Progress in Nano-Electro Optics III-Springer (2004)

Springer Series in
optical sciences 96
founded by H.K.V. Lotsch
Editor-in-Chief: W. T. Rhodes, Atlanta

Editorial Board: T. Asakura, Sapporo
K.-H. Brenner, Mannheim
T. W. Hänsch, Garching
T. Kamiya, Tokyo
F. Krausz, Wien and Garching
B. Monemar, Lingköping
H. Venghaus, Berlin
H. Weber, Berlin
H. Weinfurter, München
Springer Series in
optical sciences
The Springer Series in Optical Sciences, under the leadership of Editor-in-Chief William T. Rhodes, Georgia
Institute of Technology, USA, provides an expanding selection of research monographs in all major areas of
optics: lasers and quantum optics, ultrafast phenomena, optical spectroscopy techniques, optoelectronics,
quantum information, information optics, applied laser technology, industrial applications, and other
topics of contemporary interest.
With this broad coverage of topics, the series is of use to all research scientists and engineers who need
up-to-date reference books.
The editors encourage prospective authors to correspond with them in advance of submitting a manu-
script. Submission of manuscripts should be made to the Editor-in-Chief or one of the Editors.
Editor-in-Chief
William T. Rhodes Ferenc Krausz
Georgia Institute of Technology Vienna University of Technology
School of Electrical and Computer Engineering Photonics Institute
Atlanta, GA 30332-0250, USA Gusshausstraße 27/387
E-mail: [email protected] 1040 Wien, Austria
E-mail: [email protected]
and
Editorial Board Max-Planck-Institut für Quantenoptik
Hans-Kopfermann-Straße 1
Toshimitsu Asakura 85748 Garching, Germany
Hokkai-Gakuen University
Faculty of Engineering Bo Monemar
1-1, Minami-26, Nishi 11, Chuo-ku Department of Physics
Sapporo, Hokkaido 064-0926, Japan and Measurement Technology
E-mail: [email protected] Materials Science Division
Linköping University
58183 Linköping, Sweden
Karl-Heinz Brenner E-mail: [email protected]
Chair of Optoelectronics
University of Mannheim Herbert Venghaus
Institute of Computer Engineering Heinrich-Hertz-Institut
B6, 26 für Nachrichtentechnik Berlin GmbH
68131 Mannheim, Germany Einsteinufer 37
E-mail: [email protected] 10587 Berlin, Germany
Theodor W. Hänsch
Horst Weber
Max-Planck-Institut für Quantenoptik
Hans-Kopfermann-Straße 1 Technische Universität Berlin
85748 Garching, Germany Optisches Institut
E-mail: [email protected] Straße des 17. Juni 135
10623 Berlin, Germany
Takeshi Kamiya
Ministry of Education, Culture, Sports Harald Weinfurter
Science and Technology Ludwig-Maximilians-Universität München
National Institution for Academic Degrees Sektion Physik
3-29-1 Otsuka, Bunkyo-ku Schellingstraße 4/III
Tokyo 112-0012, Japan 80799 München, Germany
E-mail: [email protected] E-mail: [email protected]
Motoichi Ohtsu (Ed.)
Progress
in Nano-Electro-Optics III
Industrial Applications and Dynamics
of the Nano-Optical System
With 186 Figures and 8 Tables
123
Professor Dr. Motoichi Ohtsu
The University of Tokyo
Department of Electronics Engineering
School of Engineering
7-3-1 Hongo, Bunkyo-ku
Tokyo 113-8656
Japan
ISSN 0342-4111
ISBN 3-540-21050-4 Springer Berlin Heidelberg New York
Library of Congress Cataloging-in-Publication Data

Progress in nano-electro-optics III : industrial applications and dynamics of the nano-optical system /
Motoichi Ohtsu (ed.). p.cm. – (Springer series in optical sciences ; v. 96)
Includes bibliographical references and index.
ISBN 3-540-21050-4 (alk. paper)
1. Electrooptics. 2. Nanotechnology. 3. Near-field microscopy. I. Ohtsu, Motoichi. II. Series.
TA1750 .P75 2002 621.381’045–dc21 2002030321
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or
parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in
its current version, and permission for use must always be obtained from Springer. Violations are liable for
prosecution under the German Copyright Law.
Springer is a part of Springer Science+Business Media
springeronline.com
© Springer-Verlag Berlin Heidelberg 2005
Printed in Germany
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
Data prepared by the author using a Springer TEX macropackage
Data conversion by EDV-Beratung F. Herweg, Hirschberg
Cover concept by eStudio Calamar Steinen using a background picture from The Optics Project. Courtesy of
John T. Foley, Professor, Department of Physics and Astronomy, Mississippi State University, USA.
Cover production: design & production GmbH, Heidelberg
Printed on acid-free paper SPIN 10985588 57/3141/di 543210
Preface to Progress in Nano-Electro-Optics
Recent advances in electro-optical systems demand drastic increases in the

degree of integration of photonic and electronic devices for large-capacity and
ultrahigh-speed signal transmission and information processing. Device size
has to be scaled down to nanometric dimensions to meet this requirement,
which will become even stricter in the future. In the case of photonic devices,
this requirement cannot be met only by decreasing the sizes of materials. It is
indispensable to decrease the size of the electromagnetic field used as a carrier
for signal transmission. Such a decrease in the size of the electromagnetic field
beyond the diffraction limit of the propagating field can be realized in optical
near fields.
Near-field optics has progressed rapidly in elucidating the science and
technology of such fields. Exploiting an essential feature of optical near fields,
i.e., the resonant interaction between electromagnetic fields and matter in
nanometric regions, important applications and new directions such as stud-
ies in spatially resolved spectroscopy, nanofabrication, nanophotonic devices,
ultrahigh-density optical memory, and atom manipulation have been realized
and significant progress has been reported. Since nanotechnology for fabri-
cating nanometric materials has progressed simultaneously, combining the
products of these studies can open new fields to meet the above-described
requirements of future technologies.
This unique monograph series entitled “Progress in Nano-Electro-Optics”
is being introduced to review the results of advanced studies in the field of
electro-optics at nanometric scales and covers the most recent topics of the-
oretical and experimental interest in relevant fields of study (e.g., classical
and quantum optics, organic and inorganic material science and technology,
surface science, spectroscopy, atom manipulation, photonics, and electron-
ics). Each chapter is written by leading scientists in the relevant field. Thus,
high-quality scientific and technical information is provided to scientists, en-
gineers, and students who are and will be engaged in nano-electro-optics and
nanophotonics research.
I gratefully thank the members of the editorial advisory board for valu-
able suggestions and comments on organizing this monograph series. I wish to
express my special thanks to Dr. T. Asakura, Editor of the Springer Series in
Optical Sciences, Professor Emeritus, Hokkaido University for recommending
VI Preface to Progress in Nano-Electro-Optics
me to publish this monograph series. Finally, I extend an acknowledgement

to Dr. Claus Ascheron of Springer-Verlag, for his guidance and suggestions,
and to Dr. H. Ito, an associate editor, for his assistance throughout the prepa-
ration of this monograph series.
Yokohama, October 2002 Motoichi Ohtsu

Preface to Volume III
This volume contains five review articles focusing on different aspects of

nanoelectro-optics. The first article reviews fabrications of advanced fiber
probes, i.e., application-oriented fiber probes, and applications to microscopy
and spectroscopy.
The second article is devoted to reviewing a localized surface plasmon
sensor using uniform surface-adsorbed metal particles. The unique feature of
this sensor is exploited to make contributions to the field of life sciences, cul-
minating in productive interactions between nano-optics and biotechnology.
The third article concerns industrial applications to ultrahigh-density op-
tical storage. A highly efficient near-field optical head, using a wedge-shaped
metallic plate, is described.
The fourth article also deals with ultrahigh-density optical storage. Cir-
cumferential magnetic patterned media are reviewed, which are prepared by
an artificially assisted self-assembling method. These media can be used to
realize a novel storage system with high contrast and high signal-to-noise
ratio.
The last article reviews the study on quantum optical near-field inter-
action of nanometric particles, i.e., transfer of the dipole moments of the
systems and dynamics of localized photons in an open system. Through this
study, two kinds of phases are predicted: storage and nonstorage modes of
localized photons in the system. Finally, chaotic behavior, due to the nonlin-
earity of the equations of motion, is described.
As was the case of Volumes I and II, this volume is published with the
support of an associate editor and members of editorial advisory board. They
are:
Associate editor: Ito, H. (Tokyo Inst. Tech., Japan)
Editorial advisory board: Barbara, P.F. (Univ. of Texas, USA)
Bernt, R. (Univ. of Kiel, Germany)
Courjon, D. (Univ. de Franche-Comte, France)
Hori, H. (Yamanashi Univ., Japan)
Kawata, S. (Osaka Univ., Japan)
Pohl, D. (Univ. of Basel, Switzerland)
Tsukada, M. (Univ. of Tokyo, Japan)
Zhu, X. (Peking Univ., China)
VIII Preface to Volume III
I hope that this volume will be a valuable resource for the readers and future
specialists.
Yokohama, July 2004 Motoichi Ohtsu

Contents
Near-Field Optical Fiber Probes

and the Imaging Applications
S. Mononobe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Basic Techniques for Tapering and Metallizing Optical Fibers . . . . . 3
2.1 Heating-and-pulling and Metallization Techniques . . . . . . . . . . . 3
2.2 Meniscus Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Selective Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 Protrusion-type Probe and Its Imaging Applications . . . . . . . . . . . . . 12
3.1 Protrusion-type Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Fabricating Protrusion-type Probes
by Selective Resin-Coating Method . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 c-mode SNOM Imaging of Salmonella Flagellar Filaments
in Air and Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 c-mode SNOM Images of Microtubules . . . . . . . . . . . . . . . . . . . . 18
3.5 Near-Field Spectroscopic Investigation
of Semiconductor Quantum Dots
Under Extremely Low Temperature . . . . . . . . . . . . . . . . . . . . . . . 19
3.6 Transmission Efficiencies of the Protrusion-type Probes . . . . . . 22
4 Metal-Dielectric-Metal-Coated Fiber Probe
and Near-Field Imaging of DNA Molecules . . . . . . . . . . . . . . . . . . . . . 23
4.1 Ag-MgF2 -Al-Coated Fiber Probe . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 DNA Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Apertured Probes for Near-Field Imaging of Dye-Doped Samples . . 26
6 Double-Tapered Fiber Probe and Spectroscopic Applications . . . . . . 30
6.1 Double-Tapered Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.2 Near-Field Photoluminescence Image
of Lateral p-n Junctions Obtained with the i-c Mode SNOM . 33
6.3 Near-Field Raman Spectroscopy of Polydiacetylene . . . . . . . . . . 34
7 Pure Silica Core Fiber Probes and Ultraviolet Applications . . . . . . . 35
7.1 UV Triple-Tapered Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7.2 UV Near-Field Photoluminescence Images of Polysilane . . . . . . 40
X Contents
7.3Fabrication of a Pure Silica Fiber Probe

by Pulling and Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A Novel Method for Forming Uniform Surface-Adsorbed

Metal Particles and Development of a Localized Surface-
Plasmon Resonance Sensor
H. Takei, M. Himmelhaus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1 A General Method for Preparing Surface-Bound Metal Particles . . . 57
1.1 Surface-Bound Metal Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.2 Use of a Monolayer of Monodisperse Dielectric Spheres
as a Template . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
1.3 Formation of the Monolayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
1.4 Physical Parameters Under Control . . . . . . . . . . . . . . . . . . . . . . . 62
1.5 Optical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.6 General Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2 Optical Biosensing Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.1 Significance of Molecular Interactions in Life Science . . . . . . . . 78
2.2 Measurement Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.3 General Characteristics of the Sensor . . . . . . . . . . . . . . . . . . . . . . 83
3 Curious Observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Near-Field Optical-Head Technology
for High-Density, Near-Field Optical Recording
T. Matsumoto . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2 Review of Near-Field Optical Recording . . . . . . . . . . . . . . . . . . . . . . . . 94
2.1 The Limit of Conventional Optical Recording . . . . . . . . . . . . . . 94
2.2 Near-Field Optical Recording Method . . . . . . . . . . . . . . . . . . . . . 95
2.3 Hybrid Recording Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3 Technical Issues Regarding the Near-Field Optical Head . . . . . . . . . . 96
3.1 Precise Control of the Spacing Between the Head
and the Recording Medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.2 Integration of Peripheral Components
into the Near-Field Optical Head . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.3 Obtaining High Efficiency
in Generating the Optical Near Field . . . . . . . . . . . . . . . . . . . . . . 99
4 Novel Design of a Near-Field Optical Head Using a Plasmon . . . . . . 103
4.1 Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.2 Simulation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Contents XI
4.3Use of an Aperture and a Circular Metallic Plate . . . . . . . . . . . 105

4.4Use of a Wedge-Shaped Metallic Plate Placed in Air . . . . . . . . 108
4.5Use of a Wedge-Shaped Metallic Plate
Placed near the Recording Medium . . . . . . . . . . . . . . . . . . . . . . . 116
4.6 Near-Field Optical Head Using Two Metallic Plates . . . . . . . . . 120
4.7 Fabrication of a Near-Field Optical Head
with a Wedge-Shaped Metallic Plate . . . . . . . . . . . . . . . . . . . . . . 122
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Nano-Optical Media for Ultrahigh-Density Storage

K. Naito, H. Hieda, T. Ishino, K. Tanaka, M. Sakurai, Y. Kamata,
S. Morita, A. Kikitsu, K. Asakawa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
2 Magnetic Patterned Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.1 Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.2 Magnetic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3 Organic-Dye-Patterned Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.1 Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.2 Electrical and Optical Measurements . . . . . . . . . . . . . . . . . . . . . . 138
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
A Phenomenological Description of Optical Near Fields

and Optical Properties of N Two-Level Systems Interacting
with Optical Near Fields
A. Shojiguchi, K. Kobayashi, S. Sangu, K. Kitahara, M. Ohtsu . . . . . . . 145
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
1.1 What are Optical Near Fields? . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
1.2 Theoretical Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
1.3 Difference Between Optical Near Fields
and Propagating Fields: Individual vs. Global Excitation . . . . 148
1.4 Two-Level System Interacting with Radiation Fields:
Dicke’s Superradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
1.5 Chapter Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
2 Model Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3 Dynamics of Bosonic Excitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.1 Boson Approximation and Diagonalization
of the Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.2 Dipole Dynamics Driven by Local Excitation . . . . . . . . . . . . . . . 155
4 Dynamics of Fermionic Excitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
4.1 Perturbative Expansion of Time-Evolution Operator . . . . . . . . 160
4.2 Numerical Results and Dynamical Properties . . . . . . . . . . . . . . . 162
XII Contents
4.3Dynamics of Dipole-Forbidden States

via Optical Near-Field Interaction . . . . . . . . . . . . . . . . . . . . . . . . 164
4.4 Semiclassical Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5 Effective Hamiltonian and the Dipole Ordering . . . . . . . . . . . . . . . . . . 168
5.1 Effective Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.2 Classification of Quasisteady States . . . . . . . . . . . . . . . . . . . . . . . 170
5.3 Response to the Initial Input of a Localized Photon:
the Robustness of Quasisteady States . . . . . . . . . . . . . . . . . . . . . 174
6 Dicke’s Superradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
6.1 Dicke States and Superradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
6.2 Dicke Master Equation and Solutions for a Small System . . . . 180
6.3 Effect of the Dipole–Dipole Interaction . . . . . . . . . . . . . . . . . . . . 188
6.4 Large-Sample Superradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
7 Radiation from the Dipole-Ordered States . . . . . . . . . . . . . . . . . . . . . . 195
7.1 Radiation Property of the Dipole-Ordered States . . . . . . . . . . . 195
8 Radiation from a Dissipative System . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
8.1 Semiclassical Description with the Effective Hamiltonian . . . . . 199
8.2 Quantum Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9 Dynamics of Localized Photons:
the Storage Mode of Localized Photons . . . . . . . . . . . . . . . . . . . . . . . . 204
9.1 The Transportation of Localized Photons:
the Storage and Through Flowing Modes . . . . . . . . . . . . . . . . . . 204
9.2 The Emergence of Dynamical Nonlinearity . . . . . . . . . . . . . . . . . 208
9.3 Two-Site Open System with Intermittent Chaotic Behavior . . 211
10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
List of Contributors
Koji Asakawa Yoshiyuki Kamata

Advanced Materials Storage Materials
and Devices Laboratory and Devices Laboratory
Corporate Research Corporate Research
and Development Center and Development Center
Toshiba Corporation Toshiba Corporation
1 Komukai-Toshiba-cho, Saiwai-ku 1 Komukai-Toshiba-cho, Saiwai-ku
Kawasaki 212-8582, Japan Kawasaki 212-8582, Japan
[email protected] [email protected]
Hiroyuki Hieda
Storage Materials Akira Kikitsu
and Devices Laboratory Storage Materials
Corporate Research and Devices Laboratory
and Development Center Corporate Research
Toshiba Corporation and Development Center
1 Komukai-Toshiba-cho, Saiwai-ku Toshiba Corporation
Kawasaki 212-8582, Japan 1 Komukai-Toshiba-cho, Saiwai-ku
[email protected] Kawasaki 212-8582, Japan
[email protected]
Michael Himmelhaus
Angewandte Physikalische Chemie
Universitaet Heidelberg Kazuo Kitahara
Im Neuenheimer Feld 253 Division of Natural Sciences
D-69120 Heidelberg Germany International Christian University
[email protected] Tokyo 181-8585, Japan
[email protected]
Takashi Ishino
Storage Materials
and Devices Laboratory Kiyoshi Kobayashi
Corporate Research ERATO Localized Photon Project
and Development Center Japan Science
Toshiba Corporation and Technology Corporation
1 Komukai-Toshiba-cho, Saiwai-ku 687-1 Tsuruma Machida
Kawasaki 212-8582, Japan Tokyo 194-0004, Japan
XIV List of Contributors
Takuya Matsumoto Masatoshi Sakurai

Storage Technology Research Center Storage Materials
Research and Development Group and Devices Laboratory
Hitachi, Ltd. Corporate Research
1-280 Higashi-Koigakubo, Kokubunji and Development Center
Tokyo 185-8601, Japan Toshiba Corporation
[email protected] 1 Komukai-Toshiba-cho, Saiwai-ku
Kawasaki 212-8582, Japan
Shuji Mononobe [email protected]
PRESTO, Japan Science
and Technology Agency
Suguru Sangu
Kanagawa Academy of Science
ERATO Localized Photon Project
and Technology
Japan Science
KSP East 408, 3-2-1 Sakado,
and Technology Corporation
Takatsu-ku
687-1 Tsuruma Machida
Tokyo 194-0004, Japan
[email protected]
[email protected]
Seiji Morita
Storage Materials Akira Shojiguchi
and Devices Laboratory ERATO Localized Photon Project
Corporate Research Japan Science
and Development Center and Technology Corporation
Toshiba Corporation 687-1 Tsuruma Machida
1 Komukai-Toshiba-cho, Saiwai-ku Tokyo 194-0004, Japan
Kawasaki 212-8582, Japan [email protected]
[email protected]
Hiroyuki Takei
Katsuyuki Naito
Mechanical Engineering Research
Storage Materials
Laboratory
and Devices Laboratory
Hitachi, Ltd.
Corporate Research
502 Kandatsu, Tsuchiura
and Development Center
Ibaraki 300-0013, Japan
Toshiba Corporation
[email protected]
1 Komukai-Toshiba-cho, Saiwai-ku
[email protected] Kuniyoshi Tanaka
Storage Materials
Motoichi Ohtsu and Devices Laboratory
Interdisciplinary Graduate School Corporate Research
of Science and Technology and Development Center
Tokyo Institute of Technology Toshiba Corporation
4259 Nagatsuta-cho, Midori-ku 1 Komukai-Toshiba-cho, Saiwai-ku
Yokohama 226-8502, Japan Kawasaki 212-8582, Japan
Near-Field Optical Fiber Probes
and the Imaging Applications
S. Mononobe
1 Introduction
Recently, scanning near-field optical microscopy (SNOM) employing an aper-
tured probe has been seen as a powerful tool for nano-optical applications
due to its high spatial resolution capability down to a few tens of nanometers
or less. The resolution is mainly determined by the aperture size of the probe
and the sample–probe distance because the SNOM is a scanning probe mi-
croscope based on the short-range electromagnetic interaction between the
probe and sample, which are much smaller than the optical wavelength. It
is apparent that fabricating such probes and controlling the sample–probe
distance have been the most important factors in the development of high-
resolution SNOM.
A near-field optical interaction system can be operated in two comple-
mentary ways called illumination mode and collection mode, where the probe
acts as a selective generator of a localized optical near-field and a sensitive
scatterer of an optical near-field generated over the sample, respectively. Fig-
ures 1a and b show schematic illustrations of i-mode SNOM and c-mode
SNOM, respectively. In the c-mode SNOM where the light is incident to
the total internal reflection, the sample–probe distance can be controlled by
using near-field optical (NFO) variation as a feedback control signal. For i-
mode SNOM, a shear-force feedback technique [1, 2] has been widely used to
regulate the sample–probe distance.
In order to realize such a scattering probe and a generating probe, forming
a dielectric taper and metallizing the taper has been used as an effective
method. Tapered probes have been fabricated by the following methods:
– Etching a quartz crystal rod [3]
– Heating and pulling a glass pipette by a micropipette puller [4]
– Meniscus etching of an optical fiber [5–13]
– Heating and pulling an optical fiber [14–18]
– Selective etching of an optical fiber [19–23]
– Heating and pulling and etching of an optical fiber [24, 25]
– Meniscus etching and selective etching of an optical fiber [26, 27]
For metallizing the tapered probes, a vacuum evaporation method [28] has
been applied, where the dielectric taper is coated with an aluminum or gold
2 S. Mononobe
Fig. 1. Schematic illustrations of (a) the i-mode, (b) the c-mode, and (c) the i-
c mode. In the c-mode SNOM, the light is incident to the total internal reflection.
The three-dimensional optical near field generated and localized on the sample
surface is scattered by the probe, and part of the scattered field is collected and
detected through the probe. The principle of operation of i-mode is similar except
that the probe works as a generator of the optical near field to illuminate the sample
surface. The scattered field by the sample is collected by conventional optics. In i-c
mode, a sample is excited by an optical near field on the probe. The light generated
on the sample is scattered and collected by the probe
film by a vacuum evaporation unit except for its apex region. Among the
metallized tapered probes, fiber-type probes have high transmission efficien-
cies due to their waveguide structure, and moreover, have been demonstrated
to have the molecular sensitivity, the nanometric spatial resolution, and the
locally spectroscopic capability in i-mode SNOM, c-mode SNOM, and hy-
brid SNOM called illumination-collection mode (i-c mode). Figure 1c shows
schematic illustration of i-c mode SNOM where the apertured probe func-
Near-Field Optical Fiber Probes and the Imaging Applications 3
tions as both an optical generator exciting the sample and a scatterer of the
excited optical near-fields.
i-mode SNOM and i-c mode SNOM are often employed for near-field
imaging of dye-doped samples and spectroscopic investigation of semicon-
ductor devices. In such spectroscopic applications where one must cope with
extremely low detected power, the probe should have high throughput in i-
mode SNOM, and be highly sensitive in c-mode SNOM. Furthermore, in the
i-mode fashion, to avoid thermal damage to the sample and the probe, the
probe should be used with as low an input power as possible. Therefore, the
resolution capability and throughput of the tapered probe have to be opti-
mized depending on SNOM applications. This optimization should be done
by varying the probe shape, i.e., the cone angle. Furthermore, to obtain a
highly resolved image, one must fabricate a metallized probe with an apex
region emerging from a metal film.
We have developed several application-oriented fiber probes called protru-
sion-type probe, double-tapered probe, pure silica fiber probe, and performed
near-field imaging of biomolecules and dye-doped samples, spectroscopic
study of semiconductors, and ultraviolet applications. In this chapter, the
near-field optical fiber probes and near-field microscopy applications are de-
scribed. Section 2 discusses the basic techniques for tapering and metallizing
fibers. The protrusion-type probe and the biological imaging are described in
Sect. 3. Section 4 describes apertured probes and their application to near-
field imaging of dye-doped samples. The double-tapered probe and pure silica
fiber probe are described in Sects. 5 and 6, respectively. Table 1 summarizes
the SNOM images appeared in this chapter.
2 Basic Techniques for Tapering

and Metallizing Optical Fibers
For tapering an optical fiber, three basic techniques, i.e., heating and pulling,
meniscus etching, and selective etching have been used. The characteristics
of these techniques are summarized in Table 2.
2.1 Heating-and-pulling and Metallization Techniques
In the heating-and-pulling technique [14], a silica-based optical fiber is heated

and pulled by a micropipette puller combined with a CO2 -gas laser as shown
in Fig. 2. One can fabricate a tapered fiber with an apex diameter of 50 nm
and a cone angle of 20–40◦ by a commercial micropipette puller. This tapering
can be applied to any optical fiber with a diameter of more than 125 µm
by adjusting the laser power, the strength of the pull, and the delay time
between the end of the heating and the beginning of the pulling. However,
it is difficult to control the cone angle while maintaining an apex diameter
4 S. Mononobe
Table 1. List of the SNOM images appeared in this chapter
Fig. Image Probe type Mode Feedback Ref.

15c Near-field image of Protrusion c-mode NFO intensity [29]
salmonella flagellar
filaments in air
15d Near-field image of Protrusion c-mode NFO intensity [30]
salmonella flagellar
filaments in water
18c Near-field image micro- Protrusion c-mode Shear-force [31]
tubules of a pig brain
20d–f Photoluminescence im- Protrusion i-mode Shear-force [32]
ages of GaAs quantum
dots
24b Near-field image of Ag-MgF2 -Al- c-mode NFO intensity [33]
deoxyribo-nucleic acid coated
26b, c Near-field images of neu- Apertured i-mode Shear-force [34]
rons and microtubules la-
beled with toluidine blue
27b Fluorescence image of Double-tapered i-c mode Shear-force [35]
Rhodamin 6G molecules
31b Photoluminescence imag- Double-tapered i-c mode Shear-force [36]
ing lateral p-n junction
based on GaAs
32b Raman spectroscopic Double-tapered i-mode Shear-force [37]
mapping of tabular
polydyacetylene single
crystal
40b UV photoluminescence Triple-tapered i-mode Shear-force [23]
image of polydihexyl- with a pure
silane silica core
41b UV photoluminescence Triple-tapered i-mode Shear-force [38]
image of n-decyl-(s)-2- with a pure
methylbutyl silane silica core
45 UV photoluminescence Pulled and i-mode Shear-force [39]
image of polydihexyl- etched with a
silane pure silica core
Table 2. Characteristics of the three basic techniques for tapering

an optical fiber
Technique Cone angle θ Apex diameter d Reproducibility
◦
Meniscus etching 9–40 60 nm or more 80% or less
Selective etching 14–180◦ 10 nm Almost 100%
Pulling 20–40◦ 50 nm Around 80%
Fig. 2. Schematic illustration of the micropipette puller used for tapering an optical
fiber with a diameter of 125 µm
Fig. 3. Schematic illustrations of (a) the vacuum evaporation unit for metallizing
the pulled fiber and (b) the metallized fiber
as small as 50 nm. In the tapered portion, strong optical leaky modes are
generated due to the varied core diameter.
For i-mode SNOM, the pulled fiber with an apex diameter of about 50 nm
must be metallized except for its apex region. To metallize the fiber, the
pulled fiber is rotated while evaporating aluminum in vacuum as shown in
Fig. 3a so that the metallized probe has a thickness profile as shown in
Fig. 3b. Here, the typical radial thickness is around 150 nm. The apex region
is aluminized with a thickness smaller than the half-radial thickness due to
the throwing of evaporated vapor. The metal thickness covering the apex
region can be reduced to a quarter of the radial thickness by inclining the
rotating fiber in vacuum.
6 S. Mononobe
2.2 Meniscus Etching
In the meniscus etching1 , a single-mode fiber is immersed in HF acid with a

surface layer of an organic solution such as silicone oil as shown in Fig. 4a. It
is tapered with a cone angle since the height of the meniscus formed around
the fiber is reduced depending on the fiber diameter (Fig. 4b). When the
fiber is completely tapered, the etching stops automatically (Fig. 4c). The
cone angle can be increased up to 35–40◦ . However, the obtained tapered
fiber has a geometrically eccentric apex with an elliptical cross section. The
longer and shorter principal diameters of this elliptical apex take values of
200 nm and 10–20 nm, respectively.
The vacuum evaporation method as shown in Fig. 3a can be applied to
metallize the meniscus-etched fibers. Prior to metallizing, one has to remove
organic dirt from the surface of the etched fibers by cleaning techniques, i.e.,
immersing in a H2 SO4 solution.
Fig. 4. Schematic illustrations of meniscus etching of a fiber at (a) the start, (b)
tapering, and (c) stop
2.3 Selective Etching
By applying the selective-etching method to a highly GeO2 -doped fiber, one

can obtain probe tips with a small apex diameter less than 10 nm. By varying
the concentration of etching solutions based on hydrogen fluoride (HF) and
1
The meniscus-etching technique was originally developed to fabricate a fiber-
optic microlens.
Fig. 5. (a) Cross-sectional profile of a refractive index of a silica fiber. Here, n1

and n2 are the refractive indices of the core and clad, respectively; r1 and r2 are
the radii of the core and clad, respectively. (b) Top, a geometrical model for the
tapering process; Here, τ is the etching time required for making the apex diameter
zero. θ is the cone angle of the tapered core. L is the length of the tapered core.
Bottom, cross-sectional profiles of the dissolution rates R1 and R2 of the core and
clad
ammonium fluoride (NH4 F), the cone angle can be controlled in a wide region
from 20◦ to 180◦ for an apex diameter less than 10 nm. Furthermore, selective
etching is the most highly reproducible technique among the three tapering
techniques. This method can be applied to any single-mode fiber produced
by vapor-phase axial deposition (VAD) [40].
Figure 5a shows a cross-sectional profile of the refractive index of a silica
fiber with a GeO2 -doped core and a pure silica clad. Here, n1 and n2 are the
refractive indexes of the core and clad, respectively. r1 and r2 are the radii
of the core and clad, respectively. On immersing the fiber in a buffered HF
solution (BHF) with a volume ratio of [40%–NH4 F aqueous sol.]: [50%–HF
acid]: [H2 O]= X:1:1 at 25◦ C, the core is hollowed at X = 0 and is tapered at
X = 10. Figure 5b shows a schematic explanation of the geometrical model
for the tapering process based on selective etching. Bright shading and dark
shading in the upper part represent the cross-sectional profiles of the fiber
before and after the etching with an etching time τ , respectively. τ is the
etching time required for making the apex diameter zero, θ and d are the cone
angle and apex diameter of the probe, respectively. The lower part shows the
dissolution rates R1 and R2 of the core and clad, respectively. Here, R1 < R2 .
Assuming that the dissolution rates R1 and R2 are constant within the core
and clad regions, respectively, relations between the cone angle θ, the length
L of the tapered core, and the apex diameter d are represented by
8 S. Mononobe
θ R1
sin = , (1)
2 R2
r1 − d/2
L= , (2)
tan(θ/2)
and

2r1 (1 − T /τ ) (T < τ ) ,
d(T ) = (3)
0 (T ≥τ ) .
Here, the etching time τ that is required for making the apex diameter zero
is expressed as

r1 R 1 + R 2
τ= . (4)
R1 R2 − R1
GeO2 -Doped Fiber
When a fiber with a GeO2 -doped core and a pure silica clad is immersed in
BHFs with volume ratios of X:1:1, the core region is hollowed in X < 1.7,
and is tapered in X > 1.7 as shown in Fig. 5. Based on (1), the cone angle
is determined by the dissolution rate ratio of the core and clad. Figure 6a
shows variations of the dissolution rates R1 and R2 and the cone angle θ
as a function of X. Here, the fiber used was produced by vapor-phase axial
deposition (VAD) so as to have an index difference of 2.5%. The dissolution
rate ratio R1 /R2 decreases with increasing NH4 F volume ratio of X, and
Fig. 6. (a) Dependencies of the dissolution rate R1 (broken curve) of the GeO2 -
doped core, R2 (dotted line) of the pure silica clad, and the cone angle θ (solid curve)
on X. Here, the value of ∆n defined as (n21 − n22 )/2n21 is 2.5%; (b) Dependency of
θ on ∆n at X = 10
approaches a constant value at X = 10–30. The cone angle that is determined

by the ratio R1 /R2 takes a minimum value of 20◦ at X = 10. When X is
fixed, the cone angle is determined by the index difference. Figure 6b shows
the dependence of the cone angle on the index difference at X = 10. The
index difference is increased by increasing the GeO2 doping ratio.
The VAD is an effective method to produce a GeO2 -doped fiber with an
index difference as large as 2.5–3.0%. A dispersion-compensating fiber (DCF)
with an index difference of ∆n = 2.5% and core diameter of 2 µm can be
applied to various selective etching methods of fabricating a protrusion-type
probe, a double-tapered probe, etc. This DCF was originally produced as a
device for controlling the optical dispersion of a 1500-nm optical transmission
system by VAD, and has a cutoff wavelength of around 0.8 µm.
By immersing the DCF in BHF with a volume ratio of 10:1:1 at 25◦ C, a
tapered fiber probe with a small cone angle of 20◦ and an apex diameter less
than 10 nm is fabricated with almost 100% reproducibility. Furthermore, the
cone angle can be controlled as 20◦ ≤ θ < 180◦ by varying the volume ratio
X of BHF, as shown in Fig. 6a. Such high controllability of the cone angle
is indispensable for tailoring a high-throughput probe and a high-resolution
probe.
Pure Silica Core Fiber with a Fluorine-Doped Clad
Pure silica core fiber has high transmittance in a wide region from near
ultraviolet to near infrared (see Fig. 33). To make an index difference between
the pure silica core and the clad, the clad is often made of fluorine-doped
silica. We produced a pure silica core fiber with the fluorine-doped clad and
an index difference −0.7%.2 By immersing the fiber in BHFs with volume
ratios 1.7:1:1 and 10:1:1, the obtained probe has cone angles of 62◦ and 77◦ ,
respectively. The fiber could not be hollowed at any volume ratios of BHFs.
In selective etching of a pure silica fiber, varying the index difference or the
fluorine-doping ratio controls the cone angle. Among single-mode fibers, only
1500-nm pure silica fiber (PSF) with an index difference of −0.3% and core
diameter of 10 µm is commercially available.
Controlling the Clad Diameter
For the shear-force feedback technique, the clad diameter of the probe is
one of the main parameters governing the resonance frequency of dithering a
probe. In our experience, tapered probes with a clad diameter of 20–30 µm
are usually needed to obtain a proper resonance frequency, i.e., 20–40 kHz.
Figure 7 schematically shows an etching method to fabricate tapered probes
2
For the following discussion, we define the relative refractive index difference ∆n
of doped glass to pure silica, which is expressed as (n22 −n21 )/2n22 and (n21 −n22 )/2n21
for a pure silica core fiber and a pure silica clad fiber, respectively.
10 S. Mononobe
Fig. 7. Etching method to fabricate a shoulder-shaped probe. r1 , r2 , radii of the

core and clad, respectively. D, reduced clad diameter. θ, cone angle of the tapered
core. L, length of the tapered core
with the reduced clad diameter. Here, r2 is the clad radius before etching, D
is the clad diameter of the probe, θ is the cone angle of the tapered core, d
is the apex diameter, and L is the length of the tapered core. This method
involves two steps: (A) reducing the clad thickness and (B) tapering the core.
In step A, the fiber is immersed in BHF with volume ratio of [NH4 F
aqueous solution (40%)]: [HF acid (50%)]: [H2 O]=1.7:1:1 for an etching time
TA . Then, the clad diameter is reduced to [2r2 − 2R2A TA ]. The core end is
kept flat since the dissolution rates R1A and R2A of the core and clad are
equal.
In step B, the fiber is selectively etched in XB :1:1 (where XB >1.7) for
etching time TB . The core is tapered with the cone angle θ represented by
θ R1B
sin = . (5)
2 R2B
Fig. 8. SEM micrographs of (a) a shoulder-shaped probe and (b) the magnified
apex region. D = 25 µm; θ = 20◦ ; d < 10 nm
To make the apex diameter zero, the etching time TB must be longer than
the time τ , which is given by

r1 R1B + R2B
τ= . (6)
R1B R2B − R1B
The clad diameter D is represented by
D = 2r2 − 2(R2A TA + R2B TB ) , (7)
which is proportional to the etching time TB .

Using a DCF with a core diameter of 2 µm, a clad diameter of 125 µm, and
an index difference of 2.5%, we obtained a tapered probe with a cone angle
of θ = 20◦ and a clad diameter of D = 25 µm. Figures 8a and 8b show SEM
micrographs of the probe and the magnified top region, respectively. The
conditions of the etching process are summarized in Table 3. By substituting
r1 = 1 µm, R1B = 1.1 µm h−1 , and R2B = 6.5 µm h−1 into (6), an estimated
value of τ = 64 min is obtained. For TB = 75 min (> τ ), the clad diameter
D is represented by D = −60TA + 106 [µm]. The cone angle is controlled by
varying the NH4 F volume ratio in step B. For the dependence of the cone
angle on the volume ratio, refer to Fig. 6a.
Table 3. Conditions for fabricating the probe as shown in Fig. 8a
Step Volume ratio of BHF Etching time (min) Dissolution rates (µm h−1 )
A 1.7:1:1 86 R1 = R2 = 30
B 10:1:1 75 R1 = 1.1, R2 = 6.5
12 S. Mononobe
3 Protrusion-type Probe and Its Imaging Applications

3.1 Protrusion-type Probe
The uncoated fiber probe with a nanometric apex can be employed for the
c-mode SNOM as shown in Fig. 1b. Although the principal factor governing
the resolution is the small apex size, the resolution is affected by propagat-
ing components and the scattering of low spatial frequency components of
the optical near field. To obtain highly resolved c-mode images, such un-
wanted scattered light must be suppressed. To perform high-resolution near-
field imaging, we developed a fiber probe with a protruding tip from a metal
film, which is called a protrusion-type probe [41]. Figure 9a shows the cross-
sectional schematic illustration of the protrusion-type probe. Here, θ, d, df ,
and tM represent the cone angle, the apex diameter, the foot diameter, and
the thickness of the metal film, respectively. This probe scatters frequency
components between (1/df ) and (1/d) and suppresses the generation of com-
ponents of less than (1/df ). To realize such a probe, a selective resin-coating
(SRC) method [41] was developed, which can be applied to tapered probes
fabricated by the etching method as shown in Fig. 7. Apertured probes are
also effective for suppressing unwanted light. However, it is difficult to pro-
duce apertured probes with an apex diameter of a few tens of nanometers by
a vacuum evaporation unit. If applying the vacuum evaporation method to
a fiber tip with a cone angle of 20◦ , an entirely coated fiber probe as shown
in Fig. 9b is obtained. The thickness tA of metal covering the apex could not
be made less than 40 nm for tM = 150 nm.
Fig. 9. (a) Schematic cross-sectional profiles of a protrusion-type probe and (b)

entirely coated probe. θ, cone angle of the taper; d, df , apex and foot diameters of
the protruding tip, respectively; tM , thickness of the metallic film; tA , thickness of
metal covering the apex
3.2 Fabricating Protrusion-type Probes

by Selective Resin-Coating Method
Figure 10 shows a schematic diagram of the SRC involving four steps: (A)
metal coating, (B) resin coating, (C) preferential etching of metal covering
the apex region, (D) removing the resin. Figure 11 shows a scanning electron
micrograph of the top region of the fabricated protrusion-type probe. The
probe has a protrusion from a gold film and a foot diameter df of less than
30 nm.
Before applying the SRC method, a tapered probe was fabricated with a
clad diameter D = 45 µm and a cone angle θ = 20◦ based on selective etching.
In step A, the probe was coated with 120-nm thick gold by a magnetron
sputtering unit. In step B, the probe is dipped in an acrylic resin solution
and then removed from the acrylic solution with a withdrawal speed of VD =
5 cm s−1 . Using an acrylic resin solution with a low viscosity coefficient of
11 cP, we succeed in making a fiber tip protruding from resin film. The density
of the acrylic solution is 0.85 g cm−3 at 25◦ C. In step C, the probe was etched
for 2 min in a solution KI-I2 -H2 O, mixed with a weight ratio of 20:1:400 and
diluted 50 times with deionized water. In step D, the acrylic film was removed
by immersing the probe in acetone.
Fig. 10. Schematic diagram of the selective resin-coating method for fabricating
a protrusion-type probe; D, clad diameter of the probe; r1 , core radius; θ, L, cone
angle and length of the tapered core, respectively; VD , withdrawal speed of the fiber
from the resin solution; LR , length of the tapered core where the resin is coated;
df , foot diameter of the protrusion
14 S. Mononobe
Fig. 11. Scanning electron micrograph of a protrusion-type probe fabricated by

the SRC method as shown in Fig. 10; θ = 20◦ ; d < 10 nm; df = 30 nm
Fig. 12. Dependencies of the foot diameter df on (a) the clad diameter D for a
withdrawal speed of 5 cm s−1 and (b) the withdrawal speed VD for a clad diameter
of 45 µm
Figures 12a and 12b show the dependencies of the foot diameter on the
clad diameter and the withdrawal speed, respectively. The foot diameter can
be controlled by varying the viscosity of the resin solution, withdrawal speed,
and the clad diameter.
3.3 c-mode SNOM Imaging of Salmonella Flagellar Filaments

in Air and Water
Using a protrusion-type probe fabricated by selective etching and selective

resin coating, Naya et al. [29, 30] succeeded in obtaining high-resolution c-
mode SNOM images of Salmonella flagellar filaments in both air and water.
Fig. 13. Schematic illustration of c-mode SNOM
Instrument of the c-mode SNOM
Figure 13 shows a schematic illustration of the optical system of c-mode

SNOM under near-field optical intensity feedback control. The sample sub-
strate is mounted on the xyz piezo-stage. The piezo parameters are chosen
depending on features such as the lateral and longitudinal variations of the
sample. Here, the piezo-stage has a total lateral span of 100 µm × 100 µm.
The z stage has a span of 5 µm. A right-angle prism or dove prism is used
for generating the optical near field on the sample surface. Here, care should
be taken to avoid unwanted scattering light generated from the prism surface
due to the dust and scratches. The probe is mounted on a xyz stage.
The light from a laser diode of 680 nm wavelength was used as the light
source. The light is incident on the prism and is totally internally reflected.
Then, the optical near field is generated on the sample surface. As shown
by the shaded region in Fig. 13, the optical near field is localized three-
dimensionally, and follows the spatial variations of the sample. When the
probe approaches closer to the sample surface, the optical near field is de-
tected through the probe by a highly sensitive detector such as a photomulti-
plier tube. By scanning the sample, information about the sample feature can
be extracted. A typical value of the detected power is of the order of 1 nW
to 100 pW. In this system, the signal over the background can be effectively
16 S. Mononobe
Fig. 14. Variation of the optical near-field intensity as a function of the sample–
probe distance, obtained with c-mode SNOM employing a protrusion-type probe
extracted by modulating the light and by amplifying the modulated signal

with a lock-in amplifier.
Figure 14 shows the variation of the optical near-field intensity as a func-
tion of the sample–probe distance, which is obtained with c-mode SNOM
employing a protrusion-type probe with a foot diameter of 30 nm as shown
in Fig. 15a. The optical near-field intensity sharply increases by decreasing
the sample–probe distance. This sharp variation of the near-field intensity can
be used as a feedback control signal to regulate the sample–probe distance.
Sample
The samples were fabricated by fixing the separated filaments from the body
on a hydrophilic glass substrate. A single filament has a typical length of a
few micrometers and a diameter of around 25 nm as shown in Fig. 15b. The
sample substrate was mounted on the prism of the optical system as Fig. 13.
Near-Field Optical Images of Salmonella Flagellar Filaments
Figures 15c shows a near-field image of Salmonella flagellar filaments, ob-

tained by the c-mode SNOM under near-field optical (NFO) intensity feed-
back control. Here, the incident light was s-polarized. The pixel size is
25 nm × 25 nm. The sample–probe distance was kept at less than 15 nm.
The full width at half-maximum (FWHM) of the region as indicated by ar-
rows is 50 nm. This value is comparable to the diameter obtained from the
TEM image (Fig. 15a), and indicates the high-resolution capability of SNOM
employing the protrusion-type probe.
Figure 15d shows the near-field optical image of Salmonella flagellar fil-
aments obtained in water. The pixel size is 10 nm × 10 nm. This near-field
Fig. 15. (a) Schematic illustration of the fabricated protrusion-type fiber probe
with cone angle of 20◦ , a foot diameter of 30 nm, and apex diameter less than 10 nm;
(b) A transmission electron micrograph of the Salmonella flagellar filaments. The
diameter of a flagellar filament, as indicated by arrows, are 25 nm; (c) A c-mode
SNOM image of Salmonella flagellar filaments. The arrows at the top right-hand side
represent the directions of wavevector k and electric field vector E; (d) Perspective
view of the near-field image of Salmonella flagellar filaments in water. The bright
region, as indicated by arrows, is 50 nm. The arrows at the top right-hand side are
defined as the directions of wavevector k and electric field vector E
optical imaging in water was also done by c-mode SNOM under NFO in-
tensity feedback control as in air. The filament sample was fixed on a glass
substrate and was fitted with an acrylic ring. The columnar bath with an
inner diameter of 10 mm and a height of 2 mm was filled with water during
the observation. The sample–probe distance was estimated to be less than
30 nm. The bright segments correspond to the filaments. The FWHM of the
18 S. Mononobe
Fig. 16. (a) constant-height mode and (b) constant-distance mode
bright region as indicated by arrows is 50 nm. The c-mode SNOM employ-

ing the probe with a protruding tip is effective in the near-field imaging of
biological sample in water.
Figures 15c and d were obtained with c-mode SNOM operated in the
constant-height mode as schematically shown in Fig. 16a. The monotonous
variation as shown in Fig. 14 was used to select the initial distance. To com-
pensate the tilt effects of the substrate, the time constant of the feedback loop
was carefully adjusted so that the probe followed only sample features larger
than 1 µm. When the c-mode SNOM was operated in the constant-distance
mode (Fig. 16b), such highly resolved images could not be obtained.
3.4 c-mode SNOM Images of Microtubules
Zvyagin et al. [31] isolated microtubules from the brain of a pig, and observed
with c-mode SNOM using a protrusion-type probe with a foot diameter of
30 nm. For this observation, a c-mode SNOM with s-polarized incident light
was used. To regulate the sample–probe distance, a shear-force feedback sys-
tem [1, 2] was added to c-mode SNOM. In the shear-force feedback system,
the differential force and frictional force are detected by dithering the probe
at its resonance frequency at some amplitude, and the amplitude of dithering
is measured by an optical interference technique. Figure 17 shows the varia-
tion of the amplitude of dithering of the protrusion-type probe as a function
of the sample–probe distance. The inset is the resonance frequency spectrum.
Figure 18a shows the transmission electron micrograph of microtubules.
The dark region seen in the middle of a strand corresponds to microtubules.
The width of the microtubules, as indicated by arrows, is 25 nm. Figure 18b
shows a shear-force topographic image of microtubules. The sample–probe
distance was kept at less than 5 nm during the scanning. Here, the long
bright structure and the round ones correspond to a microtubule and pro-
tein aggregates, respectively. From this high-resolution imaging result, the
Fig. 17. Variation of the dithering amplitude of the probe as a function of the
sample–probe distance. The inset is the resonance frequency spectrum of the dither-
ing probe
protrusion-type probe is very effective for high-resolution shear-force mi-

croscopy. Figure 18c shows the near-field optical image of the rectangular
region in Fig. 18b, which is obtained in the constant height mode without
using shear-force feedback. The sample–probe distance was about 30 nm.
3.5 Near-Field Spectroscopic Investigation of Semiconductor

Quantum Dots Under Extremely Low Temperature
Toda et al. [32] carried out a spectroscopic study of GaAs quantum dots
by low-temperature SNOM employing a protrusion-type probe. Figures 19a
shows schematic illustration of the microscope, the probe, and the GaAs
quantum dots grown on SiO2 -patterned GaAs substrates with metalorganic
chemical vapor deposition. Here, the probe has a foot diameter of 100 nm and
a cone angle of 40◦ . The length of the protruding tip is about 130 nm. The
quantum-dot pattern has a size of 190 nm ×160 nm ×12 nm and a separation
of 2 µm. The excitation light is delivered via the fiber probe with a length
greater than 2 m from the outside of the cryostat to the sample. The pho-
toluminescence spectrum as shown in Fig. 20a was obtained by positioning
the protruding tip 200 nm above the top of the quantum-dot at a low tem-
perature of 18 K. This corresponds to the spectrum from the carriers excited
in the whole quantum dots structure including the regions A–C. Figures 20b
and c show photoluminescence spectrums with tip position less than 10 nm
above the region of the quantum dots and that of SiO2 mask, respectively.
Figures 20d–f show monochromatic photoluminescence images obtained
in the constant distance-mode at 18 K with the energy regions indicated by
the arrows labeled A–C, which originated from the three regions of the GaAs
bulk, GaAs quantum dots, and GaAs quantum well, respectively. Figure 20g
is a shear-force topographic image obtained simultaneously. The protrusion-
type fiber probe with the long protruding length is convenient for scanning
such bumpy sample surfaces.
20 S. Mononobe
Fig. 18. (a) Transmission electron micrograph of a microtubule sample; (b) Shear-
force topographic image of the microtubule sample; (c) c-mode SNOM image in the
rectangular region in (b). The FWHM of the dark-line structure is approximately
40 nm
For this near-field imaging, other probe types were also used, which we
fabricated by pulling and etching the highly GeO2 -doped fiber. In the heating-
and-pulling process, the fiber was tapered with the highest pulling strength
by a commercial puller (Sutter, P-2000) so as to have a dip structure at the
taper end. By the etching in a BHF with a volume ratio of 10:1:1, the dip
was flattened, and then, the core is protruded from the pure silica clad. The
tapered fibers were aluminized by the vacuum evaporation method. Some of
the probes had throughputs higher than the protrusion-type probes. However,
we could not fabricate high-throughput probes with reproducibility because it
is difficult to control the sizes of the dip generated at the pulling. Figures 20a
Fig. 19. Schematic illustrations of i-mode SNOM developed for photoluminescence

spectroscopy of GaAs quantum dots at a low temperature of 18 K
Fig. 20. Photoluminescence spectra on maintaining the protruding tip (a) 200 nm
above the quantum dots, (b) less than 10 nm above the quantum dots, (d) less than
10 nm above the SiO2 mask; (c–f ) Monochromatic photoluminescence images at the
energy regions labeled A–C, respectively, and simultaneously obtained topographic
image
and b show scanning electron micrographs of a pulled-and-etched probe and

its magnified view. The dotted lines represent the cross section of the GeO2 -
doped core.
22 S. Mononobe
Fig. 21. Scanning electron micrographs of (a) a pulled and etched probe and (d)
its magnified view
3.6 Transmission Efficiencies of the Protrusion-type Probes
In near-field spectroscopic applications, one has to cope with near-field signals

that are too small. In i-mode SNOM, the transmission efficiency of the probe
is defined as the ratio of the output power to input power coupled into the
fiber probe. The low efficiency is caused by a large guiding loss, which is
generated along the tapered core due to optical interaction with the metal
clad. Therefore, the throughput is affected by the length of the metallized
tapered core or the cone angle. Figure 22 shows variations of the measured
throughput values of protrusion-type probes for the different cone angles of
20◦ and 50◦ as a function of the foot diameter of the protrusion. Here, the
Fig. 22. Dependencies of the transmission efficiency of two protrusion-type probes

with a cone angle of θ = 20◦ (closed circles) and θ = 50◦ (open triangles) on the
foot diameter
closed circles and open triangles correspond to the variations for the cone
angles of 20◦ and 50◦ , respectively. To collect the scattered light from the
apex of the probe, an objective lens with a numerical aperture of 0.4 was
inserted in front of the photodetector. At dF = 200 nm, The throughput of
the probe with a large cone angle of 50◦ is 10 times larger than that with
θ = 20◦ . Increasing the cone angle enhances the throughput. For a near-field
spectroscopic application requiring high throughput, the tapered probe with
a large cone angle, i.e., 50–90◦ should be fabricated by a selective-etching
method.
4 Metal-Dielectric-Metal-Coated Fiber Probe

and Near-Field Imaging of DNA Molecules
For imaging application of single biomolecules, i.e., a single DNA string with
a lateral width of 4 nm and a height of 2 nm, one should fabricate a near-field
probe with high-resolution capability beyond 10 nm. Recently, we fabricated a
specially designed probe with a metal-dielectric-metal coating for nanometer-
level-resolving SNOM, and succeeded in obtaining a c-mode SNOM image of
deoxyribonucleic acid (DNA) molecules [33].
4.1 Ag-MgF2 -Al-Coated Fiber Probe
Figure 23a shows a schematic image of a fabricated probe with metal-

dielectric-metal coating. The probe consists of two tapered structures with
the initial taper with a cone angle of 62◦ and a tapered apex region with
Fig. 23. (a) Ag-MgF2 -Al-coated fiber probe; (b) Variation of the optical near-field
intensity as a function of the sample–probe distance
24 S. Mononobe
a cone angle of 30◦ . We call this tapered fiber a pencil-shaped probe. The
method for fabricating the pencil-shaped probe with a metal-dielectric-metal
coating involves two steps: first sharpening of an optical fiber a into pencil-
shaped structure and then coating in a vacuum evaporation unit.
Fabrication of a Pencil-shaped Probe
The pencil-shaped probe with a cone angle of 30◦ and an apex diameter
less than 10 nm was fabricated with almost 100% reproducibility based on
selective etching of a multistep index fiber with the GeO2 -doped silica core,
pure silica clad, and a F-doped silica support. This fiber is specially designed
and fabricated by the VAD method to produce pencil-shaped probes and
triple-tapered probe. For details of the etching process, refer to the Appendix.
Ag-MgF2 -Al Coating
The pencil-shaped probe is coated in a vacuum evaporation unit. The coating

consists of three layers with an inner thin coating of silver (Ag), a middle
layer of magnesium fluoride (MgF2 ), and an outer aluminum coat. These
materials were chosen due to the following facts:
– The inner thin layer of Ag is mainly for enhancing the scattering efficiency
of the three-dimensionally localized near field around the sample. Ag has
a high reflectivity at the region of the wavelength used, and has a higher
dielectric constant than silica. Furthermore, based on calculated results
of Novotony et al. [42], the field confinement for a tapered probe with a
thin silver coating is highly concentrated in both the lateral and vertical
or z directions. The field confinement of the Ag-MgF2 -Al-coated probe in
the z direction could be inferred directly from Fig. 23b, which indicates
strong near-field intensity enhancement very close to the sample surface.
– The middle layer of MgF2 has a lower refractive index of 1.3 compared
with 1.5 of the sharpened core and forms an efficient clad. It not only
prevents the deterioration of the inner layer of silver on exposure to air
but also plays some role in suppressing the propagating light. Based on
experimental experience, adding MgF2 improves the scattering efficiency
of the probe and also the robustness of the probe due to its soft nature.
– The outer Al coating with a fairly large thickness of 200 nm in comparison
with the skin depth, which is a few tens of nanometers, is coated in order
to suppress background unwanted scattered light.
4.2 Sample Preparation
The double-stranded plasmid DNA with a ring structure (pUC18 and 2868
base pairs) was used as the sample. It was diluted with distilled water to a
final concentration of 5 ng/µL. A quantity of 2 µL of this was dropped onto
the center of a wet-treated sapphire substrate with an ultrasmooth surface.

Next, the wet-treated surface was blown dry with compressed air.
The sapphire substrate was produced by an epitaxial growth method [43],
which is stable both in air and in water for a considerable duration. The
surface has a staircase-like step structure whose step height and terrace width
can be carefully controlled. As the sapphire surface itself is hydrophobic, it is
necessary to make it hydrophilic for a stronger adsorption of organic DNA to
the surface. For this purpose, the sapphire surface was treated with a sodium
diphosphate solution with a concentration of 0.5 mol/L for a few minutes, and
then rinsed with distilled water. After being dried in air, the DNA solution
was dropped onto the substrate.
4.3 DNA Images
Figure 24a shows the image of DNA on the ultrasmooth sapphire surface
obtained using a noncontact-mode atomic force microscope [43]. The bright-
ring structures correspond to DNA molecules. The observed width of a single
strand, as indicated by open arrows, is estimated to be around 10 nm.
The experimental system of c-mode as appears in Fig. 13 was used for
near-field optical imaging of single DNA molecules. The DNA sample is
mounted on the prism, and the optical near field is sensitively picked up by
the Ag-MgF2 -Al-coated probe. Figure 24b shows the c-mode SNOM image,
obtained under near-field optical intensity feedback control. The pixel size is
5 nm × 5 nm. The arrows at the top left-hand side of the figure indicate the
propagation k and electric field E vectors of the incident light with a 680 nm
wavelength. The coiled loop and single strands of DNA are observed in this
figure. The FWHM of the intensity variation at the portion, as indicated by
the line, is around 20 nm. Figure 24c shows the magnified view of the single
strands of DNA. The pixel size is 2 nm × 2 nm. The FWHM, as indicated
by the arrows, is about 4 nm. This nanometric spatial resolution reflects the
special care taken during preparation of the sample and of the probe to avoid
unwanted scattered light.
We tried to obtain near-field optical images of DNA molecules by em-
ploying the protrusion-type probe with a foot diameter of 30 nm. However,
successful results could not be obtained. The metal-dielectric-metal-coated
probe tip with a layer of thin silver coating acts as a scattering probe much
more sensitive than a dielectric tip protruding from metal.
26 S. Mononobe
Fig. 24. (a) AFM image and (b, c) near-field optical images of a DNA sample,
obtained with c-mode SNOM employing the fiber probe in Fig. 23a
5 Apertured Probes for Near-Field Imaging

of Dye-Doped Samples
In investigating biological samples with i-mode SNOM, it is an effective
method of enhancing the optical contrast to attach the sample with absorb-
ing and fluorescing dye labels. For high-resolution imaging of a dye-labeled
sample, one has to fabricate the apertured probe with an aperture diameter
of a few tens of nanometers. Maheswari et al. produced a fiber probe with an
Fig. 25. Schematic illustration of an i-mode SNOM
aperture diameter of around 15 nm and obtained near-field images of dye-

labeled neurons of rats with an i-mode system under the shear-force feedback
control. Figure 25 shows the schematic illustration of the used i-mode sys-
tem. Here, the light from an argon ion laser with a 488 nm wavelength is
coupled into the fiber probe to generate a localized optical near field around
the apex region of the probe. The sample is mounted on the xyz piezo-stage.
On approaching the sample surface very close to the apex of the probe, the
optical near field generated around the probe is scattered and detected by the
sample fine structures. The scattered light is detected by a photomultiplier
tube through collection optics such as the objective lens with a high numer-
ical aperture. To regulate the sample–probe distance, a shear-force feedback
system is added consisting of the 685-nm laser diode, dithering probe, and
photodiode.
Figure 26a shows the schematic illustration of the fabricated probe. The
cone angle is 20◦ . To fabricate the probe, a dispersion-compensating fiber
was consecutively etched in three BHFs with volume ratios of 1.7:1:1, 10:1:1,
and 10:1:120, for 60, 75, 2 min, respectively, and then was metallized with
5-nm thick chromium and gold with a thickness of 120 nm by a vacuum
evaporation unit.
Figure 26b shows the i-mode near-field image of a neuron labeled with
toluidine blue. Here, the sample–probe distance is less than 50 nm. The wide
28 S. Mononobe
Fig. 26. (a) Cross-sectional illustration of the fabricated metallized fiber by the
vacuum evaporation method. Its top region enclosed by the dotted rectangular was
removed by rubbing with the substrate under shear-force feedback control before
the near-field imaging; (b) near-field optical image of neurons of rats. The magnified
view of (c) shear-force image and (d) i-mode SNOM images of the same area
dark arrows and the bright arrow correspond to the neural process and the
cell body, respectively. Before this imaging, the metal covering the top region
of the probe, enclosed by the dotted rectangular in Fig. 26a, was removed
by rubbing with the sample substrate under shear-force feedback control.
Figures 26c and 26d correspond to the simultaneous shear-force and near-
field images of the magnified region, as shown by the long dark arrow in the
neuron process in Fig. 26b. The fringe-like structures are seen in the near-field
image of Fig. 26d, however, they do not appear in the topographic image.
This shows that the fringe-like structures labeled with toluidine blue lie just
underneath the cell membrane on the surface. The structures are identified as
microtubules, which are the main constituent elements of the neural process.
The FWHM of near-field intensity variation at the portion, as indicated by
the line in Fig. 26d, is estimated to be 26 nm, which agrees with the typical
diameter of 25 nm of a single microtubule.
Recently, another apertured probe with a large cone angle, i.e., 50–90◦
has been chosen for performing i-mode and i-c mode imaging applications
including single-dye-molecule detection because the probe has a much higher
throughput than the tapered probe as shown in Fig. 26a. Hosaka et al. [35]
fabricated a double-tapered probe [44] with a cone angle of 90◦ and an aper-
ture diameter of a few tens of nanometers as shown in Fig. 27a, and succeeded
in obtaining near-field fluorescence imaging of dye molecules with a high reso-
lution of around 10 nm. Figure 27b shows the near-field fluorescence image of
single Rhodamin 6G molecules dispersed on the quartz glass at 700 ± 20 nm.
In this experiment, the dye was excited by a He-Ne laser with a 633 nm
wavelength. The bright spots correspond to emitting single dye molecules.
The estimated resolution is around 10 nm. For details of the double-tapered
probe see Sect. 6.
Fig. 27. (a) Schematic illustration of the double-tapered probe with a cone angle
of 90◦ ; (b) i-c mode near-field fluorescence image of single Rhodamin 6G molecules
dispersed on a quartz substrate at 700 ± 20 nm
30 S. Mononobe
6 Double-Tapered Fiber Probe

and Spectroscopic Applications
SNOM is a powerful tool for studying semiconductor devices with submicrom-
eter and subnanometer sized structures by imaging and spectroscopy. For
photoluminescence investigation of a semiconductor device, i-c mode SNOM
whose resolution is not affected by carrier diffusion effect should be chosen.
In i-c mode SNOM (Fig. 1c), the sample is effectively excited by the optical
near field generated around the probe in the i-mode fashion, and the near-
field photoluminescence of the sample is scattered by the probe in the c-mode
fashion. To perform imaging and spectroscopic applications by illumination-
collection hybrid mode SNOM, high throughput probes must be fabricated.
Saiki et al. [44] have developed a double-tapered probe with high through-
put, and have performed a spectroscopic study of semiconductor devices by
i-mode SNOM and i-c mode SNOM employing the probes. In this section,
double-tapered probe and spectroscopic applications are described.
6.1 Double-Tapered Probe
Figures 28a and b show the cross-sectional profiles of the protrusion-type

double-tapered probe and apertured-type double-tapered probe. Here, θ1 and
θ2 (< θ1 ) are the cone angles of the first and second taper, respectively. λ
is the cross-sectional diameter, which agrees with the optical wavelength in
the core. dB is the base diameter of the first taper, which is larger than
Fig. 28. Schematic cross-sectional profiles of double-tapered probes with (a) a

protruding tip from metal and with (b) an aperture. θ1 , θ2 , cone angle of the first
and second tapers, respectively; d, apex diameter; df , foot diameter in (a) and
aperture diameter in (b); dB , base diameter of the first taper; λ, optical wavelength
in the core; L, length of the lossy part between the cross-sectional diameters of λ
and df
the wavelength size λ, and d and df are the apex and foot diameters of the
protrusion, respectively.
As discussed in Sect. 3.6, the first cone angle of the double-tapered probe
has to be increased to enhance the throughput of the probe. Since the light
entering the tapered core is strongly attenuated by the metal film in the
subwavelength cross-sectional portion, by making the base diameter the value
of λ or more, the double-tapered probe with the first cone angle of 90◦ has a
transmission coefficient as high as the single-tapered probe with a cone angle
of 90◦ . Based on FDTD simulation [45], the double-tapered probe can have a
collection efficiency higher than the single-tapered probe. The double-tapered
structure is suitable for coupling the scattered light into the core.
Forming a Double-Tapered Tip
The method for fabricating the probe involves two steps: (A) controlling the
clad diameter and forming the first taper; (B) forming the second taper,
which are based on selective etching as described in Sect. 2.3. Figure 29a
shows the method schematically. In step A, by immersing the fiber in BHF
with XA :1:1 (where XA > 1.7), the core is protruded from the clad with the
cone angle θ1 represented by
θ1 R1A
sin = . (8)
2 R2A
In step B, immersing the fiber in BHF with 10:1:1 the core is tapered with
cone angles of θ1 and θ2 , where
θ2 R1B
sin = . (9)
2 R2B
The base diameter dB is represented by
TB
dB = 2r1 (1 − ) (TB < τ ) , (10)
τ
which is controlled by varying the etching time TC . Here, τ is defined by (4).
The clad diameter D is given by (7).
In actual conditions as shown in Table 4, a probe with a double-tapered
core with cone angles of θ1 = 90◦ and θ2 = 20◦ , and a base diameter of
dB = 500 nm was obtained. Figures 29b and c show SEM micrographs of
the tapered core region of a double-tapered probe and its magnified top
region.
32 S. Mononobe
Fig. 29. (a) Etching method for fabricating a double-tapered probe. The fiber is
consecutively etched in buffered HF with volume ratios of XA :1:1 (1.7 < XA < 10)
and 10:1:1 in steps A and B, respectively. Here, 2r1 and 2r2 are the core and
clad diameters, respectively. θ1 and θ2 are the first and second cone angles of the
double-tapered probe; SEM micrographs of (b) a double-tapered core and (c) the
magnified top region. θ1 = 90◦ ; θ2 = 20◦ ; dB = 500 nm
Table 4. Conditions for fabricating a probe as

in Fig. 29
Step Volume ratio of BHF Etching time (min)

A 1.8:1:1 70
B 10:1:1 40
6.2 Near-Field Photoluminescence Image

of Lateral p-n Junctions Obtained with the i-c Mode SNOM
With SNOM employing a double-tapered probe, Saiki et al. [36] investigated

a GaAs-based lateral p-n junction, and obtained its near-field photolumines-
cence image with i-c mode SNOM. Figures 30a and b show the schematic
illustrations of the double-tapered probe and experimental setup of the i-
c mode SNOM, respectively. Here, the probe was produced based on selective
resin coating. The foot diameter and cone angle of the double-tapered probe
is 200 nm and 50◦ , respectively. The GaAs sample is fixed on the inclined
substrate to avoid the bumpy sample surface with the edge of the metal-
lized clad on scanning. The sample is excited by laser light with 633 nm
through the double-tapered probe. The laser power coupled into the fiber is
0.5 mW. The sample–probe distance was controlled by the shear-force feed-
back technique to scan the sample. The sample is fabricated by growing a
silicon-doped GaAs layer on a semi-insulating GaAs (111)A substrate with
(311)A slopes [46]. The thickness of the silicon-doped layer is around 1 µm.
Figures 31a and b show the shear-force topographic image and near-field
photoluminescence image, which were simultaneously obtained. Because pho-
toexcited electrons and holes are separated and drifted by the internal elec-
tric field in the upper and lower junctions, the emission intensity is lower
than those of the other regions as seen in Fig. 31b. Furthermore, in the
lower junction region, some bright portions where the photoluminescence
intensity increased locally is observed. From the intensity variation in this
region, the spatial resolution is estimated to be 200 nm or less. Recently,
near-field photoluminescence images of single GaAs dots and a GaNAs layer
Fig. 30. Schematic illustrations of (a) a double-tapered probe and (b) a i-c mode
SNOM for investigating the lateral p-n junction
34 S. Mononobe
Fig. 31. (a) Shear-force topographic and (b) i-c mode SNOM images of the GaAs
lateral p-n junction
have been obtained with i-c mode SNOM employing a double-tapered probe
with an aperture diameter of 20–30 nm, and a resolution of 20–30 nm has
been achieved. For details of these studies, see Saiki [47].
6.3 Near-Field Raman Spectroscopy of Polydiacetylene

With i-mode SNOM employing a double-tapered probe with an aperture di-
ameter of 100 nm, Narita et al. [37] succeeded in obtaining two-dimensional
mapping of Raman-signal features of a tabular polydiacetylene single crystal.
The sample was excited by 532-nm Nd:YAG laser with a power of 10 nW.
Near-field Raman spectra were measured at 10 × 10 points in a portion with
an area of 1 µm × 1 µm. Figure 32a shows a near-field Raman spectra of poly-
diacetylene. Here, two Raman peaks L and S are observed at 1457 cm−1 and
1520 cm−1 , which relate to the C=C bond and reflect the difference in the
number of successive cis-bonds. Peaks L and S originate from a longer succes-
sive bond and shorter ones, respectively. Figure 32b shows a two-dimensional
Fig. 32. (a) Near-field Raman spectra of a tabular polydiacetylene single crystal;
(b) Two-dimensional mapping the relative near-field intensity at peak S to peak L
mapping of the relative Raman intensity of peak S at 1520 nm to peak L at

1457 nm.
7 Pure Silica Core Fiber Probes

and Ultraviolet Applications
Using GeO2 -doped fibers such as the dispersion-compensating fiber and the
double-clad fiber, one can fabricate near-field probes as appeared in Figs. 15a,
19a, 21a, 23a, 26a, 27a, and 30a and carry out SNOM applications in the
visible and infrared regions. However, it is difficult to employ such probes
for ultraviolet (UV) near-field optical microscopy applications because the
GeO2 -doped silica fiber has strong guiding loss based on Rayleigh scattering
and moreover, has UV absorption, luminescence, and Raman scattering that
originate from defects in the GeO2 . In order to overcome this difficulty and to
realize UV-SNOM, we have proposed fiber probes with the pure silica core.
Figure 33 shows optical transmission spectra of a GeO2 -doped fiber with an
index difference of 1% and a fiber with the pure silica core and the fluorine-
doped clad. It is found that the pure silica fiber probe has high transmittance
in a wide region from near UV to near infrared. To realize pure silica fiber
probes, we have developed methods to fabricate the following pure silica core
probes:
– Single-tapered probe with a large cone angle of 120◦
– Pulled and etched probe with the tapered core and clad [25]
– UV triple-tapered fiber probe [48]
36 S. Mononobe
Fig. 33. Optical transmission spectrums of a GeO2 -doped core fiber with a re-
fractive index difference of 1% (dotted curve) and a pure silica core fiber (solid
curve)
Fig. 34. Pure silica core probes with (a) the conical core protruded from the flat
clad end (b) tapered clad and core, and (c) triple-taper structure
as schematically shown in Fig. 34a–c, respectively. The single-tapered fiber

is fabricated by immersing a 1300-nm pure silica fiber (PSF) (Sumitomo,
PS1) with a core diameter of 10 µm in a BHF as described in Sect. 2.3.
Figure 35 shows a scanning electron micrograph of the single-tapered fiber
with the pure silica core and a cone angle of 120◦ . We have tried to develop
UV single-mode fibers with a cutoff wavelength of less than 400 nm in order
to fabricate UV single-tapered fiber. Details of these fibers will be discussed
elsewhere. Using the pulled and etched probe and triple-tapered probe, we
have obtained UV near-field photoluminescence images of polysilane and a
near-field Raman spectrum of a silicon crystal. In the following, triple-tapered
and pulled and etched probes are described.
Fig. 35. Scanning electron micrograph of a tapered fiber with a pure silica core,
obtained by immersing a 1300-nm fiber with a core diameter of 10 µm for 240 min
in a BHF with a volume ratio of 10:1:1. The cone angle is 120◦
7.1 UV Triple-Tapered Probe
To fabricate a UV triple-tapered fiber probe, we recently developed a UV

multistep index fiber [48] with the double core involving a subwavelength
core and a pure silica core. In the following text, this fiber is called a double-
core fiber. For triple-tapered probes with the GeO2 -doped core, refer to the
Appendix.
Fabrication of UV Triple-Tapered Probe

Based on Selective Etching of a Double-Core Fiber
The inset in Fig. 36 shows the index-difference profile of the double-core fiber.
The diameters of sections 1–5 are 100 nm, 2.9 µm, 40 µm, 42 µm, and 125 µm,
respectively. Sections 3–5 correspond to the clad region. Sections 3 and 5
are made of low fluorine-doped silica with an index difference of −0.35%.
Section 4 is made of the high fluorine-doped silica with an index difference of
−0.94%. The core region consists of sections 1 and 2, which are made of GeO2 -
doped silica and pure silica, respectively. The index difference of section 1 to
section 2 is estimated to be around 0.7%. The diameter of section 1 is only
100 nm. Since 99.9% of the cross-sectional area of the core region is occupied
by the pure silica, the double-core fiber has transmittance as high as that of
the pure silica fiber as shown in Fig. 33.
To fabricate this double-core fiber, we fabricated a preform glass rod by a
combination of vapor-phase axial deposition and plasma-activated chemical
vapor deposition and then drew the 125-µm diameter fiber by heating the
preform. Table 5 summarizes the parameters for drawing the fiber.
As shown in Fig. 36, the double-core fiber is tapered by a selective etching
method, which involves three steps: (A) hollowing of sections 1 and 4, (B)
38 S. Mononobe
Fig. 36. Etching process to fabricate a triple-tapered probe. The inset shows the
cross-sectional profile of the relative refractive index difference ∆n of a multistep
index fiber produced to fabricate the probe
tapering of sections 2 and 3, and (C) sharpening of section 1. Here, a BHF

with a volume ratio of NH4 F solution (40%): HF acid (50%): H2 O is denoted
as X:1:1. The values of X in steps A–C are XA = 0.6, XB = 1.7, and
XC = 10, respectively.
Figures 37a and b show scanning electron micrographs of the fabricated
triple-tapered probe and its top region, respectively. The probe has an apex
diameter of less than 10 nm. The values of the base diameters of dB1 and
dB2 are 100 nm and 2 µm, respectively. The first cone angle θ1 , the second
Fig. 37. Cross-sectional profile of the index difference ∆n of a multistep index fiber
produced to fabricate the probe; SEM micrographs of (a) a triple-tapered probe
and (b) its magnified apex region. θ1 = 60◦ ; θ2 = 120◦ ; θ3 = 60◦ ; d < 10 nm
Table 5. Parameters for drawing process to pro-

duce the double-core fiber
Drawing speed 150 m/s
Temperature 2106◦ C
Drawing tension 75 g
Clad diameter of the preform glass rod 25.5 mm
Clad diameter of the drawn fiber 125 µm
cone angle θ2 , and the third cone angle θ3 are 60◦ (±1◦ ), 120◦ (±1◦ ), and
60◦ (±1◦ ), respectively. Table 6 summarizes the parameters of the etching
process for fabricating the probe of Fig. 37a. After the etching, the triple-
tapered probe was coated with a 200-nm thick aluminum film by a vacuum
evaporation method.
40 S. Mononobe
Table 6. Parameters of etching process for fabricating

the triple-tapered probe
Step Volume ratio of BHF Etching time Temperature
A 0.6:1:1 40 min 25◦ C
B 1.7:1:1 20 min 25◦ C
C 10:1:1 10 min 25◦ C
7.2 UV Near-Field Photoluminescence Images of Polysilane
To demonstrate the UV triple-tapered probe, we performed UV near-field

photoluminescence imaging of polysilanes, which can be regarded as quan-
tum wires based on silicon, and are expected to be applied to UV-light emit-
ting devices. By changing the conformation of the silicon backbones, the
optical properties can be controlled. Figures 38a and b shows absorption
(dotted curves) and photoluminescence (solid curves) spectra of two polysi-
lanes named polydihexylsilane (PDHS) and n-decyl-(s)-2-methylbutyl silane
(Chiral-PS), respectively. They have different backbone structures called
transplanar and 7/3 helical, respectively. For the first UV photoluminescence
imaging, PDHS was chosen, which has absorption and photoluminescence
peaks at 370 nm and 380 nm, respectively. The quantum efficiency is eval-
uated to be 5%. PDHS was fixed on a glass substrate. Figure 39 shows a
schematic diagram of the i-mode UV-SNOM employing the triple-tapered
probe. Here, an argon ion laser with a 351 nm wavelength is coupled into
the fiber probe, and the UV optical near field generated around the probe
excites the sample. Photoluminescence of PDHS is collected by an objective
lens or silica ball lens. The triple-tapered probe has three cone angles of 60◦ ,
120◦ , and 60◦ as shown in Fig. 40a. Figure 40b shows a UV near-field photo-
Fig. 38. Absorption (dotted curves) and photoluminescence (solid curves) spectra
of polydihexylsilane (PDHS) and n-decyl-(s)-2-methylbutyl silane (Chiral-PS), re-
spectively. These polysilanes have different backbone structures called transplanar
and 7/3 helical, respectively
Fig. 39. Schematic diagram of the experimental setup for UV spectroscopic applica-
tions. Here, the photon-counting detection is performed at i-mode SNOM imaging.
The spectrum of the sample is obtained by using the CCD camera and monochro-
mator
Fig. 40. (a) Triple-tapered fiber probe with a pure silica core and (b) UV near-
fild photoluminescnece image of transplanar-type polydihexylsilane obtained with
i-mode SNOM and probe as shown in (a) and (b), respectively. The top region of
the probe, surrounded by the dotted rectangular in (b), was removed before the
imaging
42 S. Mononobe
Fig. 41. (a) Topographic image and (b) i-mode near-field photoluminescence image
of chiral-PS molecules at 325 nm
luminescence image of PDHS at 380 nm. The FWHM of the bright portion is
estimated to be around 100 nm. Such a near-field signal could not be detected
through double-tapered probes with the GeO2 -doped core due to background
originated from the luminescence of the core.
Next, by applying the above-described UV-SNOM technique to chiral-
PS, we succeeded in obtaining i-mode near-field photoluminescence images
of chiral-PS molecules dispersed on ultraflat sapphire substrate. Figure 41a
shows a topographic image of chiral-PS. The height of the observed fibril-like
structure is 7–50 nm. Since the diameter of Chiral-PS molecules has been esti-
mated to be around 1 nm, we guess the fibril-like structures are constructed
of several molecules. Figure 41b shows a UV near-field photoluminescence
image of chiral-PS molecules at 325 nm in the same area as Fig. 41a. For
the excitation of the sample, UV light with a wavelength of 300 nm was
used. The achieved spatial resolution is around 60 nm. The probe used was a
UV triple-tapered probe with a small core diameter of around 2 µm. It was
fabricated based on selective etching of a new double-core fiber, which was
produced by only vapor-phase axial deposition. Details of this fiber will be
discussed elsewhere.
Near-Field Raman Spectroscopy of Silicon Crystal
Near-field Raman spectroscopy of silicon is considered a suitable method

for analyzing silicon-based electronic devices. The Raman signal of silicon
has a rather weaker sattering efficiency and a smaller wave number than
that of polydiacetylene, and competes with the background signal of Raman
scattering from the GeO2 -doped silica core. To suppress the background, a
fiber probe with a pure silica core should be used. The background of the
probe is suppressed to less than 10% compared with that of a double-tapered
probe.
Fig. 42. Near-field Raman spectrum of a silicon crystal obtained by an aliuminized

fiber probe with a pure silica core and an aperture diameter of 100 nm
Figure 42 shows a near-field Raman spectrum of a silicon crystal, obtained

by i-mode SNOM employing the triple-tapered probe with an aperture diam-
eter of 100 nm. A sharp peak from silicon is clearly seen at 520 cm−1 . The
broad signal around 400 cm−1 corresponds to the Raman scattering from the
SiO2 core. It is found that the fiber probe is very effective in suppressing the
background originated from the GeO2 .
7.3 Fabrication of a Pure Silica Fiber Probe

by Pulling and Etching
Among commercial single-mode fibers, only 1.3-µm fiber (PSF) with a pure
silica core has a high transmittance at near UV. We developed a new method
of tapering a 1.3-µm PSF based on a combination of pulling and etching.
The method involves two steps: (A) heating-and-pulling the fiber by a mi-
cropipette puller and (B) etching the fiber in buffered hydrogen fluoride so-
lution as shown schematically in Fig. 43a. Figure 43b shows the magnified
top region of the tapered shape formed by step B. Here, θ and 2r1E are the
cone angle in the apex region of the fiber and the reduced core diameter at
the end of the tapered core, respectively. Figure 43c shows the cross-sectional
profile of the tapered probe obtained by increasing the etching time in step B.
Details of this profile are discussed later. The fabricated tapered fiber has the
tapered shape and the magnified top region as seen in the scanning electron
micrographs of Figs. 44a and b, respectively. Figures 44c and d correspond
to the magnified view of regions indicated by squares in b and c, respectively.
The cone angle and the apex diameter are θ = 65◦ and 10–20 nm, respec-
tively. By investigating 20 fiber samples, we obtained 80% reproducibility for
a cone angle θ = 60◦ (±5◦ ) and an apex diameter of 10–20 nm. We now
describe the method for producing this pure silica core fiber.
44 S. Mononobe
Fig. 43. (a) Method for fabricating a tapered fiber with the pure silica core. (b)
Cross-sectional profile of the top of the fiber formed by step A in (a). r1E , reduced
core diameter at the end of the tapered core; θB1 , θB2 , cone angles of the core and
clad, respectively. (c) Cross-sectional profile obtained by increasing the etching time
in step B
Table 7. Parameters of the puller for a pure silica core fiber
Heat H, CO2 laser power 350

Filament F , length of the fiber scanned with the laser beam 0
Velocity V, velocity of the puller bar at the end of the heating time 1
Delay D, delay time between the heating and pulling 130
Pull P, strength of pull 150
In step A, a PSF with core diameter of 10 µm, a clad diameter of 125 µm,
and an index difference of −0.3% was heated and pulled by a micropipette
puller (Sutter Instrument, P-2000) combined with a CO2 laser. This puller
has been mechanically adjusted to fabricate a micropipette with a diameter of
1 mm. To pull the fiber, the two puller bars were exchanged to optional ones
on which a 125-µm bare fiber can be attached. In our case, to produce a probe
with a bare-portion length as small as 1 cm, a plastic-coated portion of the
fiber sample was carefully attached on one bar, and a bare portion was fixed
to another bar. The puller is adjusted with its parameters shown in Table 7.
Here, the heat parameter H (0≤H≤999) fixes the CO2 laser power. The
filament parameter F is the length of the fiber that is scanned with the CO2
laser beam. The velocity parameter V (1≤V≤255) shows the velocity of the
puller bar at the end of the heating time. The delay parameter D (0≤D≤255)
represents the delay time between the end of the heating and the beginning
of the pulling (in milliseconds). The puller is mechanically adjusted to make
the delay zero at D = 125. The pull parameter P decides the pull strength
and is controlled in a region of 0≤P≤255. If step A is repeated using the
same puller and the same fiber, some of the parameters, such as H and D,
may be changed to fabricate a pulled fiber as seen in Fig. 44a.
In step B, the fiber was etched by immersing for 10 min in a buffered
hydrogen fluoride solution (BHF) with a volume ratio of 40%-NH4 F aqueous
solution : 50%-HF acid : deionized water = 1.7:1:1. The temperature of the
BHF was 25◦ C (±0.1◦ C). The fluorine-doped clad and the pure silica core
have dissolution rates of R1 = 6.6 µm h−1 and R2 = 7.6 µm h−1 , respectively.
The pulled and etched probe with a few tens of nanometers was obtained by
stopping the etching before the core protruded from the fluoride-doped clad.
The cone angle θ is increased by increasing the etching time T . In our exper-
iments, the cone angles were 35◦ and 65◦ for etching times of 5 and 10 min,
respectively. However, we could not realize the pulled and etched probe with
a reproducibility of more than 80% due to the mechanical misalignment of
the puller.
On the other hand, once the core is exposed from the fluorine-doped silica,
the core is selectively etched due to the difference of the dissolution rate R1
of the core and the dissolution rate R2 of the clad (> R1 ). Then, the fiber
has two cone angles of θB1 and θB2 , as shown in Fig. 43c, where dB is the
base diameter of the conical core, with the angle θB1 given by
θB1 R1 θB2
sin = sin . (11)
2 R2 2
The pulled and etched probe as shown in Fig. 21a was fabricated by selective
etching and has a GeO2 tip protruding from the silica clad. In the case of
using a PSF, the cross section of the apex region was elliptical, with a longer
principal diameter of around 200 nm. Therefore, we could not employ the
selectively etched PSF as a probe. The value of dB is equal to the reduced
core diameter 2r1E of the fiber as shown in Fig. 43a. The reduced core diam-
eter is estimated to be 2 µm from a SEM image of selectively etched fibers
that were immersed in BHF with 1.7:1:1 for times longer than 10 min. It is
straightforward to obtain a minimum ratio of r1E /r1 = 1/5.
46 S. Mononobe
Fig. 44. Scanning electron micrographs of (a) the pulled and etched probe with
the pure silica core and (b) its magnified top region; (c, d) The magnified apex
regions as indicated by the squares in (b) and (c), respectively. The cone angle and
apex diameter are 65◦ and 20 nm, respectively
UV Application Performed with i-mode SNOM

Employing a Pulled- and Etched-Probe
The pure silica tapered probes were aluminized by a vacuum evaporation

method, and applied to a UV near-field photoluminescence image of polydi-
hexylsilane. Figures 45a and b show a shear-force topographic image and
near-field photoluminescence image at 380 nm. In Fig. 45b, the value of
Fig. 45. (a) Schematic illustration of the probe fabricated by pulling and etching
a commercial fiber with the pure silica core; (b) Shear-force topographic image and
(c) i-mode near-field photoluminescence images of polydihexylsilane aggregates at
380 nm
FWHM of the bright portion as indicated by the arrow is estimated to be

around 100 nm. The background signal in the dark region is almost zero,
in contrast to the maximum counting rate of the photoluminescence signal.
Since the pulled and ethced fiber can be fabricated by pulling and etching
commercial fiber, the probe has been widely used as a low-noise probe for
near-field fluorescent imaging in the visible region [49] and UV nanofabrica-
tion [50] based on photoenhanced chemical vapor deposition.
48 S. Mononobe
8 Outlook
The resolution capability and throughput of SNOM have been greatly im-
proved by employing application-oriented probes such as a double-tapered
probe with an aperture diameter of a few tens of nanometers and a pure
silica fiber probe, and various imaging and spectroscopic applications have
been successfully carried out. However, the techniques for mass producing
these probes is not yet established due to the low reproducibility of the met-
allizing process for forming a nanometric aperture. Furthermore, single-mode
fiber probes should be newly developed for some imaging applications such
as polarization measurements in the ultraviolet and visible regions. If these
techniques are established, SNOM and related techniques will be further used
widely by many researchers and engineers. To realize the mass production of
near-field optical probes, the author has concentrated on the development of
a novel metallization technique based on electroless plating [51] and on the
fabrication of new single-mode fibers.
Acknowledgments
The author wishes to thank M. Ohtsu, H. Ito, R. Uma Maheswari, T. Saiki,

M. Naya, Y. Toda, and N. Hosaka for their assistance and fruitful discussions.
Appendix
Fabrication of Tapered Fibers Based on Hybrid Selective Etching
of a Double-Clad Fiber
The double-clad fiber [22] was developed for fabricating pencil-shaped probes
and triple-tapered probes. These probes can be fabricated with almost 100%
reproducibility based on selective etching. In this section, we describe the
structure of a triple-tapered probe and design/fabrication of the fiber.
Triple-Tapered Probe
Figures 46a and b show cross-sectional profiles of the single-tapered probe

and triple-tapered probe. In Fig. 46a, θ is the cone angle and λ is the optical
wavelength in the fiber. In Fig. 46b, L is defined as the length of the portion
with a cross-sectional diameter of λ to the apex. L1 is the length of the first
taper, θ1 , θ2 , and θ3 are the cone angles of the first, second, and third tapers,
respectively, and dB1 and dB2 are the base diameters of the first and second
tapers, respectively. As described in Sect. 3.6, the light entering the single-
tapered probe is strongly attenuated by metal from the portion with diameter
λ to the apex. To reduce the attenuation, the length L must be decreased by
Fig. 46. (a) Cross-sectional profiles of the single-tapered probe and (b) the triple-
tapered probe; θ, cone angle of the tapered core; λ, optical wavelength; L, length
of the tapered core with the foot diameter equal to λ; tS , skin depth of the metal;
θ1 , θ2 , θ3 , cone angles of the first, second, and third tapers, respectively; dB1 , dB2 ,
base diameters of the first and second tapers, respectively; L1 , length of the first
taper
increasing the cone angle. In the case of fabricating the tapered probe with
a large cone angle, i.e., 120◦ , one must accept the limited resolution affected
by optical leaking from the metal around the apex region of the probe. In
the triple-tapered probe, we can decrease the length L by increasing θ2 in
order to enhance the throughput. Thus, the resolution capability is increased
by simultaneously decreasing the cone angle and decreasing the first taper
length L1 to a few hundreds of nanometers, which corresponds to several
times the skin depth of metal.
Selective Etching of a Double-Clad Fiber
Figure 47a shows the cross-sectional profile of the index difference of the
double-clad fiber. Here, the fiber involves three sections: (1) a GeO2 -doped
silica core, (2) a pure silica clad, and (3) a fluorine-doped silica support. The
index differences of sections 1 and 3 with respect to section 2 are 1.2% and
−0.7%, respectively. The radii of sections 1–3 are r1 = 0.65 µm, r2 = 13.5 µm,
and r3 = 62.5 µm, respectively. Figure 47b shows a schematic diagram of the
etching process consisting of three steps A–C. By defining the dissolution
rate of section i in step j as Rij (where i = 1, 2, 3; j=A, B, C), the relative
dissolution rates Rij /R3j is represented by Fig. 47c. As described in Sect. 2.3,
the relative dissolution rate of GeO2 -doped glass to pure silica glass depends
on the concentration of buffered HF mixed with a volume ratio of 40%NH4 F
sol. : 50%HF acid: H2 O = X:1:Y . We now use 1.7:1:1, 10:1:1, and 1.7:1:5, in
50 S. Mononobe
Fig. 47. (a) Cross-sectional profile of the index difference ∆n. r1 , radius of the
GeO2 -doped silica core; r2 , radius of the pure silica clad; r3 , radius of a fluorine-
doped silica support; (b) Etching method for fabricating application-oriented
probes; (c) Cross-sectional profile of the relative dissolution rate
which the dissolution rates of sections 1 and 2 are experimentally found to

satisfy the relation
R1B R1A R1C
= 0.29 < = 1.0 < = 1.48 . (12)
R2B R2A R2C
On the other hand, the relative dissolution rate of the fluorine-doped section
to the pure silica section is approximately fixed for the concentration of BHF.
So,
R2j /R3j = 0.51 (j = A, B, C) . (13)
We now discuss the etching process using 1.7:1:1, 10:1:1, and 1.7:1:5 in
steps A–C, respectively.
In step A, the fiber is tapered to an angle of θA represented by
sin(θA /2) = R2j /R3j (where j = A, B, C) . (14)
If the fiber diameter is equal to [2r2 ] after step A, the etching time TA is
represented by
TA = (r3 − r2 )/R3A . (15)
The tapered fiber will have an apex diameter smaller than 2r1 if TA ≥τA ,
where τA is the time required to make an apex diameter of 2r1 and is

r2 − r1 R2A + R3A
τA = . (16)
R3A R3A − R2A
Thus, it is straightforward to find that the radius r2 of section 2 must be
smaller than the critical radius r2P expressed as

r3 + ξ·r1 R2A /R3A + 1
r2P = , where ξ = . (17)
1+ξ 1 − R2A /R3A
In step B, section 1 is sharpened with a different angle θB , given by
sin(θB /2) = R1B /R3B . (18)
We obtain pencil-shaped probes with zero apex diameter and cone angles θA
and θB when the etching time TB is larger than τB , expressed as

r1 R1B + R3B
τB = . (19)
R1B R3B − R1B
Furthermore, to obtain a triple-tapered probe, we perform step C, when the
largest cone angle θC2 is given by
θC2 R1C
sin = (where R1C > R2C ) . (20)
2 R3C
The cone angle θC1 is increased from θB by increasing the etching time TC ,
and is equal to θC2 at TC > τC , where τC is given by

r1 R1C + R3C
τC = . (21)
R1C R3C − R1C
Therefore, we can obtain a triple-tapered probe and a pencil-shaped probe
with a cone angle θB = θC2 at 0 < TC < τC and TC > τC , respectively.
To obtain a cone angle θA as small as 62◦ , the index difference of section 3
is estimated from (15) and (19) to be as high as 0.7%. To obtain a cutoff
wavelength of around 400 nm, section 1 is tailored with an index difference
of 1.2% and a core radius of r1 = 0.65 µm. Then we obtain an estimated
value of θA = 17◦ from (13) and (19). Furthermore, when the clad radius r3
is a standard value of 62.5 µm, we obtain a critical radius of r2P = 23 µm
from (14) and (18). We make the radius r2 have a value of 13.5 µm, which is
smaller than the critical radius.
To realize the fiber as shown in Fig. 47a, we fabricated a preform glass
rod by vapor-phase axial deposition and drew the fiber. To suppress the
52 S. Mononobe
Fig. 48. (a–c) SEM micrographs of a pencil-shaped probe, its magnified tapered
core, and its magnified apex region. θA = 62◦ ; θB = 17◦ ; d < 10 nm; (d) SEM
micrograph of a triple-tapered probe. θA = 62◦ ; θC1 = 50◦ ; θC2 = 85◦ . (e) SEM
micrograph of a pencil-shaped probe. θA = 62◦ ; θB = 85◦
diffusion of GeO2 and fluorine, the drawing tension should be as high as

possible. However, in the case of drawing the fiber with a high tension of
60 g, we could not reproducibly cleave the fiber to obtain a flat facet with
a commercial fiber cleaver. We consider that the low reproducibility can be
attributed to the remaining stress between sections 2 and 3. To suppress this
remained stress, we kept a low tension of less than 28 g during the drawing.
To demonstrate the tailoring capability of the different types of probes, we
performed the etching process using the fabricated fiber. We prepared 30 fiber
samples with flat ends. The fibers were etched consecutively for TA = 40 min
in a 1.7:1:1 solution and for TB = 20 min in a 10:1:1 solution. We obtained a
pencil-shaped probe with a small cone angle for high resolution. Figures 48a–
c show SEM micrographs of the probe, the magnified tapered core, and the
Fig. 49. (a) Dependencies of the taper angle θB on the volume ratio XB of NH4 F
aqueous solution in the etching solution; (b) Dependency of the taper angle θC2 on
the volume ratio of H2 O in the etching solution
magnified apex region, respectively. Here, the apex diameter is less than
10 nm. The cone angles are θA = 62◦ and θB = 17◦ , which agree with the
estimated values from (15) and (19).
Furthermore, by etching the pencil-shaped probe in a 1.7:1:5 solution
for TC = 2 min, we obtained a triple-tapered probe with a high-resolution
capability and high throughput. Figure 48d shows a SEM micrograph of
the triple-tapered probe. The probe has three cone angles of θC1 = 50◦ ,
θC2 = 85◦ , and θA = 62◦ and an apex diameter of less than 10 nm. The two
base diameters of tapers with θC1 = 50◦ and θC2 = 85◦ are dB1 = 250 nm
and dB2 = 1.28 µm, respectively. At TC = 2.75 min, we obtained another
type of pencil-shaped probe with a large angle of 85◦ near the apex region as
shown in Fig. 48e. Figures 49a and b show the dependencies of θB and θC2 ,
respectively, on the volume ratios of BHF. Here, 1.7:1:1, XB :1:1, and 1.7:1:YC
are used in steps A–C. The cone angles θB and θC2 are controlled by varying
XB and YC , respectively.
References
1. E. Betzig, P.L. Finn, J. S. Weiner: Appl. Phys. Lett. 60, 2484 (1992)
2. R. Toledo-Crow, P.C. Yang, Y. Chen, Vaez-Iravani: Appl. Phys. Lett. 60, 2957
(1992)
3. D.W. Pohl, W. Denk, M. Lanz: Appl. Phys. Lett. 44, 651 (1984)
4. E. Betzig, M. Isaacson, A. Lewis: Appl. Phys. Lett. 51, 2088 (1987)
5. D.R. Turner: US Patent, 4,469,554 (1983)
6. D. Courjon, K. Sarayeddine, M. Spajer: Opt. Commun. 71 23 (1989)
7. R.C. Reddick, R.J. Warmack, T L. Ferrel: Phys. Rev. B 39 767 (1989)
8. K.M. Takahashi: J. Colloid Interface Sci. 134, 181 (1990)
54 S. Mononobe
9. T. Hartmann, R. Gatz, W. Wiegräbe, A. Kramer, A. Hillebrand, K. Lieberman,

W. Baumeister, R. Guckenberger: ‘A Scanning Near-Field Optical Microscope
(SNOM) for Biological Applications’, In: Near Field Optics, 242, NATO ASI
series E, (Kluwer, Dordrecht 1993) 35
10. P. Hoffmann, B. Dutoit, R.-P. Salathé: Ultramicroscopy 61, 165 (1995)
11. S. Mononobe, R. Uma Maheswari, M. Ohtsu: Opt. Express 1, 229 (1997),
http://epubs.osa.org/opticsexpress/
12. H. Muramatsu, N. Chiba, N. Yamamoto, K. Honma, T. Ataka, M. Shigeno,
H. Monobe, M. Fujihira: Ultramicroscopy 71, 73 (1998)
13. P. Lambelet, A. Sayah, M. Pfeffer, C. Philipona, F. Marquis-Weible: Appl. Opt.
37, 7289 (1998)
14. E. Betzig: ‘Principles and Application of Near-Field Scanning Optical Mi-
croscopy (NSOM)’, In: Near Field Optics, 242, NATO ASI series E (Kluwer,
Dordrecht 1993) 7
15. E. Betzig, J.K. Trautman: Science 257, 189 (1992)
16. B.I. Yakobson, A. LaRosa, H.D. Hallen, M.A. Paesler: Ultramicroscopy 61, 179
(1995)
17. G.A. Valaskovic, M. Holton, G.H. Morrison: Appl. Opt. 34, 1215 (1995)
18. M. Garcia-Parajo, T. Tate, Y. Chen: Ultramicroscopy 61, 155 (1995)
19. S. Jiang, H. Ohsawa, K. Yamada, T. Pangaribuan, M. Ohtsu, K. Imai, A. Ikai:
Jpn. J. Appl. Phys. 31, 2282 (1992)
20. R. Uma Maheswari, S. Mononobe, M. Ohtsu: J. Lightwave Technol. 13, 2308
(1995)
21. P. Tomanek: ‘Fiber Tips for Reflection Scanning Near-Field Optical Mi-
croscopy’, In: Near Field Optics, 242, NATO ASI series E, (Kluwer, Dordrecht,
1993) 295
22. S. Mononobe, M. Ohtsu: IEEE Photonics Technol. Lett. 10, 99 (1998)
23. S. Mononobe, T. Saiki, T. Suzuki, S. Koshihara, M. Ohtsu: Opt. Commun. 146,
45 (1998)
24. M. Spajor, A. Jalocha: ’The Reflection Near Field Optical Microscope: An Al-
ternative to STOM’, In: Near Field Optics, 242, NATO ASI series E, (Kluwer,
Dordrecht 1993) 87
25. S. Mononobe: ‘Probe fabrication’, In: Near-Field Nano/Atom Optics and Tech-
nology (Springer, Berlin 1998) 31
26. S. Mononobe, M. Ohtsu: J. Lightwave Technol. 14, 2231 (1996); Erratum
J. Lightwave Technol. 15, 162 (1997)
27. S. Mononobe, M. Ohtsu: J. Lightwave Technol. 15, 1051 (1997)
28. E. Betzig, J.K. Trautman, T.D. Harris, J.S. Weiner, Kostelak: Science 251,
1468 (1991)
29. M. Naya, S. Mononobe, R. Uma Maheswari, T. Saiki, M. Ohtsu: Opt. Com-
mun. 124, 9 (1996)
30. M. Naya, R. Micheletto, S. Mononobe, R. Uma Maheswari, M. Ohtsu:
Appl. Opt. 36, 1681 (1997)
31. A.V. Zvyagin, J.D. White, M. Ohtsu: Opt. Lett. 22, 955 (1997)
32. Y. Toda, M. Kourogi, M. Ohtsu, Y. Nagamune, Y. Arakawa:
Appl. Phys. Lett. 69, 827 (1996)
33. R. Uma Maheswari, S. Mononobe, K. Yoshida, M. Yoshimoto, M. Ohtsu:
Jpn. J. Appl. Phys. 38, 6713 (1999)
34. R. Uma Maheswari, H. Tatsumi, Y. Katayama, M. Ohtsu: Opt. Commun. 120,
325 (1995)
35. N. Hosaka, T. Saiki: J. Microscopy 202, Part 2, 362 (2001).

36. T. Saiki, S. Mononobe, M. Ohtsu, N. Saito, J. Kusano: Appl. Phys. Lett. 67,
2191 (1995)
37. Y. Narita, T. Tadokoro, T. Ikeda, T. Saiki, S. Mononobe, M. Ohtsu: Appl. Spec-
tros. 52, 1141 (1998)
38. M. Arai, S. Koshihara, M. Ueda, M. Yoshimoto, T. Saiki, S. Mononobe,
M. Ohtsu, T. Miyazawa, M. Kira: J. Lumines. 87–89, 951 (2000)
39. T. Saiki: ‘Diagnosing Semiconductor Nano-Materials and Devices’, In: Near-
Field Nano/Atom Optics and Technology (Springer, Berlin 1998) 153
40. T. Izawa, S. Sudo: Optical Fibers: Materials and Fabrication (KTK Scientific,
Tokyo, 1987)
41. S. Mononobe, M. Naya, T. Saiki, M. Ohtsu: Appl. Opt. 36, 1496 (1997)
42. L. Novotony, D. W. Pohl, B. Hecht: Ultramicroscopy 61, 1 (1995)
43. K. Yoshida, M. Yoshimoto, K. Sasaki, T. Ohnishi, T. Ushiki, J. Hitomi, S. Ya-
mamoto, M. Sigeno: Biophy. J. 74, 1654 (1998)
2612 (1996)
45. H. Nakamura, T. Sato, H. Kanbe, K. Sawada, T. Saiki: J. Microscopy 202, 50
(2001)
46. N. Saito, M. Yamaga, F. Sato, I. Fujimoto, M. Inai, T. Yamamoto, T. Watan-
abe: Inst. Phys. Conf. Ser. 136, 601 (1993)
47. T. Saiki: ‘Nano-Optical Imaging and Spectroscopy of Single Quantum Con-
stituents’. In: Progress in Nano-Electro-Optics II, ed. by M. Ohtsu (Springer,
Berlin 2003) 111
48. S. Mononobe, T. Saiki, T. Suzuki, S. Koshihara, M. Ohtsu: Opt. Commun. 126,
45 (1998)
49. H. Aoki, Y. Sakurai, S. Ito, T. Nakagawa: J. Phys. Chem. B 103, 10553 (1999)
50. V.V. Polonski, Y. Yamamoto, M. Kourogi, H. Fukuda, M. Ohtsu: J. Mi-
croscopy 194, 545 (1999)
51. S. Mononobe, Y. Saito, M. Ohtsu, H. Honma: Jpn. J. Appl. Phys. 43, (2004),
in press
A Novel Method for Forming Uniform
Surface-Adsorbed Metal Particles and
Development of a Localized Surface-Plasmon
Resonance Sensor
H. Takei and M. Himmelhaus
1 A General Method
for Preparing Surface-Bound Metal Particles
1.1 Surface-Bound Metal Particles
There is a vast number of studies directed toward surface-adsorbed metal par-

ticles, of gold and silver in particular. In contrast to isolated spherical metal
particles, analysis of surface-adsorbed particles is made complex because of
the interaction between the particle and the solid substrate [1] as well as
the interaction among adjacent particles. Moreover, the process of forming
metal particles on a substrate quite often leads to deformation of the parti-
cle as well as loss of monodispersity in the size [2, 3]. Despite the daunting
complexity, surface-bound metal particles continue to receive attention, both
experimentally and theoretically, because of a number of fascinating appli-
cations. These range from simple applications such as optical filter/absorber
and substrates for surface-enhanced spectroscopy to more futuristic photonic
applications relying on precise arrangement and interactions among patterned
particles [4]. In order to fully develop and exploit these applications, it goes
without saying that further progress on both experimental and theoretical
fronts is necessary.
In this chapter, we first introduce a novel way of producing surface-
adsorbed metal particles; with this method, while uniform samples can be
readily formed, a number of important physical parameters can also be con-
trolled precisely, allowing preparation of samples with various optical charac-
teristics. We also mention how these particles can be patterned, which might
prove important in regulating interparticle interactions in the future. In the
second half of the chapter, we describe a novel optical biosensor that is based
on localized surface plasmons associated with the sample so prepared. We will
give some measurement examples, followed by general characterization of the
sensor system. We then suggest how the unique feature of our sensor can be
exploited to make contributions to the field of life sciences, culminating in
productive interactions between nano-optics and biotechnology.
58 H. Takei and M. Himmelhaus
1.2 Use of a Monolayer of Monodisperse Dielectric Spheres

as a Template
Traditionally, metal particles are formed on a solid substrate by evaporating

a minute amount of metal; when the deposited amount is limited, a gran-
ular metal film is formed instead of a continuous film [2, 5, 6]. In the case
of gold, if a granular film is thermally annealed, gold becomes more or less
spherical in shape due to its high surface energy. Because of its simplicity,
this system has been widely investigated, but poor control has been a se-
rious impediment. Alternatively, chemically synthesized gold particles can
be attached to a substrate [7, 8]; high monodispersity in size can be readily
accomplished, but quite often coagulation in the process of surface adsorp-
tion can be a serious problem, making it particularly difficult to produce
large surface area samples. Either way, it is generally not possible to form
well-controlled surface-adsorbed particles without resorting to special exotic
techniques such as electron-beam technology [9].
We have devised a novel method for forming uniform surface-adsorbed
metal particles at high density. The method originated from our observa-
tion that by carefully controlling repulsive forces among suspended polymer
nanoparticles and bringing them into contact with a gold substrate [10,11], it
is possible to form a random but uniform monolayer of nanoparticles on the
substrate. Such a monolayer can serve as a template for formation of metal
particles; upon evaporation, the top half of individual polymer nanoparti-
cles becomes coated with the metal. This is one approach that allows us to
form a completely new class of surface-adsorbed metal particles; these par-
ticles are cap shaped, and by exploiting the large selection of monodisperse
spheres available commercially, it is possible to form samples of different di-
ameters endowed with various optical properties. The deposition thickness
also affects the behavior of individual particles. Controlling the average gap
distance between adjacent particles alters the collective properties. Here is an
opportunity to form novel classes of particles with various optical properties.
So called latex spheres, polymer spheres, have continued to fascinate re-
searchers because of the unparalleled monodispersity. In recent years, it is
in the field of photonic bandgap materials that researchers have tried to ex-
ploit monodispersity; a two dimensional photonic bandgap material can be
obtained by forming a regular layer of spheres through a number of different
techniques [4, 12–14]. One can use a technique of raising a substrate out of
a solution containing spheres. By careful control of the lifting rate, spheres
self-assemble on the substrate, resulting in a regular array. Such an array can
be formed also by a flowing sphere suspension against a shallow edge. There
are many other means to form a regular array, but reproducible preparation
of macroscopic samples with high uniformity remains difficult.
While all these methods lead to many interesting results, one simple
method, with potentially many new applications, has not been pursued. This
is what we present in this chapter. Our method starts, as with the example
A Novel Method for Forming Uniform Surface-Adsorbed Metal Particles 59
presented immediately above, with formation of a monolayer of spheres on

a substrate. This monolayer then serves as a template for subsequent evap-
oration of metals. Evaporated metal adheres only on the side of the sphere
facing the evaporation source so that the shape is characterized as a cap [15].
Moreover, the monolayer serves both as an insulator or a spacer. Thus, by
using this simple method, we end up with a large number of cap-shaped metal
particles that are, electrically speaking, floating over the substrate. By capi-
talizing on excellent monodispersity, it is easy to envision that a large variety
of samples can be readily formed. Interestingly, there are many reports on
the use of surface-adsorbed spheres as a temporary mask [16–18], but little
attention has been paid to the optical property with the spheres still on the
surface so far.
We should refer to some earlier examples of related studies. Fischer and
Pohl [19] evaporated gold on top of sparse surface-adsorbed microspheres to
form a metal point to study the near-field effect in its vicinity. Haginoya
et al. [16] have formed a regular array of microspheres, which served as a
mask for subsequent evaporation; here, because the final object of interest
was the metal evaporated on the substrate, the spheres were subsequently re-
moved from the substrate. While not dealing with surface-adsorbed particles,
polystyrene spheres coated with silver have nonetheless been prepared [20].
We should also mention that use of porous materials as a template for syn-
thesis is becoming popular [21].
1.3 Formation of the Monolayer
The usual method for stabilizing colloidal particles is to chemically tailor their
surface in such a way that the repulsive force among the spheres prevents
them from sticking together. By gradually decreasing the repulsive force,
for example, through an increase in the salt concentration, in the presence
of a high-energy surface, spheres begin to adsorb on the surface, forming a
monolayer. This is the basis of our methodology [10]. This simple scheme can
be expanded along a number of different themes.
We have purchased monodisperse polystyrene spheres, from Polysciences,
Warrington, PA, USA (Polystyrene bead Select Certified Size Standard Mi-
crosphere) or Dyno Particles, Norway (Dynospheres) and used them as re-
ceived. For most experiments we used various polystyrene (PS) labware for
life-science experiments as substrates. These were purchased from Nunc,
Nalgene etc. and used without washing. Prior to forming a monolayer of
polystyrene spheres, the PS substrate was coated with 20 nm gold in a vac-
uum evaporator (Type EBV-6 DA, ULVAC, Kanagawa, Japan). Gold, purity
of 99.95%, was purchased from Furuya Metal, Tokyo, Japan and was evapo-
rated from a tungsten boat at a base pressure of 2 × 10−6 Torr at an evapora-
tion rate of 0.1 nm s−1 ; PS substrates were held at least 30 cm away from the
tungsten boat in order to prevent melting of the surface. PS was chosen as the
substrate material because it is optically clear and flat PS pieces are readily
available. As will be discussed later, microscopic irregularities causing visibly

noticeable scattering have to be avoided, but slight warping on a macroscopic
scale, as typically happens with molded plastic, could be tolerated quite well.
Moreover, it was found that gold evaporated on PS showed much greater ad-
hesion than that on a glass substrate, saving us the extra step of introducing a
buffer layer to enhance gold adhesion on; we also noticed that gold evaporated
on a PS substrate was significantly free of granules typically formed when gold
is evaporated on a glass substrate. PS was obviously not the only suitable
polymer; polycarbonate and PMMA apparently serve as the substrate well.
To form a monolayer of PS spheres on a gold-coated PS substrate, partial
aggregation of the PS spheres was induced with addition of carbodiimide,
1-ethyl-3-(3-(dimethylamino)propyl) carbodiimide (EDC), Dojindo Labora-
tories, Kumamoto, Japan, to the PS sphere suspension. While the exact con-
centration of carbodiimide was a deciding factor on the optical property of
the final sample, the concentration is typically in the 10 mM range. As soon
as the mixture of the PS spheres and carbodiimide was prepared, the suspen-
sion was placed on a gold-coated PS substrate; spheres begin to preferentially
adsorb on the substrate, and within a minute, the surface coverage is almost
complete. The role played by carbodiimide is fundamentally not chemical in
nature, and simple salts such as NaCl and KCl have been found to have a
similar effect. It will be explained later, however, that carbodiimide exhibits
some characteristics not found with simple salts. To assure reproducibility,
the incubation was allowed to proceed for ten minutes, after which nonad-
sorbed PS spheres were washed off with a copious amount of distilled water,
leaving only a monolayer on the substrate. While the sample preparation
procedure, summarized in Fig. 1, is rather straightforward, extra care was
needed during drying to obtain optically uniform samples. After the sample
was dried completely, gold was further evaporated to varying thicknesses.
Sample Characterization
For optical characterization, we used various types of spectrometers includ-

ing a Hitachi U–3400 (Tokyo, Japan), an Otsuka Electronics Photal IMUC–
7000 (Osaka, Japan) and an Ocean Optics, S2000 (Florida, USA). The Hi-
tachi spectrometer was equipped with an integrating sphere. IMUC–7000
and S2000 spectrometers were equipped with CCD linear arrays, allowing
instantaneous measurement. The S2000 in particular was configured with a
custom-ordered software so that the spectrum could be fitted nonlinearly with
a pseudo-Voigt function to give the peak wavelength in real time; the peak
wavelength can be displayed as a function of time and stored for later analy-
sis. For electron microscopic characterization, a scanning electron microscope
(SEM), Hitachi SEM Model S–800 and a transmission electron microscope
(TEM), Hitachi TEH Model HF–2000, were employed.
Figure 2 shows a scanning electron micrograph of a typical sample. Al-
though spheres are randomly adsorbed, the overall uniformity is rather note-
Fig. 1. Schematic diagram showing the procedure for formation of surface-

adsorbed, cap-shaped metal particles. (a) Suspension of a sphere/carbodiimide
mixture placed on a substrate. (b) Formation of a monolayer of spheres. (c) Evapo-
ration of a metal on top of the monolayer of spheres. (d) Cap-shaped metal particle
Fig. 2. Scanning electron microscope (SEM) micrograph of a surface-adsorbed

sphere monolayer (sphere diameter 110 nm)
Fig. 3. Transmission electron microscope (TEM) microgram of a dislodged sphere,

with twenty nanometers of gold covering one half of a sphere (sphere diameter
110 nm). Surface-adsorbed spheres were dislodged from the substrate by applica-
tion of ultrasound with a bench-top sonicator. Dislodged spheres were placed on a
substrate and observed with TEM, at courtesy of T. Matsumoto
worthy and extends to the macroscopic scale. If a critical carbodiimide con-

centration is exceeded, however, additional layers begin to form on top of the
first layer. As we have found that formation of the second or more layer has a
detrimental effect on the absorption spectrum of the final sample [11]; while
evaporation of gold on a monolayer of spheres results in a sharp absorption
spectrum, the presence of additional layers of spheres reduces the peak height
as well as significantly broadening the half-width of the spectrum.
Figure 3 is a transmission electron microscope image of particles dislodged
from the substrate; the sphere size is 110 nm and the deposition thickness is
20 nm. It clearly shows that only the top half of the particle becomes covered
by gold.
1.4 Physical Parameters Under Control
We list below physical parameters that can be readily and reproducibly con-
trolled.
Sphere Diameter
A large variety of monodisperse spheres with different diameters are available

from many vendors. The diameter dictates the size of the cap-shaped particle
as well as the distance between the particle and the underlying substrate. The
surface quality of these spheres is also quite excellent.
Adsorption Density
While the dense coverage of the surface with a monolayer of spheres is over
within ten minutes, by adjusting the incubation time one can control the
adsorption density of surface-bound spheres; Fig. 4 shows SEM pictures of
samples prepared with reduced incubation times. As the adsorption density
is reduced, spheres tend to form independent clusters.
Fig. 4. SEM micrograph of adsorbed spheres (sphere diameter 209 nm) at reduced
densities. Typical density at saturation is approximately 15.3 µm−2 , but by reducing
the sphere incubation period, the density can be reduced. (a) Density of 12.2 µm−2 .
(b) Density of 10.3 µm−2
Deposition Thickness
One can readily control the deposition thickness of the metal layers. There
are two layers involved, the one deposited directly on the substrate, and the
other layer deposited on top of the sphere layer. Varying the thickness of each
layer affects the optical property of the sample, but it is the layer deposited
on top of the sphere monolayer that has the most profound effect.
Mixing Spheres of Different Diameters

While the strength of our method lies in its ability to produce monodisperse
particles, it is possible to mix more than one kind of monodisperse spheres to
obtain interesting results. Figure 5 shows such samples. The mixture consists
of 110-nm spheres and 209-nm spheres, and two samples have been prepared
from solutions of different mixture ratios.
Fig. 5. SEM micrographs of samples prepared from a solution containing spheres of

two different diameters, 209 nm and 110 nm. Two solutions with different mixture
ratios were prepared. The resulting samples have adsorbed spheres with different
densities (a) 110-nm spheres at 44.7 µm−2 and 209-nm spheres at 5 µm−2 . (b)
110-nm spheres at 20.4 µm−2 and 209-nm spheres at 10.1 µm−2
Gap Control
Our method allows controlled formation of a gap between adjacent particles.

As is well known from various studies on interactions among particles, the gap
plays an important role on the optical properties. We can form a controlled
gap by slight etching of PS spheres prior to deposition of gold on top of
the sphere. The etching of surface-adsorbed PS spheres was first reported by
Haginoya et al. [16] though their interest was primarily use of the etched PS
spheres as a mask for subsequent evaporation of magnetic materials. As with
Haginoya et al., we have also etched PS spheres by reactive ion etching (RIE)
by using oxygen. The etching duration was adjusted to control the gap size.
Figure 6 shows scanning electron micrographs of samples etched for different
durations.
Fig. 6. SEM micrograph of spheres etched in oxygen plasma. The sphere size is
209 nm. (a) Etching duration of 60 s. (b) Etching duration of 50 s
Heat Treatment
To a lesser degree, we can control the shape with temperature. By raising the
temperature slightly above the onset of polymer deformation, it is possible
to deform the particle. Figure 7 is an SEM picture of a sample that has been
heated up to 120◦ C for a brief period. The sphere has become deformed and
now shows a greater contact area to the substrate. Interestingly, the presence
of the gold cap has the effect of retaining the original shape of the sphere. In
contrast, when surface-adsorbed PS spheres without gold are heated under
the same condition, spheres simply turn into shapeless globs, as shown in
Fig. 8.
Fig. 7. SEM micrograph of a sample temporarily heated up to 120◦ C. 209-nm

spheres are covered by 20 nm of gold. The bottom half of the sphere becomes
deformed, resulting in an increased contact area with the substrate. Image was
taken at an angle of 45◦ from the surface normal
Fig. 8. SEM micrograph showing the effect of heating bare polystyrene spheres
without gold on top. As with Fig. 7, the temperature was raised temporarily up to
120◦ C; in contrast to Fig. 7 lack of gold cap leads to complete merging of adjacent
particles. Image was taken at an angle of 45◦ from the surface normal
Patterned Particles
We have found that we can control the adsorption process of spheres with
patterns of alkanethiol molecules [11]. We prepared a gold-coated substrate
consisting of regions of different alkanethiol molecules. By controlling the
functionality of the alkanethiol molecule, it was possible to form patterned
layers of spheres. To gain control on the adsorption process a distinct interac-
tion between the spheres in suspension and selected areas of the surface needs
to be introduced. While it is possible to utilize physical phenomena, such as
changes in polarity, surface energy, or charge distribution in different regions
of the surface, we have obtained the best results for chemically patterned
surfaces in connection with the use of 1-ethyl-3-(3-(dimethylamino)propyl)
carbodiimide (EDC). Here, the role of EDC in the adsorption process of the
spheres is threefold. Due to its amino group, which is protonated in a neu-
tral pH environment, the molecule is positively charged and thus behaves as a
positively charged surfactant: it adsorbs on the negatively charged sphere sur-
face, thereby changing the total charge of the spheres, i.e., their ζ-potential,
and thus reducing interparticle repulsion. At the same time, the spheres are
converted into carbodiimide functionalized particles and thus provide poten-
tial reaction sites for chemical bonding. Furthermore, dissolved excess EDC
molecules act as macromolecules in the colloidal suspension and force aggre-
gation due to depletion forces. We have found that we have to adjust the EDC
concentration in suspension in such a way, that the body of EDC molecules
adsorbs on the spheres and the remaining dissolved molecules are too few
to exert depletion forces beyond kT . In this regime, which corresponds to
0.1–0.6 mM EDC (for suspensions with 2.6 vol.% spheres and particle sizes
of some 100 nm) adsorption turned out to be sensitive to the chemistry of
the surface. Note that this concentration is significantly lower than the value
for preparing uniform, non-patterned samples. At higher EDC concentration,
this sensitivity is screened by physisorption of the spheres due to depletion
forces. We have proved this concept by replacing EDC with a poly(ethylene
glycol) (PEG) molecule of similar size and molecular weight. PEG is non-
adsorbing and chemically inert, so that it solely exerts depletion forces on
the suspended spheres. At high concentration (above 0.6 mM), where the
depletion forces exceed the Brownian motion of the suspended particles, we
obtained similar results to those found for EDC at similar concentration.
At low concentration, sphere adsorption failed independent of the chemical
functionalities present at the surface. However, by using EDC and proper
surface functionalization, e.g., by means of carboxyl-terminated alkanethiols,
the carbodiimide functionalized spheres adsorb at low concentration due to
chemical bonding. We have combined this selected chemical reactivity with
the technique of microcontact printing to form patterned sphere layers. For
that, a nonfunctionalized thiol (e.g., octadecanethiol; C18 ) is printed onto a
native gold surface by means of a microstructured elastomer stamp (Fig. 9).
In this way, the C18 adsorbs only on the gold areas in contact with the stamp.
e la s to m e r s ta m p
C 1 8
=
A u
+ & P S
T G s o lu tio n C 1 8 T G C 1 8
> A u A u
P S P S
s p h e r e s u s p e n s io n
T G C 1 8
T G C 1 8
? A u A u
P S P S
Fig. 9. Procedure for formation of patterned surface-adsorbed particles. (a) Im-

mersion of a PDMS (poly dimethyl siloxane) stamp into an alkanethiol solution
(octadecanethiol; C18 ), and transfer of the thiol molecules onto a gold-coated poly-
mer substrate. (b) Exposure of a second alkanethiol solution (thioglycolate; TG) to
the substrate results in a pattern consisting of two types of alkanethiol molecules.
(c) Exposure to a sphere suspension, and patterned sphere adsorption
In a second step, the gold substrate is immersed into a solution containing the
functionalized alkanethiol molecule (e.g., carboxyl-terminated such as thio-
glycolate; TG), filling the remaining native areas on the surface. In this way,
a chemical pattern of functionalized and nonfunctionalized areas has formed.
Exposure of this pattern to an EDC-activated sphere suspension at low EDC
concentration yields the formation of patterned sphere layers. Figure 10 is a
SEM picture of a patterned sample.
1.5 Optical Properties
Figure 11 is a visible illustration of gold particles formed with our method.

The entire region consists of a gold-coated (20 nm thick) PS substrate covered
by a monolayer of 110-nm PS spheres; only selected regions were subjected
to additional evaporation of 20 nm gold. The latter regions appear as dark
letters due to strong absorption/scattering by cap-shaped gold particles; this
Fig. 10. SEM micrograph of patterned spheres. Spheres are selectively adsorbed on
regions covered by thioglycolate separated by regions covered by octadecanethiol
Fig. 11. Photograph of gold particles in the shape of letters. Twenty nanometers
of gold were evaporated onto a polystyrene substrate, followed by formation of a
uniform monolayer of spheres (110 nm in diameter). By using a stencil in the shape
of letters, an additional 20 nm of gold were evaporated within well-defined regions.
This is to illustrate visually the optical contrast formed by the gold particle
Fig. 12. Photograph of gold particles formed with PS spheres of different diam-
eters. Spheres are adsorbed within circles defined by the rim while the number
superimposed indicates the sphere diameter in nanometers. The spheres on the
left-hand half of the sample are left bare whereas twenty nanometers of gold were
evaporated on top of the spheres on the right-hand half
was accomplished by using a stencil with openings in the shape of letters.

Although the monochromatic picture fails to convey the actual color, it is a
striking violet. In Fig. 12, spheres of different diameters were subjected to
the same treatment. Two identical sets of PS sphere monolayers, with the
diameters being 110, 152, 209, 324, 480, 800 and 1100 nm, were prepared on
the right and left halves of the substrate. Those on the left were left bare, and
those on the right were subjected to evaporation of 20 nm gold. Although
the monochromatic picture does not do it justice, in general, monolayers of
larger PS spheres exhibit slight absorption without gold coating on top while
not showing enhanced absorption after evaporation. With smaller spheres,
diameters under a few hundred nm, formation of cap-shaped gold particles
leads to significant enhancement in absorption.
Absorption Spectra in the Reflection Mode
The sphere diameter has a rather profound effect on optical properties. We

have varied the sphere diameter between 55 nm and 1.1 µm but it is with
spheres of smaller sizes, under a few hundred nm in diameter, where deposi-
tion of the top layer results in formation of markedly colorful particles.
Fig. 13. Absorption spectra of surface-adsorbed gold particles in the reflection

mode. Twenty nanometers of gold were evaporated onto spheres of different diam-
eters. The reference is a 40-nm thick gold film.
Figure 13 shows absorption spectra of smaller PS particles in the re-

flection mode. While contributions from scattering and absorption have not
been determined precisely, measurement of these samples within an integrat-
ing sphere indicates that absorption accounts for well over half of the effect.
In contrast to the reflection mode, when absorption spectra of the same set
of samples were measured in the transmission mode, there is hardly any ab-
sorption peak as shown in Fig. 14. For example, for 110-nm spheres, with a
Fig. 14. Transmission spectra of surface-adsorbed gold particles. The reference is

a 40-nm thick gold film; all samples have a 20-nm thick layer on the substrate and
a 20-nm thick top layer on top of the sphere
prominent absorption peak around 550 nm in the reflection mode, there is

no trace of reduced transmission around the same wavelength. Moreover, the
presence of the PS particle layer can enhance transmission in some cases over
a solid film of an equivalent total thickness; if a sample is prepared with both
bottom and top gold layers being 20 nm thick, it can produce a higher trans-
mission spectrum than a single 40 nm film even though the former exhibits
a pronounced absorption spectrum in the reflection mode. The sample pre-
pared with 209-nm spheres has a 65% enhancement in transmission around
720 nm. Even at 790 nm where it has the absorption peak in the reflection
mode, the transmission is approximately 100 %. Thus, effectively, absorption
by spheres more or less matches reflection by a plain gold film.
Deposition Thickness Dependence
The deposition thickness has been found to have a completely different effect
depending on whether it is the top or bottom layer that is affected. While the
bottom layer must be present in order to enhance the absorption spectrum,
its exact thickness matters relatively little. On the other hand, the amount of
deposited gold on top of the PS sphere has a great bearing on the absorption
spectrum, as shown in Fig. 15 for 110-nm spheres. For other sphere diameters
as well, initially a broad peak begins to form at longer wavelengths, and as the
deposition thickness is increased, the peak height increases and shifts toward
shorter wavelengths. Eventually, further increases fail to produce additional
shifts. The saturation limit begins to set in at around 20 nm.
Fig. 15. Effect of the deposition thickness on the absorption spectrum. The sphere
diameter is 110 nm for all the samples
Sphere Density
Figure 16 shows spectra of samples with different sphere densities. The den-
sity corresponds to 10.3 µm−2 , 12.2 µm−2 , and 15.3 µm−2 from sparse,
moderate and dense samples, respectively, for 209-nm spheres in (a), and
Fig. 16. Effect of the adsorption density on the absorption spectrum. (a) 209-nm
spheres: the density is 10.3 µm−2 , 12.2 µm−2 , and 15.3 µm−2 from sparse, moderate
and dense samples, respectively. (b) 110-nm spheres: the density is 39.8 µm−2 ,
41.9 µm−2 , and 61.9 µm−2 from sparse, moderate and dense samples, respectively
39.8 µm−2 , 41.9 µm−2 , and 61.9 µm−2 from sparse, moderate and dense
samples, respectively, for 110-nm spheres in (b).
Mixed Samples
Figure 17 shows spectra of mixed samples. Mixing two spheres of different

diameters does not result in formation of two original peaks. Rather a single
peak appears whose peak wavelength depends on the mixture ratio.
Fig. 17. Effect of mixing two sphere sizes at different mixture ratios. (a) 60.9 µm−2
for 110 nm. (b) 44.7 µm−2 for 110 nm and 5 µm−2 for 209 nm. (c) 20.4 µm−2 for
110 nm and 10.1 µm−2 for 209 nm. (d) 15.5 µm−2 for 209 nm
Material Dependence
We have concentrated our effort with gold particles simply due to its stabil-
ity in air, but on occasion we have also experimented with other metals such
as silver, copper and platinum. The spectra for silver samples are shown in
Fig. 18. Twenty nanometers of silver were evaporated on spheres with diam-
eters 55, 110, 152, 209 nm. They were quite susceptible to the atmosphere,
as has been reported before in different systems [22].
Gap Dependence
The ability to control the gap distance at will between adjacent particles is
one of the unique features of the current particle system. It is precisely this
Fig. 18. Absorption spectra of silver particles. Twenty nanometers of silver were
deposited on spheres of various diameters adsorbed on a 20-nm thick gold film
Fig. 19. Absorption spectra of samples prepared from spheres treated to oxygen
plasma etching prior to deposition of the second gold layer. It can be seen that with
increasingly longer etching periods, the absorption maximum shifts toward shorter
wavelengths
kind of structure that is expected to affect the near-field effect in a profound

way. Spectra of etched samples are shown in Fig. 19.
1.6 General Discussion
The present chapter deals mostly with a new methodology. We have presented
some examples of what can be prepared and characterized them only in terms
of absorption spectra and SEM micrographs. Most of the physical parameters
that have been mentioned are well known to affect absorption spectra of
noble metal particles; what is new here is the fact that these parameters can
be, in any combination, modified reproducibly, and resulting samples are of
sufficient size that makes macroscopic characterization an easy task. This is
important not only for practical application of nano-optics based on metal
particles, but also for rapid confirmation of theories with experimental data.
The size dependence is something that has always been part of the dis-
cussion over noble metal particles. While analysis is the most stringent for
an isolated spherical solid particle of limited size, it has been extended to
particles of different shapes such as ellipsoid, hemisphere, rod and cube [2].
In general, the fundamental mode of collective free-electron oscillation dom-
inates in the quasistatic limit where the size is significantly smaller than the
wavelength; modeling with free electrons applies well only for alkali metals,
while noble metals, particularly gold, show pronounced deviations. For shapes
other than a sphere such as the cap-shaped particle in this chapter, one would
expect more than one fundamental mode even in the quasistatic limit. As it is,
the sizes of particles prepared are beyond this limit, so that if the cap-shaped
particle were suspended in isolation, one would expect to observe more than
one peak. One might expect that if particles are oriented under an external
force, it would be possible to selectively observe these peaks, but studies on
cubic and hemispherical particles have revealed that the presence of edges
and corners would preclude homogeneous polarization even when particles
are oriented. Such an experiment is currently planned.
As the deposition thickness is increased, the diameter stays more or less
constant as determined by the cross section of the polystyrene sphere, but
the shape undergoes changes as the cap becomes thicker. In contrast to hemi-
spherical particles and core-shell particles [2,20,23], cap-shaped particles have
rarely been studied as far as we are aware. The core-shell particles, particu-
larly with a dielectric core and a gold shell, have been receiving much interest
in recent years [24–26], as they exhibit absorption peaks in the near infrared
region, and the precise peak position can be controlled by the relative thick-
ness of the shell. It can be imagined that as the deposition thickness is in-
creased for the cap-shaped particle, the characteristics would shift from that
of a core-shell particle to a hemispherical particle, if observed in isolation.
Such an investigation is currently underway.
Difficulties of analyzing the actual samples prepared stem from the fact
that particles are adsorbed on a substrate at high densities. Thus it would
be necessary to take into account image interactions due to the substrate
and interactions among adjacent particles. According to earlier studies on
the effect of the substrate [23], multipole effects are enhanced and many
secondary modes appear. In addition, with the sample under discussion here,
the underlying gold film apparently constitutes part of an optical cavity, thus
enhancing the overall spectrum [10, 27].
The effect of interactions among adjacent particles is observed in more
than one way. Varying the sphere density is one way, and the shift in the
peak position as seen in Fig. 16 is one indication (the peak height is not
a straightforward indicator of the sphere density as reduction in the density
increases the area of pure gold surface that by reflection reduces the apparent
peak height). Mixed samples are another indication. Particle systems with a
certain size distribution are quite common, and they are characterized as
mixtures of independent components. The spectra shown in Fig. 17 show,
however, that spheres of two different sizes interact to give rise to a single
new absorption peak. It should be mentioned that control over the sphere
adsorption density has received attention recently [28].
Samples where a clear gap is formed among adjacent particles provide the
clearest evidence for interactions. In recent years, there has been a revival
of interest in coupled particles, particularly in connection with the enhanced
near field in the gap region. There have been a number of theoretical re-
ports [29, 30], but experimentally it has proven difficult to create samples
consisting of coupled particles with a well-controlled gap distance; it should
be mentioned that Garcia-Santamaria et al. [31] have made an attempt with
silica-coated gold nanoparticles. To our knowledge, we are reporting here
for the first time an experimental technique that allows preparation of such
samples. For now, we limit ourselves to the fact that the gap distance can be
controlled by varying the etching duration, and the extent of the blue shift
becomes greater with longer etching durations. There is no doubt that the
etching also reduces the thickness of PS spheres so that this needs to be taken
into account for future analysis. Moreover, to facilitate comparison with the-
oretical analysis, it would be desirable to prepare samples in such a way that
the axis connecting all coupled particles would be the same because the rela-
tive angle between the axis and the polarization direction of the probing light
plays an important role in the type of excited modes. One such sample would
consist of parallel lines of stringed particles; it should be possible to prepare
such samples by exploiting the patterning technique mentioned earlier.
It is quite exciting to speculate on and eventually verify experimentally
the strength of the near field in the gap region. Theoretical work suggests [30]
that the gap of coupled spheres is an ideal location to generate an intense
“hot” spot that can be utilized for surface-enhanced Raman spectroscopy [32–
34]. This might be verifiable by a combination of Raman spectroscopy and
atomic force microscopy in the fashion reported by [35]; they attempted to
correlate between the morphology of particle clusters and the strength of
observed Raman signal but could not draw a conclusion on the shape of
the ideal “hot” particle. Their conclusion, was however, that the intrinsic
Raman enhancement factors of “hot” particle were on the order of 1014 to
1015 and only a very small number of particles contribute to enhancement.

If we identify the characteristics necessary for enormous enhancement and
selectively prepare them in a large number, we might be better able to exploit
surface-enhanced Raman spectroscopy.
Besides surface-enhanced Raman spectroscopy, transmission of light
through a film consisting of subwavelength structures has been receiving at-
tention. In particular, an opaque metallic film consisting of a periodic array of
subwavelength holes was found to transmit light of significantly longer wave-
lengths [36]. The transmission enhancement has been explained by a number
of models including excitation of surface plasmon polaritons at the holes.
2 Optical Biosensing Application

Noble metal nanoparticles have long been known to exhibit an absorption
spectrum that depends sensitively on the refractive index of the surround-
ing environment [2]. In recent years, as interest in life science has increased,
many workers have proposed using this principle for nonlabeling detection
of biomolecular interactions in a fashion similar to what has been accom-
plished with the surface plasmon resonance (SPR) sensor. Some have ex-
ploited changes in optical properties of metal particles upon aggregation in-
duced by binding of biomolecules [37, 38], while others have made use of
surface-adsorbed particles [5, 8, 18]. We will describe our biosensing tech-
nique [39, 40], starting with surface modifications with biomolecules and ex-
amples of actual measurement. Finally, we will emphasize one significant
property of this novel biosensor; the “bulk” effect is rather diminished. The
“bulk” effect refers to the fact although the SPR sensor is characterized as
a surface-sensitive technique, it is actually influenced by bulk refractive in-
dex of the adjacent layer of fluid some few hundreds of nanometers in depth.
Thus, when used as a biosensor, the sensor is sensitive toward both adsorp-
tion of molecules on the surface and presence of unbound molecules in the
adjacent layer as well as fluctuations in salt concentration, making it not so
trivial to separate these two effects. A diminished “bulk” effect of the current
sensor, however, makes the novel biosensor rather robust against fluctuations
in temperature and salt concentrations of the buffer used, while exhibiting a
comparable sensitivity toward molecular binding.
2.1 Significance of Molecular Interactions in Life Science
There are many opportunities, ranging from medical applications in diag-

nostics and drug development to research applications in genomics and pro-
teomics. Life, from the viewpoint of molecular biology, is fundamentally a
manifestation of a myriad of biomolecular interactions in a wonderfully well-
orchestrated fashion. Binding of a low molecular weight compound to a re-
ceptor molecule on a cell surface may trigger a series of reactions within the
cell while binding of a promotor molecule to a DNA fragment initiates the

expression of a gene. As the complete sequencing of the human genome comes
to fruition, there will be even greater needs to screen binding events involving
the products of these genes, in other words, proteins. With this in mind, we
have selected biosensing as our principle target of application.
Experimental Setup
Figure 20 shows the sensor system consisting of a gold particle sensor, an

optical fiber set up for illuminating the sensor with white light and col-
lecting reflected light to be sent to a spectrometer, a custom software for
peak-fitting and monitoring the peak position of the spectrum in real time.
The gold particles were formed at the bottom of wells, 10 µl in volume,
of the miniplates purchased from Nunc. The tungsten halogen lamp, LS-1,
and the spectrometer, S2000, were both purchased from Ocean Optics. The
fiber bundle equipped with SAM 905 connectors was fabricated by Tsuchida
Seisakusho (Tokyo, Japan). The custom software is capable of fitting the
spectrum with a pseudo-Voigt function in real time and displays the peak
wavelength as a function of time. Four independent channels are provided so
that four samples can be measured independently of each other. The setup is
provided with a mixing mechanism to ensure that proper kinetic data can be
obtained. The measurement protocol consists of the following steps. Figure 21
is a photograph of the setup.
1. The bottom of each well is coated with a capture biomolecule such as an
antibody, a receptor protein, or a DNA primer.
2. A sample potentially containing the antigen is injected into the well.
Fig. 20. Schematic diagram of the experimental setup. Gold particles were formed
at the bottom of a polystyrene well with a volume of 10 µl. Light from a tungsten
halogen lamp is led to the well bottom via a fiber bundle, and the reflected light
is collected for real-time analysis by a spectrometer. The sample is mechanically
agitated to assure immediate and uniform mixing
Fig. 21. Photograph of a four-channel model. A personal computer, not shown,

completes the system
3. The reflectance spectrum is fitted by a pseudo-Voigt function in real-time,

and the peak wavelength is displayed as a function of time.
4. The time change of the peak wavelength is fitted to an exponential func-
tion, and the time constant is obtained.
5. The time constant is obtained from samples with different concentrations,
and the affinity is calculated.
Surface Coating
There are many methods for adsorbing biomolecules onto the sensor surface.
(1) Direct physisorption
It is well known that many biomolecules bind irreversibly to a bare gold
surface. It is thus possible to inject a buffer solution containing a biomolecule.
The selection of the buffer solution is not crucial, but we often use phosphate
buffer saline (PBS) maintained at pH 7.4. Although the concentration of
1 mg/ml quickly saturates the surface in a matter of a few minutes, the
concentration of 100 µg/ml is enough with a slightly longer incubation period.
(2) Coupling via protein A or protein G
Direct physisorption described above is rather straightforward, but there is
some concern that the antibody may partially lose its ability to bind its
antigen through denaturation. This possibility is reduced if the antibody is
adsorbed indirectly via an intermediate layer. Protein A and protein G can
serve such a purpose; they both have a multiple number of binding sites for
antibodies (four for protein A and two for protein G) so that physisorbed
protein A or protein G is likely to retain its function. As in (1), protein A or
protein G at the concentration of 100 µg/ml can be injected into the sensor
well, and after some 10 minutes, the coverage is more or less complete. Then,
a solution containing an antibody is injected. The origin of the antibody,
whether from mouse, goat, human, etc., has some effect on the affinity of
protein A or protein G toward the antibody.
(3) Coupling via biotin–avidin (streptavidin) reaction
The great affinity between biotin and avidin (or similarly for streptavidin) is
well known, and most biomolecules can be readily modified by a biotin moi-
ety; thus physisorbed avidin can serve as the binding layer for a biotinylated
biomolecule.
(4) Activated carboxyl group
Formation of a covalent bond between an activated carboxyl group and an
amino group is widely used as a linking method; most biomolecules possess
an amino group so that a surface-bound carboxyl group can serve as a start-
ing point. Gold is a convenient surface because an alkanethiol molecule is
known to readily form a self-assembled monolayer (SAM) and alkanethiol
molecules with carboxyl functional groups are commercially available. An
ethanol solution containing 10-carboxy-1-decane thiol (10-carbo.), purchased
from Dojindo, is prepared at the concentration of 1 mM. Exposure of the
sensor well to the solution for a period of 10 minutes is enough to prime the
sensor surface with a carboxyl group. A freshly prepared mixture of EDC
and N-hydroxy succinimide (NHS), 0.1 M, is added for some 10 minutes and
immediately afterwards the biomolecule to be adsorbed is injected.
2.2 Measurement Examples
Antigen–Antibody Reaction 1
One of the most widespread uses of a biosensor is to detect an antigen with a

corresponding antibody. Here we show in Fig. 22 a result of detecting a low
molecular weight peptide (molecular weight of 841 dalton), a polypeptide of
6 histidine (6 His) amino acid residues, with an anti-6 His antibody. First,
streptavidin protein is directly physisorbed onto the bare gold surface. Then,
biotinylated protein A is coupled to streptavidin via biotin/streptavidin bind-
ing. Finally, the anti-6 His antibody is anchored to protein A via the Fc region.
Note that each time a molecule binds to the sensor surface, the peak wave-
length increases. Once the sensor is coated with an antibody, it is ready to
detect an antigen. When 6 His is added to the sensor at the concentration of
0.5 mg/ml, despite its low molecular weight, there is an observable increase
Fig. 22. Antigen–antibody reaction using a polypeptide (6 His: a string of six

histidine amino acid residues) as the antigen. In order to bind an anti-6 His antibody
(IgG: immunoglobulin G), the sensor surface is exposed to streptavidin, biotinylated
protein A and anti-6 His antibody sequentially. When captured by protein A, the
antibody exposes its binding sites toward the liquid phase, away from the substrate.
The surface-adsorbed antibody can be used a number of times; a capture antigen
molecule, 6 His in this example, can be released by exposure to diluted 10 mM
hydrochloric acid, and the antibody is ready to capture another antigen molecule
again. Here, the sequence is repeated 8 times
in the peak wavelength. Once the binding becomes saturated, the unbound
antigen is washed off with phosphate buffer. To remove the antigen, addition
of diluted hydrochloric acid, 10 mM, is enough to break up the antigen–
antibody reaction. This is reflected in a reduction in the peak wavelength.
Addition of diluted acid, however, leaves the antibody intact so that when
the antigen is added again, the surface-bound antibody can capture it. This
process can be repeated.
Antigen–Antibody Reaction 2
Here we show another example of an antigen–antibody reaction. This time

streptavidin protein is directly physisorbed onto the sensor as in the pre-
vious example. However, surface-bound streptavidin is used to capture an
anti-streptavidin antibody. As the molecular weight of an antibody (IgG: im-
munoglobulin G) is rather high at 150 kilodalton, it can be detected at sig-
nificantly reduced concentrations. Samples of different concentrations were
added to the sensor at time t = 200 s, and all the results were superimposed.
After 600 s, the sensor was rinsed with phosphate buffer saline, followed by
Fig. 23. Dependence of the sensorgram on the sample concentration. The covered
range is from 0.25 µg/ml to 32 µg/ml. At t = 200 s, anti-streptavidin antibody
is injected onto a sensor covered by streptavidin. At t = 800 s, phosphate buffer
saline (PBS) is used to rinse off unbound antibody. After t = 900 s, the sensor is
exposed to diluted hydrochloric acid to break the bond between streptavidin and
anti-streptavidin
injection of diluted hydrochloric acid to remove the antibody completely. This

was carried out for samples of different concentrations as shown in Fig. 23.
The purpose of injecting samples of different concentrations was to obtain
kinetic data. Quite often the main goal of measuring binding reactions is to
obtain the equilibrium binding constant kD . This can be obtained by repeat-
ing the same reaction at different concentrations and each binding curve is
fitted by an exponential function. The time constant is plotted against the
concentration. The slope is equivalent to kass and the y intercept is equivalent
to kdiss . kD is simply kdiss divided by kass . Figure 24 shows an example of
such a plot. The quality of fitted data is comparable to that obtained with a
commercial instrument with significantly increased complexity.
Protein–DNA Interactions
Figure 25 shows a binding event between a surface-adsorbed double-stranded

DNA and a DNA-binding protein.
2.3 General Characteristics of the Sensor
While there are many types of biosensors [6, 8, 41–44], one of the most im-
portant parameters for a biosensor is sensitivity. Naturally it is roughly de-
Fig. 24. Obtaining kinetic data. Sensorgrams from Fig. 23 are fitted to a single
exponential function, A − B exp(−kt), and the binding constant, k, was obtained
for various concentrations. All the binding constants were plotted as a function of
the concentration; linear regression was used for fitting. The slope of the fitted data
is the association time constant, kass , and the y intercept is the dissociation time
constant, kdiss . The equilibrium dissociation constant, kD , is kdiss /kass . kD of 63 nM
is fairly typical for an antigen–antibody reaction. The quality of linear regression
also indicates the reliability of our sensor
Fig. 25. Sensorgram showing DNA hybridization followed by binding of a DNA-

binding protein. A biotinylated 60 mer was first attached via surface-bound strep-
tavidin, and the second DNA strand, 57 mer, was hybridized in a high salt con-
centration buffer. After rinsing with HEPES buffer, a nuclear receptor, RXR, was
injected
Fig. 26. Sensorgram showing formation of a self-assembled monolayer (SAM) of

octadecanethiol. Because SAM has been well characterized, it is used here as a
calibration standard
termined by the amount of shift induced when a molecular layer of certain

density and thickness forms and the minimum shift that can be detected.
Figure 26 shows a sensorgram when a self-assembled monolayer (SAM)
of octadecanethiol forms on the sensor with the full shift of approximately
5.5 nm in the peak wavelength. According to the literature [45, 46], the full
thickness of such a SAM is 2 nm. The noise level obviously depends on many
factors such as the specification of the spectrometer, intensity of the light
source, throughput of the optical fiber system, as well as processing of the
signal. Thus, what is shown in Fig. 27, stability of the baseline when no
reaction is taking place, is only a rough indication of what the current system
is capable of; if the shift in the peak wavelength is more than 0.02 nm, it can
be detected without difficulty. Thus, it is safe to say that a submonolayer
with an averaged thickness of 0.01 nm can be detected.
Refractive Index Dependence
The sensor is exposed to various fluids of different refractive indices. This was
achieved by preparing mixtures of water and glycerol with different mixture
ratios as shown in Fig. 28. With refractive indices of 1.33 for water and 1.47
for glycerol, the sensorgram shows that the change in the peak wavelength
for a unit change in the refractive index is 61 nm. In contrast, the litera-
ture figure for the traditional SPR, when measured in terms of wavelength,
is typically 3000 nm per unit change in the refractive index [42, 47]. This
Fig. 27. Sensorgram showing stability of the baseline
Fig. 28. Sensorgram showing sensor response toward water/glycerol mixtures. The
volume concentration of glycerol is incrementally increased by 10%; the first mixture
contains 10% glycerol, and the last mixture added at t = 1900 s has 80% glycerol.
With refractive indices, n = 1.33 for water and 1.47 for glycerol, it is roughly
estimated that the shift is 61 nm per one refractive index unit, δn = 1
would suggest that the sensitivity of our sensor is significantly lower. Fig-
ure 29, however, shows an interesting comparison between the SPR sensor
and our sensor subjected to an identical protocol; the first half consists of a
Fig. 29. The following protocol was used to illustrate the reduced “bulk” effect of
our LSPR instrument with respect to the SPR instrument; at t = 100 s, 1 mg/ml
avidin was added to the sensor to initiate physisorption. After saturation, unbound
avidin was rinsed with PBS buffer at t = 400 s. 100 µg/ml of anti-avidin antibody
was injected to start the antigen–antibody reaction; the overall shift in the wave-
length is typical. The portion of the sensorgram after t = 1200 s is identical to the
initial portion of Fig. 28. (a) Response of an SPR sensor from Texas Instruments,
Spreeta, that was chosen as a representative SPR sensor. (b) Our instrument. Note
how the shifts induced by injection of glycerol mixtures are suppressed in (b)
regular antigen–antibody reaction followed by an incremental increase in the

refractive index of the fluid. At 100 s, avidin was physisorbed on the sensor,
followed by injection of an anti-avidin antibody at 500 s. Then, starting at
1200 s, water/glycerol mixtures are injected every 100 s, with increasingly
higher mixture ratios. Note that shifts in the latter half of the protocol, in
comparison to the first half, are significantly greater with the SRP than with
our sensor. This shows that our sensor has a significantly enhanced response
toward binding of a biomolecule over that induced by change in the refractive
index of the surrounding medium [48].
Temperature Dependence
Figure 30 shows the response of the sensor when water at various temper-
atures is injected. Initially it is at room temperature (20◦ C), and the tem-
perature of water injected is increased incrementally from 1 to 63◦ C; each
injection results in a peak because the temperature returns to room temper-
ature rather rapidly. There are two noteworthy features. One is that even
though the refractive index is higher for water at lower temperature in this
temperature range, the shift in the peak wavelength is opposite; an increase in
Fig. 30. Temperature dependence of the sensor. Water of various temperatures

was injected. The experiment starts at room temperature, 20◦ C. Temperature of
water injected was 1◦ C at t = 100 s, 5◦ C at t = 200 s, 10◦ C at t = 250 s, 16◦ C at
t = 300 s, 20◦ C at t = 350 s, 26◦ C at t = 400 s, 30◦ C at t = 450 s, 35◦ C at t = 500 s,
39◦ C at t = 550 s, 43◦ C at t = 600 s, 48◦ C at t = 650 s, 50◦ C at t = 700 s, 53◦ C at
t = 750 s, 57◦ C at t = 800 s, 60◦ C at t = 850 s, 63◦ C at t = 900 s The temperature
returns to room temperature within tens of seconds. Between t = 1000 and 1050 s,
water at 1◦ C was injected three times
the density results in a decrease in the peak wavelength. This suggests that a
change in temperature has an effect on the sensor through a mechanism other
than a simple change in the refractive index, possibly a structural change in
the nanostructure of the sensor itself. The second noteworthy feature is that
there is hysteresis; the peak wavelength at room temperature, after the se-
ries of temperature jumps, is permanently changed. Again, this might reflect
a structural change in the nanostructure of the sensor itself. While this is-
sue might call for attention in the future, the fact that temperature-induced
shifts are quite small and most if not all binding experiments will not subject
the sensor to such large jumps in temperature renders this a minor nuisance
rather than a problem. In contrast, a typical SPR instrument is characterized
by a jump of 0.01 angular degree (e.g., 100 resonance units in the industry
parlance) per 1 degree Celsius [49]. In terms of the wavelength of our instru-
ment, it would loosely translate to 0.5 nm/degree Celsius.
In summary, we have developed a novel biosensor based on optical proper-
ties of gold nanoparticles. There have been numerous reports in recent years
on similar sensors also based on localized surface plasmon resonance, LSPR,
of gold nanoparticles [6, 8]. Advantages cited for the LSPR sensor over the
SPR sensor are simplicity of the instrument, low costs, ability to give a result
that can be read off visually, etc. On the other hand, there have been few
reports on the actual performance. It has been observed that the dependence
of the peak wavelength on the refractive index of the surrounding material is
significantly lower, by more than one order of magnitude. We also make sim-
ilar observations, but we emphasize that as far as detection of biomolecules
is concerned, it has a sensitivity significantly higher than what the result
of the bulk dependence leads one to believe. The reduced “bulk” effect is a
tremendous advantage because molecular-binding events can be monitored
stably in the presence of fluctuations in temperature and salt concentrations
of the buffer solution.
One outstanding characteristic of our LSPR sensor is the extraordinarily
high optical density and the narrow width of the absorption peak. Typically
the optical density of surface-adsorbed gold nanoparticles is less than 0.5,
but the figure for our sensor is more than four times greater. This facilitates
accurate determination of the peak wavelength, better than 0.02 nm. As the
shift induced upon formation of a SAM, corresponding to a mass loading of
1 ng/mm2 , is typically 5.5 nm, under the assumption that linearity holds,
this leads to the detection limit of under 5 pg/mm2 . This compares well with
the reported practical limit of 10 pg/mm2 for the SPR sensor.
3 Curious Observation
While in most cases, the response of our sensor is quite comparable to that
of the SPR sensor with the generally observed difference mentioned earlier,
some protocols have led to curious observations. One such case is shown in
Fig. 31. Binding of amino dextran (MW 10 kilodalton) onto the sensor coated with
10-carbo. Ostensibly, the negative charge of the thiol molecule attracts electrostat-
ically positively charged amino dextran suspended in water. Remarkably, the shift
is well over 120 nm. If the refractive-index dependence in Fig. 28 is any indication,
this would reflect change of more than two refractive index units. Obviously, this
cannot be so, thus inviting further exploration for an explanation
Fig. 31. Here, The sensor was first exposed to an ethanol solution containing
10-carbo at the concentration of 1 mM. Then, excess 10-carbo was rinsed with
pure water, resulting in reduction in the peak wavelength. Then, a solution
containing 10 kilodalton amino dextran was injected. Opposite charges on the
surface-bound carboxyl group and amino dextran are expected to lead to for-
mation of an amino dextran layer. Indeed this was what we observed, but the
shift was phenomenal at well over 120 nm. Although the exact shift depends
on the molecular weight of the amino dextran used and its concentration,
such phenomenal shifts are routinely observed with this particular protocol.
When we recall that a shift due to a SAM formation was well under 10 nm,
and a shift caused by a change in the refractive index was some 61 nm per
unit change, a shift of more than 100 nm is surprising. While we suspected
a permanent change in the sensor structure, observation of the sensor with
a scanning electron microscope did not reveal any noticeable change, not to
mention removal of the spheres.
4 Conclusion
We have shown a new method for forming surface-adsorbed metal parti-
cles; The beauty of this method is its simplicity and practicality. While this
method will certainly give rise to many applications, we have concentrated on

its application to biosensing where currently there is a great demand for the
emerging field of proteomics. In comparison to existing analytical techniques,
the LSPR sensor based on gold nanoparticles produced with the new method
has turned out to possess a number of advantages. It is characterized by a sig-
nificantly reduced “bulk” effect, which is a great plus as it allows stable read
out in the presence of fluctuations in temperatures and salt concentrations.
Acknowledgements
This work has been made possible by a grant MF-16 from the Ministry
of Health and Labor. We express much appreciation to Dr. T. Fujimura,
Dr. J. Pipper, Dr. Z. Cao, Dr. T. Okamoto, Dr. H. Sugiyama, Dr. K. Ni-
nomiya, Dr. T. Iwayanagi, and Dr. H. Taguchi in the course of this project.
Without their help, the project would not have been possible. We also thank
the Ministry of Health and Labor, and Dr. Inoue of the National Institute
of Health Science of Japan for their partial support of this project. Professor
M. Furuya and Dr. H. Kambara have been constant sources of encourage-
ment.
References
1. T. Kume, N. Nakagawa, S. Hayashi, K. Yamamoto: Solid State Commun. 93,
171 (1995)
2. Optical Properties of Metal Clusters by U. Kreibig, M. Vollmer (Springer, Berlin
1995)
3. Nanoparticles and Nanostructured Films ed. by J.H. Fendler (Wiley-VCH,
Weimheim 1998)
4. R. Micheletto, H. Fukuda, M. Ohtsu: Langmuir 11, 3333 (1995)
5. G. Kalyuzhny, M.A. Schneeweiss, A. Shanzer, A. Vaskevich, I. Rubinstein:
J. Am. Chem. Soc. 123, 3177 (2001)
6. F. Meriaudeau, T. Downey, A. Wig, A. Passian, M. Buncick, T.L. Ferrell: Sens.
Actuators B 54, 106 (1999)
7. M. Brust, D. Bethell, C.J. Kiely, D.J. Schiffrin: Langmuir 14, 5425 (1998)
8. T. Okamoto, I. Yamaguchi, T. Kobayashi: Opt. Lett. 25, 372 (2000)
9. G.A. Niklasson, H.G. Craighead: Thin Solid Films 125, 165 (1985)
10. H. Takei: J. Vac. Sci. Technol. B 17, 1906 (1999)
11. M. Himmelhaus, H. Takei: Phys. Chem. Chem. Phys. 4, 496 (2002)
12. A.S. Dimitrov, C.D. Dushkin, H. Yoshimura, K. Nagayama: Langmuir 10, 432
(1994)
13. T. Fujimura, T. Itoh, A. Imada, R. Shimada, T. Koda, N. Chiba, H. Mura-
matsu, H. Miyazaki, K. Ohtaka: J. Lumino. 954, 87 (2000)
14. Y. Masuda, M. Itoh, T. Yonezawa, K. Koumoto: Langmuir 18, 4155 (2002)
15. H. Takei, N. Shimizu: Langmuir 13, 1865 (1997)
16. C. Haginoya, M. Ishibashi, K. Koike: Appl. Phys. Lett. 71, 2934 (1997)
17. J.C. Riboh, A.J. Haes, A.D. McFarland, C.R. Yonzon, R.P. Van Duyne:
J. Phys. Chem. B 107, 1772 (2003)
18. M.D. Malinsky, K.L. Kelly, G.C. Schatz, R.P. Van Duyne: J. Am. Chem. Soc.
123, 1471 (2001)
19. U.C. Fischer, D.W. Pohl: Phys. Rev. Lett. 62, 458 (1989)
20. P. Barnickel, A. Wokaun: Mol. Phys. 67, 1353 (1989)
21. G.L. Hornyak, C.R. Martin: Thin Solid Films 303, 84 (1997)
22. Y.W. Cao, R. Jin, C.A. Mirkin: J. Am. Chem. Soc. 123, 7961 (2001)
23. R.J. Warmack, S.L. Humphrey: Phys. Rev. B 34, 2246 (1986)
24. H.S. Zhou, I. Homma, H. Komiyama, J.W. Haus: Phys. Rev. B 50, 12052 (1994)
25. R.D. Averitt, D. Sarkar, N.J. Halas: Phys. Rev. Lett. 78, 4217 (1997)
26. T. Pham, J.B. Jackson, N.J. Halas, T.R. Lee: Langmuir 18, 4915 (2002)
27. A. Leitner, Z. Zhao, H. Brunner, F.R. Aussenegg, A. Wokaun: Appl. Opt. 32,
102 (1993)
28. R.R. Bhat, D.A. Fischer, J. Genzer: Langmuir 18, 5640 (2002)
29. J.P. Kottmann, O.J.F. Martin: Opt. Lett. 26, 1096 (2001)
30. H. Xu, M. Käll: Phys. Rev. Lett. 89, 246802 (2002)
31. F. Garcia-Santamaria, V. Salgeuirino-Maerira, C. Lopez, L.M. Liz-Marzan:
Langmuir 18, 4519 (2002)
32. P.F. Liao, M.B. Stern: Opt. Lett. 7, 483 (1982)
33. R.G. Freeman, K.C. Grabar, K.J. Allison, R.M. Bright, J.A. Davis,
A.P. Guthrie, M.B. Hommer, M.A. Jackson, P.C. Smith, D.G. Walter,
M.J. Natan: Science 267 1629 (1995)
34. K. Kneipp, Y. Wang, H. Kneipp, L.T. Perelman, I. Itzkan, R.R. Dasari,
M.S. Feld: Phys. Rev. Lett. 78, 1667 (1997)
35. S. Nie, S.R. Emory: Science 275, 1102 (1997)
36. A. Dogariu, T. Thio, L.J. Wang, T.W. Ebbesen, H.J. Lezec: Opt. Lett. 26, 450
(2001)
37. R. Elghanian, J.J. Storhoff, R.C. Mucic, R.L. Letsinger, C.A. Mirkin: Science
277, 1078 (1997)
38. P. Bao, A.G. Frutos, C. Greef, J. Lahiri, U. Muller, T.C. Peterson, L. Warden,
X. Xie: Anal. Chem. 74, 1792 (2002)
39. H. Takei: In: SPIE Microfluidics Devices and Systems (Santa Clara, California,
1998) 278
40. M. Himmelhaus, H. Takei: Sens. Actuators B 63, 24 (2000)
41. C.R. Musil, D. Jeggle, H.W. Lehmann, L. Scandella, J. Gobrecht, M. Döbeli:
J. Vac. Sci. Technol. B 13, 2781 (1995)
42. J. Homola, S.S. Yee, G. Gauglitz: Sens. Acutators B 54, 3 (1999)
43. N.-P. Huang, J. Voros, S.M. De Paul, M. Textor, N.D. Spencer: Langmuir 18,
220 (2002)
44. A.J. Thiel, A.G. Frutos, C.E. Jordan, R.M. Corn, L.M. Smith: J. Vac. Sci. Tech-
nol. B 13, 2781 (1995)
45. A. Ulman: Chem. Rev. 96, 1533 (1996)
46. S. Maltis, I. Rubinstein: Langmuir 14, 1116 (1998)
47. L.S. Jung, C.T. Campbell, T.M. Chinowsky, M.N. Mar, S.S. Yee: Langmuir 14,
5636 (1998)
48. H. Takei, M. Himmelhaus, T. Okamoto: Opt. Lett. 27, 342 (2002)
49. Biomolecular Sensors ed. by E. Gizeli, C.R. Lowe (Taylor & Francis, London
2002)
Near-Field Optical-Head Technology for
High-Density, Near-Field Optical Recording
T. Matsumoto
1 Introduction
Increased recording density is required in the storage devices used in personal

computers, network servers, home servers, and other multimedia devices as
users increasingly want to save large amount of images, movies, music files,
and so on. To meet this demand, the recording density of storage devices has
been steadily increased (Fig. 1). For example, the recording density of optical
recording devices reached 19.5 Gb/in2 in the case of the Blu-ray disk drive
introduced to the market in 2003, while that of hard disk drives (HDDs) on
the market reached 70 Gb/in2 in 2002.
However, the rise in recording densities seems to be encountering barriers
in both optical and magnetic recording. In the case of optical recording, the
recording density depends on the spot size of the focused laser. However, the
minimum spot size is limited by the diffraction of light, and the recording
Fig. 1. Growth of the recording density of recording devices

94 T. Matsumoto
density cannot be increased beyond this limit. In the case of magnetic record-
ing, the recorded information is expected to decay spontaneously because of
thermal fluctuation in the magnetic domains when the recording density ap-
proaches 1 Tb/in2 . A solution is to increase the coercivity of the magnetic
recording medium. However, conventional writing heads cannot generate a
magnetic field large enough to write data on such recording medium because
of a fundamental limit on the saturation moment of the writing head material.
As a method to overcome these limits, near-field optical recording is now
drawing much attention. In this recording, the data bits are written and
read by using an optical near field generated near a nanometer-scale object.
Because the size of the optical spot is not limited by the diffraction of light,
and can be reduced to less than a few tens of nanometers [1], this method is
expected to enable a recording density of over 1 Tb/in2 .
In a near-field optical recording device, the key component is a near-field
optical head that generates the optical near field. In this chapter, after briefly
reviewing near-field optical recording, we describe technical issues regarding
the near-field optical head and introduce some solutions to resolve them.
Among these issues, improved efficiency in generating the optical near field
is the most important. We describe a highly efficient near-field optical head
that uses a metallic plate; the intensity of the optical near field is enhanced
by a plasmon generated in the metallic plate.
2 Review of Near-Field Optical Recording
2.1 The Limit of Conventional Optical Recording
The recording density of conventional optical recording drives has been in-
creased by reducing the spot size of the focused laser. However, the minimum
spot size, D, is governed by the law of diffraction and is approximately
D = λ/N A , (1)
where λ is the light wavelength and N A is the numerical aperture of the

focusing lens. As an example, in the most advanced optical recording device,
the Blu-ray disk drive, the laser wavelength is 405 nm, and the N A is 0.85,
thus the spot size is approximately 480 nm. To reduce the spot size further,
we must decrease the laser wavelength or increase the N A of the focusing
lens. However, a wavelength of less than 405 nm cannot be obtained from
currently available semiconductor laser diodes, and it is difficult to make the
N A larger than 0.85 because a larger N A reduces the tolerance for recording
disk fluctuation. Therefore, it will be difficult to further reduce the spot size.
One way to reduce the spot size is to place a hemispherical or a Weierstrass-
sphere shaped lens, called a solid immersion lens (SIL), near the recording
medium (Fig. 2) [2, 3]. When an SIL is placed near the recording medium,
Near-Field Optical-Head Technology 95
Fig. 2. Light focusing methods in optical recording devices
the wavelength inside the sphere is reduced by a factor of n, where n is the

refractive index of the lens; thus, the spot size is reduced to
D = λ/nN A . (2)
In addition, in the case of the Weierstrass-sphere lens, the incident rays are
refracted at the sphere’s surface. This corresponds to an increased N A, and
the spot size is further reduced to
D = λ/n2 N A . (3)
This small optical spot enables increased recording density. For example, a
recording density of about 70 Gb/in2 has been realized using an SIL made of
Bi4 Ge3 O12 with a refractive index of 2.23 [4]. This type of recording method is
sometimes classified as near-field optical recording because it uses an evanes-
cent wave generated by a ray whose incident angle is larger than the critical
angle. However, the spot size is still limited by the diffraction limit of light,
and no dramatic increase of recording density is expected.
2.2 Near-Field Optical Recording Method
In near-field optical recording, the data bits are written and read by using
the optical near field. When light is introduced into a nanometer-scale object,
such as a subwavelength aperture or scatterer, a localized electromagnetic
field – the optical near field – is generated near the object. The distribution
of the optical near field is determined by the object’s shape and size, and
does not depend on the light wavelength. For example, when the optical
near field is generated by the subwavelength aperture, the width of the near-
field distribution is as small as the aperture diameter [1]. Therefore, a high
96 T. Matsumoto
recording density beyond the diffraction limit can be realized by writing and
reading data bits using the optical near field.
Betzig et al. [5] have demonstrated near-field optical recording using a
magneto-optical recording medium. They generated the optical near field us-
ing a fiber probe, which is widely used in scanning near-field optical mi-
croscopy (SNOM). In the writing process, a Pt/Co multilayer recording
medium was heated to the Curie temperature by the optical near field to
reverse the magnetization of the recording medium. In the reading process,
transmitted light from the recording medium was collected by the fiber probe
and the rotation of the light polarization caused by the Faraday effect was
detected. They have written and read recording marks with a diameter of
60 nm, which corresponds to a recording density of 170 Gb/in2 .
Near-field optical recording using a phase-change recording medium or a
photochromic recording medium has been reported by, respectively, Hosaka
et al. [6] and Jiang et al. [7]. In the phase-change recording, the recording
medium is heated by light to change its phase from crystalline to amorphous,
or vice versa. Hosaka et al. used GeSbTe recording medium, and wrote and
read recording marks with a diameter of 60 nm using a fiber probe. In the
photochromic recording, the recording marks are written through a photo-
chemical reaction. Jiang et al. wrote and read recording marks with a diam-
eter of 130 nm on a Langmuir–Blodgett film of photochromic material.
2.3 Hybrid Recording Method
A hybrid recording method, known as thermally assisted magnetic recording

or heat-assisted magnetic recording, has also been proposed [8, 9]. In this
method, data bits are written on a magnetic recording medium by applying
a magnetic field while the medium is heated by the optical near field, and
read out with a reading head used for hard disk drives, such as a giant
magnetoresistive (GMR) head or a tunneling magneto-resistive (TMR) head.
Because GMR and TMR heads are highly sensitive to magnetic fields, a high
signal-to-noise ratio in the reading signal can be achieved. Note that the
problem of thermal fluctuation in the magnetic domains can be solved when
data bits are written using the optical near field. During the writing process,
the recording medium is heated by light, hence its coercivity is decreased.
Therefore, a recording medium with high coercivity can be used to make the
magnetic domains stable.
3 Technical Issues
Regarding the Near-Field Optical Head
In the primary experiments done on near-field optical recording, a fiber probe
was used as the near-field optical head. However, this probe was designed
mainly for near-field optical microscopy and is not suitable for practical
recording devices. To realize a practical near-field optical head, the following
requirements must be met.
1. Precise control of the spacing between the head and the recording medium
2. Integration of peripheral components into the near-field optical head
3. Highly efficient generation of the optical near field
3.1 Precise Control of the Spacing Between the Head

and the Recording Medium
The optical near field is localized near the aperture or the scatterer and its
intensity decreases exponentially as the distance from the surface of the head
increases. Because the decay length is about 10–20 nm, the spacing between
the head and the recording medium must be maintained within a range of
10–20 nm.
With the fiber probe, the spacing is controlled by using the shear force
acting between the fiber probe and the recording medium [10, 11]. In this
method, the fiber probe is dithered parallel to the surface of the recording
medium. Because the dithering amplitude is changed by the shear force, a
feedback loop can be constructed by detecting the change in the dithering
amplitude. However, in a practical recording device, the recording disk rotates
at a very high speed. For example, when the recording density is 1 Tb/in2 ,
and the data transmission rate is 1 Gbps, the linear velocity of the recording
medium is about 30 m/s. If the fiber probe is placed near recording medium
rotating at such speed, the fiber probe soon crashes because the response
based on shear-force feedback is too slow.
To solve this problem, planar near-field optical heads have been proposed.
In this type of head, the structure for generating the optical near field is
mounted on the slider used for the hard disk drive. In the hard disk drive,
the spacing between the head and the recording medium is maintained at a
constant value by air pressure acting between the slider and the recording
medium. (The air pressure is determined by the design of pads formed on
the slider surface.) The head–medium spacing of hard disk drives is currently
near 10 nm, and is expected to decrease to less than 10 nm within a few years.
By applying this technology to the near-field optical head, we can precisely
control and maintain the spacing between the near-field optical head and the
recording medium.
Figure 3 shows an example of a planar near-field optical head.
Lee et al. [12] have proposed a silicon planar near-field optical head that
has an array of square apertures (Fig. 3a). The 60 × 60 nm apertures were
fabricated by an isotropic etching process with high reproducibility, and a line
pattern with linewidths of 250 nm was read out with this near-field optical
head.
98 T. Matsumoto
Fig. 3. Examples of planar near-field optical heads: (a) Head with an array of
apertures formed in a Si substrate, (b) Head with an aperture formed on a glass
protrusion, (c) Head with an array of apertures formed on Si protrusions
Isshiki et al. [13] have proposed a planar near-field optical head with an
aperture formed on a pyramidal protrusion (Fig. 3b). They fabricated the
pyramidal protrusion by cutting the glass substrate with a focused ion beam,
and fabricated the aperture by coating the protrusion with metal and slicing
the apex with a focused ion beam. They read out a 250-nm line pattern on
a disk rotating at a linear velocity of 0.38 m/s. The data transmission rate
and carrier-to-noise ratio were, respectively, 1.5 Mbps and 37 dB (bandwidth:
10 kHz).
Yatsui et al. [14] have also proposed a planar near-field optical head with
an array of apertures formed on silicon pyramidal protrusions (Fig. 3c). They
fabricated the array of silicon pyramidal protrusions through an isotropic
etching process, and formed the apertures at the apices of the protrusions.
They used a phase-change recording medium, and succeeded in writing and
reading data bits using the apertures. The recording medium was rotated at
a linear velocity of 0.43 m/s, and the minimum mark length and the data
transmission rate were, respectively, 110 nm and 2.0 MHz. Note that parallel
writing and reading using the aperture array is expected to enable a high
data transmission rate.
3.2 Integration of Peripheral Components

into the Near-Field Optical Head
To realize practical recording devices, several optical components – such as

a light source, lens, and detector – must be integrated into a solid near-field
optical head. If such components are separated from each other, servo control
is needed to adjust the positions of these components. However, such control
makes the device very complicated mechanically and electrically. Further-
more, for hybrid recording, integration of a magnetic transducer and mag-
netic sensor is also required.
Partovi et al. [15] have proposed a near-field optical head where an aper-
ture is formed on a metal-coated facet of the laser diode. They demonstrated
writing of recording marks with a size of 250 nm on a phase-change recording
medium. In addition, they succeeded in reading out the recording marks by

detecting the change in the impedance of the laser caused by the reflected
light.
A near-field optical head with subwavelength apertures formed on an ar-
ray of vertical cavity surface emitting lasers (VCSELs) has been proposed by
Goto [16]. A large-scale two-dimensional laser array can be easily fabricated
as the VCSEL array. Therefore, this method is advantageous for parallel writ-
ing and reading of data bits. In addition, low power consumption is expected
because the threshold current is quite low.
Instead of integrating the laser diode into the solid near-field head, light
may be delivered to the head by using an optical fiber fixed on the slider. For
example, Kato et al. [17] have developed a near-field optical head with an
optical fiber fixed horizontally to the slider surface; the light from the optical
fiber is guided to an aperture with a micromirror and microlens formed on
the slider.
3.3 Obtaining High Efficiency

in Generating the Optical Near Field
When a subwavelength aperture is used to generate the optical near field, the
efficiency (defined as the ratio between the power of the optical near field and
that of the incident light) is quite low. For example, the efficiency is 10−4 –
10−5 when the aperture diameter is 60 nm [5], and it decreases further as the
aperture diameter becomes smaller. This is because a cutoff diameter exists
for mode propagation in the waveguide. When the diameter of the tapered
waveguide becomes smaller than the cutoff diameter, the propagating light
suffers a huge loss and its intensity decays exponentially.
This low efficiency is a serious hindrance to realizing practical near-field
optical recording devices. For example, in the case of hybrid recording, the
temperature rise at the heated point in the recording medium is 40◦ C when
a recording medium moving at a linear velocity of 15 m/s is irradiated by
light with a spot size of 50 nm, peak power of 1 mW, and a pulsewidth of
2 ns [18]. Given these values and the output power of available laser diodes,
the required efficiency for writing is 1–10% if we assume that the temperature
rise required for writing is 200◦ C.
To raise efficiency, several methods have been proposed, such as
1. Optimization of the waveguide shape
2. Use of a surface plasmon
3. Use of a metallic waveguide
4. Use of a planar antenna
Low efficiency has also been a problem for SNOM when observing a weak
optical signal, such as photoluminescence, or a nonlinear signal. Although
most of these methods have been developed for SNOM applications, they
100 T. Matsumoto
may also be applied to near-field optical recording. Here, we explain these

methods briefly.
Optimization of the Waveguide Shape
A double- and triple-tapered fiber probe (Fig. 4) have been proposed to

increase the efficiency by, respectively, Saiki et al. [19] and Yatsui et al. [20].
In this type of fiber probe, efficiency is increased by optimizing the waveguide
shape. As mentioned, when light is introduced into the tapered waveguide,
the light power decreases in the loss region where the waveguide diameter is
smaller than the cutoff diameter. In double- and triple-tapered fiber probes,
high efficiency is realized by shortening the length of the loss region. In the
case of the triple-tapered probe, efficiency is also increased because the edge
Fig. 4. Examples of methods to increase efficiency

between the second and third tapers excites the HE-plasmon mode, which
has a smaller cutoff diameter. The efficiency of the double-tapered probe is
10 times higher than that of a conventional probe, and the efficiency of the
triple-tapered probe is 1000 times higher than that of a conventional probe.
Use of a Surface Plasmon
Fischer et al. [21,22] have proposed a tetrahedral probe that uses the surface
plasmon generated on a surface of a metallic film. The surface plasmon is a
collective oscillation of electrons generated on a surface of a thin metallic film
by irradiating the film with light [23]. When the surface plasmon is excited,
the electric field near the metallic field is enhanced. In the tetrahedral probe,
the surface plasmon is excited on the metallic film on the tetrahedral glass
chip, and propagates toward the apex. Thus, the electric field at the apex is
enhanced.
Thio et al. [24] have proposed a structure that consists of a subwavelength
aperture created in a metallic film and a set of concentric circular grooves
surrounding the aperture. When light is introduced into this structure, the
surface plasmon is excited on the metallic film by the circular grooves, and
the optical near-field intensity at the aperture is enhanced because the sur-
face plasmon assists the light transmitting through the aperture. In their
experiment, the transmission of the aperture with the grooves was 50 times
that of an aperture without grooves.
Fischer and Pohl [25] have also proposed a probe that uses a surface
plasmon generated at a subwavelength metallic sphere.
When light is introduced into the subwavelength metallic sphere, the
dipole moment p inside the sphere is given by
ε(ω) − 1
p = 4πε0 a3 E0 , (4)
ε(ω) + 2
where ε(ω) is the dielectric constant of the sphere, ω is the angular frequency
of the light, a is the radius of the sphere, ε0 is the dielectric constant of
vacuum, and E0 is the magnitude of the applied electric field. The field Es
at the surface of the sphere is the superposition of the field generated by the
dipole and the applied field E0 , and is given by
ε(ω)
Es = 3 E0 . (5)
ε(ω) + 2
As shown in this equation, Es is enhanced when the light frequency satisfies
the equation
Re[ε(ω)] = −2 . (6)
As an example, Fig. 5 shows the relation between the wavelength and an

enhancement factor T given by
102 T. Matsumoto
Fig. 5. Relation between the enhancement factor and the wavelength for a gold
sphere calculated by a dipole model
T = |Es /E0 |2 (7)
for a gold sphere. In this calculation, the experimental value given in [34] was
used as the dielectric constant of the gold. As shown in this figure, the field
at the surface became greater when the wavelength was near 500 nm.
The peak corresponds to a resonance condition of the collective oscillation
of the electrons in the metallic sphere. When the sphere is illuminated by
light, charges in the sphere move to the surface (Fig. 6). These charges give
rise to an internal field Ei , hence a restoring force acts on the charges. This
force causes a charge oscillation that has a resonance frequency satisfying
(6). If the light frequency corresponds to the resonance frequency of the
charge oscillation, light energy is strongly absorbed by the sphere and the
optical near field at the surface is enhanced. The quantum of the collective
oscillation of the charges in the small metallic particle is called a localized
surface plasmon or a localized plasmon.
Fig. 6. Schematic of a localized plasmon excited in a metallic particle

In the experiment by Fischer and Pohl, a small polystyrene sphere coated

with gold film was formed on a prism surface, and it was illuminated by light
under a total internal reflection condition to excite the localized plasmon. The
field-enhancement effect by the localized plasmon was observed by placing a
substrate near the particle and detecting the scattered light caused by the
interaction with the substrate.
A fiber probe having a metallic particle at its apex has also been proposed
by Matsumoto et al. [26]. An aperture was formed at the apex of a tapered
optical fiber, and the particle was created at the center of the aperture.
Because the light was introduced and collected through the aperture, the
amount of background light in the detected light was reduced.
Use of a Metallic Waveguide
A coaxial probe and a metallic pin probe have been proposed by, respectively,
Fischer [27] and Takahara et al. [28]. These probes have metallic cores, and
high efficiency is expected because there exists a mode propagating at the
surface of the metallic core, and this mode does not have a cutoff diameter.
Use of a Planar Antenna
Grober et al. [29] have proposed the use of a bow-tie antenna, which has been
used for microwave applications, as an optical near-field generator. When an
electromagnetic wave is incident upon such an antenna, current is induced in
the two arms, and charges accumulate at the apices. These charges work as
a dipole and a strong electromagnetic field is generated at the gap. To check
the feasibility of this approach, they used microwaves with a wavelength of
13.6 cm (2.2 GHz), and used a dipole antenna to observe the electromagnetic
field localized at the gap.
4 Novel Design of a Near-Field Optical Head

Using a Plasmon
To realize a highly efficient near-field optical head, we have developed a near-
field optical head using a plasmon generated by a wedge-shaped metallic
plate [30, 31]. In this section, we describe its principle, simulation results,
and its fabrication results.
4.1 Principle
This head consists of a transparent substrate and a wedge-shaped metallic

plate – a metallic plate in the shape of a sector or a triangle, formed at the
bottom of the substrate (Fig. 7). When the metallic plate is illuminated by
104 T. Matsumoto
Fig. 7. Schematic of the near-field optical head with a wedge-shaped metallic plate
light polarized in the x direction, charges oscillate in the x direction, and

concentrate at the apex. These concentrated charges generate a localized
electromagnetic field: i.e., an optical near field close to the apex of the metallic
plate.
The optical near field is enhanced when the plasmon is excited in the
metallic plate. That is, the displacement of charges creates a depolarization
field inside the metallic plate. Because of this field, the charges act like an
oscillator system as is the case with a metallic sphere. When the light fre-
quency corresponds to the resonance frequency of this oscillation, light energy
is strongly absorbed and the intensity at the apex is enhanced.
This type of near-field optical head has the following advantages:
1. High efficiency
The field-enhancement effect of the localized plasmon enables generation
of a strong optical near field. Although the localized plasmon is also gener-
ated in the metallic sphere, a stronger optical near field can be generated
by using the wedged-shaped metallic plate as described in Sect. 4.4.
2. Simple structure
Because it is a planar structure, the head can be easily fabricated by tech-
niques used in the fabrication of semiconductor devices, such as electron
beam lithography. This is beneficial for integration with other compo-
nents such as the laser diode and the lens. Furthermore, if the metallic
plate is embedded in the slider surface as described in Sect. 4.7, high
durability can be realized because there is no protruding structure.
4.2 Simulation Method
To check the feasibility of the proposed head, the distribution of the optical
near field was calculated through a finite difference time-domain (FDTD)
calculation [32, 33].
In this calculation, the incident light was assumed to be a plane wave.
Experimental values given in [34] were used as the dielectric constants of the
metal. The cell size was made large near the boundary and small near the
object to be calculated to minimize the computer memory required. When
the object was the wedge-shaped metallic plate, the cell size was designed
to have the smallest value near the apex, and the minimum cell size was
1/10 of the radius of the apex. As the boundary condition, Mur’s absorption
boundary condition was used. A parallel computer with 24 processors was
used for the calculation.
4.3 Use of an Aperture and a Circular Metallic Plate
To begin with, for reference we will show the distribution generated near a
subwavelength aperture and a circular metallic plate.
Aperture
Figure 9 shows the intensity distribution of an optical near field generated

near an aperture created in a metallic film as shown in Fig. 8. For this calcu-
lation, we assumed that the metallic film was a gold film with a thickness t
of 100 nm, the diameter a of the aperture was 30 nm, and the wavelength of
the incident light was 650 nm. The polarization of the incident light was in
the direction of the arrows in Fig. 9. The distance d between the observation
plane and the aperture was 5 nm. The unit of intensity was the ratio between
the intensity of the optical near field and that of the incident light. As shown
in this figure, the intensity at the aperture became lower than the incident
light intensity. The intensity at the peak was about 0.04 times that of the
incident light.
Fig. 8. Simulation model for an aperture

106 T. Matsumoto
Fig. 9. Intensity distribution of the optical near field generated near an aperture
created in a metallic film. It was assumed that the metal was gold, the aperture
diameter was 30 nm, and the thickness of the metallic film was 100 nm. The dis-
tribution was calculated on a plane 5 nm from the metallic plate
Fig. 10. Simulation model for a circular metallic plate
Circular Metallic Plate
Figure 11 shows the intensity distribution of the optical near field near a
circular metallic plate as shown in Fig. 10. For this calculation, we assumed
that both the diameter a and the thickness t were 30 nm, the metal was
gold, and the wavelength of the incident light was 650 nm. The polarization
of the incident light was in the direction of the arrows shown in Fig. 11.
The distance d between the observation plane and the aperture was 5 nm.
As shown in this figure, the near-field intensity became high at the edges A
Fig. 11. Intensity distribution of the optical near field generated near a circular
metallic plate. It was assumed that the metal was gold, and the circle diameter and
the metallic film thickness were both 30 nm. The distribution was calculated on a
plane 5 nm from the metallic plate
Fig. 12. Intensity of the optical near-field generated near the circular metallic plate
as a function of the incident light wavelength. The intensity was calculated on a
plane 2 nm from the metallic plate
and A of the metallic plate. This is because the charges in the metallic plate
accumulated at these edges. The intensity at the peak was 3.5 times that of
the incident light. Figure 12 shows the dependence of the optical near-field
108 T. Matsumoto
intensity (intensity at the peak) on the wavelength of the incident light. In

this calculation, the near-field intensity was calculated on a plane 2 nm from
the metallic plate. As shown in this figure, the curve was almost the same
as that of the metallic sphere calculated by (7), and the optical near-field
intensity became high near a wavelength of 550 nm, which corresponds to
the resonance wavelength of the localized plasmon.
4.4 Use of a Wedge-Shaped Metallic Plate Placed in Air
In this subsection, we will present our simulation results for the wedge-shaped
metallic plate placed in air, and describe the fundamental properties of the
optical near field generated near the metallic plate, such as the intensity
distribution, efficiency, and resonance property of the plasmon.
Near-Field Distribution
Figure 14 shows the near-field distribution near the metallic plate in the shape
of a sector as shown in Fig. 13a. For this simulation, we assumed that the
metal was gold, the radius r of the apex was 10 nm, the length l was 150 nm,
the thickness t was 30 nm, the apex angle θ was 60◦ , and the wavelength of
the incident light was 690 nm. The distribution was calculated on a plane
5 nm from the metallic plate. The unit of intensity was the ratio between the
intensity (power density) of the optical near field and that of the incident
light. As shown in this figure, a strong optical near field was generated at
the apex, and the peak intensity was 350 times that of the incident light.
This intensity was 8750 times that of the 30-nm aperture created in the
metallic film, and 85 times that of the circular metallic plate. The distribution
measured at half-maximum was 19 nm wide in the x direction and 25 nm
wide in the y direction.
For this type of near-field optical head, the polarization must be in the
x direction as shown in Fig. 13. Figure 15 shows a near-field distribution
when the polarization of the incident light was in the y direction. As shown
Fig. 13. Simulation model for the wedge-shaped metallic plate: (a) metallic plate
in the shape of a sector, (b) metallic plate in the shape of a triangle, (c) side view
Fig. 14. Intensity distribution of the optical near field generated near a metallic
plate in the shape of a sector. It was assumed that the radius r of the apex was
10 nm, the length l was 150 nm, the thickness t was 30 nm, the apex angle θ was
60◦ , and the incident light wavelength was 690 nm. The unit of intensity was the
ratio between the intensity of the optical near field and that of the incident light
Fig. 15. Intensity distribution of the optical near field when the polarization was
in the y direction
110 T. Matsumoto
in this figure, the optical near field became strong along the side edges of the
metallic plate, but weak at the apex. This is because the charges oscillate in
the direction parallel to the y axis; hence the charges accumulate along the
side edges instead of accumulating at the apex.
plate in the shape of a triangle. It was assumed that the metal was gold, the radius
r of apex A was 10 nm, the radius r of apices B was 30 nm, the apex angle θ was
60◦ , the thickness t was 30 nm and the incident light wavelength was 690 nm. The
distribution was calculated on a plane 5 nm from the metallic plate
Figure 16 shows the near-field distribution when the shape of the metallic
plate was a triangle. For this calculation, we assumed that the metal was
gold, the radius r of apex A was 10 nm, the radius r of apices B was 30 nm,
the apex angle θ was 60◦ , the thickness t was 30 nm and the wavelength of
the incident light was 690 nm. The distribution was calculated on a plane
5 nm from the metallic plate. As shown in this figure, a strong optical near-
field was also generated at apex A as with the metallic plate in the shape of
the sector. The distribution at the apex was the same as that of the metallic
plate in the shape of the sector. Note that the radius r of apices B should
be larger than the radius r of the apex A, otherwise charges also concentrate
at apices B, and a high optical near-field intensity is also generated there.
Dependence of the Spot Size on the Distance

from the Metallic Plate
Because the optical near field is localized near the metallic plate, the size of
the optical spot depends on the distance from the metallic plate d. Figure 17
shows the spot size as a function of the distance from the metallic plate for a
metallic plate in the shape of the sector, and Fig. 18 shows the distributions
for distances of 2, 5, and 10 nm. In Fig. 17, the solid line represents the
width measured at half-maximum in the x direction dx, and the broken line
represents the width measured at half maximum in the y direction dy. As
shown, the spot size decreased as the distance became smaller. For example,
the spot size was dx = 9 nm, dy = 16 nm when the distance was 2 nm, while
it was dx = 28 nm, dy = 30 nm when the distance was 10 nm.
70
60
50
Spot size [nm]
40 dy
30
20 dx
10
0
0 5 10 15 20
Distance [nm]
Fig. 17. Dependence of the spot size on the distance from the metallic plate when
the radius of the apex was 10 nm
Dependence of the Spot Size on the Radius of the Apex
The size of the optical spot also depends on the radius r of the apex. Figure 19
shows the size of the optical spot as a function of the radius of the apex
simulated for the metallic plate in the shape of the sector. The solid line
shows the width measured at half-maximum in the x direction dx, and the
broken line shows the width measured at half-maximum in the y direction
dy. The distribution was calculated on a plane 5 nm from the metallic plate.
As shown, the spot size decreased linearly as the radius of the apex became
smaller.
112 T. Matsumoto
plate in the shape of a sector for a distance of (a) 2 nm, (b) 5 nm, (c) 10 nm
50
40
dy
Spot size [nm]
30
20 dx Fig. 19. Dependence of the spot

size on the radius of the apex.
10 The distance between the ob-
servation plane and the metallic
0
plate was 5 nm
0 5 10 15 20 25
Radius [nm]
Efficiency
Here, we estimate the efficiency η by defining it as

pnear dS
η = S , (8)
p dS
S in
where pnear is the power density (intensity) of the optical near field, pin is
that of the incident light, and S and S are the areas where the power density
is higher than half-maximum. The efficiency calculated for the near-field dis-
tribution in Fig. 14 was about 15% if we assume that the incident beam was
a Gaussian beam with a full width at half-maximum of 1 µm (corresponding
to an optical spot focused by a lens with an N A of 0.35). When we calculated
the near-field distribution, we assumed the incident light was a plane wave.
Therefore, in this estimation, the diameter of the incident beam was assumed
to be 1 µm so that the incident light intensity would be nearly constant at the
metallic plate. However, under such a condition, the energy loss is large be-
cause much of the light is not incident on the metallic plate. Higher efficiency
should be attainable by reducing the diameter of the incident light.
In the case of the aperture, the near-field intensity became lower, and the
efficiency decreased rapidly as the optical spot became smaller. However, for
the head using the metallic plate, the near-field intensity increased when the
spot size became smaller. As a consequence, the efficiency did not decrease
as rapidly as the aperture. The solid line in Fig. 20 shows the intensity as a
function of the apex radius simulated for the metallic plate in the shape of the
sector. The intensity represents the peak intensity in a distribution calculated
on a plane 5 nm from the metallic plate. As shown, the near-field intensity
rose as the apex radius became smaller. This is because the oscillating charges
concentrated in a smaller area, and the charge density increased when the
apex radius became smaller. The broken line in Fig. 20 shows efficiency as a
function of the apex radius. The efficiency was about 10% even if the radius
was 5 nm (corresponding to an optical spot 15 nm by 18 nm).
Tuning of the Resonance Wavelength of the Plasmon
The wavelength of the incident light must correspond to the resonance wave-
length of the plasmon to generate a strong optical near field. However, the
available wavelength is limited if we use a semiconductor laser as the light
source. Therefore, we need to tune the resonance wavelength of the plasmon
to the wavelength of the semiconductor laser. For the wedge-shaped metallic
plate, the resonance wavelength of the plasmon can be adjusted by changing
the material or the length l of the metallic plate. In this subsection, we will
describe how the resonance wavelength depends on these parameters.
114 T. Matsumoto
600 25
500 Efficiency
20
Efficiency [%]
400
Intensity
15
300
Intensity 10
200
5
100
0 0
0 10 20
Radius [nm]
Fig. 20. Dependence of the optical near-field intensity and efficiency on the apex
radius. It was assumed that the length l, the apex angle θ, and the incident light
wavelength were 150 nm, 60◦ , and 690 nm, respectively. The intensity represents the
peak intensity in a distribution measured on a plane 2 nm from the metallic plate.
For the calculation of efficiency, the incident light was assumed to be a Gaussian
beam with a full width at half-maximum of 1 µm
Dependence on the Material
Figure 21 shows the relation between the optical near-field intensity and
the incident light wavelength for metallic plates made of gold, silver, and
aluminum. For this calculation, we assumed the shape of the metallic plate
was a sector with an apex radius r of 20 nm, a length l of 100 nm, a thickness
Fig. 21. Relation between the near-field intensity and the wavelength for metallic
plates made of silver, gold, and aluminum. The shape of the metallic plate was
assumed to be a sector with an apex radius r of 20 nm, a length l of 100 nm, a
thickness t of 30 nm, and an apex angle θ of 60◦ . The closed triangles, closed circles,
and closed squares represent the values for gold, silver, and aluminum, respectively.
The intensity represents the peak intensity in the distribution calculated on a plane
2 nm from the metallic plate
t of 30 nm, and an apex angle θ of 60◦ . The intensity represents the peak
intensity in a distribution calculated on a plane 2 nm from the metallic plate.
The closed triangles, closed circles, and closed squares represent the values for
gold, silver, and aluminum, respectively. The resonance wavelength of gold
was about 650 nm, which is near the wavelengths of a red semiconductor laser
(635 nm, 650 nm, and 670 nm). The resonance wavelength of aluminum was
370 nm, which is near the wavelength of a blue semiconductor laser (405 nm).
In the case of silver, the resonance wavelength was 550 nm. Although, there
is no semiconductor laser of such wavelength, the resonance wavelength can
be shifted toward the red or near-infrared regions by changing the length l
as we explain next.
The near-field intensity at the resonance wavelength became larger in the
order Al < Au < Ag. This indicates that the near-field intensity becomes
large when the imaginary part of the dielectric constant becomes small. This
is because the damping of the charge oscillation, which causes energy loss,
becomes small when the imaginary part of the dielectric constant is small.
To realize high efficiency, we therefore need to use a metal whose dielectric
constant has a small imaginary part.
Dependence on the Length of the Metallic Plate
Figure 22 shows the resonance curves when the length l was varied from
100 nm to 600 nm. For this calculation, we assumed that the shape of the
metallic plate was a sector with an apex radius r of 20 nm, a thickness t of
30 nm, and an apex angle θ of 60◦ , and that the metal was silver. The intensity
Fig. 22. Relation between the near-field intensity and the wavelength for metallic
plates of various lengths. It was assumed that the shape of the metallic plate was
a sector with an apex radius r of 20 nm, a thickness t of 30 nm, and an apex angle
θ of 60◦ , and that the metal was silver. The intensity represents the peak intensity
in the distribution calculated on a plane 2 nm from the metallic plate
116 T. Matsumoto
represents the peak intensity in the distribution calculated on a plane 2 nm

from the metallic plate. As shown, the resonance wavelength of the plasmon
shifted toward a longer wavelength when the metallic plate became longer. We
expected the restoring force acting on the charges to become weaker when the
plate became longer, and so the resonance frequency would decrease. When
the length became greater than 400 nm, a second peak appeared at a short
wavelength. This was because another resonance mode was excited at this
peak. The broadening of the resonance curve when the length increases was
probably due to the radiation damping [35].
4.5 Use of a Wedge-Shaped Metallic Plate

Placed near the Recording Medium
In the previous calculation, the metallic plate was placed in air. However, in
recording devices, the metallic plate is formed on the slider (a transparent
substrate), and placed near the recording medium. In this subsection, we will
describe the resonance property of the plasmon and the near-field distribution
in such a situation.
Simulation Model
To realize high durability, the metallic plate should be embedded in the slider
surface as shown in Fig. 23. Thus, we assumed that the metallic plate was
embedded in a glass substrate with a refractive index of 1.45. The recording
medium was assumed to be a thin film of TbFeCo or GeSbTe with a thickness
of 20 nm, and the separation between the recording medium and the metallic
plate was assumed to be 10 nm. For the metallic plate, we assumed that the
metal was gold, the apex radius r was 10 nm, the length l was 150 nm, the
thickness t was 30 nm, and the apex angle θ was 60◦ .
Fig. 23. Simulation model when the recording medium and slider were placed near
the metallic plate: (a) when there was no recording medium, (b) when there was a
recording medium
Resonance Property of the Plasmon
First, we show the influence of the glass substrate on the resonance prop-
erty of the plasmon. Figure 24 shows the resonance curve when there was
no recording medium. The solid line represents the resonance curve for the
metallic plate embedded in the glass substrate and the broken line represents
the resonance curve when the metallic plate was placed in air. The intensity
represents the peak intensity in the distribution calculated on a plane 2 nm
from the metallic plate. As shown, the resonance wavelength shifted by about
50 nm toward a longer wavelength when the metallic plate was embedded in
the glass substrate. We believe this was because the dipoles in the dielec-
tric substrate interact with the charges in the metallic plate, and weaken the
restoring force working on the charges.
Figure 25 shows the resonance curves when the recording medium was
placed near the metallic plate embedded in the glass substrate. The solid line
represents the resonance curve when the TbFeCo recording medium was used,
the broken line represents the resonance curve when the GeSbTe medium
was used, and the dash-dot line represents the resonance curve when there
was no recording medium. The intensity represents the peak intensity in the
distribution calculated on a plane 2 nm from the metallic plate. As shown, the
resonance wavelength was shifted toward a longer wavelength by interaction
with the recording medium. The shift was about 50 nm for both TbFeCo and
GeSbTe.
Fig. 24. Relation between the near-field intensity and the wavelength for the metal-
lic plates embedded in a glass substrate. The solid line represents the resonance
curve for the metallic plate embedded in the glass substrate and the broken line
represents the resonance curve when the metallic plate was placed in air. The in-
tensity represents the peak intensity in the distribution calculated on a plane 2 nm
from the metallic plate
118 T. Matsumoto
Fig. 25. Relation between the near-field intensity and the wavelength when a
recording medium was placed near the metallic plate. The solid line represents the
resonance curve for a TbFeCo recording medium, the broken line represents the res-
onance curve for a GeSbTe recording medium, and the dash-dot line represents the
resonance curve when there was no recording medium. The separation between the
recording medium and the metallic plate was assumed to be 10 nm. The intensity
represents the peak intensity in the distribution calculated on a plane 2 nm from
the metallic plate
Near-Field Distribution
Figure 26 shows the intensity distribution of the optical near-field when there
was no recording medium, and Fig. 27 shows the intensity distribution when
the TbFeCo recording medium was placed near the metallic plate. The wave-
length of the incident light was assumed to be at the resonance wavelength
of the plasmon – 780 nm when there was no recording medium, and 830 nm
when there was a recording medium. Both distributions were calculated on a
plane 10 nm from the metallic plate (on the surface of the recording medium
when a recording medium was used). As shown in these figures, the opti-
cal spot was semicircular when the recording medium was used, but was
crescent-shaped when there was no recording medium. With the recording
medium, the distribution measured at half-maximum was about 30 nm wide
in both the x and y directions. If we compare this spot size with that of a 4.7-
GB, DVD-RAM (full width at half-maximum of the optical spot = 580 nm,
recording density = 3.7 Gb/in2 ), this spot size corresponds to a recording
density of approximately 1 Tb/in2 . The calculated efficiency for this distri-
bution was 20% when we assumed that the incident beam was a Gaussian
beam with a full width at half-maximum of 1 µm.
When we placed the recording medium near the metallic plate, the in-
tensity of the optical near field increased by a factor of about 2.5. This was
because the decay length of the optical near field changes when the recording
Fig. 26. Intensity distribu-

tion of the optical near field
generated near the metallic
plate embedded in a glass
substrate. The distribution
was calculated on a plane
10 nm from the metallic plate
Fig. 27. Intensity distribu-

tion of the optical near field
when a TbFeCo recording
medium was placed near the
metallic plate. The distri-
bution was calculated on a
plane 10 nm from the metallic
plate (on the surface of the
recording medium)
120 T. Matsumoto
Fig. 28. Intensity of the optical near field as a function of the distance from the
metallic plate. The solid line represents the curve when a recording medium was
used, and the broken line represents the curve when there was no recording medium
medium is placed near the metallic plate. Figure 28 shows the dependence of
the near-field intensity on the distance from the surface of the metallic plate.
The solid line represents the curve when a recording medium was used, and
the broken line represents the curve when there was no recording medium.
The near-field intensity decreased exponentially as the distance increased,
and its decay length increased when the recording medium was placed near
the metallic plate. We believe this was because image charges were induced in
the recording medium, and the intensity of the optical near field between the
head and the recording medium became greater because of the interaction of
the charges induced at both the apex of the metallic plate and the recording
medium.
4.6 Near-Field Optical Head Using Two Metallic Plates
When a wedge-shaped metallic plate is used, the spot size can be further
decreased by reducing the apex radius or the spacing between the head and
the recording medium. An alternative way to decrease the spot size is to use
two metallic plates (Fig. 29). When the two metallic plates are illuminated by
light, an optical near field is generated between the two apices because of the
interaction of charges concentrated at these apices. This mechanism is similar
to that of the bow-tie antenna used at microwave frequencies. However, at
optical frequencies, the intensity of the optical near field can be enhanced
when the plasmon is excited in the metallic plates by optimizing the material
and the size of the metallic plate.
Figure 30 shows the intensity distribution of the optical near field gener-
ated near the two metallic plates. For this calculation, we assumed that the
shape of the metallic plates was a sector, the metal was silver, the gap width
Fig. 29. Schematic of the double metallic plate
Fig. 30. Intensity distribution of the optical near field generated near the double
metallic plate. It was assumed that the metallic plates had a sector shape, the metal
was silver, the gap width g was 5 nm, the apex radius r was 20 nm, the length l was
200 nm, the thickness t was 30 nm, the apex angle θ was 60◦ , and the wavelength
was 830 nm. The distribution was calculated on a plane 2 nm from the metallic
plates
g was 5 nm, the radius r of each apex was 20 nm, the length l was 200 nm,
the thickness t was 30 nm, the apex angle θ was 60◦ , and the wavelength was
830 nm. The distribution was calculated on a plane 2 nm from the metallic
plates. As shown, a strong optical near field was generated at the gap. The
peak intensity was about 2000 times that of the incident light. The distribu-
122 T. Matsumoto
Fig. 31. Relation between the near-field intensity and the wavelength for the double
metallic plates with a gap width of 24 nm, 16 nm, and 8 nm
tion measured at half-maximum was 5 nm by 5 nm, which was the same as

the gap width.
Figure 31 shows the intensity of the optical near field as a function of
the incident light wavelength when the gap width of two metallic plates was
varied from 24 nm to 8 nm. For this calculation, we assumed that the metal
was silver, the length l was 100 nm, the radius r of each apex was 20 nm, the
thickness t was 30 nm, the apex angle θ was 60◦ . As shown, the intensity of
the optical near field increased when the plasmon was excited in the metallic
plates, as was the case with a single metallic plate. The resonance wavelength
shifted toward a longer wavelength, and the intensity rose as the gap width
became smaller because of the interaction of plasmons generated on both
sides of the metallic plates.
4.7 Fabrication of a Near-Field Optical Head

with a Wedge-Shaped Metallic Plate
As described in Sect. 4.1, a near-field optical head with a metallic plate can
be easily fabricated by using electron beam lithography. Figure 32 shows an
example of fabrication process. In this process, the head is fabricated using a
sacrificial substrate to embed the metallic plate in the slider. First, an array of
metallic plates is fabricated, using electron beam lithography, on a sacrificial
substrate coated with a sacrificial film (a thin metallic film) in steps (a) to
(c). Secondly, a photoresist pattern is formed on the sacrificial substrate to
make pads near the metallic plates, and a dielectric film (Al2 O3 ) is formed
on the substrate by sputtering. Thirdly, a glass substrate was glued to the
Fig. 32. Fabrication process of a near-field optical head with a wedge-shaped

metallic plate: (a) Coating resist and exposure with a electron beam exposure ma-
chine, (b) coating metal, (c) removing resist, (d) coating photoresist, (e) exposure,
(f ) coating dielectric film, (g) gluing a glass substrate onto the structure, and (h)
removing the sacrificial substrate
Fig. 33. SEM image of the metallic plates: (a) one metallic plate, (b) two metallic
plates
top of the structure, and then the photoresist and the sacrificial substrate are
removed. Figure 33 shows a scanning electron microscope image of metallic
plates fabricated by the method described above. In this fabrication, gold
was used as the metal, and the thickness was 30 nm. As shown in this figure,
the radius of the apex was about 10 nm, and the gap width was 20 nm for
the two metallic plates. Based on our simulation results, a recording density
of about 1 Tb/in2 should be achievable by using the fabricated near-field
optical head.
124 T. Matsumoto
5 Summary
Near-field optical recording is a promising way to realize a recording density
of over 1 Tb/in2 . In this chapter, we focused on the near-field optical head,
which is a key device for near-field optical recording. First, we explained the
technical issues regarding the near-field optical head and introduced some
solutions to these issues. We focused on a highly efficient near-field optical
head that uses a wedge-shaped metallic plate, and described its optical prop-
erties based on a simulation using a finite-difference time-domain method.
The simulation results confirmed that a strong optical near field is generated
at the apex of the metallic plate when a plasmon is excited in the metallic
plate. When a TbFeCo recording medium was placed 10 nm from the near-
field optical head, the size of the optical spot was 30 nm, which corresponds
to an areal recording density of approximately 1 Tb/in2 . The efficiency was
20% if we assume that the incident beam was a Gaussian beam with a full
width at half-maximum of 1 µm. Furthermore, we discussed an optical head
using two metallic plates. We confirmed through our simulation that a highly
localized optical near field was generated at the gap when the plasmon was
excited in the metallic plates. The distribution was 5 nm by 5 nm when the
two apices were separated by 5 nm. These results indicate that the recording-
density limit now being encountered in conventional recording devices can be
overcome with this type of near-field optical head.
To realize such advanced recording devices, development of a recording
medium suitable for near-field optical recording is also important. For in-
stance, in the case of magnetic recording medium, a recording medium with
a small grain size and thin cover layer will be needed, and tuning of the mag-
netic and thermal characteristics will be necessary. High-precision tracking
servo technology will also be required. For example, a track pitch of less than
40 nm is required for a recording density of more than 1 Tb/in2 ; that of the
current Blu-ray disk is 320 nm. We believe that these technologies can be
realized on the basis of technology that has been developed for conventional
optical or magnetic recording devices. However, various technical challenges
will have to be overcome to realize such technology because the required size
and precision are less than 1/10 those of today’s conventional technology.
Acknowledgements
The author would like to thank H. Sukeda, T. Shimano, M. Kiguchi, H. Saga,

T. Shintani, Y. Anzai and K. Ishikawa of the Research and Development
Group, Hitachi Ltd. for their helpful discussions. This work was supported
by the “Terabyte optical storage technology” project, which the Optoelec-
tronic Industry and Technology Development Association contracted with
The Ministry of Economy, Trade, and Industry of Japan (METI) in 2002
based on funds provided by METI.
References
1. Near-field Nano/Atom Optics and Technology, ed. by M. Ohtsu (Springer,
Tokyo 1998)
2. S.M. Mansfield, G. S. Kino: Appl. Phys. Lett. 57, 2615 (1990)
3. B.D. Terris, H.J. Mamin, D. Rugar, W.R. Studenmund, G.S. Kino:
Appl. Phys. Lett. 65, 388 (1994)
4. M. Shinoda, K. Saito, T. Ishimoto, T. Kondo, A. Nakaoki, M. Furuki,
M. Takeda, M. Yamamoto: In: Technical Digest of Optical Data Storage 2003
(Canada 2003) 120
5. E. Betzig, J.K. Trautman, R. Wolfe, E.M. Gyorgy, P.L. Finn, M.H. Kryder,
C.-H. Chang: Appl. Phys. Lett. 61, 142 (1992)
6. S. Hosaka, T. Shintani, M. Miyamoto, A. Kikukawa, A. Hirtsune, M. Terao,
M. Yoshida, K. Fujita, S. Krammer: J. Appl. Phys. 79, 8082 (1996)
7. S. Jiang, J. Ichihashi, H. Monobe, M. Fujihira, M. Ohtsu: Opt. Commun. 106,
173 (1994)
8. H. Saga, H. Nemoto, H. Sukeda, M. Takahashi: Jpn. J. Appl. Phys. 38, Part
1, 1839 (1999)
9. H. Sukeda, H. Saga, H. Nemoto, Y. Ito, C. Haginoya: IEEE Trans. Magn. 37,
1234 (2001)
10. M. Vasz-Iravani, R. Toledo-Crow, Y. Chen: J. Vac. Sci. Technol. A 11, 742
(1993)
11. E. Betzig, P.L. Finn, J.S. Weiner: Appl. Phys. Lett. 60, 2484 (1995)
12. M.B. Lee, M. Kourogi, T. Yatsui, K. Tsutsui, N. Atoda, M. Ohtsu: Appl. Opt.
38, 3566 (1999)
13. F. Isshiki, K. Ito, K. Etoh, S. Hosaka: Appl. Phys. Lett. 76, 804 (2000)
14. T. Yatsui, M. Kourogi, K. Tsutsui, J. Takahashi, M. Ohtsu: Opt. Lett. 25, 1279
(2000)
15. A. Partovi, D. Peale, M. Wutting, C.A. Murray, G. Zydzik, L. Hopkins, K. Bald-
win. W.S. Hobson, J. Wynn, J. Lopata, L. Dhar, R. Chichester, J.H.-J. Yeh:
Appl. Phys. Lett. 75, 1515 (1999)
16. K. Goto: Jpn. J. Appl. Phys. 37, Part 1, 2274 (1998)
17. K. Kato, S. Ichihara, M. Oumi, H. Maeda, T. Niwa, Y. Mitsuoka, K. Nakajima,
T. Ohkubo, K. Itao: Jpn. J. Appl. Phys. 42, Part 1, 5102 (2003)
18. T.W. McDaniel, W.A. Challener: In: Technical Digest of MORIS’02 (France
2002) 117
2612 (1996)
20. T. Yatsui, M. Kourogi, M. Ohtsu: Appl. Phys. Lett. 73, 2090 (1998)
21. U.C. Fischer et al.: In: Near-Field Optics ed. by D.W. Pohl, D. Courjon (Kluwer
Academic, Dordrecht 1993) 255
22. J. Koglin, U.C. Fischer, H. Fuchs: Phys. Rev. B 55, 7977 (1997)
23. H. Raether: Surface Plasmon on Smooth and Rough Surface and on Grating
(Springer-Verlag, Berlin 1988)
24. T. Thio, K.M. Pellerin, R.A. Linke, H.J. Lezec, T.W. Ebbesen: Opt. Lett. 26,
1972 (2001)
25. U.C. Fischer, D.W. Pohl: Phys. Rev. Lett. 62, 458 (1989)
26. T. Matsumoto, T. Ichimura, T. Yatsui, M. Kourogi, T. Saiki, M. Ohtsu:
Opt. Rev. 5, 369 (1998)
126 T. Matsumoto
27. U.C. Fischer: Ultramicroscopy 42, 393 (1992)

28. J. Takahara, S. Yamagishi, H. Taki, A. Morimoto, T. Kobayashi: Opt. Lett.
22, 475 (1997)
29. R. Grober, R. Schoelkopf, D. Prober: Appl. Phys. Lett. 70, 1354 (1997)
30. T. Matsumoto, T. Shimano, S. Hosaka: In: Technical digest of 6th international
conference on near field optics and related techniques (Netherlands 2000) 55
31. T. Matsumoto, T. Shimano, H. Saga, H. Sukeda, M. Kiguchi: J. Appl. Phys.
95, 3901 (2004)
32. J.B. Judkins, R.W. Ziolkwski: J. Opt. Soc. Am. A 12, 1974 (1995)
33. H. Furukawa, S. Kawata: Opt. Commun. 132, 170 (1996)
34. Handbook of Optical Constants of Solid, ed. by E.D. Palik (Academic Press,
Orlando 1985)
35. A. Wokaun, J.P. Gordon, P.F. Liao: Phys. Rev. Lett, 48, 957 (1982)
Nano-Optical Media
for Ultrahigh-Density Storage
K. Naito, H. Hieda, T. Ishino, K. Tanaka, M. Sakurai, Y. Kamata,

S. Morita, A. Kikitsu, and K. Asakawa
1 Introduction
The data storage density is advancing because of the requirement of infor-
mation data growth as shown in Fig. 1. A 1 Tb/inch2 density will be re-
quired around 2010. Physical limits, however, are predicted both for magnetic
recording and for optical means. To overcome the optical limit (diffraction
limit), near-field optics is required [1].
The magnetic recording density is currently increasing at rates of up to
100 % per year. To obtain the higher densities, grain sizes of the conventional
continuous magnetic films should be reduced to maintain a necessary signal-
to-noise ratio. The small grain sizes, however, reduce the thermal stability
of the magnetization of each bit. This, referred to as superparamagnetism,
can be overcome by increasing KuV /kT , where Ku is an energy barrier to
reversal per grain volume (anisotropic energy), V is a volume per grain, k
is the Boltzmann constant, and T is temperature. High Ku materials such
Fig. 1. Roadmap of storage technologies

128 K. Naito et al.
Fig. 2. Schematic explanation of hybrid recording
as FePt are preferable to overcome the thermal fluctuation. Higher magnetic

fields are required to write the signals on the high-Ku media. Magnetic fields
generated by the thin-film magnetic head have almost reached their limits.
Hybrid recording, consisting of near-field light and a magnetic field, is one of
the promising candidates to solve the paradox as shown in Fig. 2.
Magnetic patterned media, which consist of islands of magnetic mate-
rial surrounded by a non magnetic matrix as shown in Fig. 3, increase an
effective V resulting in bit thermal stability [2]. Magnetic patterned media
can decrease media transition noise, track edge noise and nonlinear bit shift
effects. Additionally, precise servo information can be embedded in the pat-
terned media. This is very advantageous for narrower track widths of future
HDDs [3]. Ultrahigh density beyond 1 Tb/inch2 should be obtained by the
combination of hybrid recording and the patterned media.
Some serious problems must be resolved such as the high cost of nano-
patterning and write synchronization [4]. Magnetic patterned media have
been prepared previously, for example, by electron-beam or focused-ion-beam
patterning [5], ion-beam modification [6], interferometric lithography [7],
nanoimprint lithography [8], self-assembling template lithography [9] and
plating in anodized aluminum pores [10]. The latter four methods can produce
patterned media with a large area. To our knowledge, however, there is no
example of preparing circumferential patterned media necessary for HDDs.
In this chapter, we show circumferential magnetic patterned media, pre-
pared by an artificially assisted (or aligned) self-assembling (AASA) method
[11], which includes simple nanopatterning using a nanoimprint and fine
nanopatterning using self-assembling diblock polymers [12]. The fine nanopat-
terns created by the self-assembling method can be aligned in a large area by
the artificial nanoimprint.
We have aimed at a novel write/read principle media on the nm scale:
organic-dye-patterned media. In the previous near-field optical storage demon-
strations, conventional recording media such as a magnetic-optical recording
(MO) medium [13], a phase-change (PC) medium [14], or photochromic ma-
terials [15, 16] have been used. Using an MO medium and a PC medium,
Nano-Optical Media for Ultrahigh-Density Storage 129
Fig. 3. Schematic explanation of patterned media
signal contrasts are 1% and 10%, respectively. In future near-field storage de-
vices, signal contrast will be insufficient to achieve practical read-out speeds.
As for photochromic materials, although better signal contrast is expected,
read out without destroying the recorded information has been a major issue.
We showed that amorphous organic dye molecules formed droplet-like nm-
scale structures and were aligned by the AASA method [17]. It is possible to
inject and confine charges into a single dot using AFM [18, 19]. Fluorescence
changes in dye molecules associated with charge injection into their thin films
was observed [20]. Fluorescence measurement has such a high signal-to-noise
ratio that it can be observed even in a single molecule. Therefore, read out
using fluorescence is expected to achieve high-speed scanning. Furthermore,
electrical writing, namely charge confinement, will prevent record rupture at
read out.
2 Magnetic Patterned Media

2.1 Preparation
Figure 4 shows a schematic explanation of the preparation method. The self-

assembly method has attracted much attention because of its ability to pro-
duce ordered fine nanopatterns with a large area. We adopted diblock copoly-
Fig. 4. Scheme of the preparation method of patterned media

130 K. Naito et al.
Fig. 5. Schematic explanation of phase separation of diblock copolymer
Fig. 6. Type of phase separation for diblock coplymer
mers as the self-assembly system. Figure 5 shows the explanation of phase-

separation of diblock coplymers. There are several types of phase-separated
states as shown in Fig. 6.
The diblock copolymer molecule used here consists of a hydrophobic
polystyrene chain and a hydrophilic polymethylmethacrylate chain [21].
These chains are not mixed resulting in a phase-separation state. The
dot diameter and the dot pitch are determined by the molecular weights.
For the diblock copolymer [polymethylmethacrylate (Mw 41 500)-polystyrene
(Mw 17 200)], 40 nm diameter and 80 nm pitch were obtained. Figure 7 shows
the dot patterns of the diblock copolymers on a flat surface. A polycrys-
talline structure was observed for the phase-separation state with polymethyl-
methacrylate dots in a polystyrene surrounding.
Figure 8 shows the consept of the artificially assisted (or aligned) self-
assembly (AASA) method. Phase separation takes place in the grooves to
produce aligned dot patterns.
Fig. 7. Phase-separation dot patterns of the diblockpolymer on a flat Si substrate
Fig. 8. Concept of AASA method
Fig. 9. The whole image of the Ni master
A Ni master disk possessing spiral patterns with 60-, 100-, 150-, 200- and
250-nm width lands and a 400-nm width groove (110-nm depth) was pressed
to a resist film on a CoCrPt film to transfer the spiral patterns at room
temperature and at a pressure of 1000 bar. The master disk was prepared by
the photoprocess as shown in Fig. 9.
132 K. Naito et al.
Fig. 10. AFM images of spiral patterns of the Ni master (upper images) and of
the resist film (lower images). The lengths are the land widths of the master disk
Fig. 11. Phase-separation dot patterns of the diblock copolymers in grooves with
different widths: (a) 60 nm, (b) 150 nm, (c) 200 nm, (d) 250 nm
The AFM images of spiral patterns transferred to the resist film are shown
in Fig. 10. The land-groove spiral patterns were precisely transferred.
A diblock copolymer solution was cast into the obtained grooves, and then
annealed to prepare self-assembling dot structures aligned along the grooves.
Figure 11 indicates dot structures in grooves with different widths. One to
four dot lines could be obtained by changing the groove width. The groove
width should be controlled to obtain the regularly aligned dot lines along the
grooves. In the figure, the groove widths were considered rather suitable for
the two and four dot lines. Figure 12 shows the phase-separation dot patterns
of another diblock copolymer in grooves. The dot size was 15 nm and the dot
pitch was 35 nm. This small dots were obtained by using the low molecular
weight diblock copolymer.
The PMMA dots were selectively removed by the oxygen plasma treat-
ment [12]. The resulting holes were filled by spin-on-glass (SOG). The lower
magnetic films were patterned by ion milling using the SOG dots as a mask.
Fig. 12. Phase-separation dot patterns of another diblock copolymers in grooves
Fig. 13. (a) SEM image of the patterned magnetic media with a 40-nm diameter,
(b) the whole image of the patterned media disk prepared on a 2.5-inch HDD glass
plate
Figure 13a shows the SEM image of the patterned magnetic media with
40 nm diameter. There are fluctuations in size and in position of the magnetic
dots. The irregularities are probably caused by the diblock copolymer dot
disorder and by the nanopatterning process fluctuation. The whole image of
the patterned media disk prepared on a 2.5-inch diameter HDD glass plate
is shown in Fig. 13b. Interference colors based on the wide groove lines were
observed. Figure 14 shows the TEM image of the patterned dot with about
40 nm diameter. This image indicates that the magnetic dot is completely
separated.
134 K. Naito et al.
Fig. 14. TEM image of the patterned CoPt dot
Fig. 15. Schematic explanation of ordering of FePt alloy
FePt is an ordered alloy that has a high Ku. High temperature is required
to obtain the ordered structure form the sputtered film as shown in Fig. 15.
We found that addition of Cu decreases the temperature [22].
Using an electron-beam mastering apparatus recently developed will en-
able reduction of the groove width to increase the density of magnetic cells.
It may be impossible to produce circumferential dot patterns with a 40 nm
diameter like those in the figure even by the present electron beam apparatus.
We used the AASA method to prepare the etching mask for the magnetic
films. For cheaper preparation, we prepared the Ni master disk itself by the
AASA method. The details will be shown elsewhere.
2.2 Magnetic Properties
Figure 16 shows the M –H loops of the raw continuous film (CoCrPt) and
the patterned media. A coercive force and squareness ratio of the patterned
film increased compared to the continuous film. The magnetization of the
patterned media decreased due to the decrement of the material volume.
The squareness ratio was almost unity.
Figure 17 indicates the MFM images of the patterned magnetic media
after AC erase and after DC erase. Three magnetic dot lines are observed
corresponding to Fig. 11c and Fig. 13a. Bright and dark dots are randomly
Fig. 16. M –H loops of the raw continuous film (CoCrPt) and the patterned media
Fig. 17. MFM images of the patterned media
distributed after AC erase. The DC-erased dots are observed to be magne-

tized in one direction in Fig. 17b. Single magnetic domains with an almost
perpendicular orientation were confirmed in each magnetic dot.
136 K. Naito et al.
3 Organic-Dye-Patterned Media
3.1 Preparation
The organic molecule mainly used here was TTPAE or TPD, shown in Fig. 18.
These molecules give amorphous films as a hole transport layer in organic EL
devices [23]. The films were formed by vacuum evaporation onto thermally
oxidized p-Si substrates with a resistivity of 0.001–0.01 Ω cm. The thickness
of the SiO2 layer was 20 nm. The electrodes were fabricated on the back
of the Si substrates by evaporating Cr and Au. Before the evaporation of
molecules, the SiO2 /Si substrates were made hydrophobic by treatment in
an atmosphere of hexamethyldisilazane.
Figure 19 shows the contact-mode topographic image of the TTPAE film
observed using an AFM. It is clearly seen that the film is constructed of
spatially isolated droplet-like domain structures. The size and the density of
domains were varied according to the amount of deposition. The diameter,
height and density were varied from 20 to 300 nm, from 10 to 100 nm, and
from 1011 to 109 cm−2 , respectively. When the amount of deposition was
Fig. 18. Molecular structures of TTPAE and TPD
Fig. 19. AFM image of TTPAE films

increased to the film thickness of more than 40 nm, the film structure changed
into a continuous layer with a flat surface. We verified that thin films of several
amorphous molecules, including TPD, also showed similar domain structures.
The observed uniform droplet-like domain structures suggest that the
film growth is based on the Volmer–Weber mechanism. In this mechanism,
growth characteristics are known to depend strongly on the surface energies
of the film and of the substrate. When the films were formed onto hydrophilic
SiO2 /Si substrates, it was found that the domain size was much larger than
films on the hydrophobic substrates.
The droplet size and alignment could be controlled by the AASA method.
The hole arrays were prepared on polymethylmethacrylate film by using
AFM. The pitch was 30 nm. TPD molecules were evaporated onto the ar-
rays. TPD amorphous dots were selectively formed on the holes with the same
size as shown in Fig. 20. Using a nanoimprint method should produce the
large-area hole arrays to obtain the large-area organic-dye-patterned media
as shown in Fig. 21.
Fig. 20. TPD dot patterns prepared by AASA method
Fig. 21. Schematic explanation of TPD dot

alignment
138 K. Naito et al.
3.2 Electrical and Optical Measurements
Figure 22 shows a schematic diagram of the structure of the device used in this
chapter. To apply voltage to the organic films, Au films of 10 nm thickness
were fabricated on the substrates following the deposition of organic films. Au
films of 10 nm were selected as the thinnest providing good electrical contact
and sufficient fluorescence intensity from the underlying organic films.
Scanning of the applied voltage started at 0 V. Fluorescence measure-
ments were first obtained in the positive-voltage regions, and then in the
negative regions. The voltage scan speed was 0.02 Hz. All fluorescence mea-
surements were carried out in air and at room temperature.
In order to investigate the electrical properties of TTPAE films, displace-
ment-current measurements were performed. Since details of displacement-
current measurements were described in [24], only distinctive features are
shown here. Figure 23 shows a schematic diagram of the displacement-current
measurement. The measured current is expressed by the capacitance and scan
speed of the applied voltage. In the simplest case, the organic film behaves
as a dielectric and the measured current reflects the capacitance of the entire
dielectric film, that is, organic and SiO2 (CSiO2 +Org ). As for the donor or
acceptor molecules, carrier injection or ejection through metal/org interfaces
Fig. 22. Schematic diagram of device structure and fluorescence measurement
Fig. 23. Schematic diagram of displacement-current measurement

occur. If carrier injection through the metal/org interface occurs, the mea-
sured current will reflect the capacitance of the SiO2 (CSiO2 ) film. Therefore,
the displacement current is sensitive to carrier injection into the organic film.
Figure 24 shows the applied-voltage dependence of fluorescence in the
TTPAE film. These data were obtained in the first scan of voltage for positive
and negative voltages, respectively. As shown later, large hysteresis effects
were observed in fluorescence quenching. As shown in Fig. 24, the fluorescence
intensity decreases as applied voltage V increases at positive voltages. On the
other hand, at negative voltages, no variations with V are observed. Changes
in the fluorescence spectrum and excitation spectrum are shown in Fig. 25. It
Fig. 24. Applied-voltage dependence of fluorescence (PL) of Au/TTPAE/SiO2 /Si

with TTPAE thickness of 15 nm and SiO2 thickness of 100 nm. Excitation wave-
length: 360 nm. Fluorescence wavelength: 540 nm
Fig. 25. Spectra of Au/TTPAE/SiO2 /Si with TTPAE thickness of 15 nm and

SiO2 thickness of 100 nm: (a) fluorescence spectra with bias voltage of 0 V, exci-
tation wavelength: 360 nm; (b) fluorescence spectra with bias voltage of +50 V,
excitation wavelength: 360 nm; (c) excitation spectra with bias voltage of 0 V, flu-
orescence wavelength of 540 nm; (d) excitation spectra with bias voltage of +50 V,
fluorescence wavelength of 540 nm
140 K. Naito et al.
Fig. 26. Displacement-current diagram of Au/TTPAE/SiO2 /Si with TTPAE

thickness of 50 nm and SiO2 thickness of 50 nm under white-light irradiation
was found that the observed fluorescence quenching did not follow any shift
of spectral peaks.
Figure 26 shows the displacement-current characteristic of a TTPAE film
under visible light irradiation. At positive bias voltages, hole injection and
ejection were clearly observed, and at negative bias voltages, no carrier injec-
tion was observed. These findings are consistent with the fact that TTPAE
is an electron-donor material. Figure 27 shows a comparison between fluores-
cence measurements and displacement-current measurements under UV ir-
radiation. Since both fluorescence variations and displacement-currents show
large hysteresis especially at negative bias voltages, the results for the first
cycle of voltage are shown separately from those for the second and third cy-
cles. At positive bias voltages, fluorescence and current into the TTPAE film
were observed to vary in the same manner. Both fluorescence variations and
displacement currents are reproducible at positive bias voltages. On the other
hand, at negative bias voltages, large hysteretic variations were observed in
both measurements. On the way to −25 V in the first cycle, no fluorescence
change or negative current into the TTPAE film was observed. On the way
back to 0 V, a positive current and fluorescence quenching were observed
at the same voltage. In the second or third cycle, two current peaks were
observed on the way to −25 V and one peak was observed on the way back
to 0 V. Fluorescence quenching was observed at lower voltages where current
was observed on the way to −25 V, and at the same voltage as the current
observed on the way back to 0 V.
Figure 28 shows the fluorescence-quenching efficiency for various thick-
nesses of TTPAE films. The fluorescence-quenching efficiency increased as
the film thickness decreased and had a peak value at a thickness of 15 nm.
In the case of films thicker than 15 nm (c), uniform film structures were
Fig. 27. Comparison between

applied-voltage dependence of
fluorescence and displacement-
current diagram under UV
irradiation (excitation wave-
length: 360 nm). Sample was
Au/TTPAE/SiO2 /Si with
TTPAE thickness of 50 nm
and SiO2 thickness of 50 nm.
(a) first scan of voltage. (b)
second (solid line) and third
scans (dashed line) of voltage
142 K. Naito et al.
Fig. 28. TTPAE film thickness dependence of fluorescence quenching efficiency.

Sample was Au/TTPAE/SiO2 /Si with SiO2 thickness of 100 nm. Excitation wave-
length: 360 nm. Fluorescence wavelength: 540 nm
Fig. 29. NSOM images of fluorescence-intensity reduction caused by charge injec-

tion
observed. On the other hand, in the case of thinner films, separated film
structures (b) or isolated dot structures (a) were observed. As for the sepa-
rated structures or the dot structures, the effective thickness of the film was
found to be thicker than the case for uniform films.
We directly observed fluorescence quenching for the TTPAE dots by a
near-field optical microscope. Charge injection is carried out through contact
electrification by a gold-coated AFM tip with a positive bias. The charges
were found to be very stable over a long period. This phenomena will be
the same as in a flash silicon memory, where electrons are stored in a float-
ing gate. Injected holes can be extracted by a negatively biased tip. Holes
were injected to diamine dots at 3 V. These holes were extracted at −2 V.
Therefore, the organic dot structure can be operated as a rewritable memory.

However, it takes a time to detect surface potential values. Optical read-out
procedures are required for high-speed scanning. Figure 29 shows that fluo-
rescence intensity is reduced for the dot where the charge was injected.
We hypothesize that Coulomb interaction between photoexcited molecules
and carriers (holes), as suggested by Deussen et al. [25] in an earlier explana-
tion of the effect of current flow on fluorescence quenching, plays an essential
role in the fluorescence quenching described here.
4 Conclusion
We have described the circumferential magnetic patterned media with a 40-
nm diameter on a 2.5-inch diameter glass plate. The media were prepared
by an artificially assisted self-assembling (AASA) method, which includes
simple nanopatterning of land-groove spirals using a nanoimprint and self-
assembling of the diblock polymer aligned in the grooves. Magnetic films were
etched by ion milling using the nanodot structures of the diblock copolymer
as a mask. Fluctuations of size and position of the patterned cell, observed for
the present media, should be reduced for the realization of patterned media.
We have also investigated organic-dye-patterned media prepared by the
AASA method. Fluorescence quenching of TTPAE dots with charge injection
applying bias voltages. Displacement-current measurements show that the
quenching efficiencies depend strongly on the polarities of injected charges
and that quenching cannot be explained by the external-field-induced exciton
dissociation. Charge injection properties show that the hole injections play
an essential role in fluorescence quenching. A 60% fluorescence-quenching
efficiency is achieved. This will be a novel optical storage medium with high
contrast and high signal-to-noise ratio.
These nm-scale dot structures should prove to be effective in the applica-
tion of near-field optics to ultrahigh-density storage media.
Acknowledgements
The authors are very grateful to T. Maeda for providing them with the sam-
ples of the magnetic films. They also acknowledge useful discussions with
H. Yoda, J. Akiyama, T. Hiraoka and Y. Yanagita.
References
1. Modified by K.N. from the roadmap (Fig. 2.3.1.1) In: Research survey on near-
field light technology, OITDA (2000)
2. S.Y. Chou, M.S. Wei, P.R. Krauss, P.B. Fisher: J. Appl. Phys. 76, 6673 (1994)
3. R.L. White, R.M.H. New, RE.F.W. Pease: IEEE Trans. Magn. 33, 990 (1997)
144 K. Naito et al.
4. R.L. White: J. Magn. Magn. Mater. 209, 1 (2000)

5. C.T. Rettner, M.E. Best, B.D. Terris: IEEE Trans. Magn. 37, 1649 (2001)
6. C. Chappert, H. Bernas, J. Ferre, V. Kottler, J.-P. Jamet, Y. Chen, E. Cambril,
T. Devolder, F. Rousseaux, V. Mathet, H. Launois: Science 280, 1919 (1998)
7. C.A. Ross, H.I. Smith, T. Savas, M. Schattenburg, M. Farhoud, M. Hwang,
M. Walsh, M.C. Abraham, R.J. Ram: J. Vac. Sci. Technol. B. 17, 3168 (1999)
8. B. Cui, W. Wu, L. Kong, X. Sun, S.Y. Chou: L. Appl. Phys. 85, 5534 (1999)
9. T. Thurn-Albrecht, J. Schotter, G.A. Kästle, N. Emley, T. Shibauchi, L. Krusin-
Elbaum, K. Guarini, C.T. Black, M.T. Tuominen, T.P. Russell: Science 290,
2126 (2000)
10. K. Nielsch, F. Muller, A.-P Li, U. Gosele: Adv. Mater. 12, 582 (2000)
11. K. Naito, H. Hieda, M. Sakurai, Y. Kamata, K. Asakawa: IEEE Trans. Magn.
38, 1949 (2002)
12. R.A. Segalman, H. Yokoyama, E.J. Kramer: Adv. Mater. 13, 1152 (2001)
13. E. Betzig, J.K. Trautman, R. Wolfe, E.M. Gyorgy, P.L. Finn, M.H. Kryder,
C.-H. Chang: Appl. Phys. Lett. 61, 142 (1992)
14. S. Hosaka, T. Shintani, M. Miyamoto, A. Hirotsune, M. Terao, M. Yoshida,
S. Honma, S. Kammer: Thin Solid Films 273, 122 (1996)
15. S. Jiang, J. Ichihashi, H. Monobe, M. Fujihira, M. Ohtsu: Opt. Commun. 106,
173 (1994)
16. M. Hamano, M. Irie: Rev. Laser Eng. 24, 1045 (1996) (in Japanese)
17. H. Hieda, K. Tanaka, N. Gemma: J. Appl. Phys. 40, 1071 (2001)
18. N. Gemma, H. Hieda, K. Tanaka, S. Egusa: Jpn. J. Appl. Phys. 34, L859 (1995)
19. H. Hieda, K. Tanaka, N. Gemma: J. Vac. Sci. Technol. B 14, 1234 (1996)
20. H. Hieda, K. Tanaka, K. Naito, N. Gemma: Thin Solid Films 331, 152 (1998)
21. K. Asakawa, T. Hiraoka: Annual APS march 2000 Meeting, I33-2
22. T. Maeda T. Kai, A. Kikitsu, T. Nagase, J. Akiyama: Appl. Phys. Lett. 80,
2147 (2002)
23. K. Naito, M. Sakurai, S. Egusa: J. Phys. Chem. A 101, 2350 (1997)
24. S. Egusa, N. Gemma, A. Miura, K. Mizushima, M. Azuma: J. Appl. Phys. 71,
2042 (1992)
25. M. Deussen, M. Scheidler, H. Bassler: Synth. Met. 73, 123 (1995)
A Phenomenological Description of Optical
Near Fields and Optical Properties of N
Two-Level Systems Interacting with Optical
Near Fields
A. Shojiguchi, K. Kobayashi, S. Sangu, K. Kitahara, and M. Ohtsu
1 Introduction
1.1 What are Optical Near Fields?
Since the insight of Synge in 1928 [1], the localization of light near a material
system has been recognized as one of the unique properties of an optical near
field, and for more than two decades a variety of theoretical and experimental
studies have been intensively conducted [2–9]. At the beginning of this chap-
ter, we explain the spatial localization property of light, using two examples,
in order to give a physical insight into the optical near-field phenomenon.
The first example is evanescent light, which is generated on a dielectric
surface, as is well known, when the incident angle θ of a plane wave of light
exceeds the critical angle θc and total internal reflection occurs as shown in
Fig. 1 [2, 10]. The penetration or attenuation depth normal to the surface
(along the z axis) is of the order of the wavelength of light, which indicates
the spatial localization of light along the z axis. This differs from ordinary
propagating light, and much attention has been paid to its characteristics [11–
27]. Note that the total internal reflection is a macroscopic phenomenon, and
that it is valid only on a scale at which the refractive index n of bulk material
has meaning.
Another example is illustrated in Fig. 2 [2,9]. Figure 2a shows a plane wave
of light incident on a plane with a small hole of radius a that is much smaller
than the wavelength λ of light. Little of the propagating far-field light passes
through the hole, but light exists as an optical near field near the aperture on
the other side. The localization range of the optical near field is much smaller
than the wavelength of light and is determined by the aperture size a. This
kind of field, which a scanning near-field optical microscope (SNOM) uses at
the tip of an optical near-field probe, has been investigated in detail [28–32].
A more general situation is depicted in Fig. 2b. By irradiating a dielectric
that is much smaller than the wavelength of the incident light, optical near
fields coiling around the object are generated together with scattering far
fields. The localization range of the optical near fields is independent of the
wavelength of the incident light and is determined by the size of the object:
it depends on the radius a of the dielectric sphere as shown in Fig. 2b. Such
spatial localization reflecting the structure of the source of the fields (i.e.,
146 A. Shojiguchi et al.
Fig. 1. (a) The infinite planar boundary between a dielectric surface and a vacuum.
Incident light from the dielectric surface is generally reflected and refracted. (b)
Evanescent light on the dielectric surface generated by total internal reflection of
the incident light
Fig. 2. Optical near fields in various boundary conditions. (a) An optical near field
leaking from a hole that is much smaller than the wavelength of light. The leaking
field attenuates as it leaves the aperture. The localization range of light is much
smaller than the wavelength of light, depending on the size of the aperture. (b) An
optical near field surrounding a small dielectric sphere and the scattered optical far
field produced by an incident optical far field. The localization range of the optical
near field is determined by the sphere radius and is independent of the wavelength
of light
the induced polarization in matter) provides SNOM and other nanophotonic

devices with spatial resolution much higher than the diffraction limit [2–4,9].
Readers interested in a variety of SNOM applications and the state-of-the-
art of nano- and atom photonics are recommended to refer to the literature
[2–4, 7–9, 33–37].
Theoretically, we need a reasonable explanation of why light waves or
photons are localized, and present an example of an empirical method here.
The evanescent field generated on a planar dielectric surface can be obtained
by analytical continuation of the wavevector k⊥ normal to the surface as
A Phenomenological Description of Optical Near Fields 147
exp(ik · r − γz) with k⊥ = iγ (γ: real), whose localization range is about

that of the wavelength of the incident light. Similarly, optical fields around a
tiny dielectric sphere that is much smaller than the wavelength are described
by the spherical Hankel function with analytic continuation of the radial
component of the wavevector k⊥ to the pure imaginary k⊥ = iγ; the first
kind of Hankel function of order 0 gives the typical form of an optical near
field around a sphere as
e−γr
φ= . (1)
γr
The fields satisfying the spherical boundary condition can be expanded in
terms of plane waves, or can be expressed using angular spectrum repre-
sentation. From a numerical analysis, it follows that the distribution of the
angular spectrum has a peak at γ = 1/a [2]. Applying the result to (1), we see
that the localization range of the optical near fields around a sphere is of the
order of the radius a, i.e., the size of the material system. Here, it is obvious
that an optical near field should be derived consistently from microscopic-
induced polarization in matter because they are mutually related, and several
approaches in this direction are outlined in the following subsection 1.2.
1.2 Theoretical Approaches
Self-consistent descriptions of optical near fields are required, as mentioned

above, and the recent advances in nanoscience and technology have promoted
investigation of the dynamics of nanometric material systems interacting with
optical near fields, which is a critical and fundamental base for nanophoton-
ics. However, the situation is not simple; an optical near field involves both
an optical field and the excitation of matter on a nanometer scale, which is
considered as elementary excitation. Moreover, the existence of a boundary
or surface strongly affects the elementary excitation mode. It is mandatory
to find a proper normal mode, or a suitable basis function to satisfy such
boundary conditions for a nanometric system. Such a basis function defi-
nitely differs from the plane-wave basis used in bulk theories, but it is very
difficult to evaluate it in a rigorous manner except for the simple boundary
condition. On a nanometric scale, a quantum nature appears in an optical
field–material system, which requires a quantum description of the system
creating additional fundamental difficulties.
One possible approach is the semiclassical, self-consistent theory proposed
by Cho et al. [38,39], which quantum-mechanically describes microscopic po-
larization in matter to include a nonlocal response, while the electromagnetic
field is classically treated by the Helmholtz equation. They predicted the size
dependence and allowance of a dipole-forbidden transition in a nanometric
quantum-dot system numerically [40]. Nevertheless, the principal difficulty
of quantizing the electromagnetic field using a nonlocal approach remains,
as does the difficulty in treating the dynamics of a nanometric material sys-

tem coupled via optical near fields, because of the complexity involved in a
self-consistent procedure.
An alternative approach is to use a normal mode, or the Carniglia–Mandel
mode as a complete and orthogonal set that satisfies the infinite planar
boundary condition between the dielectric and a vacuum [11,12,23]. This for-
mulation allows us to quantum-field-theoretically describe an evanescent field
that is localized along the normal to the surface and possesses an anomalous
dispersion relation different from free photons, both of which are important
properties of an optical near field. This approach has revealed interesting phe-
nomena that occur near the surface [11, 16–26, 41, 42], but it unfortunately
provides neither a self-consistent description of a material system nor a nor-
mal mode to satisfy more general or realistic boundary conditions, such as
an optical near-field probe–sample system.
Kobayashi et al. [43–45] proposed another approach for focusing on the
relevant subsystem that we are interested in, and to extract a character-
istic feature of the subsystem after renormalizing the other irrelevant sub-
system. Using a quasiparticle representation for a macroscopic material–
electromagnetic field system, which only considers the near-field configuration
of the system without rigorously taking into account the boundary condi-
tion, they showed the localization of optical near fields around a nanometric
material system. They have also discussed the dynamics of a single atom,
molecule, or nanometric quantum-dot system interacting with the optical
near field [46–48].
In this chapter, we present a simple phenomenological model of an opti-
cal near field based on this last approach, in order to discuss the dynamic
properties of a nanometric material system interacting with an optical near
field. The model uses a concept of the “localized photon” to represent the
localization property of the optical near field phenomenologically, and two-
level systems as representative of a nanometric material system. To clarify
the difference between optical near fields and far fields, we compare the re-
sults obtained using our model with those of the Dicke model, which consists
of a two-level system interacting with radiation field and has been studied
extensively since originally presented by Dicke in 1954 [49].
1.3 Difference Between Optical Near Fields

and Propagating Fields: Individual vs. Global Excitation
The main difference between a material system coupled with optical near
fields and a material system coupled with propagating far fields, is that one
is coupled locally and the other is coupled globally, as depicted schematically
in Fig. 3.
Since the propagating field is usually expanded in terms of a plane-wave
basis, photons emitted as plane waves propagate from one site to distant
Fig. 3. Schematic drawing of (a) a global and (b) a local coupling systems. Each
element of the system in (a) can interact directly with arbitrary elements at a
distant site, while the elements at the near site can interact only with each other
in (b)
sites in the material system and can also excite the far sites (global excita-
tion [50]), as shown in Fig. 3a. By contrast, as shown in Fig. 3b, “near-field
photons” or “localized photons” can move only to nearby sites and excite
them (individual excitation). Generally speaking, the dynamics of a locally
coupled system differ from those of a globally coupled system with respect
to the relaxation speed towards equilibrium [51]. Therefore, we expect the
dynamics and equilibrium states for a locally coupled system to differ from
those of an ordinary globally coupled system, such as a propagating field–
matter system [52]. From this perspective, it is important to investigate the
nanometric material system that is interacting with optical near fields via
individual excitations [46–48, 53].
This chapter presents a model of a material system, i.e., for a quantum-
dot system, interacting with optical near fields, and discusses the dynamics
of the electric dipole moment and the radiation properties of the system in
detail. For comparison, a material system interacting with a radiation field
is also examined. We study the propagation of an initial excitation signal
in a quantum dot that is prepared by an optical near field locally and in-
dividually, and describe it in terms of excitons in N two-level systems. In a
dilute limit, excitons are approximated as bosons, and a rigorous solution of
the Heisenberg equation shows that the dipole moments representing quan-
tum coherence between any two levels follow linear dynamics. In a nondilute
case, excitons at one site obey the fermion commutation relations, while ex-
citons at other sites satisfy the boson commutation relations; this results in
nonlinear equations of motion. We predict a coherent oscillation of all the
dipoles in the system (dipole ordering), and strong radiation from some of
the dipole-ordered states, which are close to Dicke’s superradiance [49].
1.4 Two-Level System Interacting with Radiation Fields:

Dicke’s Superradiance
Dicke originally discussed superradiance, the cooperative emission of radia-

tion from a collection of excited two-level systems, in 1954. He found that
under certain conditions the radiation rate is proportional to the square of
the number of two-level systems involved, and an intense pulse is emitted [49].
Superradiant phenomena attracted considerable attention from a large num-
ber of authors in the 1970s [54–65]. For a small system that is much smaller
than the wavelength of the radiation, the equation of motion for the two-level
system coupled to the radiation field (Dicke model) is given as follows:
∂ρA
= −γ0 (R+ R− ρA − 2R− ρA R+ + ρA R+ R− ) , (2)
∂t
where ρA is the density operator after eliminating the radiation field’s degrees
of freedom from the total density operator of the system, R± is the collective
raising and lowering operators of two-level systems [49], and 2γ0 is the inverse
of the spontaneous emission lifetime of each two-level system (Einstein’s A
coefficient). From (2), one obtains differential equations for the observables
of the two-level system. By using a semiclassical approximation of neglecting
quantum correlations in the two-level system, an analytical solution [55, 58]
for the radiation intensity I can be written as
I ∝ N 2 sech2 [γ0 N (t − t0 )] , (3)
which has a cowbell-shaped radiation profile, and shows a radiation pulse
emitted under certain initial conditions whose height and width are propor-
tional to N 2 and 1/N , respectively. Note that the semiclassical approximation
does not correctly predict the dynamics starting from completely inverted
states [58, 59] with no dipole moments because the quantum fluctuations of
the dipole moments are essential in the radiation process.
In a large system, the superradiant state is affected by both inhomo-
geneity of the two-level system in a sample and the reabsorption of emitted
photons [61], which results in decoherence; namely, some part of the coher-
ence in the system is destroyed by the induced dipole–dipole interaction.
This reduces peak height and extends the tail of the radiation profile [61,62].
For two-level systems confined within a long cylindrical sample, Bonifacio et
al. [60,63] predicted multiple peaks in the radiation profile that stem from the
stimulated absorptions and emissions in the long active region. Skribanowitz
et al. [64] made experimental observations using a cylindrical sample, and
their experimental results were analyzed by Bonifacio et al. [65]. For a small
two-level system, the emission spectra of CuCl quantum dots were measured,
and Nakamura et al. [66] reported a single peak of pulse emission. Recently,
Chen et al. [67] proposed a way to detect a current of superradiance in a
double-dot system, while Piovella et al. [68] discussed how to detect a photon
echo in the superradiance from a Bose–Einstein condensate.
Tokihiro et al. [69] examined a linear excitonic system, in which exci-

tons can hop from one site to its nearest neighbors via the dipole–dipole
interaction, and showed that radiation from a totally inverted state as the
initial condition has a reduced peak intensity with extension of the tail, as
discussed by Coffey and Friedberg [62]. They also claimed that the radiation
profile from a partially excited state shows oscillatory behavior, as indicated
by Bonifacio and Lugiato [63]. In addition, T. Brades et al. discussed the oscil-
latory behavior of a superradiating system coupled to electron reservoirs [70].
It is clear that the peak intensity reduction of radiation in both cases comes
from the dipole–dipole decoherence, but it is not obvious whether the multi-
ple pulses so generated have the same origin. Our model also shows multiple
pulse emission, and we discuss its mechanism.
1.5 Chapter Outline
The chapter is organized as follows. Section 2 presents a model of N two-

level quantum-dot systems interacting with optical near fields represented in
terms of localized photons, and describes the model Hamiltonian. Section 3
obtains a rigorous solution of the equations of motion for the system within a
boson approximation. Using the solution, we investigate transfer of the dipole
moments of the system. In Sect. 4 we obtain the second-order perturbative
solution without using the boson approximation to show the dynamics of
the system. Section 5 introduces an effective Hamiltonian by renormalizing
the degrees of freedom of localized photons. With the help of the effective
Hamiltonian, we classify the quasisteady states of the dipole distribution in
order to discuss the origin of the dipole ordering. Section 6 is devoted to a
review of the Dicke model and superradiance for comparison in the following
sections. Section 7 examines the radiation property of the system in a weak
limit, where the radiation field does not affect the dynamics of the system. In
order to discuss the radiation property of the dissipative system, a quantum
master equation is derived in Sect. 8 after introducing a radiation reservoir,
and this is solved with and without including quantum correlations between
excitons and localized photons. Section 9 deals with the dynamics of localized
photons in an open system, which predicts two kinds of phases: storage and
nonstorage modes of localized photons in the system. Due to the nonlinearity
of the equations of motion, chaotic behavior is seen in the coupling constant
dependence of the total number of localized photons. Finally, in Sect. 10
concluding remarks are provided.
2 Model Hamiltonian
One of the most important features of optical near fields is the localization
property, where it is not suitable to use a broad spreading wave as a basis
function of quantization of the fields. Therefore, it is important to find a
good normal mode of the electromagnetic fields to satisfy a boundary con-

dition specific to the problem, by which the optical near fields are produced
and quantized [11, 43, 44]. It is difficult, however, to find a general and ap-
propriate normal mode satisfying any arbitrary boundary conditions. As an
alternative approach, it is possible to model optical near fields and their im-
portant characters phenomenologically. Such an approach is adopted here to
formulate the problem and to discuss the dynamics of a nanoscale material
system.
To describe the localizability of the optical near fields it is very effective to
use a localized basis function such as a Wannier function or a delta function
instead of a plane wave. However, at the same time we need to describe the
property of a short-range interaction among material sites via optical near
fields, or a steep gradient of optical near fields. Taking into account such
circumstances we model optical near fields phenomenologically in terms of
localized photons that are described as on harmonic oscillator localized in
each quantum-dot site, and are only allowed to hop from one site to the
nearest neighbors [71]. Figure 4 schematically describes our model system.
We suppose a closely located quantum-dot chain as a nanoscale material
system that is expressed as a one-dimensional N two-level system, or an
excitonic system with a periodic boundary condition. Since each exciton can
only interact with localized photons in the same QD, each quantum dot
indirectly interacts with one another via localized photons, as illustrated
in Fig. 4. This model is based on the unique property of localization or
nonpropagation of optical near fields mentioned above [44], and is preferable
for an intuitive understanding of coherent excitation transfer between the
Fig. 4. Schematic diagram of a model system: N two-level QDs labelled by the site
number are closely located in a ring, and are interacting with localized photons
QDs and its manipulation by the localized photons. The model Hamiltonian
of the system can be written as
H = Ha + Hb + Hint , (4)
where Ha represents localized photons, Hb describes excitons, and Hint rep-
resents the localized photon–exciton interaction. Each Hamiltonian can be
expressed as
N
Ha = εa†n an + V (a†n+1 an + a†n an+1 ) , (5a)
n=1
N

Hb = E b†n bn , (5b)
n=1
N

Hint = U (a†n bn + b†n an ) , (5c)
n=1
where n indicates the site number, and an (a†n ) and bn (b†n ) represent annihi-
lation (creation) operators of a localized photon and an exciton, respectively.
The periodic boundary condition requires that the (N +1)-th site corresponds
to the first site. The constant energies of the localized photons and excitons
are represented by ε = ω and E = Ω, respectively. The hopping energy
of the localized photons is represented as V = v, and U = g gives the
strength of the conventional dipolar coupling between a localized photon and
an exciton in the rotating-wave approximation.
We apply the boson commutation relations to the localized photons as
[an , a†n ] = δnn , [an , an ] = [a†n , a†n ] = 0 . (6)
The creation operator of an exciton, b†n ,
can be written in terms of the anni-
hilation operator of a valence electron of the n-th site, cn0 , and the creation
operator of a conducting electron of n-th site, c†nf , as b†n = c†nf cn0 . In the
same way, the annihilation operator of an exciton, bn , can be written as
bn = c†n0 cnf . Assuming that only one exciton is generated in a quantum dot
site, we obtain the relation as Nf + N0 = c†nf cnf + c†n0 cn0 = 1. With the help
of these expressions and the fermi commutation relations for electrons, the
commutation relations for excitons [72, 73] are derived as
[bn , b†n ] = δnn (1 − 2b†n bn ) , (7)
which shows that excitons behave as fermions at intrasite locations and as
bosons at intersite ones. It follows from (7) that excitons are approximated
as bosons in a dilute limit of the exciton density, b†n bn ≡ Nn 1.
With the boson approximation, the Heisenberg equation of motion can
be rigorously solved. First, we obtain a rigorous solution of the Heisenberg
equation for bosonic excitons to investigate the dynamical properties. Then
we investigate the dynamics of fermionic excitons, solving the Heisenberg
equation both perturbatively and numerically.
3 Dynamics of Bosonic Excitons

3.1 Boson Approximation and Diagonalization
of the Hamiltonian
To solve the Heisenberg equation we introduce the spatial Fourier transfor-

mation for an and bn as
N
1 ikn
Ak = √ e an , (8a)
N n=1
N
1 ikn
Bk = √ e bn , (8b)
N n=1
where we set k = 2πl/N for l = 1, · · · , N and the site distance as 1. Fourier
transforms Ak satisfy the following commutation relations
[Ak , A†k ] = δkk , [Ak , Ak ] = [A†k , A†k ] = 0 , (9)
and the commutation relations of Fourier transforms Bk are similarly ob-

tained as
1 i(kn−k n ) 1 i(k−k )n
[Bk , Bk† ] = e [bn , b†n ] = e (1 − 2b†n bn ) . (10)
N
N n
n,n
It follows from this expression that when the expectation value of the number
of excitons is small as b†n bn N , the second term on the right-hand side
of (10) can be neglected, and excitons are approximated as bosons as
[Bk , Bk† ] = δkk , [Bk , Bk ] = [Bk† , Bk† ] = 0 , (11)
where the following relation

1 i(k−k )n
e = δkk (12)
N n
is used. Fourier inverse transforms are also given as

1 −ikn
an = √ e Ak , (13a)
N k∈1BZ
1 −ikn
bn = √ e Bk , (13b)
N k∈1BZ
where the summation of k runs over the first Brillouin zone (1BZ). Using
(8a) and (8b), and the following relation
1 ik(n−n )
e = δnn , (14)
N
k∈1BZ
we can transform the Hamiltonian (4) into a Fourier transform representation

H= {(ε + 2V cos k)A†k Ak + EBk† Bk + U (Ak Bk† + A†k Bk )}
k

≡ Hk . (15)
k
Note that the dispersion relation of localized photons is changed from ε to

(ε + 2V cos k). The transformed Hamiltonian Hk can be diagonalized into a
simple quadratic form

ε + 2V cos k U Ak
Hk = (A†k , Bk† )
U E Bk

2
† † e U Ak
≡ (Ak , Bk ) ≡ Mij X † (k)i X(k)j , (16)
U E Bk
i,j=1
where the abbreviations e = ε + 2V cos k and X † (k) = (A†k , Bk† ) are used.
The matrix kernel of the Hamiltonian, M , is diagonalized as

T λ+ 0
S MS = , (17)
0 λ−
by an orthogonal matrix S as
⎛ ⎞
L+K L−K
S=⎝ 2L 2L ⎠ . (18)
L−K L+K
2L − 2L
Here, eigenvalues λ± are written as

e+E 1
λ± = ± (e − E)2 + 4U 2 (19)
2 2
≡ K + E ± L,
√
with K = (e − E)/2 and L = K 2 + U 2 . Using the polariton transformation
as

αk T Ak
≡S , (20)
βk Bk
we can finally diagonalize the Hamiltonian (15) as follows:

H= Hk = (λ+ αk† αk + λ− βk† βk ) . (21)
k k
3.2 Dipole Dynamics Driven by Local Excitation
With the help of the diagonalized Hamiltonian (21), we can immediately

solve the Heisenberg equation and can express the time evolution of exciton-
polariton operators αk (t) and βk (t) as

αk (t) e−iλ+ t/ αk
= , (22)
βk (t) e−iλ− t/ βk
where αk and βk denote the operators in the Schrödinger representation. It

is assumed in this chapter that an operator with no indication of the time
dependence is expressed in the Schrödinger picture as a time-independent op-
erator. From the inverse transformation of (20), we obtain a time-dependent
solution of the Fourier transforms Ak (t) and Bk (t) as
−iλ t/
Ak (t) αk (t) e + αk
=S =S . (23)
Bk (t) βk (t) e−iλ− t/ βk
By transforming the exciton-polariton operators (αk , βk ) into (Ak , Bk ) again,

explicit time-dependent solutions Ak (t) and Bk (t) of the Heisenberg equa-
tions are written as

L K L
Ak (t) = e−i(e+E)t/2 cos t − i sin t Ak
L

U −i(e+E)t/2 L
−i e sin t Bk , (24a)
L

U L
Bk (t) = −i e−i(e+E)t/2 sin t Ak
L

L K L
+ e−i(e+E)t/2 cos t + i sin t Bk . (24b)
L
Using the Fourier inverse transformation, we obtain a final expression of the
time evolution of the exciton operator bn (t) as follows:
1 ik(m−n)−i(e+E)t/2
bn (t) = e
N
k,m

U L L K L
× −iam sin t + bm cos t + i sin t . (25)
L L
Here we introduce a physical quantity of the electric dipole moment for our
two-level system defined as
µn (t) = µ(bn (t) + b†n (t)) = µPn (t) , (26)
where Pn (t) = bn (t) + b†n (t) denotes the dipole operator that describes the
electric dipole moment of each quantum dot. Moreover, two other indepen-
dent operators are defined as
Vn (t) = i(bn (t) − b†n (t)) , Wn (t) = b†n (t)bn (t) − bn (t)b†n (t) , (27)
where the latter represents the population difference of the system. The unit
operator and these three operators form a basis for any two-level system,
and the set (Pn , Vn , Wn ) corresponds to the Bloch vector in a collection of

two-level atom systems. However, it should be noted that the population
difference Wn is always −1 for bosonic excitons, and the number operators
of excitons b†n bn are the independent operators in this case.
In order to investigate the coherent excitation dynamics of the system,
we examine the time evolution of the expectation value of the dipole moment
at an arbitrary QD n, Pn (t), under a variety of initial conditions. The
expression given by (25) and its Hermitian conjugate provide a useful result
as
N
1 L
Pn (t) = (Pm cos ξmn + Vm sin ξmn ) cos t
N
k∈1BZ m=1

K L
+ (Vm cos ξmn − Pm sin ξmn ) sin t , (28)
L
where the operation · · · = Trρ · · · means the expectation value of an arbi-
tray operator. The notation ξmn ≡ k(m − n) − t(e + E)/2 is used. Localized
photons are assumed to be initially in the vacuum state. In particular, it is
intriguing to investigate the dynamics with the initial condition of a locally or
individually excited state, i.e., how the initial excitation prepared only at one
site is transferred in the system. Setting Pn (0) = δn1 P1 and Vn (0) = 0
for all n, we obtain from (28) an explicit solution as

1 L K L
Pn (t) = cos ξ1n cos t − sin ξ1n sin t P1 . (29)
N L
k
In Fig. 5 we present one of the numerical results of the time evolution of

the dipole moment distribution calculated from (29), where the total number
of sites is eight, and the parameter values E = 2, ε = 1, and V = 1 are used
in a unit of a typical exciton energy in the QD.
Although the dynamics seems complicated at first sight, we can find a
characteristic behavior that the dipole of the first site moves to the fifth site
(the opposite site) and returns to the first site again. To see the behavior
more precisely we show plots of the time evolution of the dipole moments at
the first site and the fifth site in Fig. 6a and Fig. 6b, respectively.
The result clearly shows as a dominant behavior that when the dipole
of the first site is active, that of the fifth site is inactive and vice versa, or
the dipole moves like a seesaw between the first site and the fifth site. The
recurrence behavior is reasonable from the fact that the Heisenberg equation
of the system is linear in the boson approximation. Since the dipole moment
Pn is quantum coherence between the ground and excited states of a two-
level system, quantum coherence is transported to the farthest site of the
chain and then comes back to the original site. When the total number of
sites is odd, the seesaw motion arises between an initially excited site and its
farthest pair of sites of the chain. An interesting point in this case is that we
Fig. 5. Time evolution of the dipole-moment distribution when the total number
of sites is eight. The parameters E = 2, ε = 1, and V = 1 are used in the
calculation. The vertical and horizontal axes represent the dipole amplitude and
the site number, respectively. The dipole moment is initially set only at the first
site
can obtain two copies of coherence at the farthest pair of sites as illustrated
in Fig. 7b.
Moreover, it might be possible to use the system as a nanophotonic device
of transporting or splitting quantum coherence.
Fig. 6. Time evolution of the dipole moment of (a) the first site and (b) the fifth
site. When the dipoles are large in (a), corresponding dipoles in (b) are small, and
vice versa
Fig. 7. Transportation of quantum coherence in (a) an even site number system

and (b) an odd-site number system. In the even-site system (a), output signals are
obtained from the opposite site to the input site. In the odd-site system (b), output
signals are split into the two-pair site opposite to the input site
4 Dynamics of Fermionic Excitons

If we rigorously adopt the commutation relation for excitons given by (7),
the Heisenberg equations for the system read
dbn (t) i i i
= [H, bn (t)] = − Ebn (t) + U an (t)Wn (t) , (30a)
dt
dan (t) i
= [H, an (t)]
dt
i i i
= − εan (t) − V (an+1 (t) + an−1 (t)) − U bn (t) . (30b)

Since the higher-order terms are produced by a mode–mode coupling such
as an Wn = an (b†n bn − bn b†n ), the equations become nonlinear and are hardly
solved analytically. Thus we first solve the equations perturbatively in order
to investigate the dynamics discussed in the preceding section.
4.1 Perturbative Expansion of Time-Evolution Operator
Noticing that the Hamiltonian for localized photons Ha given by (5a) can be
written in a quadratic form as
⎛ ⎞
⎛ ⎞ a1
ε V V ⎜ ⎟
⎜V ε V ⎟⎜ ⎟
⎜ ⎟ ⎜ .. ⎟
⎜ . . . . . . ⎟ ⎜ . ⎟
Ha = (a†1 , · · · , a†N ) ⎜
⎜
⎟⎜
⎟⎜ . ⎟
⎟ (31)
⎜ . . . . . . ⎟ ⎜ .. ⎟
⎝ V ε V ⎠⎜ ⎝
⎟
⎠
V V ε
aN

†
≡ an Rnm am , (32)
nm
we obtain an orthogonal matrix P to diagonarize the matrix kernel R of the

Hamiltonian as
(P −1 RP )ij = λi δij . (33)

Then new annihilation operators vj = n Pnj an and creation operator vj† =
†
n Pnj an of new modes of localized photons are introduced. Here note that
the new modes are extended over a whole system, since all the elements of
the diagonalization matrix P are not zero. The commutation relation of the
new operators can be given as [vi , vj† ] = δij . With the help of (33) and the
modes vi , the Hamiltonian in (32), or (5a) is diagonalized as

Ha = λj vj† vj , (34)
j
and the localized photon–exciton interaction Hint in (5c) is written in terms

of vi as

Hint = g Pnj (vj† bn + vj b†n ) . (35)
nj
In order to derive a perturbative expansion of the time evolution of an arbi-

trary operator, we define the Liouvillians as
1 1 1
L ··· = [H, · · · ] , L0 · · · = [H0 , · · · ] , Lint · · · = [Hint , · · · ] . (36)

Then the Heisenberg equation of an arbitrary operator O is written as
i
Ȯ = [H, O] = iLO , (37)

and we can obtain a formal solution as
O(t) = G(t)O(0) , (38)
with the time evolution operator G(t) = eiLt that satisfies the following
equation
Ġ(t) = iLG(t) = iL0 G(t) + iLint G(t) . (39)
Treating the interaction term Hint as a perturbation, we solve (39) pertur-
batively up to the second order of the perturbation as [73]
t
G(2) (t) = G0 (t) + i G0 (t − s)Lint G0 (s)ds
0
t s
− ds duG0 (t − s)Lint G0 (s − u)Lint G0 (u) , (40)
0 0
where the notation G0 (t) = eiL0 t is used. Substituting (40) into (38) and
using (36), we can obtain the time evolution of the exciton operator bn (t) =
G(2) (t)bn . Suppose that localized photons are initially in the vacuum and
Vn = 0, then the expectation value of bn (t) is expressed as
N

bn (t) = e−iΩt bn + g 2 Γj (t)
j=1
N

× Pnj Pmj (1 − δnm )Wn bm − 2
Pnj bn , (41)
m=1
where the notation

t s
−iΩt
Γj (t) = e dse is(Ω−λj )
due−iu(Ω−λj )
0 0
e−iΩt − e−iλj t ite −iΩt
= + ≡ cj (t) + idj (t) (42)
(Ω − λj )2 Ω − λj
is used, and cj (t) and dj (t) represent the real and imaginary parts of Γj (t),
respectively. The expression given by (41) and its Hermitian conjugate pro-
vide a time evolution of the expectation value of the dipole at an arbitrary
QD site n as
⎛ ⎞

Pn (t) = Pn ⎝cos Ωt − g 2 2 ⎠
cj (t)Pnj
j

+g 2 cj (t)Pnj Pmj Pm Wn . (43)
j m=n
Corresponding to the discussion developed in Sect. 3.2, we set initially Pn =

δn1 and obtain the perturbative solution that describes the dipole dynamics
driven by a local excitation of QD site one as follows:
⎛ ⎞

Pn (t) = δn1 ⎝cos Ωt − g 2 2 ⎠
cj (t)Pnj P1
j

+(1 − δn1 )g 2 cj (t)Pnj P1j P1 Wn . (44)
j
In the next section we discuss the dynamical properties of the system, on the
basis of the solution given by (44).
4.2 Numerical Results and Dynamical Properties
As discussed in Sect. 3.2, we numerically examine the dynamics of the dipole

of the system depending on the initial conditions. Suppose that localized
photons are initially in the vacuum, Vn = 0 for all n, the dipole is initially
set only at the first site as Pn = δn1 , and all the populations are in the
ground states except for the first site as Wn = −(1 − δn1 ). As mentioned
before, all the population differences are automatically −1 for bosonic exci-
tons. In addition, we impose that the length of the Bloch vector (Pn , Vn , Wn )
is normalized as
Pn (t)2 + Vn (t)2 + Wn (t)2 = 1 . (45)
In Fig. 8 the time evolution of the dipole distribution is plotted. The

parameter values used are the same as in Sect. 3.2 (E = 2, ε = 1, and
V = 1).
It follows from the figure that the dipole excitation of the first site moves
to the opposite site as shown in the boson case (see Fig. 5). The reason is
that the initial exciton density is so dilute as to validate the boson approx-
imation. As time advances, the amplitude of each dipole increases because
the perturbative solution violates the unitarity.
To find a new feature of fermionic excitons, we vary the initial conditions
for the population differences from Wn = −(1 − δn1 ). Since the initial
Fig. 8. Time evolution of the dipole distributions (E = 2, ε = 1, V = 1). Initially

all of the sites except the first site are in the ground state, while the dipole moment
is initially set only at the first site
population differences at site n contributes to the perturbative solution of

the dipole at the same site, (44), in a product of Wn P1 , the following
hypothesis is proposed.
Flip Hypothesis
If the sign of the initial population difference of the n-th site, Wn (0), is
inverted, then the direction of the dipole moment of the n-th site at arbitrary
time, Pn (t), is flipped.
Since the dipoles at sites 4, 5 and 6 in Fig. 8 are directed opposite to the
others, we invert the sign of Wn for n = 4, 5, 6. Figure 9 shows the result
that the direction of the dipoles at sites 4, 5, and 6 are flipped, and that all
the dipoles oscillate with the same phase but with different amplitudes. Thus
we observe that the system is transferred from a locally excited state to a
coherently oscillating state of the dipoles, in other words, to a dipole-ordered
state. In the ordered state the total dipole is N times larger than each single
dipole. Moreover, since radiation from an oscillating dipole is proportional to
Fig. 9. Time evolution of the dipole distribution (E = 2, ε = 1, V = 1): the initial

population differences of sites 4, 5 and 6 are inverted from Wn = −(1 − δn1 )
the square of the dipole moment, we can expect a high intensity of radiation
from the dipole-ordered state. The radiation property of the system will be
discussed in Sect. 7.1.
4.3 Dynamics of Dipole-Forbidden States

via Optical Near-Field Interaction
It is well known that the electric interaction between molecules with no dipole
moments is weak when they are separated in a macroscopic distance, and
that the interaction becomes strong when they are very close to each other.
Taking this into account, we investigate the dynamics of the system driven
by localized photons, from an initial state that alternating dipoles are set
and thus the total dipole of the system vanishes for an even number of sites
(see Fig. 10a). Such a dipole-forbidden state can be manipulated by localized
photons, not by propagating far fields.
As an initial condition we set Pn = (−1)n , and for simplicity, localized
photons are in the vacuum, Vn = 0, and Wn = 0. The system parameters
E = 2, ε = 1, and V = 1 are used as before. Figure 11 shows the time
evolution of such a dipole-forbidden state.
Fig. 10. Schematic illustration of (a) an alternating dipole distribution (a dipole-

forbidden state) and (b) a dipole-ordered state. The state with alternating dipoles
that result in the total dipole of zero cannot be coupled by the radiation field with
the dipole interaction
Fig. 11. Time evolution of a dipole-forbidden state (alternating dipole distribu-

tion). All the population differences are initially set as 0. We observe an oscillation
of the alternating dipoles. The system remains in the dipole-forbidden state
From Fig. 11 it follows that the system oscillates as schematically shown

in Fig. 10a, and that it remains in the initial dipole-forbidden state. This
kind of dynamics is achieved in the system whose initial distribution of the
√
population differences is uniform. For example, if we set Pn = (−1)n / 2
Fig. 12. Time evolution of a dipole-forbidden state (E = 2, ε = 1, V = 1). The

population differences are initially set as Wn = −Pn . The system is converted
from a dipole-forbidden state to a dipole-ordered state
√
and Wn = 1/ 2 for all n, the result is the same as shown in Fig. 11 except
for its amplitude of the oscillation.
Next, we manipulate the distribution of the population difference nonuni-
formly, so that the signs are set as opposite to those of the corresponding
dipoles:
√
Pn = (−, +, −, +, −, +, −, +)/ 2 ,
√
Wn = (+, −, +, −, +, −, +, −)/ 2 .
Figure 12 presents the result that a dipole-forbidden state as shown in
Fig. 10a is converted to a dipole-ordered state as illustrated in Fig. 10b.
Since the dipole ordering has occurred by manipulating the initial distri-
bution of the population differences, the result can be interpreted by the “flip
hypothesis” proposed in Sect. 4.2. As a collective oscillation of the dipoles oc-
curred, the system evolved from a nonradiative state to a radiative state
through the localized photon–exciton interaction.
4.4 Semiclassical Approximation
Here we should note that the second-order perturbative solutions break the
unitarity, and that the long-time behaviors of the dynamics given by the
solutions are not necessarily the same as the exact ones. Thus we have to
return to the Heisenberg equations, (30a) and (30b), to evaluate the time
evolution of physical observables. The Heisenberg equations are first solved
by neglecting the quantum correlations between excitons and localized pho-
tons such as Wn yn = Wn yn , and later the quantum correlations are
estimated in Sect. 8.2. In this semiclassical approximation, (30a) and (30b)
can be converted to the following coupled differential equations
⎧
⎨ ∂t Pn = −ΩVn + gWn yn ,
∂t Vn = ΩPn − gWn xn , (46)
⎩
∂t Wn = g (Vn xn − Pn yn ) ,

∂t xn = −ωyn − v (yn−1 + yn+1) − gVn ,
(47)
∂t yn = ωxn + v xn−1 + xn+1 + gPn ,
where the notations xn = an + a†n , yn = i(an − a†n ), and v = V / are used.
To check the reliability of the semiclassical approximation, we again inves-
tigate the system examined in Sect. 4.2. Figure 13 shows the time evolution
of the dipole distribution obtained from the semiclassical approach for the
system corresponding to Fig. 8.
Comparing Fig. 13 with Fig. 8, we find that both profiles are the same, and
that the amplitude of each dipole obtained from the Heisenberg equations is
less than 1, which means that the unitarity of the time evolution is conserved.
In Fig. 14, we show the time evolution of the dipole distribution obtained
from the semiclassical approach when the initial population differences of
sites 4, 5, and 6 are inverted. It follows from the figure that each dipole is
ordered as the flip hypothesis predicts, and that the flip hypothesis is still
valid in the numerical solution of the semiclassical Heisenberg equation.
Numerical results obtained from (46) and (47) with a semiclassical approx-
imation predict the dipole dynamics with qualitative similarities as shown
from the perturbative solution (44). The flip hypothesis proposed in Sect. 4.2
is also verified by using the numerical solutions of (46) and (47), or those of
the Heisenberg equations without quantum correlations. These results show
that the approach discussed in this section, maintaining an advantage to con-
serve the unitarity, qualitatively describe the dipole dynamics of the system
in a similar way as the other two approaches. The validity of the approxi-
mation employed here, which we will examine in Sect. 8.2, is reported in the
Dicke model that superradiant phenomena can be described with a semiclas-
sical approximation that neglects quantum correlations among the atoms [57],
and that the contribution from the quantum correlations is of the order of
1/N for large N [58].
Fig. 13. Time evolution of the dipole distribution for the system (E = 2, ε = 1, V =
1) obtained from the Heisenberg equation with the semiclassical approximation.
Initially all of the sites except the first site are in the ground state
5 Effective Hamiltonian and the Dipole Ordering

In order to investigate the origin of the dipole ordering discussed above,
we first introduce an effective Hamiltonian to renormalize the degrees of
freedom of localized photons. Using the effective Hamiltonian, we then classify
quasisteady states with respect to the dipole distribution.
5.1 Effective Hamiltonian
A unitary transformation with an anti-Hermitian operator S is applied to

the Hamiltonian (4) as [74–76]
1
H̃ = e−S HeS = H + [H, S] + [[H, S], S] + · · · . (48)
2
If the Hamiltonian (4) is divided into the unperturbed part H0 = Ha + Hb
and perturbed part Hint , and if S is chosen to satisfy
Hint + [H0 , S] = 0 , (49)

Fig. 14. Time evolution of the dipole distribution for the system (E = 2, ε = 1, V =
1) obtained from the Heisenberg equation with the semiclassical approximation. The
initial population differences of sites 4, 5, and 6 are inverted
the terms linear in the coupling constant g = U/ vanish as

1
H̃ = H0 + [Hint , S] + O(g 3 ) . (50)
2
The transformed Hamiltonian H̃ contains the terms in the second order of
the exciton–localized photon interaction that describe exciton–exciton inter-
actions via localized photons. Taking the expectation value of H̃ in terms of
the vacuum of localized photons |vac, we obtain the effective Hamiltonian
Heff as
Heff = vac|H̃|vac = Hb + Hb−b , (51a)

Hb = Ωb†n bn , (51b)
n
g 2
Hb−b = b†n bm , (51c)
n,m
ΩI − R/ nm
where the exciton energy E is denoted as Ω. With the help of the compo-
nents of the Bloch vector (Pn , Vn , Wn ) the effective Hamiltonian (51a) can
be rewritten as
1 + Wn
Heff = (Ω + ∆Ωn ) + ∆Ωnm (Pn Pm + Vn Vm ) , (52)
n
2 4 n
m=n
where the interaction energy or coupling strength between excitons ∆Ωnm

are given as
g2
g2
∆Ωnm = = Pnj Pmj , (53)
ΩI − R/ nm j
Ω − λj
and the abbreviation ∆Ωn ≡ ∆Ωnn is used for a special case.
5.2 Classification of Quasisteady States
Using the effective Hamiltonian Heff , we can obtain the time evolution of the
dipole moment at site n in a similar way as discussed in Sect. 4.1:

Pn (t) = Pn cos[(Ω + ∆Ωn )t] + ∆Ωnm tWn Pm sin(Ωt) . (54)
m=n
If the dipole at site 1, in particular, is only excited at t = 0, (54) reads

Pn (t) = δn1 P1 cos[(Ω + ∆Ωn )t]
+(1 − δn1 )∆Ωn1 tWn P1 sin(Ωt) . (55)
It follows from this expression that the sign of the dipole at site n depends
on the coefficient ∆Ωn1 as well as the initial values of P1 (0) and Wn (0),
and that the coefficient ∆Ωn1 determines what kind of quasisteady states of
the dipole distribution the system reaches when the initial values are fixed.
Therefore we examine the matrix ∆Ω to classify the quasisteady states of the
dipole distribution of the system. First, as an example, we show the matrix
elements of ∆Ω for the system examined in Sect. 3 and Sect. 4 whose material
parameters are E = 2, ε = 1, and V = 1 as follows:
⎛ 1 2 1 2 ⎞
3 −3 −3 −3 3
1 2 1 1
3 3 3
⎜ 2 1 2 1 −1 −2 −1 1 ⎟
⎜ 3 3 3 3 3 3 3 3 ⎟
⎜ 1 2 1 2 1 ⎟
⎜ 3 3 3 3 3 − 13 − 23 − 13 ⎟
⎜ 1 1 2 1 2 1 ⎟
⎜− 3 3 3 3 3 3 − 13 − 23 ⎟
∆Ω = ⎜ ⎜− 2 − 1 1 2 1 2 1 − 1 ⎟ .
⎟
⎜ 3 3 3 3 3 3 3 3⎟
⎜ 1 2 1 1 2 1 2 1 ⎟
⎜− 3 − 3 − 3 3 3 3 3 3 ⎟
⎜ 1 ⎟
⎝ 3 − 13 − 23 − 13 13 23 13 32 ⎠
2
3
1
3 − 13 − 23 − 13 1
3
2
3
1
3
The matrix is symmetric and has a cyclic structure as

Table 1. Classification of quasisteady states
Group Signs of ∆Ω1n for n = 1, 2, 3, · · · , 8

1st (−, −, −, +, +, +, −, −), (+, +, +, −, −, −, +, +)
2nd (−, +, −, −, +, −, −, +), (+, −, +, +, −, +, +, −)
3rd (−, −, −, −, −, −, −, −), (+, +, +, +, +, +, +, +)
4th (−, +, −, +, −, +, −, +), (+, −, +, −, +, −, +, −)
∆Ωn+1,m+1 = ∆Ωn,m .
This is due to the fact that the periodic boundary condition is imposed on the
system, and that the Hamiltonian given in (4) has translational invariance.
Owing to the symmetric structure as
(∆Ω22 , ∆Ω23 , ∆Ω24 , · · · ) = (∆Ω11 , ∆Ω12 , ∆Ω13 , · · · ) ,
the dipole distribution of the system initiated by the dipole excitation at site
2 can be discussed in terms of ∆Ω1n that governs the problem driven by
the site-one excitation. In general the problem is reduced to examining the
structure of the coefficients ∆Ω1n that depend on the material parameters
Ω and λj , or E, ε, and V . Since the parameter sets crossing the resonant
points (Ω = λj ) change the pattern of the dipole distribution in the system,
we can classify quasisteady states of the dipole distribution into four groups
in Table 1.
The system with E = 2, ε = 1, and V = 1, which has been investigated
in Sect. 4, belongs to the first group. Since the coefficients ∆Ω1n in the first
group have different signs at sites 4, 5, and 6 from the others, the quasisteady
state of the dipole distribution that is predicted by the effective Hamiltonian
is the one described by the second-order perturbation solution (see (44) and
Fig. 8). The third and fourth groups are, in particular, interesting among
the four groups of the quasisteady states. The third group corresponds to a
“ferromagnetic” system, in which all the electric dipole moments are aligned
in the same direction, while the fourth group corresponds to an “antiferro-
magnetic” system, where the dipoles are aligned to have alternating signs.
Using numerical results of (46) and (47), we examine the dynamics of the
third and fourth groups whether they behave as predicted by the effective
Hamiltonian: “ferromagnetic,” or “antiferromagnetic.” The parameter set of
E = 3.01, ε = 1, and V = 1 gives
∆Ω1n = (−13.1, −12.7, −12.4, −12.2, −12.2, −12.2, −12.4, −12.7) ,
and thus the system belongs to the third group. The result is shown in Fig. 15,
which predicts the “ferromagnetic” behavior as expected.
When the parameter set of E = 1, ε = 3.01, and V = 1 is chosen, the
system corresponds to the fourth group. Figure 16 presents the time evolution
Fig. 15. Time evolution of the dipole distribution of the system belonging to the
third group. The initial conditions are set as Pn = δn1 and Wn = 1 − δn1 . All
the electric dipole moments are aligned in the same direction, and a dipole-ordered
state appears
of the dipole distribution of the system that shows alternating changes in

signs, as expected to be similar to antiferromagnetics.
In an even site number “antiferromagnetic” system, the dipole distribu-
tion from an initial state evolves to align the dipoles alternatingly up and
down, as shown in Fig. 16. On the other hand, an odd site number sys-
tem becomes frustrated because the dipole distribution cannot be perfectly
aligned alternatingly. Thus it is interesting to examine the time evolution
of the dipole distribution in an odd site number system. The result for the
N = 9 “antiferromagnetic” system is presented in Fig. 17. It follows from
the figure that there are ordered dipoles near the one initially prepared at
site one, and alternating dipoles including a pair at the opposite site to site
one. Since strong frustration, for example as occurred in spin glass, should
produce a random distribution, the figure indicates that the odd site number
system considered here is not frustrated so strongly.
With the help of the flip hypothesis proposed in Sect. 4.2, we infer that the
distribution of the dipoles can be changed by individual manipulation of the
initial population differences, and that an “antiferromagnetic” state can be
Fig. 16. Time evolution of the dipole distribution of the system belonging to the
fourth group. The initial conditions are set as Pn = δn1 and Wn = 1 − δn1 .
The dipole distribution similar to the “antiferromagnetic” state is shown, where
the dipoles alternatingly change the directions
Fig. 17. Quasisteady state of the dipole distribution in the “antiferromagnetic”

system belonging to the fourth group. The total number of sites is assumed to be
nine
Fig. 18. Time evolution of the dipole distribution of the system belong-
ing to the fourth group. The initial population differences are set as Wn =
(0, −, +, −, +, −, +, −). According to the flip hypothesis, the alternating dipoles
change their directions and are aligned to form a dipole-ordered state
transformed into a dipole-ordered state. To verify this, the initial population

differences are set in the zigzag form as Wn = (0, −, +, −, +, −, +, −) for
the “antiferromagnetic” system (E = 1, ε = 3.01, and V = 1), and the
time evolution of the dipole distribution is examined. The result, as shown in
Fig. 18, clearly indicates that a dipole-ordered state manifests itself owing to
the local manipulation of the initial population differences. It is noteworthy
that the amplitude of each dipole is close to one, and that the total dipole of
the system is of the order of N , whose radiation property is investigated in
the following sections.
5.3 Response to the Initial Input of a Localized Photon:

the Robustness of Quasisteady States
Until now, the state of localized photons at each site was initially taken to be
a vacuum. Since localized photons in the Heisenberg equations have different
degrees of freedom, this section examines the temporal evolution given an
initial distribution of localized photons. The initial conditions take the state
given below, in which the distribution of localized photons is given for site 1
only.
xn = x1 δn1 , yn = 0 , Pn = Vn = 0, Wn = 1 . (56)
This state can be achieved by having the localized photons at site 1 take on
a coherent state in which the eigenvalue is a real number. The coherent state
is the eigenstate of the annihilation operator of a simple harmonic oscillator.
Therefore, the initial state satisfying the following condition needs to be
established at site 1:
a1 |x1 = x1 |x1 x1 is real. (57)
The temporal evolution of the dipole distribution is examined by solving

the semiclassical Heisenberg equations (46) and (47) numerically with these
initial conditions for each of the systems categorized in this section. Figure 19
shows the dynamics of systems in the first group.
Fig. 19. The temporal evolution of the dipole distribution of the system for the
first group in which the localized photonic states are initially in the coherent state
at the first site and the other sites are in the vacuum state and all the excitonic
states are in the totally inverted state
Fig. 20. Quasisteady states arise from the initial condition in which there is lo-
calized photon excitation at a single site when the system parameters have various
values (, V, E). The results in (a), (b), and (c) correspond to the first, third, and
fourth groups in Table 1, respectively
A quasisteady state emerges that is similar to the case in which dipoles

were given as the initial condition (e.g. Fig. 13). Figure 20 shows the qua-
sisteady states corresponding to each of the three systems considered in this
section.
These results demonstrate that the states of each group behave exactly
as predicted by the effective Hamiltonian. In other words, while the state
derived for each group originally involved a dipole (instead of a localized
photon), similar states are also realized with the initial input of localized
photons. On closer examination of Fig. 19, it can be seen that dipoles are
first induced at the site during the initial time frame as a result of the input of
localized photons. Therefore, these dipoles established the basis for realizing
quasisteady states via a mechanism that is similar to the case in which dipoles
were the initial input. These results imply the robustness of each quasisteady
state predicted by the effective Hamiltonian against input.
6 Dicke’s Superradiance
In the preceding section, we have shown that electric transition dipole mo-
ments of the system are aligned to have a large value, depending on the
material parameters of the system. It is then expected that strong radiation
can be emitted due to a large total dipole moment, and it is intriguing to
examine radiation properties of the dipole-ordered states obtained from our
localized photon model. Before such an investigation, we will outline superra-
diance [49, 54–57, 60, 63] as a cooperative phenomenon, where radiation from
N two-level systems interacting with propagating light exceeds the sum of
the one individually emitted from the systems. More precisely, a radiation
pulse with a sharp width of 1/N and a high intensity of N 2 is emitted [49].
This kind of system is called the Dicke model, and was extensively studied
by the middle of the 1970s since the pioneering work of Dicke in 1954 [49].
In this section we first introduce the Dicke model and Dicke states, and
then discuss superradiant states, or superradiance, following Mandel and
Wolf [77]. In order to examine how such states are dynamically obtained, a
master equation is derived with the help of the projection operator method,
and the time evolution of radiation is studied. Numerical solutions of the
Dicke master equation are presented for a small system whose size is much
smaller than the wavelength of the radiation field. At the same time, ana-
lytic solutions are obtained within a semiclassical approximation to discuss
the importance of quantum correlation. Finally, we examine the effects of the
dipole–dipole interaction for an inhomogeneous system, as well as the effects
of reabsorption and re-emission of photons for a large system, both of which
violate the cooperative states and destroy some part of the coherence in the
system.
6.1 Dicke States and Superradiance
N Two-level Systems and Dicke States
The Dicke model consists of N two-level systems, as a representative of atoms

or quantum dots, coupled to the radiation field. The free Hamiltonian for the
two-level systems is written as
N

Htwo-level = Ω b†n bn , (58)
n=1
where Ω, b†n , and bn denote the excitation energy, creation and annihilation
operators of excitation in the n-th site, respectively. These operators can also
(n)
be defined by spin operators Ri (i = 1, 2, 3) and Bloch vectors (Pn , Vn , Wn )
as follows.
1 1
(bn + b†n ) = Pn ,
(n)
R1 = (59a)
2 2
(n) i † 1
R2 = (bn − bn ) = Vn , (59b)
2 2
(n) 1 † † 1
R3 = (bn bn − bn bn ) = Wn . (59c)
2 2
(n)
The spin operators Ri satisfy the commutation relations
(n) (n ) (n)
[Ri , Rj ] = iδnn ijk Rk , (60)
with Levi-Civita’s symbol ijk . Introducing total spin operators for the N
two-level systems as
N
N
N

(n) (n) (n)
R1 = R1 , R2 = R2 , R3 = R3 , (61)
n=1 n=1 n=1
we can rewrite the free Hamiltonian Htwo-level as
Htwo-level = ΩR3 , (62)
where the total spin operators also obey the commutation relations
[Ri , Rj ] = iijk Rk , (63)
similar to (60).
Let us consider a state in which N1 of the N two-level systems are in the
lower state |1 and N2 are in the excited state |2, where
1
N = N1 + N 2 , m= (N2 − N1 ) . (64)
2
Evidently m is a measure of the total population inversion, and it is an
eigenvalue of the spin operator R3 . Introducing the operator R2 = R12 + R22 +
R32 and using (63), we obtain the commutation relations
[Ri , R2 ] = 0 , for i = 1, 2, 3 , (65)
so that the collective operators Ri obey the same algebra as the angular mo-
mentum. Thus there exist the states |l, m that are simultaneous eigenstates
of both R3 and R2 as
1 1
R3 |l, m = m|l, m , for N ≥ m ≥ − N , (66a)
2 2
1
R2 |l, m = l(l + 1)|l, m , for N ≥ l ≥ |m| . (66b)
2
The quantum number l is called the cooperation number [49] and determines
the collectivity of cooperative phenomena. For the raising and lowering op-
erators as
N
N

R+ = b†n = R1 + iR2 , R− = bn = R1 − iR2 , (67)
n=1 n=1
it follows from the well-known properties of angular momentum operators

that

R± |l, m = (l ∓ m)(l ± m + 1)|l, m ± 1 , (68)
where the collective states |l, m are called Dicke states.
Superradiant States
In the following, let us assume that the system size is much smaller than the
wavelength of the radiation, and that the long-wave approximation is valid.
The interaction Hamiltonian between the two-level system and radiation in
the dipole coupling is given by [49]
HFint (t) = −µtot (t) · E(r = 0, t) , (69)
where the electric field is evaluated at the origin owing to the long-wave
approximation, and the total electric dipole of the system is written as
N

µtot = µ (bn + b†n ) = µ(R+ + R− ) . (70)
n=1
Here we suppose a transition from an initial state |l, m|vac to a final state
|ψ, where |l, m and |vac represent a Dicke state and the vacuum of the
radiation field, respectively. Then the transition probability can be written
as
−ψ|µ · E − (t)R− (t)|l, m|vac , (71)
where E − (t) is the e−iωt -dependent part of the electric field E(t). By squaring
the absolute value of the transition probability and summing over all possible
final states, the probability of emission of radiation photons wemit is given
as [49, 77]

wemit = |ψ|µ · E(t)R− (t)|l, m|vac|2
ψ
= l, m|R+ (t)R− (t)|l, mvac|µ · E + (t)µ · E − (t)|vac
= R+ (t)R− (t)A = (l + m)(l − m + 1)A , (72)
where (68) is used for the final expression, and A = vac|µ · E + (t)µ ·
E − (t)|vac is identified as the Einstein A coefficient for each two-level system
like an atom or a quantum dot.
If all quantum dots are in the ground state |1, then m = −N/2, and
from (66b) we must have l = N/2. Hence, from (72) the radiation rate wemit
is zero as expected. If all quantum dots are in the excited states |2, then
m = N/2 and we again have l = N/2. Hence wemit = N A, which is just
what one would expect from a group of N independently radiating quantum
dots. However, the situation is quite different if the initial state is not the
completely excited state. Let us consider a state in which half of the quantum
dots are excited and half are not, so that m = 0. Then we have
wemit = l(l + 1)A , (73)
and l can have any value between 0 and N/2. The larger the value of l,
the larger is the collective rate of radiation of the system. In particular, if
l = N/2, then
1
wemit = N (N + 2)A , (74)
4
which maximizes the value of cooperative spontaneous radiation. Since the
maximum radiation probability is proportional to the square of the site
number N , it becomes very large for large N . This phenomenon is called
superradiance, and the state |l = N/2, m = 0 is called a superradiant
Dickestate. From the
definition of the total Bloch vector (P, V, W ) as
P = n Pn , V = n Vn , W = n Wn , it follows that the Bloch vec-
tor for the superradiant state is on the P –V plane and its length reaches a
maximum, as shown in Fig. 21.
Fig. 21. Total Bloch vector in a superradiant state.

It is on the P –V plane, and its length has a maximal
value
6.2 Dicke Master Equation and Solutions for a Small System
In this section the Dicke master equation is derived for a small system to
discuss dynamical manifestation of superradiant states described in the pre-
ceding section.
Nakajima–Zwanzig’s Method and Generalized Master Equation
According to Mandel and Wolf [77, 78], we outline the derivation of a gener-
alized master equation by projection operators to project out of the density
operator for a total system that part on which our interest is focused [79–82].
The total system can be described in the interaction representation by the
Liouville–von Neumann equation for the density operator ρ(t) as
∂ρ(t) 1
= [HI (t), ρ(t)] = −iL(t)ρ(t) , (75)
∂t i
where HI (t) is the interaction Hamiltonian, and L · · · = [HI (t), · · · ]/ is the
Liouvillian operator, as before. We now introduce a time-independent pro-
jection operator P :
P2 = P , (76)
which is chosen so as to project out the most relevant part of ρ(t). Using the
complementary operator (1 − P ), we can express ρ(t) as
ρ(t) = P ρ(t) + (1 − P )ρ(t) . (77)
On multiplying both sides of (75) by P on the left and then using (77) to
substitute for ρ(t) on the right, we have
∂ρ(t)
P = −iP L(t)P ρ(t) − iP L(t)(1 − P )ρ(t) . (78)
∂t
Similarly, after multiplying (75) by (1 − P ) on the left, we obtain
∂ρ(t)
(1 − P ) = −i(1 − P )L(t)P ρ(t) − i(1 − P )L(t)(1 − P )ρ(t) . (79)
∂t
Formally integrating the first-order differential equation (79) and solving for
(1 − P )ρ(t), we find
t
(1 − P )ρ(t) = exp −i(1 − P ) L(t )dt (1 − P )ρ(0)
0
t t
−i exp −i(1 − P ) L(t )dt
0 τ
×(1 − P )L(τ )P ρ(τ )dτ . (80)
Substitution of (1 − P )ρ(t) into (78) gives
∂ρ(t)
P = −iP L(t)P ρ(t)
∂t
t
−iP L(t) exp −i(1 − P ) L(t )dt (1 − P )ρ(0)
0
t t
−P L(t) exp −i(1 − P ) L(t )dt
0 τ
×(1 − P )L(τ )P ρ(τ )dτ . (81)
If we set
t
U (t, τ ) ≡ exp −i(1 − P ) L(t )dt , (82)
τ
we finally obtain Nakajima-Zwanzig’s generalized master equation [79–82] as

follows:
∂ρ(t)
P = −iP L(t)P ρ(t) − iP L(t)U (t, 0)(1 − P )ρ(0)
∂t
t
−P L(t) U (t, τ )(1 − P )L(τ )P ρ(τ )dτ . (83)
0
Application to the Dicke problem

As an example, let us consider the application of (83) to the Dicke problem,
i.e., to the interaction of a group of N closely spaced, identical two-level
atoms with an electromagnetic field. For simplicity, we assume that two-level
atoms are at the origin, and we make the rotating-wave approximation. Then
the Hamiltonian is given as [57]
†
HD = ΩR3 + ωa akλ + (gkλ a† R− + h.c.) , (84)
kλ kλ
kλ kλ
where a† (akλ ) is the photon creation (annihilation) operator specified by
kλ
wavevector k, polarization λ, and frequency ω, and the coupling constant
gkλ is given by the formula
iΩ
gkλ = √ ( · µ) . (85)
20 ωV kλ
Here kλ represents the polarization vector, and µ denotes the matrix element
of the dipole moment in the two-level (atomic) system.
In the following, we are mainly interested in the two-level (atomic) system
as the relevant system and identify the ‘reservoir’ with the radiation field, so
that the projection operator P is given by
P · · · = |vacvac|TrF · · · , (86)
where TrF denotes that the trace is taken over the radiation field states. As
the initial state of the coupled system, we also assume the direct product
state with the radiation field in the vacuum state
ρ(0) = ρA (0) ⊗ |vacvac| , (87)
with the reduced density operator defined as ρA (t) = TrF ρ(t). In order to
work in the interaction picture, we transform the interaction Hamiltonian
into

HFint (t) = (gkλ a† R− ei(ω−Ω)t + h.c.) . (88)
kλ
kλ
Under these conditions, the first term on the right-hand side of (83),
P L(t)P ρ(t), can be written as
1
P L(t)P ρ(t) = |vacvac|TrF [HFint (t), |vacvac|ρA (t)]
⎛ ⎞

= |vacvac| ⎝ gkλ vac|a† |vacR− ei(ω−Ω)t ρA (t) − h.c.⎠
kλ
kλ
=0,
(89)
because
vac|a† |vac = 0 = vac|akλ |vac . (90)

kλ
In addition, the second term on the right-hand side of (83) can be rewritten
to
(1 − P )ρ(0) = (1 − |vacvac|TrF )ρA (0)|vacvac| = 0 . (91)
Substituting (89) and (91) into (83), we have

t
∂ρ
P = −P L(t) L(τ )P ρ(τ )dτ , (92)
∂t 0
where we approximate U (t, τ ) ∼ 1 (Born approximation). Explicit use of (86)

yields the equation
t
∂ρA
= −TrF L(t) L(τ )|vacvac|ρA (τ )dτ . (93)
∂t 0
After transforming the variable as τ = t − τ in the last integral, we can

reduce (93) to
t
∂ρA
= −TrF L(t) L(t − τ )|vacvac|ρA (t − τ )dτ . (94)
∂t 0
Further approximation of replacing ρA (t − τ ) by ρA (t), which is known as

the Markov approximation, provides the final relation
t
∂ρA
= −TrF L(t) L(t − τ )|vacvac|ρA (t)dτ
∂t 0
t
1
= − 2 TrF [HFint (t), [HFint (t − τ ), |vacvac|ρA (t)]]dτ
0
t
1
= − 2 TrF {HFint (t)HFint (t − τ )|vacvac|ρA (t)
0
−HFint (t)|vacvac|ρA (t)HFint (t − τ )}dτ + h.c. (95)
Noticing (88) and the relations

vac|akλ a† |vac = δkk δλλ ,
kλ
vac|a† a† |vac = vac|akλ ak λ |vac = vac|a† ak λ |vac = 0 , (96)
kλ k λ kλ
we have the following equation
t
∂ρA
=− |gkλ |2 {ei(Ω−ω)τ R+ R− ρA (t) + h.c.
∂t 0
kλ
−2 cos(ω − Ω)τ R− ρA (t)R+ }dτ . (97)
In the limit t 1/Ω we put
t
|gkλ |2 ei(Ω−ω)τ dτ ≡ β + iγ , (98)
0
kλ
where β is half the Einstein A coefficient and γ is a frequency shift due to
the interaction. Finally we obtain the appropriate master equation describing
spontaneous emission from a collective two-level (atomic) system, which is
called the Dicke master equation [56, 57, 59]
∂ρA
= −β(R+ R− ρA − 2R− ρA R+ + ρA R+ R− ) − iγ[R+ R− , ρA (t)] , (99)
∂t
or
∂ρA
= −β([R+ , R− ρA ] + [ρA R+ , R− ]) − iγ[R+ R− , ρA (t)] . (100)
∂t
Numerical Solutions of Dicke Master Equation
The rate of photon emission was obtained perturbatively in Sect. 6.1. In this
section, we examine the time evolution of this emission rate, using the Dicke
master equation (99). Since the rate of energy dissipation from the two-level
(atomic) system is proportional to the emission rate, −∂R3 /∂t should be
evaluated. Multiplying (99) by R3 on the right and taking the trace, one finds
∂
R3 = β{−R3 R+ R− + 2R+ R3 R− − R+ R− R3 }
∂t
−iγ{R3 R+ R− − R+ R− R3 } . (101)
In order to simplify the above equation, we use the following relations
[R+ , R3 ] = −R+ , (102)
1 †
R+ R− = R3 + N bn bm , (103)
2 n m=n
which follow from the collective operators defined in (67) and the commuta-
tion relations in (63). In addition, using (103), we can show that

[R+ R− , R3 ] = [b†n bm , R3 ]
n m=n

= [b†n bm − b†n bm ] = 0 , (104)
n m=n
and rewrite (101) as
∂
R3 = 2β[R+ R3 R− − R3 R+ R− ] . (105)
∂t
With the help of (102), this leads to the final equation
∂
R3 = 2β[(R3 R+ − R+ )R− − R3 R+ R− ]
∂t
= −2βR+ R− . (106)
It follows from the above equation that the time evolution of R+ R− is
required to evaluate the rate of energy dissipation. Multiplying (99) by R+ R−
on the right and taking the trace, we then find
∂
R+ R− = −2β{R+ R− R+ R− − R+ R+ R− R− }
∂t
−iγ[R+ R− , R+ R− ]
= −4β{R+ R− − R+ R− R3 } , (107)
where we used the following relation to obtain the second line from the first
line
R+ R+ R− R− = R+ R− R+ R− − 2R+ R− + 2R+ R− R3 . (108)
Neglecting the quantum correlation and putting
R+ R− R3 = R+ R− R3 , (109)
we then have
∂
R+ R− = −2βR+ R− (1 − R3 ) . (110)
∂t
Simultaneously solving (106) and (110), one can obtain the final result. Or
putting z = R3 , this leads to the following nonlinear equation
z̈ = −4β ż(1 − z) . (111)
This allows us to evaluate the rate of energy dissipation, −∂ R3 /∂t = −ż,
and thus to obtain the rate of photon emission.
Under the following initial conditions
R3 (0) = N/2 , R+ R− (0) = N , (112)

Fig. 22. Time evolution of the rate of photon emission that is numerically obtained
from (111) with the initial condition of a totally inverted state. The vertical axis is
represented in units of the Einstein A coefficient A = 2β for β = 0.05. The dotted,
dashed, and solid curves show the results for the site number N = 12, 15, and 20,
respectively
which correspond to l(0) = N/2 and m(0) = N/2 in terms of Dicke states,
we solve (111) numerically. The result is shown in Fig. 22 for the case of
β = 0.05 and the site number N = 12, 15, and 20. It follows from Fig. 22
that the peak value of the rate of emission becomes higher and the width
gets narrower, as the site number becomes larger. As will be shown in the
next subsection, a pulse of radiation whose height is proportional to N 2 and
width is proportional to 1/N is emitted from the two-level system prepared
in the initial condition.
Analytical Solutions of Dicke Master Equation

in a Semiclassical Approximation
In the preceding subsection, we obtained simultaneously coupled equations
after neglecting the quantum correlation more than third order, for example,
approximating R+ R− R3 as R+ R− R3 , and numerically examined the
emission rate of radiation to predict a characteristic feature. Here in this
section, an analytical solution will be given for time evolution of the emission
rate after making more approximations [55, 58, 59].
Using the polar angle θ of a Bloch vector (P, V, W ), we can express the
expectation value of R3 as
N
R3 = cos θ (113)
2
and that of R+ R− as
R+ R− = R+ R− = R1 2 + R2 2

N2
= R2 − R3 2 = sin2 θ , (114)
4
where we neglected the quantum correlations more than second order as
R+ R− = R+ R− , which is called a semiclassical approximation [57].
Substitution of z = cos θ into (106) leads to
N
ż = −A (1 − z 2 ) , (115)
2
where A is the Einstein A coefficient and is set to A = 2β. Similarly we can
rewrite the left-hand side of (110) as
∂ N2 N2
(1 − z 2 ) = − z ż , (116)
∂t 4 2
and the right-hand side as

N2 N N3
−2A (1 − z 2 ) 1 − z A (1 − z 2 )z , (117)
4 2 4
where we assumed N 1. Both results allow us to simplify Eq. (110), and
one finds
N
z ż = −A z(1 − z 2 ) , (118)
2
which is the same as (115). Thus, both (106) and (110) can be transformed
to the same equation,
1 dz
= −dt , (119)
Γ 1 − z2
with Γ = AN/2, and can be easily solved as
z = − tanh Γ (t − t0 ) . (120)
We can finally obtain the analytical solution of the emission probability as
N Γ2
I ∝ −∂t R3 = − ż = sech2 Γ (t − t0 ) , (121)
2 A
which represents a cowbell-shaped function with a width of 2.6/N β and a
peak position at t = t0 . This function is plotted in Fig. 23 for the case of
β = 0.05, t0 = 3, and N = 20. Comparing the result for N = 20 in Fig. 22
with the current result, one finds that a semiclassical treatment can describe
radiation profile of the Dicke model very well. However, it should be noted
that there must be nonzero total dipole moment in the initial condition in the
semiclassical description because the semiclassical approximation is not valid
for the totally inverted system [58, 59], where quantum correlation should be
essentially included.
Fig. 23. Time evolution of the rate of photon emission plotted according to the
analytic solution obtained with the semiclassical approximation. The vertical axis
is represented in units of the Einstein A coefficient A = 2β. The parameters are set
to β = 0.05, t0 = 3, and N = 20, corresponding to Fig. 22
6.3 Effect of the Dipole–Dipole Interaction
The dipole–dipole interaction generally breaks the permutation symmetry

of the atom (two-level system)–field coupling because various atoms in the
sample have different close-neighbor environments. This perturbation is anal-
ogous to an inhomogeneous dephasing of the dipoles, leading to a loss of
dipole–dipole correlation. In order to discuss such an effect, we employ the
following interaction Hamiltonian, instead of (69)
N

HFint = − {E + (r n ) · P (r n ) + E − (r n ) · P † (r n )} , (122)
n=1
where the electric field operator E is divided into the positive and negative
frequency parts E + and E − defined as

ck
E (r n ) = i
+
eλ akλ eik·r n , (123a)
20 V
kλ
E − (r n ) = (E + (r n ))† . (123b)
The creation and annihilation operators a† and akλ act on the mode of the
kλ
radiation field with wavevector k and polarization eλ , and obey the boson
commutation relation. An arbitrary quantization volume is denoted as V .
The dipole operator is similarly defined as
P (r n ) = µ(b†n + bn ) . (124)
Using this interaction Hamiltonian, we can obtain the following master equa-
tion in the Born–Markov approximation [62, 83]
dρ
= −i Ωnm [b†n bm , ρ] − γnm [b†n bm ρ + ρb†n bm − 2bm ρb†n ] , (125)
dt nm
n=m
with
µ2 Ω 3
γnm = F (Ωrnm /c) , (126a)
16π 2 c3 0
∞
µ2 P
Ωnm = − dkk 3 F (krnm ) , (126b)
3
16π 0 0 k − Ω/c

F (krnm ) = dΩk (e · )2 eik·r nm , (126c)
where Ω is the angular frequency corresponding to the excitation energy of

the two-level system, and rmn is the absolute value of the vector r nm =
r n − r m . The solid angle Ωk is defined as dk = k 2 dkdΩk . Moreover, we have
used the following formula
∞
P
cdτ e±i(kc−Ω)τ = πδ(k − Ω/c) ± i , (127)
0 k − Ω/c
whose real and imaginary parts produce site-dependent relaxation coefficient

γnm and frequency shift Ωnm , respectively. The integral in (126c) can be
explicitly calculated and gives [54]

( · r nm )2 sin krnm
F (krnm ) = 4π 1 − 2
rnm krnm

( · r nm )2
cos krnm sin krnm
+4π 1 − 3 2
− , (128)
rnm (krnm )2 (krnm )3
and one readily obtains, at the limit krnm → 0

µ2 ( · r nm )2
Ωnm = 3
1−3 2
, (129)
4π0 rnm rnm
which is the static, or Coulombic dipole–dipole interaction.

In order to examine the effects of the dipole–dipole interactions, we evalu-
ate the magnitude of the total Bloch vector R2 , which is conserved in Dicke
states. According to (125), one finds [62]

d 2
R = −i [R2 , Ωnm b†n bm ]
dt
n=m

= −2i Ωnm Rz(m) (b†n bl − b†l bn ) , (130)
n=m=l
Fig. 24. Arrangement of three two-level systems. Each two-level system is config-
ured at the apex of an isosceles triangle
from which it follows that R2 is generally not conserved, but is a conserved
quantity only if the site number N is equal to 2, or only if Ωnm is constant for
all suffixes [62, 69]. Therefore the state is not expressed by Dicke states, and
quantum coherence of the two-level system is destroyed by the dipole–dipole
interactions.
Now we consider how the dipole–dipole interactions affect superradiance.
Let the site number be three and be arranged as shown in Fig. 24. For sim-
plicity, we assume Ω12 = Ω13 = Ω23 and γmn = γ.
The following eight bases are employed to take matrix elements of (125)
and to solve the simultaneous differential equations.
|s1 , s2 , s3 (si = ±) , (131)
where the sign + indicates that the upper level is occupied, while the sign −
means that the lower level is occupied. Under the initial condition of + +
+|ρ(0)| + ++ = 1, we obtain the matrix elements of the density operator
ρ(t) at time t to express the radiation intensity

I(t) = b†n bm = R+ R− . (132)
n,m
Using the Pauli matrices σµ (µ = 0, 1, 2, 3), we can write nonzero components
of the density matrices as µ xµ σµ . The vector x = (x1 , x2 , x3 ) executes fast
√
precession about a “torque” Γ = 2 2Ω12 , 0, Ω23 (see Fig. 25). When we

introduce the rotation R β, 2̂ that rotates about the 2 axis through an angle
√
2 2Ω12
tan β = , (133)
Ω23

then the vector Γ is transformed as R β, 2̂ Γ = (0, 0, Γ ) with
2
√
Γ = 2 .
2 2Ω12 + Ω23
Fig. 25. Relation between the angles β and χ. In the case of pure superradiance
(Ω12 = Ω23 ), the torque Γ is given by Γ0
The solutions of the differential equations for the density matrices can be
obtained from a similar rotation of x with cos χ defined as [62]
1
cos (χ + β) = . (134)
3
When cos χ = 1 or χ = 0, Ω12 = Ω23 is satisfied and the Dicke superradiant
states become exact eigenstates, while cos χ = 0 gives orthogonal states to
the superradiant states. This means that the deviation of χ from zero shows
the inhomogeneity of the system, or a measure of the coherence properties.
According to Coffey and Friedberg [62], we present the radiation intensity
I(t) in Fig. 26, which shows the peak and tail of I(t) become reduced and
longer, respectively, as cos χ varies from one to zero corresponding to the
sequence of a, c, d, and b in Fig. 26.
Tokihiro et al. [69] also examined the effects of the dipole–dipole interac-
tion on superradiance, using a linear excitonic system, in which excitons can
hop from one site to its nearest neighbors via the dipole–dipole interaction.
They used the following master equation in the interaction representation
dρ
= −iJ [b†n bn+1 + b†n+1 bn , ρ]
dt n
−γ[R+ R− ρ + ρR+ R− − 2R− ρR+ ] , (135)
where the hopping energy of an exciton and the radiation decay rate are
denoted as

µ2 ( · a)2
2J = 3
1−3 , (136)
4π0 a a2
µ2 Ω 3
γ= , (137)
3πc3 0
respectively. The first term in (135) corresponds to the Hamiltonian of the
XY-model in a spin system whose exact solutions are well known [84]. Tak-
Fig. 26. Normalized radiation intensity for three two-level systems with different
Ω12 /Ω23 : (a) Ω12 = Ω23 , cos χ = 1, corresponding to pure superradiance, (b)
cos χ = 0, (c) Ω12 = 8Ω23 , cos χ = 0.96, (d) Ω12 = Ω23 /8, cos χ = 0.63, (e)
I = 3e−γt , where three two-level systems radiate independently. The vertical axis
is represented in units of the Einstein A coefficient while the horizontal axis is
represented by γt
ing matrix elements of (135) in terms of the solutions as basis functions,

they numerically solved the master equation under the initial condition of
a totally inverted system. The result, similar to that of Coffey and Fried-
berg [62], shows a reduced peak intensity with extension of the tail, and a
substantial deviation from Dicke’s superradiance that is proportional to the
system size N . In addition, they examined the initial-condition dependence,
and showed the oscillation of the radiation profile obtained from a partially
excited state [83]. This resembles the effects of reabsorption and re-emission
of photons, which is discussed in the next subsection.
6.4 Large-Sample Superradiance
In the preceding section, we discussed superradiance in a small system whose

size is much smaller than the wavelength of the radiation field. Now we con-
sider large-sample superradiance, taking into consideration a finite system
size. In such a system, an inhomogeneous arrangement in the two-level sys-
tem cannot be neglected, as well as the effects of reabsorption of emitted
photons. Dipole–dipole interactions, in particular, become important for the
inhomogeneous system and correlation of dipoles in the system is degraded to
result in weakened radiation intensity [62, 69], as discussed in the preceding
subsection. Reabsorption of emitted photons causes stimulated absorption
and emission of photons in the system, which leads to oscillation of radi-
ated pulses [63, 65]. We will analyze this effect semiclassically, according to
Banifacio et al. [60, 63, 85], and comment on experimental observations.
Effect of Reabsorption and Reemission of Photons
As samples have a volume with linear dimensions much larger than the wave-
length of the radiation, some quantitative differences appear in the dynamics
of the system evolution due to the effects connected with the propagation of
the electromagnetic field along the emitting sample. According to Bonifacio
et al., we use a laser model to describe equations of motion as
ρ̇(t) = −i(LAF + iΛF )ρ(t) , (138)
where LAF denotes the Liouvillian operator, and ΛF is the dissipation term
that represents photons coming away from the cavity. They are defined as
follows:
1
LAF ρ = [HAF , ρ] , (139a)

HAF = g(aR+ + a† R− ) , (139b)
† †
ΛFX = γ{[aρ, a ] + [a, ρa ]} , (139c)
where the constants (g and γ) and the operators have their usual meanings,
as before. If we multiply (138) on the left by a, R+ , and R3 , respectively, and
if we take the trace, we obtain equations for α(t) = ia(t), r1 (t) = R+ (t),
and r3 (t) = R3 (t) as
ṙ1 = 2gαr3 , (140a)
ṙ3 = −2gαr1 , (140b)
α̇ = gr1 − γα , (140c)
where we have used the semiclassical approximation to neglect quantum cor-
relations. Introducing the polar angle θ of a Bloch vector as
N N
r1 (t) = sin θ , r3 (t) = cos θ , (141)
2 2
one finds the following differential equation that shows the damped harmonic
oscillation [60, 63]
θ̈(t) + γ θ̇(t) = N g 2 sin θ(t) , (142)
and the equations for conservation of the magnitude of the Bloch vector R2
and energy of the system
∂t (r12 + r32 ) = 0 , (143a)
∂t (α2 + r3 ) = −2γα2 . (143b)
Fig. 27. Time evolution of the rate of radiation from a large sample. The vertical
axis is represented in units of the Einstein A coefficient as before. The parameters
used are γ = 0.1, g = 1/14, and N = 20
It follows from (143b) that the dissipation energy from the system per unit
time is 2γα2 , and it corresponds to radiation energy from the system. Thus
we can write radiation intensity I(t) as
γ 2
I(t) = 2γα2 = θ̇ . (144)
2g 2
In Fig. 27, as an example, we show a numerical result obtained from (142)

and (144), where the parameters γ = 0.1, N = 20, and g = 1/14 are used.
This figure clearly illustrates an oscillation of large-sample superradiance that
was first predicted by Bonifacio et al. [60, 63].
If we assume that the reabsorption of photons emitted from the system
can be negligibly small enough to drop the θ̈(t) term, (142) can be reduced
to
N g2
θ̇(t) = sin θ(t) . (145)
γ
Multiplying sin θ on both sides and setting z = cos θ, we can derive
N g2
ż = − (1 − z 2 ) , (146)
γ
which is the same as (115), the Dicke master equation for a small system
obtained semiclassically.
At the end of this subsection, we briefly comment on experimental obser-
vation of superradiance. Skribanovitz et al. [64] experimentally showed large
sample superradiance similar to Fig. 27, using optically pumped HF atomic

gas. For a small system, the radiation profile from CuCl quantum dots was
shown to have a single pulse [66].
7 Radiation from the Dipole-Ordered States

In this section we examine the radiation property of the dipole-ordered states
discussed in Sects. 4.2 and 5.2. Our model has excluded radiation fields or
free photons from the preceding discussions about the dynamics because the
interaction energies of excitons with optical near fields are much larger in
our system than those with radiation fields. This means that radiation fields
are weakly coupled with our system so as not to disturb the dynamics of
it. On the basis of such an understanding, we investigate radiation from a
quantum-dot system whose dynamics is driven by optical near fields.
7.1 Radiation Property of the Dipole-Ordered States
According to the formulation developed in Sect. 6, we now investigate the

radiation property of the dipole-ordered states discussed in Sects. 4.2 and
5.2. From the expression (72) the radiation intensity at time t is determined
by the radiation factor, R+ R− = (l + m)(l − m + 1), i.e., the expectation
value of the operator
2 2
P V W
R+ R− = R12 + R22 + R3 = + + , (147)
2 2 2
which indicates that there are two elements mainly contributing to the ra-
diation factor; one is the collectiveness of the system measured by the co-
operation number l, and the other is the total energy of the quantum-dot
system given by R3 = m. Since the dynamics of the system is driven in our
system by localized photons, the evolution of the radiation factor is also de-
scribed in terms of l(t) and m(t) that are developed according to the localized
photon–exciton interaction.
Numerical results of (46) and (47) are shown in Figs. 28 and 29. The
upper parts in Fig. 28 show the time evolution of the radiation factors, while
the lower parts illustrate the dipole distribution when each radiation factor
has a maximal value. The total dipole in Fig. 28a is smaller than those in
Figs. 28b and 28c, and the peak value of the radiation factor is also smaller. It
follows from the figure that the radiation factor increases as the total dipole
becomes larger. The peak values of the radiation factor in Figs. 28b and 28c,
15 and 14, correspond to the value for the Dicke’s superradiance, which is
obtained as 20 for N = 8 from (74). We thus expect that quasisteady states
shown in Figs. 28b and 28c are close to the superradiant states, and that the
total Bloch vectors for such states are on the P − V plane.
Fig. 28. Time evolution of the radiation factor R+ R− (upper) and the dipole
distribution at time indicated by the arrows when the radiation factors take max-
imal values (lower) for (a) a system belonging to the first group with an initial
population difference Wn = (0 + + − − − ++), (b) a system belonging to the
third group with an initial population difference Wn = (0 + + + + + ++), and
(c) a system belonging to the fourth group with an initial population difference
Wn = (0 − + − + − +−)
Fig. 29. Time evolution of the radiation factor (left) and the polar angle of the
total Bloch vector (right) for a system belonging to the third group. The polar angle
is measured from the Wn = (− − − − − − −−) axis
In order to check whether the total Bloch vectors for the states belonging
to the third group in Table 1 (see Fig. 28b) are on the P − V plane or not, we
examine the time evolution of the polar angles of the Bloch vectors as well
as the radiation factors. As shown in Fig. 29, the polar angle of the Bloch
vector takes 90 degrees, that is, on the P − V plane when the radiation factor
has a maximal value. Therefore we conclude that the system belonging to
the third group in Table 1 is in transition to a quasisteady state close to
Fig. 30. The dependence of the width and the peak point of superradiant pulse
on localized photon–exciton couplings g: double logarithmic plot (left) and the
definitions of the pulse width δt, the pulse peak points TA , and the bottom point
of pulse valley TB (right)
the superradiant Dicke state, judging from its large radiation factor and the
polar angle of the total Bloch vector.
In order to see the mechanism of the superradiance phenomena in our
model we shall investigate the dependence of the localized photon–exciton
couplings g on the width of superradiant pulse. The result is shown in Fig. 30.
From this plot we find a relation as follows
1
T ∝ . (148)
g
Recalling the mechanism of the oscillation of superradiant pulse in the Dicke
model explained by Bonifasio (see (142)) we know that this result means
the oscillation of superradiant pulses in our model is caused by emission and
absorption of localized photons.
Figures 28b and 28c indicate that multiple peaks appear in the radiation,
or that multiple pulses are emitted from the system. One may think, as a
possible origin, that such a phenomenon stems from the recurrence inherent
in an isolated system. However, such multiple pulses may survive even if the
system becomes dissipative, which will be examined in detail in the next
section.
8 Radiation from a Dissipative System

We have assumed in the previous section that the radiation field is so weak as
not to disturb the exciton dynamics of a quantum-dot system. When radiation
pulses are emitted from the system, however, energy has to be dissipated, and
it is interesting but not clear whether multiple pulses shown in Figs. 28b and
28c are emitted from the system or not. In this section we thus study the
radiation profile, adding a radiation field to the system as a reservoir that

does not affect the dynamics but makes the system a dissipative one.
Suppose the Hamiltonian
H2 = HQDeff + HF + HFint , (149)
where HQDeff is the effective Hamiltonian given by (51a) that describes the N
two-level quantum-dot system interacting with localized photons. The Hamil-
tonians HF and HFint describe the free-radiation field and the exciton–free
photon interaction, respectively. Explicit Hamiltonians are written as

HF = ωk a† ak,λ , (150)
k,λ
k,λ
i(ω −Ω)t
HFint (t) = gk,λ a† R− e k + h.c. , (151)
k,λ
k,λ
where creation and annihilation operators of a free photon with wavevector k,
polarization λ, and frequency ωk are denoted as a† and ak,λ , respectively.
k,λ
The coupling constant between the free photon and exciton is given as

iΩ
gk,λ = √ µ · ∗k,λ , (152)
V 2ωk 0
where V , ∗k,λ , and 0 represent the quantization volume of the radiation

field, the unit polarization vector, and the dielectric constant in vacuum,
respectively. Using the Hamiltonian H2 , we write the Liouville equation as
∂ρ(t) 1
= [H2 , ρ(t)] = −iL2 ρ(t) , (153)
∂t i
and eliminate the degrees of freedom of the radiation field with the help of a
projection operator defined as [77]
P · · · = |00|TrF · · · , (154)
with the vacuum of the radiation field |0. Moving to the interaction rep-
resentation, we obtain equations of motion for the density operator ρ̃(t) as
follows:
∂ ρ̃
P = −iPL2 (t)P ρ̃(t) − iPL2 (t)U (t, 0)(1 − P)ρ̃(0)
∂t
t
−PL2 (t) dτ U (t, τ )(1 − P)L2 (τ )P ρ̃(τ ) , (155)
0
where L2 (t) is the Liouville operator associated with H2 (t), and the operator
U (t, τ ) is defined as
t
U (t, τ ) = exp −i(1 − P) L2 (t )dt . (156)
τ
With the Born–Markov approximation applied to the third term of (155), we

obtain the following equation for ρ̃A = P ρ̃ as
∂ ρ̃A
= −iLQDeff (t)ρ̃A (t)
∂t
t
−TrF LFint (t) dτ UQDeff (t − τ )LFint (t − τ )|00|ρ̃A (t) , (157)
0
with
UQDeff (t) = exp (−iLQDeff t) , (158)
where LQDeff and LFint are the Liouville operators associated with the Hamil-
tonians HQDeff and HFint , respectively. Moreover, using the Born approxima-
tion that neglects the exciton operators of higher than the second order [77],
we approximate
UQDeff (t) ∼ 1 (159)
to obtain a compact equation as

∂ ρ̃A
= −iLQDeff (t)ρ̃A (t) + β {[R− ρ̃A (t), R+ ] + [R− , ρ̃A (t)R+ ]}
∂t
−iγ[R+ R− , ρ̃A (t)] , (160)
with
t
i(Ω−ω )τ
β + iγ ≡ |gk,λ |2 e k dτ , (161)
0
k,λ
which is exactly the same form as Lindblad’s master equation [86] describing
a general Markovian dynamics for a dissipative quantum system. Here the
real and imaginary parts of the right-hand side of (161) are designated as β
and γ, respectively. In the following we neglect the energy shift as γ = 0,
for simplicity. Note that the second and third terms of the right-hand side of
(160) are known as Dicke’s master equation [57, 60, 62].
8.1 Semiclassical Description with the Effective Hamiltonian
Neglecting quantum correlation between excitons [58], we approximate the

total density operator ρ̃A as a direct product of the density operator ρn at
each site n
ρ̃A = ρn , (162)
n
and then solve (160). Noticing that the dynamics governed by the original
Hamiltonian (4) is not rigorously identical to the one described by the effec-
tive Hamiltonian (51a), we use an isolated system described by the effective
Fig. 31. Time evolution of the radiation factor for dissipative systems (upper) and
isolated systems (lower). We consider the following three cases for both systems:
(a) a “ferromagnetic” case, (b) an “antiferromagnetic” case that is turned into a
dipole-ordered state after manipulating the initial population differences, and (c) a
“dipole-forbidden” case discussed in Sect. 4.3. The parameters β = 0.05 and γ = 0
are used
Hamiltonian (51a) in order to clarify the dynamics governed by (160) for a

dissipative system whose relevant system is described by the same Hamilto-
nian (51a). By comparing the radiation factors for the isolated system and
the dissipative system, the similarity and the difference are discussed.
Figure 31 shows the time evolution of the radiation factor for a “ferro-
magnetic” system belonging to the third group of Table 1, an “antiferromag-
netic” system that belongs to the fourth group of Table 1 and is turned to a
dipole-ordered state after manipulating the initial population differences, and
a “dipole-forbidden” system that is converted to a dipole-ordered state by the
local manipulation discussed in Sect. 4.3 as Wn (0) = −Pn (0). It follows
from the lower part of Fig. 31a that superradiance is suppressed and does
not manifest itself due to the difference between the original Hamiltonian (4)
and the effective Hamiltonian (51a). On the other hand, superradiant multi-
ple pulses are generated in the isolated system and survive in the dissipative
system for both cases shown in Figs. 31b and 31c. Therefore we find with
a semiclassical approach that multiple pulses, as predicted for an isolated
system, can be emitted superradiantly from dipole-ordered states even in a
dissipative system coupled to a radiation reservoir.
8.2 Quantum Correlations
It is well known that superradiance in the Dicke model [49] occurs from a
state where all excited states of all sites are occupied. On the other hand,
the semiclassical approach discussed above cannot predict the occurrence of
superradiance of the system when the total dipole of the system is zero as
an initial condition [58]. This means that quantum fluctuations and corre-
lations should be properly included so as to correctly describe the radiation
properties of a system with no initial dipoles, and that the semiclassical ap-
proximation is not appropriate in this case. Thus we numerically solve the
master equation (160), taking quantum correlations into account, and we
investigate the radiation properties of the dissipative system. In particular,
we compare the results obtained from our model, i.e., the localized photon
model with those obtained from the Dicke model, for which the first term of
the right-hand side is dropped from (160). Some of such results are shown in
Figs. 32 and 33.
Figure 32a shows the time evolution of the radiation factor for the case
that all the populations are completely in the excited states and there are
no dipoles as an initial condition. The solid curves are the results for our
model, while the dashed curves represent the results for the Dicke model. It
follows from the figure that a single superradiant pulse is emitted in both
models, but that the peak value of the superradiant pulse is reduced, while
the tail is extended in our model. In Fig. 32b, we present the result obtained
from the initial√condition as a zigzag profile of the dipole distribution √ of
Pn = (−1)n / 2 and flat population differences of Wn = 1/ 2. The
system corresponds to a dipole-forbidden state as shown in Fig. 11, where the
total dipole is always zero. The radiation profiles obtained for both models are
qualitatively, same as found as in Fig. 32a. The time evolution of the radiation
factor is illustrated in Fig.
√ 32c for the case that initially the alternating dipole
is set as Pn = (−1)n / 2 and the sign of the population difference in each
QD is set opposite to that of the corresponding dipole as Wn = −Pn .
Owing to the flip hypothesis, a dipole-ordered state emerges in this case,
and a distinct difference between the two models is observed: our model
(solid curve) shows superradiance, while the Dicke model (dashed curve) does
not. Since we can infer that the difference stems from the dipole-ordering
phenomenon, we further examine this case in order to clarify the relation
between the radiation factor and the total dipole of the system.
Figure 33 presents the time evolution of the radiation factor (upper) and
the total dipole (lower) by changing the exciton–localized-photon coupling
constant g as g = 0.5, 0.8, and 1.2. It is found that the frequency of the
total dipole increases as the coupling constant becomes strong. As a result,
the oscillation frequency of the radiation factor increases, and thus multiple
pulses are emitted superradiantly from the dissipative system. Therefore the
difference between the two models, as we expected, originates from the occur-
Fig. 32. Time evolution of the radiation factor obtained with quantum correlations.
The solid curves are the results for our model, while the dashed curves represent the
results for the Dicke model. The dissipative system is assumed, and the parameters
N = 4, β = 0.05, and γ = 0 are used. In addition, the following
√ initial conditions
√
are used: (a) Pn = 1 and Wn = 1, (b) Pn = (−1)n / 2 and Wn = 1/ 2
which correspond
√ to a dipole-forbidden state as shown in Fig. 11, and (c) Pn =
(−1)n / 2 = − Wn , which corresponds to a dipole-ordered state as shown in
Fig. 12
rence of a dipole-ordered state or a collective dipole oscillation via localized

photon-exciton interactions.
At the end of this section, we examine the applicability of the semiclas-
sical approach that has an advantage over the quantum approach that one
can easily handle a relatively large number N system. In Fig. 34 we show
the radiation profile obtained from three different methods. The system and
the initial conditions to be considered are the same as in Fig. 33. The result
is obtained for the isolated system by the Liouvillian dynamics of the ef-
fective Hamiltonian with the semiclassical approximation, for the dissipative
system by the master equation (160) with the semiclassical approximation,
and for the dissipative system by the master equation (160) with quantum
correlations. We find from Fig. 34 that multiple pulses are generated for all
the cases though each frequency of the oscillation is different. This indicates
Fig. 33. Time evolution of the radiation factor (upper) and the total dipole moment
(lower) obtained with quantum correlations. The oscillating curves are the results
for our model, while the monotonically decreasing curves represent the results for
the Dicke model. The system is assumed to be dissipative and in a dipole-ordered
state evolved from a dipole-forbidden state, i.e., the same as in Fig. 32c. The initial
conditions are also the same as in Fig. 32c, except for the exciton–localized-photon
coupling (a) g = 0.5, (b) g = 0.8, and (c) g = 1.2
4.5 1.4
5.0
Radiation factor
Radiation factor
1.2
Radiation factor
4.0 3.5 1.0

0.8
3.0 0.6
2.5
0.4
2.0 0.2
1.5 0.0
1.0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
Time Time Time
(a) (b) (c)
Fig. 34. Radiation profile obtained from three different methods. The system and
the initial conditions to be considered are the same as in Fig. 33. The result is
obtained (a) for the isolated system by the Liouvillian dynamics of the effective
Hamiltonian with the semiclassical approximation, (b) for the dissipative system
by the master equation (160) with the semiclassical approximation, and (c) for the
dissipative system by the master equation (160) with quantum correlations. The
parameters N = 4, β = 0.05, and γ = 0 are used
that the semiclassical approach can describe qualitatively the radiation prop-
erties of both isolated and dissipative systems when the total dipole of the
system is not zero. The strong radiation coming from the dipole-ordering phe-
nomenon, or the nonlinearity and the collective phenomena of the dynamics
of the system considered in this chapter can be qualitatively predicted by the
semiclassical approach.
9 Dynamics of Localized Photons:

the Storage Mode of Localized Photons
Since the semiclassical Heisenberg equations (46) are nonlinear, neglecting
quantum correlations between quantum dots (QDs) and localized photons,
we expect such a system to display nonlinear or possibly chaotic phenomena.
Indeed, since a system with strong nonlinearity behaves chaotically and shows
deterministic randomness, chaotic systems represent macroscopic transport
phenomena. In this context, this section considers an open system for the
localized photon model and investigates the transportation of localized pho-
tons.
As depicted in Fig. 35, we lined up QDs that interact with localized
photons with a reservoir of localized photons at either end of the line. In this
setting, we investigated the transportation of localized photons through the
system.
Fig. 35. The QDs interacting with localized photons are configured in a line. At
each end of the line, we set different boundary conditions
9.1 The Transportation of Localized Photons:

the Storage and Through Flowing Modes
The numerical operator of the localized photons at site j is defined as nj =

a†j aj and the equation of motion for the expectation value nj using the
semiclassical approximation is derived as follows.
dnj v
= (xj−1 + xj+1 )yj
dt 2
v g
− (yj−1 + yj+1 )xj − (Vj − Pj )yj . (163)
2 2
Since the right-hand side of (163) does not include the localized photon num-
ber nj , the dynamics of the localized photon number are decided by the
other Heisenberg equations (46) and (47). From the Heisenberg equations,
it is clear that using the semiclassical approximation, the physical quantity
that can be input from both ends as a boundary condition is the coherence of
localized photons xj = aj + a†j and yj = i(aj − a†j ) (not the localized photon
number a†j aj ). Therefore, we define the boundary condition at both ends us-
ing (57) to achieve a coherent state in which the eigenvalue is a real number.
Fig. 36. Behavior 1: the dynamics of the distribution of localized photons of an

open system with reservoirs of localized photons in the coherent states at the zeroth
and ninth sites. In this case we observe linear propagation and the reflection of
excitation of localized photons
This boundary condition corresponds to the locally determined amplitude of

the classical electric field oscillation (at the boundaries). The electric field at
the left boundary is larger than that at the right boundary (xl > xr ). The
emerging dynamics are then examined with the focus on the distribution of
localized photons. Consider the case where there are eight sites and the val-
ues at sites 0 and 9 are fixed at x0 = 1, x9 = 0, respectively, by assuming
that a reservoir of localized photons has been attached, with all the excitons
in the initial state of excitation, and localized photons are set to 0 at all the
other locations, except at the boundaries. In this configuration, the tempo-
ral evolution of localized photons shows that the resultant behavior can be
broadly classified into the following two types: Behavior 1 is shown in Fig. 36,
in which the excitation travels from left to right, just like a wave. Behavior 2
is shown in Fig. 37, in which a structure similar to a standing wave emerges
and develops over time.
In Fig. 36, transmission of a wave can be observed in which excitation
emerges from the input side, travels to the right, reflects off the right bound-
ary, travels to the left, gets pushed back by the input on the left, and travels
Fig. 37. Behavior 2: the dynamics of the distribution of localized photons of an open
system with reservoirs of localized photons in the coherent states at the zeroth and
ninth site. In this case, we observe the growth of a structure similar to a standing
wave
to the right again. Consequently, it travels back and forth. In other words,
linear behaviors like transmission of a wave can be observed in the temporal
evolution of localized photons despite the nonlinear nature of the system’s
dynamics. In Fig. 37, there is a clear division between the sites where local-
ized photons accumulate (peaks) and the sites where localized photons are
rare (nodes). Localized photons continue to accumulate in the system and no
stationary state emerges. This can be viewed as formation
of a standing wave.
Examination of the total number of localized photons n an (t)† an (t) in
the system yields the results shown in Fig. 38.
Figure 38 clearly distinguishes between the two types of system when
looking at the number of photons. In the system shown in Fig. 36, in which
the transportation of localized photons follows a linear wave motion, the
total number of photons is approximately constant. This means that localized
photons in the system are flowing through the system. Therefore we call this
through flowing mode. In the system shown in Fig. 37, the total number
of localized photons increases monotonically and localized photons continue
to accumulate in the system. We call this phenomenon the photon storage

Fig. 38. Temporal evolution of the total localized photon number n an (t)† an (t)
(feet). (a) and (b) correspond to the systems in Fig. 36 and Fig. 37, respectively
mode. The fact that the relaxation phenomenon does not occur implies that
either photons can accumulate in the system, or that the system relaxes very
slowly and our calculations are insufficient for it to reach equilibrium.
Following the discussion given in Sect. 7 and using (147), we then cal-
culated the radiation intensity for the two systems, as shown in Fig. 39. It
follows from Fig. 39 that no cooperative effect can be seen in the system
shown in Fig. 36, while a cooperative effect (superradiance) and a time-
irreversible phenomenon (dacaying profile of radiation) is observed in the
system shown in Fig. 37. Here the system parameters used in Fig. 37 are
same as those used in Fig. 28b, but it should be noted that a superradiant
radiation intensity decays time-irreversibly in the open system as assumed in
Fig. 37 while a recursive superradiance occurs in the closed system consid-
Fig. 39. Temporal evolution of the radiation intensity of the open systems. (a) and
(b) correspond to the systems in Fig. 36 and Fig. 37, respectively. Strong radiation
occurs in the system that is accumulating the localized photons and not in the
system where there is no accumulation
ered in Fig. 28b. Nevertheless, it becomes clear that a statistical-mechanical

transportation phenomenon does not occur as a nonequilibrium steady state,
despite the nonlinear dynamics of the system. One of the causes of this is that
the system’s nonlinearity is weak. Generally, the ergodicity or mixing prop-
erty strengthens as nonlinearity increases, and relaxation is reached more
quickly towards the steady state [87]. Another cause may be that chaotic
characteristics in the classical sense do not emerge, because the system is
currently dealt with in terms of semiclassical quantum mechanics. However,
other phenomena characteristic of a nonlinear system should appear when
nonlinearity exists in the system of equations, similar to quantum chaotic
systems in which characteristic phenomena different from those of a regular
system appear indirectly in the level statistics [88, 89]. The following subsec-
tion contains observations regarding this point.
9.2 The Emergence of Dynamical Nonlinearity
Dependence of the Coupling Constant g
To study the appearance of nonlinearity in this system, the dependency of

the temporal evolution of the localized photons on the coupling constant g is
examined. The temporal evolution of the total number of localized photons
that accumulate in the entire system for the case in Fig. 38b is plotted for
various coupling constants, g, as shown in Fig. 40.
Looking at the plot, there is a characteristic sudden shift in behavior
from monotonically increasing to oscillatory at g = 0.67, and this returns
to monotonically increasing behavior at g = 0.69. Indeed, the number of
localized photons decreases by one order of magnitude at the transition from
monotonic behavior at g = 0.65 to oscillatory behavior at g = 0.67. This
is a kind of nonlinear behavior. That is to say, we can expect nonlinear,
random behavior in the dependency of the total number of localized photons
on the exciton–photon coupling constant g. To examine this in detail, Fig. 41
plots the dependency of the total number of localized photons on the coupling
constant g at time t = 100. Figure 41 shows that the appearance of nonlinear,
random behavior for the dependency of the total number of localized photons
depends on the exciton–photon coupling constant g, for the systems in both
Figs. 36 and 37. Particularly for the system in Fig. 37, in which localized
photons accumulate, the fluctuation in the dependency of the total number
of photons on the coupling constant is extremely vigorous. Examination of
the Heisenberg equations (46) and (47) shows that the coupling constant
g is included in the nonlinear terms of the equations. Therefore, it is easy
to see that the coupling constant g is one of the parameters that measures
nonlinearity. Therefore, the dynamics of systems with boson-approximated
excitons, in which nonlinear terms do not emerge in the equations of motion,
must be examined in order to confirm that this behavior originates from the
nonlinearity of the equations.
Fig. 40. Temporal evolution of the total number of localized photons in the system
corresponding to Fig. 38b when the coupling constant g equals 0.65, 0.67, and 0.69.
We observe a drastic change of the number of localized photons between g = 0.65
and g = 0.67. The difference amounts to a single digit
Linear System
The Heisenberg equations of the system become the following when an exciton
is approximated by a boson:
⎧
⎪
⎪ ∂t Pn = −ΩVn − gyn ,
⎨
∂t Vn = ΩPn + gxn ,
(164)
⎪
⎪ ∂t xn = −ωyn − v(yn−1 + yn+1 ) − gVn ,
⎩
∂t yn = ωxn + v(xn−1 + xn+1 ) + gPn .
These are clearly linear equations. Since the equation of motion of the num-
ber of localized photons an (t)† an (t) is the same as in the case of (163) dis-
cussed above, which is obviously linear, in the boson-approximated system
Fig. 41. Dependence of the total localized photon number on the coupling con-
stant g at t = 100. (a) and (b) correspond to the systems in Fig. 36 and Fig. 37,
respectively
Fig. 42. The dependence of the coupling constant g of the total localized photon
number at t = 100 for the system in which there is boson-approximation of excitons.
(a) and (b) take the same system parameters as the systems in Fig. 36 and Fig. 37,
respectively
all the equations of motion are linear. Solving these equations, Fig. 42 shows
the dependency of the total number of localized photons on the coupling
constant g. Although random behavior does not appear in this case, resonat-
ing coupling constant dependence is found. The meaning of the resonating
coupling constants is not yet clear; however, at least there is no disorderly,
unpredictable behavior. From these observations, we conclude that the dis-
orderliness of Fig. 41 is a chaotic behavior originating from the nonlinearity
of the dynamics.
Here, one must be cautious, because a system consisting of eight sites has
many modes, and it is possible that an apparently disorderly dependence has
emerged via the competition of many modes (for example, the superposition
of many types of quasiperiodic behavior). Quasiperiodic behavior does not
have a period either, so chaos does not exist simply because of a lack of
periodicity. In the next section, the phenomenon is examined by simplifying
the problem by reducing the degrees of freedom to the minimal value, while
considering the above.
9.3 Two-Site Open System with Intermittent Chaotic Behavior
This section considers the nonlinear dependency on the coupling constant

g observed in the previous section in a system with two sites (four sites
including the reservoirs at the two ends) in order to simplify the problem.
This is done to reduce the degrees of freedom so as to differentiate between the
complexities caused by the superposition of many quasi-periodic behaviors
originating from the many modes and the nonlinearity of the dynamics of the
system.
Figure 43 shows the dependency of the total number of localized pho-
tons on the coupling constant g for the two-site system. This shows that
dependency on g behaves unpredictably. Since with two sites there are few
participating modes, the vigorous fluctuations are a manifestation of chaotic
behavior, reflecting the nonlinearity of the dynamic equations. Examination
of Fig. 43a shows that there are vigorous fluctuations (chaotic regions) in
some areas, and not in others (regular windows). In other words, this system
displays ‘intermittent chaos’ [87], using the terminology to describe the tem-
poral evolution of chaos. Since chaotic systems display deterministic proba-
bilistic behavior (e.g., diffusion) [87], the number of localized photons should
obey a statistical distribution. Therefore, we examined histograms of the
number of localized photons corresponding to Figs. 43a–d. Figure 44 shows
the histograms of the numbers of localized photons in the chaotic regions
of the system corresponding to Figs. 43a–d in a double-logarithmic graph.
This figure shows that the histograms for each individual region share a com-
mon behavior. The double logarithmic plot of the histograms consists of two
straight lines; the first line has a slope of approximately −1. In other words,
when the distribution of localized photons n is approximated as P (n), the
distribution in the regions of small n is given by P (n) ∝ 1/n. The slope is
very steep in the region where there are many photons, and the graph can be
regarded as an exponential function. In conclusion, the emergence of a com-
mon statistical distribution of localized photons in regions (a) through (d) is
the evident result of a universal statistical law caused by the chaotic dynam-
ics of the system. This means that the apparent complexity does not arise
merely by chance; there is a common principle working in the background.
In other words, the nonlinear dynamics of the system indirectly cause such
behavior to emerge.
Fig. 43. The dependence of the coupling constant g on the total localized photon-
number of the two-site open system. (A) is the plot for the range from g = 0 to
4.1. (a)–(d) show the chaotic regions in (A). (a) shows the range from g = 2.40
to 2.41, (b) is from g = 2.95 to 3.02, (c) is from g = 3.09 to 3.28, and (d) is from
g = 3.76 to 3.965
As we show above, many interesting phenomena emerge when we look

at the behavior of localized photons, including the existence of the photon-
storage mode and the existence of chaotic dependency of the system’s mode
on a system parameter. Nevertheless, our discussion is based on a semiclas-
sical system of equations that ignores quantum correlations, and it remains
Fig. 44. A double logarithmic plot of the histogram of the localized photon numbers
for the regions depicted in Fig. 43. Figures (a)–(d) correspond to chaotic regions
(a) to (d) in Fig. 43
to be determined whether these phenomena actually occur, i.e., do these

phenomena survive in the case using full quantum correlations? Therefore,
we examined the behavior considering the quantum correlation of the next
order. In this case, the storage mode for photons also exists. The fact that
the behavior in the next order does not differ significantly suggests that the
semiclassical approach is sufficient.
10 Conclusions
In order to investigate the dynamics of the system, we presented a model
of N two-level quantum-dot systems interacting with optical near fields that
are explicitly expressed in terms of the localized-photonic degrees of free-
dom, which characterize the unique property of the localization of optical
near fields. At the low-density limit, excitons (N two-level systems) are ap-
proximated as bosons, and a rigorous solution of the Heisenberg equation
is obtained. Using this solution, we examined the dynamics of the excitonic
system, and showed that the dipole moments are transferred linearly within
the system. Since the dipole moments in the system represent the quantum
coherence between any two energy levels, this phenomenon might be applica-
ble to a photonic device on a nanometer scale or to the transfer of quantum
information.
For fermionic excitons, the Heisenberg equation becomes nonlinear, and
the dynamics is more complicated. We obtained a perturbative solution given
by (43) within the second order with respect to the localized photon–exciton
interaction, to investigate the dipole dynamics. The study revealed that sev-
eral oscillating quasisteady states exist, depending on the material parame-
ters. Using the effective Hamiltonian obtained from renormalization of the
localized photonic degrees of freedom, we classified such quasisteady states
into several groups: one is a “ferromagnetic” state with all the dipoles aligned
in the same direction, the other is an “antiferromagnetic” state with alter-
nating dipoles, as shown in Table 1.
Note that if the sign of the population difference Wn (0) at an arbitrary
site n and time t = 0 is reversed, then the dipole of the same site at arbitrary
time t, Pn (t), also changes sign (see (44)). Using this flip hypothesis, we can
transform an arbitrary dipole distribution of the system into a dipole-ordered
state after manipulating the initial distribution of the population differences.
This hypothesis is based on the perturbative solution (44), which determines
the sign of Pn (t) according to the sign of the product of P1 (0)Wn (0).
It also originates from the fermionic property of excitons, which gives the
Heisenberg equations of motion for bn (∝ Pn ) as

[Hb−b , bn ] = ∆Ωnm Wn bm . (165)
m=n
The right-hand side of (165) is proportional to Pm Wn for fermions, while
it is proportional to Pm for bosons. Therefore, the occurrence of this kind of
nonlinearity for fermionic excitons is a possible origin of the flip hypothesis.
Moreover, it is not trivial; this property can be applicable to results obtained
from the semiclassical Heisenberg equations (46) and (47).
We examined the radiation property of our system, solving the Heisen-
berg equations (46) and (47) using the semiclassical approximation. In the
investigation, we treated the system as isolated, since the coupling between
radiation fields and the sytem is very weak. The numerical analysis showed
that dipole-ordered states, which have large total dipole moments, show a
large radiation probability comparable to Dicke’s superradiance. In partic-
ular, it predicted that multiple pulses are emitted from the system super-
radiantly. In order to verify whether such a phenomenon is inherent in an
isolated system, we solved the master equation (160) for a dissipative system
using the semiclassical approximation, and found that such multiple pulses
in the radiation profile can even survive in a dissipative system coupled to a
radiation reservoir.
Multiple superradiant peaks have been observed experimentally in an

atomic-gas system [64], and are reported to originate from the dipole–dipole
interaction between a two-level system within the framework of the Dicke
model [65]. In our model, excitons are coupled with one another indirectly
through the interactions with localized photons; therefore, the pulse oscil-
lation in the radiation profile occurs in a mechanism similar to that in the
Dicke model.
We solved the master equation (160), considering quantum correlations,
and compared the results with those obtained from the Dicke model. When
all the populations are initially in excited states, both models produce similar
radiation profiles. They differ qualitatively in that the peak value of the ra-
diation pulse in our model is reduced and the tail is extended. This tendency
was examined by comparing atomic and Frenkel exciton systems, in which
excitons can hop via the dipole–dipole interaction, and the same qualitative
difference was reported [69]. Regarding the multiple-pulse generation that we
obtained, the Frenkel exciton model also predicts the possibility of the super-
radiance profile oscillating if the system is initially prepared with a partial
population inversion [83]. From our model, we concluded that the superra-
diant peaks of multiple pulses correspond to the occurrence of a collective
dipole oscillation, or a dipole-ordered state.
It is still not clear why the dipole distribution in our system has several
quasisteady states, which was predicted in our model using the perturba-
tive expressions, numerical solutions of the Heisenberg equations, and the
effective Hamiltonian. Since our system has several kinds of symmetry, we
expect to clarify the stability of such quasisteady states from the viewpoint
of symmetry.
The size dependence of the radiation profile has been investigated in the
Dicke model [54]. To clarify the differences between optical near fields and
propagating fields, we should also examine such a size dependence in our
model. The semiclassical approach might be useful for such a qualitative
study because it has an advantage over the quantum approach with respect
to computation time.
One of our main goals was to clarify the inherent characteristics in optical
near fields from the viewpoint of the coupling scheme, i.e., a local coupling
system versus a global coupling system. We have shown several differences
between our model (local coupling system) and the Dicke model (global cou-
pling system), but further effort is required to answer the question more
directly. In spite of some common properties, we must emphasize that our
model differs from the Frenkel exciton model of Tokihiro et al., which is an
extended Dicke model that considers the dipole–dipole interaction between
excitons. The similarities of the radiation properties in our model and the
Tokihiro model arise from the similar master equations used (see (135) and
(160)). However, our master equation (160) was derived after neglecting the
product of excitons with an order higher than two (see (159)). This approx-
imation is justified for a dilute excitonic system. Nevertheless, the precise

conditions under which this approximation is applicable and its consistency
should be investigated thoroughly.
Therefore, our localized photon model concurs with the model of Tokihiro
et al. only in the simple case, and the true dynamics in our model is more
complicated. Unfortunately, there is no straightforward way to derive a more
correct equation beyond the approximation (159). The difficulty comes from
how to treat the dynamics of three different degrees of freedom: excitons,
localized photons, and radiation photons. Rigorously speaking, the difficulty
arises from the fact that the unperturbed part H = Ha +Hb +Hint (4) is made
up of a mixture of spin operators (exitons) and bosons (localized photons).
However, it is possible to consider the next order of excitons in UQDeff in
(157) beyond the approximation (159). To proceed to higher orders, we need
to remove the time integration in (157), and may use the time-convolutionless
(TCL) formulation of the projection operator method [90, 91].
Finally, we focused on the nonlinearity of the Heisenberg equations in
our system, and investigated the transport phenomena of localized photons,
adding a reservoir of localized photons to the system. Neither normal macro-
scopic transport phenomena nor equilibrium states were found, which indi-
cates that the nonlinearity of this system is very weak, as a fully nonlin-
ear system, such as a chaotic system, shows statistical transport phenom-
ena (e.g., diffusion) [87, 92] and large-scale simulations are required to find
the equilibrium states. Instead, as a transient behavior, we found a remark-
able phenomenon in which the transfer of localized photons is jammed or
through flowing, depending on the system parameters as ε, V , E, and U .
We called the former case the storage mode of localized photons. Moreover,
the number of localized photons stored varies drastically, depending on the
localized photon–exciton coupling constant U = g. The dependence of the
total number of localized photons on the coupling constant g is chaotic and
unpredictable. Reduction of the system to the minimal size did not change
this behavior; therefore, the complex behavior arose from the intrinsic non-
linearity of the system, not from the interference of many modes due to
many degrees of freedom. Since the unpredictable coupling-constant depen-
dence had both regular windows and chaotic regions (see Fig. 43), our system
shows “intermittent chaos” in terms of usual dynamical chaos. We found that
the chaotic regions have similar distributions, but the meanings of the dis-
tributions are not clear. We also checked the dynamics using the next-order
quantum correlation and observed the storage and through-flowing modes.
The semiclassical treatment seems sufficient to produce these phenomena.
In summary, we found an interesting phenomenon arising from the
nonlinearity in our system. The nonlinearity of the dynamics originates from
the commutation relation of excitons that differs from that of bosons. This is
true in the Dicke model. Therefore, the Dicke model has the same nonlinear-
ity, which is essential for superradiance [58], and certain kinds of signatures
of chaos. Indeed, it was recently reported that the Dicke model is a quan-
tum chaotic system [93]. The term “quantum chaotic system” means that the
distribution of the energy-level spacing of the Hamiltonian differs from that
of a regular system [88]. Since there is no instability of orbits in a quantum
system, due to the uncertainty principle, we need new signatures of chaos for
a quantum system whose classical correspondence is also chaotic. One of the
quantum signatures of chaos is the level statistics [89], and the analysis of
these level statistics remains an intriguing open problem. Since our system
has a richer structure than the Dicke model, we expect our model to produce
more fruitful properties as a quantum chaotic system. Finally, we point out
that the open system discussed using our localized photon model is one of
the simplest nano-optical structures to use to examine the optical response
of nanostructures and to discuss the susceptibilities of nanometric material
systems.
Acknowledgements
The authors are grateful to H. Hori, I. Banno, N. Ohnishi, and C. Uchiyama

of Yamanashi University, T. Kawazoe of the Japan Science and Technology
Agency, and H. Nejo of the National Institute for Materials Science for fruitful
discussions. They also gratefully acknowledge valuable suggestions from F.
Shibata of Ochanomizu University on the topics in Sect. 5.1.
References
1. E.A. Synge: Philos. Mag. 6, 356 (1928)
2. M. Ohtsu, H. Hori: Near-Field Nano-Optics (Kluwer Academic/Plenum pub-
lishers, New York 1999)
3. M. Ohtsu (ed.): Near-Field Nano/Atom Optics and Technology (Springer-
Verlag, Tokyo 1998)
4. D.W. Pohl, D. Courjon: Near Field Optics (Kluwer Academic Publishers, Dor-
drecht 1993)
5. J.P. Fillard: Near Field Optics and Nanoscopy (World Scientific, Singapore
1996)
6. M.A. Paesler, P.J. Moyer: Near-Field Optics: Theory, Instrumentation, and
Application (Wiley, New York 1996)
7. M. Ohtsu (ed.): Progress in Nano Electro-Optics I (Springer-Verlag, Berlin
2002)
8. M. Ohtsu (ed.): Progress in Nano Electro-Optics II (Springer-Verlag, Berlin
2003)
9. M. Ohtsu, K. Kobayashi: Optical Near Fields (Springer-Verlag, Heidelberg
2004)
10. K. Shimoda: Introduction to Laser Physics (Springer-Verlag, Berlin and Hei-
delberg 1986)
11. C. Carniglia, L. Mandel: Phys. Rev. D 3, 280 (1971)
12. I. Bialynicki-Birula, J. B. Brojan: Phys. Rev. D 5, 485 (1972)

13. R. Asby, E. Wolf: J. Opt. Soc. Am. 61, 52 (1971)
14. E. Lalor, E. Wolf: Phys. Rev. Lett. 26, 1274 (1971)
15. O. Costa de Beauregard, C. Imvert, J. Ricard: Int. J. Theor. Phy. 4, 125 (1971)
16. J.M. Vigoureux, R. Payen: J. de Phys. 35, 617 (1974)
19. J.M. Vigoureux, L. D’Hooge, D. Van Labeke: Phys. Rev. A 21, 347 (1980)
20. S. Huard: Opt. Commun. 30, 8 (1979)
21. J.M. Vigoureux: ‘Emission and absorption of light by electrons or atoms in
optical near fields’. In Near Field Optics ed. by D.W. Pohl, D. Courjon, NATO
ASI Series 242 (Kluwer Academic Publishers, Dordrecht 1993), pp. 239–246
22. M. Janowicz, W. Zakowicz: Phys. Rev. A 50, 4350 (1994)
23. T. Inoue, H. Hori: Phys. Rev. A 63, 063805 (2001)
24. H. Khosravi, R. Loudon: Proc. Roy. Soc. London A 433, 337 (1991)
25. Y. Ohdaira, K. Kijima, K. Terasawa, M. Kawai, H. Hori, K. Kitahara: J. Micros.
202, 255 (2001)
26. H. Hori, K. Kitahara,, M. Ohtsu: In Abstracts of the 1st AP-NFO, 49 (1996)
27. A. Shojiguchi, K. Kitahara, K. Kobayashi, M. Ohtsu: J. Micros. 210, 301 (2003)
28. H. Levine, J. Schwinger: Phys. Rev. 74, 958 (1948)
29. H.A. Bethe: Phys. Rev. 66, 163 (1944)
30. C.J. Bouwkamp: Rep. Prog. Phys. 17, 35 (1954)
31. C.J. Bouwkamp: Philips Res. Rep. 5, 321 (1950)
32. C.J. Bouwkamp: Philips Res. Rep. 5, 401 (1950)
33. T.H. Stievater, Xiaoqin Li, D.G. Steel, D. Gammon, D.S. Katzer, D. Park,
C. Piermarocchi, L.J. Sham: Phys. Rev. Lett. 87, 133603 (2001)
34. H. Kamada, H. Gotoh, J. Temmyo, T. Takagahara, H. Ando: Phys. Rev. Lett.
87, 246401 (2001)
35. H. Htoon, T. Takagahara, D. Kulik, O. Baklenov, A.L. Holmes, Jr., C.K. Shih:
Phys. Rev. Lett. 88, 087401 (2002)
36. T. Kawazoe, K. Kobayashi, J. Lim, Y. Narita, M. Ohtsu: Phys. Rev. Lett. 88,
067404 (2002)
37. M. Ohtsu, K. Kobayashi, T. Kawazoe, S. Sangu, T. Yatsui: IEEE J. Selec.
Topics Quantum Electron. 8, 839 (2002)
38. K. Cho: Suppl. Prog. Theor. Phys. 106, 225 (1991)
39. H. Ishihara, K. Cho: Phys. Rev. B 48, 7960 (1993)
40. K. Cho, Y. Ohfuti, K. Arima: Surf. Sci. 363, 378 (1996)
41. R.T. Cox: Phys. Rev. 66, 106 (1944)
42. J.G. Linhart: J. Appl. Phys. 26, 527 (1955)
43. K. Kobayashi, M. Ohtsu: J. Micros. 194, 249 (1999)
44. K. Kobayashi, S. Sangu, H. Ito, M. Ohtsu: Phys. Rev. A 63, 013806 (2001)
45. K. Kobayashi, S. Sangu, M. Ohtsu: ‘Quantum theoretical approach to optical
near-fields and some related applications’. In Progress in Nano-Electro-Optics
I ed. by M. Ohtsu, Springer Series in Optical Sciences (Springer-Verlag, Berlin,
2002) pp. 119–157
46. A. Shojiguchi, K. Kobayashi, S. Sangu, K. Kitahara, M. Ohtsu: Nonlin. Opt.
29, 563 (2002)
47. S. Sangu, K. Kobayashi, A. Shojiguchi, T. Kawazoe, M. Ohtsu: J. Appl. Phys.
93, 2937 (2003)
48. K. Kobayashi, S. Sangu, A. Shojiguchi, T. Kawazoe, K. Kitahara, M. Ohtsu:

J. Micros. 210, 247 (2003)
49. R.H. Dicke: Phys. Rev. 93, 99 (1954)
50. L. Quiroga, N.F. Johnson: Phys. Rev. Lett. 83, 2270 (1999)
51. H.A. Posch, W.G. Hoover, F.J. Vesely: Phys. Rev. A 33, 4253 (1986)
52. W.G. Hoover, H.A. Posch: Phys. Rev. E 49, 1913 (1994)
53. A. Shojiguchi, K. Kobayashi, S. Sangu, K. Kitahara, M. Ohtsu:
J. Phys. Soc. Jpn. 72, 2984 (2003)
54. M. Gross, S. Haroche: Phys. Rep. 93, 301 (1982)
55. N. Rehler, J.H. Eberly: Phys. Rev. A 3, 1735 (1971)
56. R.H. Lehmberg: Phys. Rev. A 2, 883 (1970)
57. G.S. Agarwal: Phys. Rev. A 2, 2038 (1970)
58. G.S. Agarwal: Phys. Rev. A 4, 1791 (1971)
59. L.M. Narducci, C.A. Coulter, C.M. Bowden: Phys. Rev. A 9, 829 (1974)
60. R. Bonifacio, P. Schwendimann, F. Haake: Phys. Rev. A 4, 302 (1971)
61. R. Friedberg, S.R. Hartmann: Phys. Rev. A 10, 1728 (1974)
62. B. Coffey, R. Friedberg: Phys. Rev. A 17, 1033 (1978)
63. R. Bonifacio, L.A. Lugiato: Phys. Rev. A 11, 1507 (1975)
64. N. Skribanowitz, I.P. Herman, J.C. MacGillivray, M.S. Feld: Phys. Rev. Lett.
30, 309 (1973)
65. R. Bonifacio, L.A. Lugiato, A.A. Crescentini: Phys. Rev. A 13, 1507 (1976)
66. A. Nakamura, H. Yamada, T. Tokizaki: Phys. Rev. B 40, 8585 (1989)
67. Y.N. Chen, D.S. Chuu, T. Brandes: Phys. Rev. Lett. 90, 166801 (2003)
68. N. Piovella, V. Beretta, G.R.M. Robb, R. Bonifacio: Phys. Rev. A 68, 021801
(2003)
69. T. Tokihiro, Y. Manabe, E. Hanamura: Phys. Rev. B 47, 2019 (1993)
70. T. Brandes, J. Inoue, A. Shimizu: Phys. Rev. Lett. 80, 3952 (1998)
71. T. Saso, Y. Suzumura, H. Fukuyama: Suppl. Prog. Theoret. Phys. 84, 269
(1985)
72. V.M. Agranovich, M.D. Galanin: Electric Excitation Energy Transfer in Con-
densed Matter (North-Holland, Amsterdam 1982)
73. J. Knoester, S. Mukamel: Phys. Rev. A 39, 1899 (1989)
74. H. Fröhlich: Phys. Rev. 79, 845 (1950)
75. J. Bardeen, D. Pines: Phys. Rev. 99, 1140 (1955)
76. C. Kittel: Quantum Theory of Solids, 2nd revised printing (John Wiley & Sons,
New York 1987)
77. L. Mandel, E. Wolf: Optical Coherence and Quantum Optics (Cambridge Uni-
versity Press, Cambridge 1995)
78. G.S. Agarwal: Phys. Rev. 178, 2025 (1969)
79. S. Nakajima: Prog. Theor. Phys. 20, 948 (1958)
80. R. Zwanzig: J. Chem. Phys. 33, 1338 (1960)
81. R. Zwanzig: Phys. Rev. 124, 983 (1961)
82. R. Kubo, M. Toda, N. Hashitsume: Statistical Physics II (Springer-Verlag,
Berlin 1985)
83. T. Tokihiro, Y. Manabe, E. Hanamura: Phys. Rev. B 51, 7655 (1995)
84. E. Lieb, T. Schultz, D. Mattis: Annal. Phys. 16, 407 (1961)
85. R. Bonifacio, L.A. Lugiato: Phys. Rev. A 12, 587 (1975)
86. G. Lindblad: Comun. Math. Phys. 48, 119 (1976)
87. P. Gaspard: Chaos, Scattering and Statistical Mechanics (Cambridge University
Press, Cambridge 1998)
88. M.C. Gutzwiller: Chaos in Classical and Quantum Mechanics (Springer-Verlag,

Berlin 1995)
89. F. Haake: Quantum Signatures of Chaos (Springer-Verlag, Berlin 1991)
90. S. Chaturvedi, F. Shibata: Z. Phys. B 35, 297 (1979)
91. F. Shibata, T. Arimitsu: J. Phys. Soc. Jpn. 49, 891 (1980)
92. D.J. Evans, G.P. Morriss: Statistical Mechanics of Nonequilibrium Liquids (Aca-
demic Press, New York 1997)
93. C. Emary, T. Brandes: Phys. Rev. Lett. 90, 044101 (2003)
Index
AASA 128–130, 134, 137, 143 coaxial probe 103

Ag-MgF2 -Al-coated probe 24, 25 coercive force 134
alkanethiol 67, 81 coherent state 175, 204
amorphous 129, 136, 137 collection mode 1
angular spectrum representation 147 contact-mode topography 136
antiferromagnetic system 171, 172, Coulombic dipole-dipole interaction
174, 200 189
apertured probe 1, 2, 12, 26, 29 cut-off diameter 99, 100, 103
apertured-type double-tapered probe
30 data storage density 127
application-oriented probe 3, 48 DCF 9, 11
artificially assisted self-assembling diblock copolymer 130, 132, 133
method 128, 130, 143 Dicke master equation 177, 180, 184,
avidin 81, 88 194
Dicke model 148, 150, 151, 167, 177,
BHF 7–10, 20, 27, 31, 36, 38, 45, 50, 53 187, 197, 201, 215, 216
biomolecule 78–81, 88, 89 Dicke state 177, 179, 180, 186, 189,
biotin 81 190, 197
Bloch vector 157, 162, 169, 177, 180, diffraction limit 95, 96, 127
186, 189, 193, 195, 196 dipole distribution 151, 162, 167, 168,
Blu-ray disk drive 93, 94 170–172, 174, 175, 195, 201, 214,
Born-Markov approximation 189, 199 215
boson approximation 151, 153, 157, dipole moment 101
162 dipole-dipole decoherence 151
bosonic exciton 153, 157, 162 dipole-dipole interaction 150, 151,
bow-tie antenna 103, 120 177, 188–192, 215
buffered HF solution 7 dipole-forbidden state 164–166, 201
buffered hydrogen fluoride solution dipole-forbidden transition 147
43, 45 dipole-ordered state 149, 163, 166,
bulk effect 78 174, 177, 195, 200–202, 214, 215
dispersion-compensating fiber 9, 27,
c-mode SNOM 1–3, 12, 14–18, 23, 25 35
cap-shaped 58, 59, 62, 68, 70, 76 displacement current 138, 140, 143
carbodiimide 60, 62, 67 DNA 23–25
Carniglia–Mandel mode 148 double clad fiber 35, 48, 49
chaotic system 204, 211, 216 double-core fiber 37, 42
charge injection 129, 142, 143 double-tapered probe 3, 9, 29–31, 33,
Chiral-PS 40, 42 34, 42, 48
222 Index
dye 128, 129, 137, 143 i-mode near-field photoluminescence

42
eating-and-pulling 1, 3, 20, 43 i-mode SNOM 1–3, 5, 22, 26, 30, 34, 43
electron beam patterning 128 i-mode UV-SNOM 40
etching 65, 77 illumination mode 1
evanescent field 146, 148 illumination-collection mode 2
evanescent light 145 intensity distribution 105, 106, 108,
exciton 149, 151–154, 157, 160, 167, 118, 120
170, 191, 195, 198, 199, 205, 209, interferrometric lithography 128
213–216 intermittent chaos 216
exciton operator 156, 161, 199 interparticle interaction 57
exciton–exciton interaction 169 ion beam modification 128
exciton–localized photon interaction
169 kinetic data 83
exciton-free photon interaction 198
exciton-polariton operator 156 lateral p-n junction 33
latex sphere 58
FDTD 31, 105 Lindblad’s master equation 199
fermionic exciton 153, 162, 214 localized photon 148, 149, 151–153,
ferromagnetic system 171, 200 155, 157, 160–162, 164, 167–169,
fiber probe 96, 97, 100, 103 174–177, 195, 197, 198, 201, 204,
finite difference time-domain 105, 124 206, 208, 209, 211, 212, 215, 216
fluorescence 129, 138–140, 142, 143 localized photon–exciton interaction
fluorine-doped clad 9, 35, 45 153, 161, 166, 195, 202, 214
fluorine-doped silica 9, 37, 45, 49 localized surface plasmon 57, 89
focused ion beam patterning 128
Frenkel exciton 215 magnetic recording 127
Frenkel exciton model 215 magnetic-optical recording medium
128
GaAs quantum dots 19
Markov approximation 183
GaAs quantum well 19
meniscus etching 1, 3, 6
GeO2 -doped core 8, 21, 24, 37, 42
metal particle 57–59, 76, 78, 90
GeO2 -doped fiber 8, 9, 20, 35
metal-dielectric-metal coat 23, 25
GeO2 -doped silica core 42, 49
metallic pin probe 103
giant magneto-resistive head 96
metallic plate 94, 104–106, 108, 110,
GMR 96
111, 113–118, 122
gold 57–60, 62, 65, 67, 68, 72, 74, 76,
77, 79–81, 89, 91 metallized tapered probe 2
microcontact printing 67
Hankel function 147 micropipette puller 1, 3, 43, 44
heat-assisted magnetic recording 96 microtubules 18, 29
Heisenberg equations of motion 153, monolayer 58
214
Helmholtz equation 147 n-decyl-(s)-2-methylbutyl silane 40
hybrid recording 128 nano-imprint 128, 137
hydrophilic polymethylmethacrylate nano-patterning 128
chain 130 nanoimprint 143
hydrophobic polystyrene chain 130 nanopatterning 133, 143
near-field optical head 94, 96–99, 103,
i-c mode SNOM 2, 3, 30, 33, 34 104, 108, 122–124
Index 223
near-field optical microscope 142 quantum wires 40

near-field optical recording 94–96, 99, quasi-particle representation 148
100, 124 quasi-steady state 151, 168, 171, 176,
near-field photoluminescence 30, 33, 195, 196, 215
46 quasisteady state 170, 171, 176, 214
near-field Raman spectroscopy 34, 42 quenching 139, 140, 142, 143
neuron 27
noncontact-mode atomic force recording density 93–97, 118, 123, 124
microscope 25 refractive index 78, 85, 88–90
Rhodamin 6G 29
optical biosensor 57 rotating-wave approximation 153, 182
organic film 138
Salmonella flagellar filaments 14, 16
patterned media 128, 133, 134, 137, SAM 79, 81, 85, 89, 90
143 scanning electron microscope 60, 90
PDHS 40 scanning near-field optical microscope
peak wavelength 60, 74, 79–81, 85, 145
88–90 scanning near-field optical microscopy
pencil-shaped probe 24, 48, 51–53 1, 96
phase separation 130, 132 selective etching 1, 3, 6, 7, 9, 13, 14, 23,
phase-change medium 128 24, 31, 37, 42, 45, 48, 49
phase-change recording 96, 98 selective resin coating 14, 33
photochromic recording 96 selective resin etching 12
photoenhanced chemical vapor self-assembled monolayer 81, 85
deposition 47 self-assembling diblock polymer 128,
planar near-field optical head 97, 98 143
plane wave basis 148 self-assembling template lithography
polydiacetylene 34 128
polydihexylsilane 40, 46 SEM 60, 63, 65, 68, 76
polysilanes 40 semiclassical approximation 150, 167,
polystyrene sphere 59, 60, 76 177, 187, 193, 201, 202, 204, 214
protein A 81 shear-force feedback 1, 9, 18, 19, 27,
protein G 81 28, 33
protrusion-type double-tapered probe shear-force topographic image 18, 19,
30 33, 46
protrusion-type probe 3, 9, 12–14, 16, SIL 94, 95
18–20, 22, 25 single-tapered probe 31, 35, 48
PS 59, 60, 65, 68, 71, 72, 77 SNOM 1, 3, 16, 19, 23, 30, 33, 35, 48,
PSF 9, 36, 43–45 96, 99, 145
pure silica fiber 9, 36, 37 SOG 132
pure silica fiber probe 3, 35, 48 solid immersion lens 94
pure silica tapered probe 46 spatial Fourier transformation 154
sphere density 73, 77
quantum chaotic system 217 spherical Hankel function 147
quantum coherence 149, 157, 158, 190, spin on glass 132
214 spiral pattern 131, 132
quantum correlation 167, 177, 187, spot size 93–95, 99, 111, 113, 118, 120
201, 202, 215 squareness ratio 134
quantum fluctuation 150, 201 SRC 12, 13
224 Index
streptavidin 81, 82 two metallic plates 120, 122

superradiance 149–151, 177, 180, two-level system 148–150, 152, 156,
190–192, 194, 195, 197, 200, 201, 157, 177–179, 186, 188–190, 192,
214–216 213, 215
superradiant state 150, 177, 180, 191,
195 UV nanofabrication 47
surface plasmon 101, 102 UV near-field optical microscopy 35
surface-adsorbed 57–59, 65, 77, 78, 82, UV near-field photoluminescence 36,
83, 90 40, 42, 46
surface-enhanced spectroscopy 57, 77, UV single-tapered probe 36
78 UV triple-tapered probe 35, 37, 40, 42
UV-SNOM 35, 42
tapered probe 1, 3, 9, 11–13, 23, 24,
29, 43, 49 vacuum evaporation 1, 6, 12, 20, 24,
TCL 216 27, 39, 46
tetrahedral probe 101 VAD 7–9, 24
thermally assisted magnetic recording vapor-phase axial deposition 7, 8, 37,
96 42, 51
throughput 3, 9, 20, 22, 29–31, 48, 49, VCSEL 99
53 vertical cavity surface emitting laser
time-convolution-less formulation 216 99
TMR 96 Volmer–Weber mechanism 137
total internal reflection 145
triple-tapered probe 24, 36–38, 40, 43, wedge-shaped metallic plate 103, 105,
48, 49, 51, 53, 100 108, 113, 120, 124
tunneling magneto-resistive head 96 Weierstrass-sphere lens 94
Springer Series in
optical sciences
New editions of volumes prior to volume 70
1 Solid-State Laser Engineering
By W. Koechner, 5th revised and updated ed. 1999, 472 figs., 55 tabs., XII, 746 pages
14 Laser Crystals
Their Physics and Properties
By A. A. Kaminskii, 2nd ed. 1990, 89 figs., 56 tabs., XVI, 456 pages
15 X-Ray Spectroscopy
An Introduction
By B. K. Agarwal, 2nd ed. 1991, 239 figs., XV, 419 pages
36 Transmission Electron Microscopy
Physics of Image Formation and Microanalysis
By L. Reimer, 4th ed. 1997, 273 figs. XVI, 584 pages
45 Scanning Electron Microscopy
Physics of Image Formation and Microanalysis
By L. Reimer, 2nd completely revised and updated ed. 1998,
260 figs., XIV, 527 pages
Published titles since volume 70

70 Electron Holography
By A. Tonomura, 2nd, enlarged ed. 1999, 127 figs., XII, 162 pages
71 Energy-Filtering Transmission Electron Microscopy
By L. Reimer (Ed.), 1995, 199 figs., XIV, 424 pages
72 Nonlinear Optical Effects and Materials
By P. Günter (Ed.), 2000, 174 figs., 43 tabs., XIV, 540 pages
73 Evanescent Waves
From Newtonian Optics to Atomic Optics
By F. de Fornel, 2001, 277 figs., XVIII, 268 pages
74 International Trends in Optics and Photonics
ICO IV
By T. Asakura (Ed.), 1999, 190 figs., 14 tabs., XX, 426 pages
75 Advanced Optical Imaging Theory
By M. Gu, 2000, 93 figs., XII, 214 pages
76 Holographic Data Storage
By H.J. Coufal, D. Psaltis, G.T. Sincerbox (Eds.), 2000
228 figs., 64 in color, 12 tabs., XXVI, 486 pages
77 Solid-State Lasers for Materials Processing
Fundamental Relations and Technical Realizations
By R. Iffländer, 2001, 230 figs., 73 tabs., XVIII, 350 pages
78 Holography
The First 50 Years
By J.-M. Fournier (Ed.), 2001, 266 figs., XII, 460 pages
79 Mathematical Methods of Quantum Optics
By R.R. Puri, 2001, 13 figs., XIV, 285 pages
80 Optical Properties of Photonic Crystals
By K. Sakoda, 2nd ed., 2004, 107 figs., 29 tabs., XIV, 255 pages
81 Photonic Analog-to-Digital Conversion
By B.L. Shoop, 2001, 259 figs., 11 tabs., XIV, 330 pages
82 Spatial Solitons
By S. Trillo, W.E. Torruellas (Eds), 2001, 194 figs., 7 tabs., XX, 454 pages
83 Nonimaging Fresnel Lenses
Design and Performance of Solar Concentrators
By R. Leutz, A. Suzuki, 2001, 139 figs., 44 tabs., XII, 272 pages
84 Nano-Optics
By S. Kawata, M. Ohtsu, M. Irie (Eds.), 2002, 258 figs., 2 tabs., XVI, 321 pages
Springer Series in
optical sciences
85 Sensing with Terahertz Radiation
By D. Mittleman (Ed.), 2003, 207 figs., 14 tabs., XVI, 337 pages
86 Progress in Nano-Electro-Optics I
Basics and Theory of Near-Field Optics
By M. Ohtsu (Ed.), 2003, 118 figs., XIV, 161 pages
87 Optical Imaging and Microscopy
Techniques and Advanced Systems
By P. Török, F.-J. Kao (Eds.), 2003, 260 figs., XVII, 395 pages
88 Optical Interference Coatings
By N. Kaiser, H.K. Pulker (Eds.), 2003, 203 figs., 50 tabs., XVI, 504 pages
89 Progress in Nano-Electro-Optics II
Novel Devices and Atom Manipulation
By M. Ohtsu (Ed.), 2003, 115 figs., XIII, 188 pages
90/1 Raman Amplifiers for Telecommunications 1
Physical Principles
By M.N. Islam (Ed.), 2004, 488 figs., XXVIII, 328 pages
90/2 Raman Amplifiers for Telecommunications 2
Sub-Systems and Systems
By M.N. Islam (Ed.), 2004, 278 figs., XXVIII, 420 pages
91 Optical Super Resolution
By Z. Zalevsky, D. Mendlovic, 2004, 164 figs., XVIII, 232 pages
92 UV-Visible Reflection Spectroscopy of Liquids
By J.A. Räty, K.-E. Peiponen, T. Asakura, 2004, 131 figs., XII, 219 pages
93 Fundamentals of Semiconductor Lasers
By T. Numai, 2004, 166 figs., XII, 264 pages
94 Photonic Crystals
Physics, Fabrication and Applications
By K. Inoue, K. Ohtaka (Eds.), 2004, 209 figs., XV, 320 pages
95 Ultrafast Optics IV
Selected Contributions to the 4th International Conference
on Ultrafast Optics, Vienna, Austria
By F. Krausz, G. Korn, P. Corkum, I.A. Walmsley (Eds.), 2004, 281 figs., XIV, 506 pages
96 Progress in Nano-Electro Optics III
Industrial Applications and Dynamics of the Nano-Optical System
By M. Ohtsu (Ed.), 2004, 186 figs., 8 tabs., XIV, 224 pages
97 Microoptics
From Technology to Applications
By J. Jahns, K.-H. Brenner, 2004, 303 figs., XI, 335 pages
98 X-Ray Optics
High-Energy-Resolution Applications
By Y. Shvyd’ko, 2004, 181 figs., XIV, 404 pages
99 Few-Cycle Photonics and Optical Scanning Tunneling Microscopy
Route to Femtosecond Ångstrom Technology
By M. Yamashita, H. Shigekawa, R. Morita (Eds.) 2005, 241 figs., XX, 393 pages
Printing: Saladruck, Berlin

Binding: Stein+Lehmann, Berlin

Progress in Nano-Electro Optics III-Springer (2004)

Uploaded by

Copyright:

Available Formats

Progress in Nano-Electro Optics III-Springer (2004)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Progress in Nano-Electro Optics III-Springer (2004)

Uploaded by

Copyright:

Available Formats

Springer Series in

Editor-in-Chief: W. T. Rhodes, Atlanta

With 186 Figures and 8 Tables

Library of Congress Cataloging-in-Publication Data

Recent advances in electro-optical systems demand drastic increases in the

me to publish this monograph series. Finally, I extend an acknowledgement

Yokohama, October 2002 Motoichi Ohtsu

This volume contains ﬁve review articles focusing on diﬀerent aspects of

Yokohama, July 2004 Motoichi Ohtsu

Near-Field Optical Fiber Probes

7.3Fabrication of a Pure Silica Fiber Probe

A Novel Method for Forming Uniform Surface-Adsorbed

4.3Use of an Aperture and a Circular Metallic Plate . . . . . . . . . . . 105

Nano-Optical Media for Ultrahigh-Density Storage

A Phenomenological Description of Optical Near Fields

4.3Dynamics of Dipole-Forbidden States

Koji Asakawa Yoshiyuki Kamata

Takuya Matsumoto Masatoshi Sakurai

2 Basic Techniques for Tapering

2.1 Heating-and-pulling and Metallization Techniques

In the heating-and-pulling technique [14], a silica-based optical ﬁber is heated

Table 1. List of the SNOM images appeared in this chapter

Fig. Image Probe type Mode Feedback Ref.

Table 2. Characteristics of the three basic techniques for tapering

2.2 Meniscus Etching

In the meniscus etching1 , a single-mode ﬁber is immersed in HF acid with a

2.3 Selective Etching

By applying the selective-etching method to a highly GeO2 -doped ﬁber, one

Fig. 5. (a) Cross-sectional proﬁle of a refractive index of a silica ﬁber. Here, n1

GeO2 -Doped Fiber

approaches a constant value at X = 10–30. The cone angle that is determined

Pure Silica Core Fiber with a Fluorine-Doped Clad

Controlling the Clad Diameter

Fig. 7. Etching method to fabricate a shoulder-shaped probe. r1 , r2 , radii of the

D = 2r2 − 2(R2A TA + R2B TB ) , (7)

which is proportional to the etching time TB .

Table 3. Conditions for fabricating the probe as shown in Fig. 8a

3 Protrusion-type Probe and Its Imaging Applications

Fig. 9. (a) Schematic cross-sectional proﬁles of a protrusion-type probe and (b)

3.2 Fabricating Protrusion-type Probes

Fig. 11. Scanning electron micrograph of a protrusion-type probe fabricated by

3.3 c-mode SNOM Imaging of Salmonella Flagellar Filaments

Using a protrusion-type probe fabricated by selective etching and selective

Fig. 13. Schematic illustration of c-mode SNOM

Instrument of the c-mode SNOM

Figure 13 shows a schematic illustration of the optical system of c-mode

extracted by modulating the light and by amplifying the modulated signal

Near-Field Optical Images of Salmonella Flagellar Filaments

Figures 15c shows a near-field image of Salmonella flagellar filaments, ob-

Fig. 16. (a) constant-height mode and (b) constant-distance mode

bright region as indicated by arrows is 50 nm. The c-mode SNOM employ-

3.4 c-mode SNOM Images of Microtubules

protrusion-type probe is very eﬀective for high-resolution shear-force mi-

3.5 Near-Field Spectroscopic Investigation of Semiconductor

Fig. 19. Schematic illustrations of i-mode SNOM developed for photoluminescence

and b show scanning electron micrographs of a pulled-and-etched probe and

3.6 Transmission Eﬃciencies of the Protrusion-type Probes

In near-ﬁeld spectroscopic applications, one has to cope with near-ﬁeld signals

Fig. 22. Dependencies of the transmission eﬃciency of two protrusion-type probes