K., Jennewein, T ., Pan, J .-W., Resch, K., and Zeilinger, A. (2004).
Exper imenta l non linear sign shift for linear optics quantum computation,
P hys . Rev. Lett. 9 2 , p . 017902.
Sanaka, K., Resch, K. J., and Zeilinger, A. (2006) . Filtering out photonic Fack
states, Phys. Rev. Lett. 96, p . 083601.
Sanders, B. C. and Mi lburn, G. J. (1989) . Comp lementarity in a quantum non-
demo lition measurement, Phys. Rev. A 39 , p. 694 .
Santori, C ., Fatta l, D., Vuckov ic, J., Solomon, G . 8., and Yamamoto, Y. (2002).
Ind istinguishab le photons from a sing le-photon device, Nature 419 , p. 594.
Schaw low, A . L . and Townes, C. H . (1958). Infrared and optica l masers, Phys.
Rev . 112 , p . 1940.
Bibliography 407
Schleich, W .. Dowling, J. P .. and Borowicz, R . .J. (1991). Exponential decrease
in phase uncertainty, Phys. Rev. A 44, p. 3365.
Schleich. W. and Wheeler, J. A. (1987). Oscillations in photon distribution of
squeezed states and interference in phase space, Nature 326, p. 57<-l.
Schleich. W. P. (2001). Quantum Optics in Phase Space (Wiley-Yeh. Berlin).
Schrodinger, E. (1926). aturwissenschaften 14. p. 66--l.
Schrodinger, E. (1935). Naturwissenschaften 23, pp. 807, 823, 844.
Scully, l\I. 0. and Druhl, K. (1982). Quantum eraser: A proposed photon corre-
lation experiment concerning observation and "delayed choice'· in quantum
mechanics, Phys. Rev. A 25, p. 2208.
Scully, M. 0., Englert, B.-G., and Walther, H. (1991). Nature 351, p. 111.
Scully, l\I. 0. and Zubairy. M. 8. (1995). Quantum Optics (Cambridge University
Press, New York, NY).
Shafiei, F., Srinivasan, P .. and Ou, Z. Y. (2004). Generation of three-photon
entangled state by quantum interference between a coherent state and pa-
rametric down-conversion, Phys. Rev . A 70, p. 043803.
Shapiro, J. H. (1980). Optical waveguide tap with infinitesimal insertion loss,
Opt. Lett. 5, p. 351.
Shapiro, J. H., Shepard, 8 . R., and Wong, N. C. (1989). Measurements of the pro-
babi lity distribution of the operationally defined quantum phase difference,
Phys. Rev. Lett. 62, p. 2377.
Short, R. and Mandel, L. (1983). Observation of sub-Poissonian photon statistics,
Phys. Rev. Lett. 51, p. 384.
Siegman, A. E. (1986) . Lasers (University Science Books, Mill Valley, CA).
Simon, R. (2000). Peres -Horodecki separability criterion for continuous variable
systems, Phys. Rev. Lett. 84, p . 2726.
Slusher, R. E., Grangier, P ., LaPorta, A., Yurke, B., and Potasek, M . J . (1987).
Pulsed squeezed light, Phys. Rev. Lett. 59, p. 2566.
Slusher, R. E., Hollberg, L. W., Yurke, B., Mertz, J. C., and Valley, J. F. (1985) .
Observation of squeezed states generated by four -wave mixing in an optical
cav ity, Phys. Rev . Lett. 55, p. 2409 .
Smiles -Mascarenhas, K. (1991). Am. J. Phys. 59, p. 1150.
Smithey, D . T., Beck, M., Raymer, M. G., and Faridani, A. (1993). Measure -
ment of t he Wigner distribution and the density matrix of a light mode
using optical homodyne tomography: Application to squeezed states and
the vacuum, Phys. Rev. Lett. 7 0 , p. 1244 .
Sto ler, D . (1970). Equivalence classes of minimum uncertainty packets, Phys. Rev.
D 1 , p. 3217.
Sudar~an, E. C . G. (1963). Equivalence of semiclassical and quantum mechanical
desc riptions of statistical light beams, Phys. Rev. Lett. 10, p. 277.
Sun, F. W., Liu, B. H., Huang, Y. F., Ou, Z. Y., and Guo, G . C. (2006). Obser-
vat ion of the four -photon de Broglie wave length by state -project ion mea-
surement Phys. Rev. A 7 4, 033812.
Sun, F. W ., Liu, B. H., Gong, Y. X., Huang, Y. F., Ou, Z. Y., and Guo, G. C.
(2007). Stimulated emission as a resu lt of multi photon interference, Phys.
Rev. Lett. 99, p. 043601.
-108 Q11ant11
.m, Optics Fo·,· E xp erim en/.al'ist.s
Bibliography 409

Sun. F . W. and Wong , C. W. (2009). Indistinguishability of independent sing le

Yuen. H. P. (1976). Two-photon coherent states of th e radiatiou field. Phys. Rev.
photons , Phys. Rev. A 79 , p. 013824. 13, p. 2226.
Susskind, L. and Glogower. J. (1964). Physics 1. p. 49 .
Yuen , H. P. and Chan , V. W. S. (1983). Noise in homodyne and heterodyne
Takeda, S., l\Iizuta , T., Fuwa, M., van Loock, P., and Furusawa, A. (2013) . detection, Opt. Lett. 8, p. 177.
Deterministic quantum te leportation of photonic quantum bi~s by a hybrid
Yurke . B .. l\IcCall, S. L., and Klauder , J. R. (1986). SU(2) and SU(l, 1) interfe-
techn iqu e, Nature 500 , p. 315. rometers, Phys. Rev. A 33 , p. 4033.
Takesue, H. (2010 ). Single-photon frequency down-conversion experiment , Phys. Yurke, B. and Potasek, M. (1987). Obtainment of thermal no ise from a pure
Rev . A 82 , p. 013833. quantum state, Phys. Rev. A 36 , p. 3464 .
Tay lor~ (1909) . Proc. Cam b. Phil. Soc. 15 , p. 114.
Yurke. B. and Sto ler, D. (1992) . E in stein-Podo lsky -R osen effects from indepen-
Teich, M. C. and Saleh, B. E. A . ( 1988). Photon bunching an d antib un chin g, in dent particle sources , Phys. Rev. Lett. 68 , p. 1251.
E. Wolf (ed.), Progress in Optics, Vol. XXVI (Elsevier Science Publishers
Zajonc. A. G. , Wang, L. J., Zou, X. Y.. and Mandel, L. (1991). Quantum eraser.
B.V.), pp. 1- 104. Nature 353 , p. 507.
Vahlbruch, H. , Mehmet, M., Danzmann, K. , and Schnabel, R. (2016). De- Zeilinger, A. (1981). General properties of lossless beam sp litters in interferome-
tect ion of 15 dB squeezed states of light and t heir appli cat ion for the abso - try, Am. J. Phys. 49 , p. 882.
lut e ca libration of photo electr ic qu antum efficiency, Phys. Rev. Lett. 117 , Zeilinger, A. (1999). Experiment and t he foundations of quantum physics, Rev.
p. 110801. Mod. Phys. 71 , p. S288. --
Vaidman, L. (1994). Teleportation of quantum states, Phys. Rev. A 49 , p. 1473 .
Zernike, F. (1938). The concept of degree of cohere nce and its ap pli cat ion to
van Cittert, P. H. (1934). Physica 1 , p. 201. optical probl ems, Phy sica 5 , p. 785.
Vandevender, A. P. and Kwiat , P. G. (2004) . High efficiency single photon de-
Zh ang , J. and P eng , K. (2000). Qu ant um teleportation and dense coding by me ans
tection via frequenc y up-conv ersion , J. Mod. Opt. 51 , p. 1433.
of bright amplitude-squeezed light and direct m eas urement of a Bell stat e,
Walls , D. F. and Milburn , G. J. (1985). Phys . Rev. A 31 , p. 2403. Phys. Rev. A 62 , p. 064302.
Walls, D. F . and Milburn, G. J. (2008). Quantum Optics , 2nd edn. (Springer-
Zou , X. Y. , Wang , L. J. , and Mandel , L. (1991a). Violation of classical probability
Verlag, Berlin). in param et ric down-conversion , Opt. Comm. 84 , p. 351.
Wang , H. and Kobayashi, T. (2005). Phase meas urement at the Heisenberg limit
Zou , X. Y. , Wang , L. J. , and Mand el, L. (1991b). Induced coherence and indis-
with three photons, Phys. Rev. A 71, p. 021802(R). tinguishabi lity in optical interfer en ce, Phys. Rev. Lett. 67 , p. 318.
Wang, L. J. and Rhee , J. K. (1999). in Quantum Electronics and Laser Scienc e Zukowski , M., Zeilinger , A., and Weinfurter, H. (1995). in D. M. Greenberger and
Conference, OSA Technical Digest (Optical Soci ety of America, Washington
A. Zeiling er (eds.), Fundamental Probl ems in Quantum Theory (Ann. NY
DC), p. 143. Acad. Sci. vol. 755) , p. 91.
Wigner , E. P. (1932). On the quantum correction for thermodynamic equilibrium,
Phys. Rev. 40 , p. 749.
Wu, L. A. and Kimble , H. J. (1985). Interference effects in second -harm onic
generation within an optical cavity , J. Opt. Soc. Am. B 2 , p. 697.
Wu, L.-A., Kimble, H. J. , Hall, J., and Wu, H. (1986). Generation of squeezed
states by parametric down conversio n , Phys. Rev. Lett. 57 , p. 2520.
Wu, L. A., Xiao, M. , and Kimble , H. J. (1987). Squeezed states of light from an
optical parametric oscillator, J. Opt. Soc. Am. B 4 , p. 1465.
Xiao, M., Wu, L.-A., and Kimble, H. J. (1987). Precision measurement beyond
the shot-noise limit, Phys. Rev. Lett. 59, p. 278.
Xie , Z. , Zhong, T., Shrestha, S., Xu, X., Liang, J., Gong , Y., Bienfang, J. C.,
Restelli , A., Shapiro , J. H., Wong, F. N. C., and Wong, C. W. (2015). Har-
nessing high-dimensional hyperentanglement through a biphoton frequency
comb, Nat. Photon. 9 , p. 536.
Yonezawa, H. , Braunstein, S. L. , and Furusawa, A. (2007). Experimental demon-
stration of quantum teleportation of broadband squeezing, Phys. Rev. Lett.
99 , p. 110503.
 Index

accidental coincidence , 202 tim e, 206 , 239

anti-bunching effect 21 142 202
207,267 ' ' ' ' den sit y op era t or , 82
displacement operator , 65, 75, 91
beam splitter , 161 distinguishability , 230 , 248 , 347
evolution operator , 162, 391 optical coherenc e, 261
Bell states, 119, 267 photon pair, 252
bunching effect , 19, 113, 128, 142,
202, 206, 213 Einstein-Podolsky-Rosen
multi-photon, 245, 254 correlation , 124
two -photon interference , 243 entangled state , 125 , 304
entanglem ent, 308
Casimir effect, 50
paradox , 301
Cauchy-Schwarz inequality, 20, 113,
142, 147, 149, 211
field operators, 45
cavity, 169
finesse , 171, 177
Gaussian beams, 36
modes, 39
G lauber-Sudarshan P -representation ,
classical coherence theory, 16, 129
70, 94, 139, 145, 165
coherence function, 17, 18
coherence time, 9, 111
Hanbury Brown -Twiss experiment,
coherent states, 64, 96
19, 112, 134
Wigner function, 100
coincidence measurement, 134, 198, homodyne detection, 276
201,205 balanced, 285
time resolved, 205 pulsed , 290
corre lation function , 16, 18, 109, 111, self, 287
130, 137 Hong-Ou -Mande l interference , 163,
cross -in tensity, 113, 142, 212 222
intensity, 18, 112 , 134, 141, 204 , multi -ph oton, 246, 265
208 two pairs , 263

411
412 Qucmtu.m Optics For E:rperi111ental-ists

Isserlis Theorem, 15. 20, 112, 134 spectrum of squeezing, 125, 182, 282
spherical waves, 36
mixed states, 81 squeezed states, 72, 146
mode of optical fields, 32. 55 amplitude, 79. 294
generation, 155
NOON states, 119, 121 measurement, 299
normal ordering, 88. 92, 136. 147 multi-frequency mode, 125, 160,
number states, 59, 97 181, 184, 282, 303
phase, 79, 105
one-dimensional approximation, 48 pulsed, 161
optical coh er ence, 9 two-mode, 122, 169, 180
optical parametric oscillator Wigner function, 101
degenerate, 180 squeezing operator, 73, 91, 365
non-degenerate, 178, 186 two-mode, 122
threshold, 184 ste llar interferometry, 22, 136
sub-Poisson photon statistics, 145,
particle -wave dua lity, 4, 54 294
plane waves, 29
temporal mode, 41, 117, 157, 160,
quadrature -phase amplitudes, 67, 276 284, 289, 290, 292
multi-mode, 126, 281 thermal states, 87, 96, 123
quantization, 42 transform-limited pulses , 114
quantum coherence theory , 137 twin beams, 122, 147 , 169 , 305
quantum erasers, 230 experiment, 307
quasi -monochromatic field, 47 two -photon interference, 23
classical limit of visibility, 221
Schrodinger cat states, 70 time -bin entanglement, 227
Wigner function, 102 two -photon wave function, 220
second harmonic generation , 171, 187 two -photon states, 118
shot noise, 273, 278 , 279, 281, 284 frequency and time entangled, 120
spectrum, 274, 275 generation, 154
simple harmonic oscillator , 42 polarization, 119
sing le-ph oton states , 115 two -ph oton wave function , 209
wave function, 62
wave packet, 116 Wigner function, 97
Wigner function, 102 measurement, 330
 Quantum
Optics
forExperimentalists
This book on quantum optics is from the point of view of an
experimentalist . It approaches the theory of quantum optics
with the language of optical modes of classical wave theory, with
which experimentalists are most familiar. This approach makes
the transition easy from classical optics to quantum optics. The
emphasis on the multimode description of an optical system is
more realistic than in most quantum optics textbooks. After the
theoretical part, the book goes directly to the two most basic
experimental techniques in quantum optics and establishes
the connection between the experiments and the theory. The
applications include some key quantum optics experiments, and
a few more current interests that deal with quantum correlation
and entanglement, quantum noise in phase measurement and
amplification, and quantum state measurement.

World Scientific ISBN 978-98 1-3220·20 · 1 (pbk )

www.worldscientific.com
10453SC
111111111111111111
9 789813
Ill Ill
220201
llllll