Photonics, A Short Course - Degiorgio

Undergraduate Lecture Notes in Physics
Vittorio Degiorgio
Ilaria Cristiani
Photonics
A Short Course
Second Edition
Undergraduate Lecture Notes in Physics (ULNP) publishes authoritative texts covering
topics throughout pure and applied physics. Each title in the series is suitable as a basis for
undergraduate instruction, typically containing practice problems, worked examples, chapter
summaries, and suggestions for further reading.
ULNP titles must provide at least one of the following:

• An exceptionally clear and concise treatment of a standard undergraduate subject.
• A solid undergraduate-level introduction to a graduate, advanced, or non-standard subject.
• A novel perspective or an unusual approach to teaching a subject.
ULNP especially encourages new, original, and idiosyncratic approaches to physics teaching
at the undergraduate level.
The purpose of ULNP is to provide intriguing, absorbing books that will continue to be the
reader’s preferred reference throughout their academic career.
Series editors
Neil Ashby
Professor Emeritus, University of Colorado, Boulder, CO, USA
William Brantley
Professor, Furman University, Greenville, SC, USA
Matthew Deady
Professor, Bard College Physics Program, Annandale-on-Hudson, NY, USA
Michael Fowler
Professor, University of Virginia, Charlottesville, VA, USA
Morten Hjorth-Jensen
Professor, University of Oslo, Oslo, Norway
Michael Inglis
Professor, SUNY Suffolk County Community College, Long Island, NY, USA
Heinz Klose
Professor Emeritus, Humboldt University Berlin, Germany
Helmy Sherif
Professor, University of Alberta, Edmonton, AB, Canada
More information about this series at http://www.springer.com/series/8917

Vittorio Degiorgio Ilaria Cristiani
•
Photonics
A Short Course
Second Edition
123
Vittorio Degiorgio Ilaria Cristiani
Department of Electrical, Computer Department Electrical, Computer
and Biomedical Engineering and Biomedical Engineering
University of Pavia University of Pavia
Pavia Pavia
Italy Italy
ISSN 2192-4791 ISSN 2192-4805 (electronic)

ISBN 978-3-319-20626-4 ISBN 978-3-319-20627-1 (eBook)
DOI 10.1007/978-3-319-20627-1
Library of Congress Control Number: 2015947401
Springer Cham Heidelberg New York Dordrecht London

© Springer International Publishing Switzerland 2014, 2016
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, 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.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
Preface
Apart from a number of small corrections and updates, the main change with
respect to the first edition is that a set of problems is provided at the end of each
chapter, except the last one. The problems deal with numerical computations
designed to illustrate the magnitudes of important quantities, and are also meant to
reinforce the understanding of the material presented and to test the students’ ability
to apply theoretical formulas.
Pavia, Italy Vittorio Degiorgio

May 2015 Ilaria Cristiani
Preface to the First Edition
The invention of the laser in 1960 has brought new ideas and new methods to many
areas of science and technology. Starting from the consideration that, in a simplified
view, a light beam can be seen as a stream of energy quanta, known as photons, the
name “photonics” has been coined, in analogy to electronics, to indicate the gen-
eration and use of photon streams for engineering applications. The applications of
photonics are everywhere around us from long-distance communications to DVD
players, from industrial manufacturing to health-care, from lighting to image for-
mation and display.
Many university programs in electrical engineering and applied physics propose
courses on photonics. Our aim is to offer a concise, rigorous, updated book, that can
serve as a textbook for an advanced undergraduate course. We believe that the book
can be also useful for graduate students, professionals, and researchers who want to
understand the principles and the functions of photonic devices.
The description of photonic devices essentially consists in a treatment of the
interactions between optical waves and materials. When the interaction is
non-resonant, an approach using Maxwell’s equations and the macroscopic optical
v
vi Preface
properties of materials is adequate. The description of optical amplification pro-

cesses is more complex, because it becomes unavoidable to introduce quantum
concepts. Our aim is to make laser action in atomic and semiconductor systems
understandable also to an audience that has little background in quantum
mechanics. Since photonics is based on a very fruitful interplay between physics
and engineering, we have tried to keep the balance between these two aspects in all
chapters. In order to give a feeling for the correct order of magnitudes of the dis-
cussed phenomena, several numerical examples are given throughout the book.
The book is divided into eight chapters. The first two chapters provide the
background in electromagnetism and optics for the entire book. In particular, the
second chapter describes the main optical components and methods of linear
radiation–matter interactions. The third chapter covers the principles of lasers, and
the properties of continuous and pulsed laser sources. Chapters 4 and 7 deal with
nonlinear interactions that produce processes such as modulation, switching, and
frequency conversion of optical beams. Chapter 5 describes photonic devices based
on semiconductors, including not only lasers, but also light-emitting diodes and
photodetectors. The properties of optical fibers, and of different fiber-optic com-
ponents and devices are the subject of Chap. 6. The last chapter discusses the most
important applications, emphasizing the main principles, while avoiding detailed
technical descriptions that may quickly become obsolete in this rapidly changing
field. Introductory information is also given on arguments of particular practical
importance, such as solid-state lighting, displays, and photovoltaic cells.
It should be emphasized that the invention of the laser has deeply affected many
fields of research, either by putting topics, such as optical spectroscopy, light
scattering, and coherence theory on a completely different footing, or by creating
entirely new topics, such as the physics of coherent light sources, nonlinear optics
and quantum optics. Our text does not touch upon those new topics that generally
require a deeper understanding of quantum concepts. The only exception is non-
linear optics, which can be mostly discussed in classical terms.
The book systematically uses the SI metric system with the exception of the use
of the electron volt (eV) for the energy of photons and atomic levels. The system of
units is summarized in Appendix A, together with a list of values of some fun-
damental physical constants. Use of acronyms is kept to a minimum, and the list of
those used is given in Appendix B.
The book is intended to be self-contained. The prerequisites include a back-
ground knowledge from standard courses in general physics, electromagnetic fields,
and basic electronics. Some background in quantum physics would be helpful.
Considering the level of treatment, we have judged it not useful to provide refer-
ences to articles that have appeared in scientific journals. Instead, to help an
interested reader gain a deeper insight into the fundamentals of photonics or into a
particular application, a bibliography containing a selection of a few excellent
reference books is given.
Preface vii
We are especially indebted to Lee Carroll for providing us with many sugges-
tions that greatly improved the presentation. We also acknowledge useful conver-
sations with Antoniangelo Agnesi, Cosimo Lacava, Roberto Piazza, and Giancarlo
Reali.
Pavia, Italy Vittorio Degiorgio

February 2014 Ilaria Cristiani
Contents
1 Electromagnetic Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Spectrum of Electromagnetic Waves . . . . . . . . . . . . . . . . . . . . . 1
1.2 Electromagnetic Waves in Vacuum . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Polarization of Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Paraxial Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.1 Spherical Waves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Gaussian Spherical Waves . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Diffraction Fresnel Approximation . . . . . . . . . . . . . . . . . . . . . . 13
1.6 Fraunhofer Diffraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.6.1 Rectangular and Circular Apertures. . . . . . . . . . . . . . . . 19
1.6.2 Periodical Transmission Function . . . . . . . . . . . . . . . . . 21
2 Optical Components and Methods. . . . . . . . . . . . . . . . . . . . . . . . . 25

2.1 Electromagnetic Waves in Matter . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Reflection and Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.1 Dielectric Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.2 Reflection from a Metallic Surface . . . . . . . . . . . . . . . . 33
2.2.3 Anti-reflection Coating . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2.4 Multilayer Dielectric Mirror . . . . . . . . . . . . . . . . . . . . . 38
2.2.5 Beam Splitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.6 Total-Internal-Reflection Prism. . . . . . . . . . . . . . . . . . . 41
2.2.7 Evanescent Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.2.8 Thin Lens and Spherical Mirror . . . . . . . . . . . . . . . . . . 45
2.2.9 Focusing Spherical Waves . . . . . . . . . . . . . . . . . . . . . . 47
2.2.10 Focusing Gaussian Spherical Waves . . . . . . . . . . . . . . . 48
2.2.11 Matrix Optics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.3 Fourier Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.4 Spectral Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.1 Dispersive Prism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.2 Transmission Grating . . . . . . . . . . . . . . . . . . . . . . . . . 55
ix
x Contents
2.4.3 Reflection Grating . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

2.4.4 Fabry-Perot Interferometer . . . . . . . . . . . . . . . . . . . . . . 59
2.5 Waves in Anisotropic Media . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.5.1 Polarizers and Birefringent Plates . . . . . . . . . . . . . . . . . 66
2.5.2 Jones Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.5.3 Rotatory Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.5.4 Faraday Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.5.5 Optical Isolator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.6 Waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3 The Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1 Conventional Light Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2 Origins of the Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.3 Properties of Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.4 Emission and Absorption of Light . . . . . . . . . . . . . . . . . . . . . . 87
3.4.1 Einstein Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.5 Optical Amplification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.6 Scheme and Characteristics of the Laser . . . . . . . . . . . . . . . . . . 96
3.7 Rate Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.8 The Laser Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.9 Solid-State and Gas Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.10 Pulsed Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.10.1 Q-Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.10.2 Mode-Locking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.11 Properties of Laser Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.11.1 Directionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.11.2 Monochromaticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.11.3 Spectrum of Laser Pulses . . . . . . . . . . . . . . . . . . . . . . 117
4 Modulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1 Linear Electro-Optic Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1.1 Phase Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.1.2 Amplitude Modulation . . . . . . . . . . . . . . . . . . . . . . . . 126
4.2 Quadratic Electro-Optic Effect . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.2.1 Liquid Crystal Modulators . . . . . . . . . . . . . . . . . . . . . . 135
4.3 Acousto-Optic Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.3.1 Acousto-Optic Modulation. . . . . . . . . . . . . . . . . . . . . . 139
4.3.2 Acousto-Optic Deflection . . . . . . . . . . . . . . . . . . . . . . 142
Contents xi
5 Semiconductor Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

5.1 Optical Properties of Semiconductors . . . . . . . . . . . . . . . . . . . . 145
5.2 Semiconductor Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.2.1 Homojunction Laser . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.2.2 Double-Heterojunction Structures . . . . . . . . . . . . . . . . . 151
5.2.3 Emission Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5.3 Semiconductor Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.4 Electroluminescent Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.4.1 LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.4.2 OLEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.5 Photodetectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.5.1 Photoelectric Detectors . . . . . . . . . . . . . . . . . . . . . . . . 162
5.5.2 Semiconductor Photodetectors . . . . . . . . . . . . . . . . . . . 165
5.5.3 CCD Image Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.6 Electro-Absorption Modulators . . . . . . . . . . . . . . . . . . . . . . . . . 167
6 Optical Fibers . . . . . . . . . . . . . . . ............... . . . . . . . . . 171

6.1 Properties of Optical Fibers . . . ............... . . . . . . . . . 171
6.1.1 Numerical Aperture . . ............... . . . . . . . . . 173
6.1.2 Modal Properties . . . . ............... . . . . . . . . . 173
6.2 Attenuation . . . . . . . . . . . . . . ............... . . . . . . . . . 176
6.3 Dispersion . . . . . . . . . . . . . . . ............... . . . . . . . . . 178
6.3.1 Dispersive Propagation of Ultrashort Pulses . . . . . . . . . . 178
6.4 Fiber Types . . . . . . . . . . . . . . ............... . . . . . . . . . 182
6.5 Fiber-Optic Components . . . . . ............... . . . . . . . . . 185
6.6 Fiber Amplifiers . . . . . . . . . . . ............... . . . . . . . . . 187
6.7 Fiber Lasers. . . . . . . . . . . . . . ............... . . . . . . . . . 190
7 Nonlinear Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.2 Second-Harmonic Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 194
7.2.1 Second-Order Nonlinear Optical Materials . . . . . . . . . . . 198
7.2.2 Phase Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
7.2.3 SHG with Ultrashort Pulses . . . . . . . . . . . . . . . . . . . . . 203
7.3 Parametric Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.4 Third-Order Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
7.5 Stimulated Raman Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 214
7.6 Stimulated Brillouin Scattering . . . . . . . . . . . . . . . . . . . . . . . . . 218
8 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
8.1 Information and Communication Technology . . . . . . . . . . . . . . . 221
8.1.1 Optical Communications . . . . . . . . . . . . . . . . . . . . . . . 222
8.1.2 Optical Memories . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
8.1.3 Integrated Photonics . . . . . . . . . . . . . . . . . . . . . . . . . . 226
xii Contents
8.2 Optical Metrology and Sensing . . . . . . . . . . . . . . . . . . . . . . . . 227

8.2.1 Measurements of Distance . . . . . . . . . . . . . . . . . . . . . . 227
8.2.2 Velocimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
8.2.3 Optical Fiber Sensors . . . . . . . . . . . . . . . . . . . . . . . . . 233
8.3 Laser Materials Processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
8.4 Biomedical Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
8.4.1 Ophtalmology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
8.4.2 Bioimaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
8.5 Liquid Crystal Displays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
8.6 LED Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
8.7 Photovoltaic Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Appendix A: System of Units and Relevant Physical Constants . . . . . . 243
Appendix B: List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Reading List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Chapter 1
Electromagnetic Optics
Abstract Light is an electromagnetic phenomenon that is described, in common

with radio waves, microwaves and X-rays, by Maxwell equations. This chapter
briefly reviews those aspects of electromagnetic theory that are particularly rele-
vant for optics. The propagation of quasi-collimated monochromatic light beams,
Gaussian waves, is treated by using the paraxial approximation. Diffraction theory
is described as an approximate theory of wave propagation. In particular the Fres-
nel approach is shown to be equivalent to the paraxial picture. Finally, the use of
the Fraunhofer approximation allows establishing a very useful relation between the
far-field pattern of the diffracted field and the spatial Fourier transform of the field
distribution immediately after the screen plane.
1.1 Spectrum of Electromagnetic Waves
The typical nomenclature of electromagnetic waves is given in Table 1.1, along with
the wavelength ranges associated with the different spectral bands. The field of
Photonics covers the infrared, visible, and ultraviolet wavelength ranges.
The term “light” is born to indicate electromagnetic radiation having wavelength
within the interval 0.4–0.75 µm, which is the interval spanned by the radiation emitted
by the sun. As a result of biological adaptation, this is also the wavelength range to
which the retina of our eyes is sensitive. In contemporary scientific terminology, the
term “light” is also applied to infrared and ultraviolet radiation.
In this text ν expresses the frequency of light, and is measured in units of hertz (Hz).
Alternatively, ω = 2π ν, the angular frequency of light can be used, and is measured
in radians per second. The wavelength of light, expressed as λ, is usually measured
in micrometers/microns (1 µm = 10−6 m) or nanometers (1 nm = 10−9 m). The
speed of light in vacuum is expressed as c, and is related to the wavelength and
frequency by c = λν. In general, the International System (SI) of units will be used.
Appendix A gives the list of prefixes used to indicate multiples and submultiples of
the fundamental units.
Table 1.2 shows the wavelength ranges corresponding to different colors of visible
electromagnetic waves.
© Springer International Publishing Switzerland 2016 1

V. Degiorgio and I. Cristiani, Photonics, Undergraduate Lecture Notes in Physics,
DOI 10.1007/978-3-319-20627-1_1
2 1 Electromagnetic Optics
Table 1.1 Spectrum of electromagnetic waves

Name Frequency Wavelength
Ultralow frequencies <30 kHz >10 km
Low frequencies 30–300 kHz 10–1 km
Radio frequencies 0.3–3 MHz 1–0.1 km
High frequencies (HF) 3–30 MHz 100–10 m
VHF 30–300 MHz 10–1 m
UHF 0.1–1 GHz 1–0 .3 m
Microwaves 1–100 GHz 30–0.3 cm
Millimeter waves 0.1–3 THz 3–0.1 mm
Infrared 3–400 THz 100–0.75 µm
Visible 400–750 THz 0.75–0.4 µm
Ultraviolet 750–1,500 THz 0.4–0.2 µm
Vacuum UV 1.5–6 ×103 THz 200–50 nm
Soft X rays 6–300 ×103 THz 50–1 nm
X rays 0.3–30 ×106 THz 1–0.01 nm
γ rays >106 THz <0.3 nm
Table 1.2 Visible

Color λ (nm)
electromagnetic waves
Violet 400–450
Blue 450–495
Green 495–575
Yellow 575–595
Orange 595–620
Red 620–750
Historically, the nature of light has been subject to considerable debate. Some sci-
entists, like Huygens and Young, posited a “wave-like” nature of light, while others,
like Newton, proposed a “corpuscular” point of view. The highly successful descrip-
tion of light provided by Maxwell’s equations seemed to end the debate, by providing
a strong validation of the wave-like interpretation. However the controversy reopened
in the last part of the nineteenth century, because the wave-like interpretation was
inconsistent with experimental observations of certain phenomena such as the photo-
electric effect and the spectral radiance of blackbody emission. Planck and Einstein
clearly demonstrated that electromagnetic energy is emitted and absorbed in discrete
units, or fundamental quanta of energy, now called photons.
Quantum theory provides a unified view that combines both the wave-like and the
corpuscular aspects of light. Traditionally, textbooks discussing lasers and related
applications do not make full use of the quantum theory of light. Instead, they adopt an
intermediate approach by treating all wave propagation phenomena with Maxwell’s
1.1 Spectrum of Electromagnetic Waves 3
equations, and introducing light quanta only when discussing absorption and emis-
sion phenomena. This intermediate approach is also used in this book, as it is partic-
ularly suited to a concise introduction to Photonics.
Considering a monochromatic plane wave with frequency ν and wave vector k,
the corresponding photon is a quasi-particle without mass, with an energy of hν
and a momentum of hk/(2π ). The universal constant, h, called Planck constant,
has value 6.55 × 1034 J·s. Usually the photon energy is conveniently expressed in
units of electronvolt (eV): 1 eV is the energy acquired by an electron (electric charge
e = 1.6 × 10−19 C) when it is accelerated by a potential difference of 1 V, and so
1 eV = 1.6 × 10−19 J.
1.2 Electromagnetic Waves in Vacuum
The vacuum propagation of electromagnetic waves is described by Maxwell’s

equations:
∇ · E = 0, (1.1)
∇ · B = 0, (1.2)
∂E
∇ × B = εo μo , (1.3)
∂t
∂B
∇ ×E=− , (1.4)
∂t
where E is the electric field (measured in V/m), and B is the magnetic induction
(measured in T ). The constant εo = 8.854 × 10−12 F/m is called the dielectric
constant (or electric permittivity) of free space, and μo = 4π × 10−7 H/m is the
vacuum magnetic permeability.
Recalling a general property of differential operators:
∇ × (∇ × E) = ∇(∇ · E) − ∇ 2 E, (1.5)
taking the rotor of both members of (1.4), and using (1.1), it is found that:
∂(∇ × B)
∇ × (∇ × E) = −∇ 2 E = − . (1.6)
∂t
By using (1.3), one obtains a single equation, called the wave equation, in which
the unknown function is the electric field:
1 ∂ 2E
∇2E = , (1.7)
c2 ∂t 2
where c = (εo μo )−1/2 is the propagation velocity of the electromagnetic wave. It

can be shown that B satisfies an equation that is identical to (1.7).
The wave described by (1.7) transports both energy and momentum. The flow of
electromagnetic energy is governed by the Poynting vector, defined as:
SP = E × H, (1.8)
where H = B/μo is the magnetic field (units: A/m). The modulus of SP , which has
dimension W/m2 , represents the power per unit area transported by the electromag-
netic wave.
An important particular solution of (1.7) is the monochromatic plane wave:
E = Eo cos (ωt − k · r + φ) = Eo Re{exp[−i(ωt − k · r + φ])}, (1.9)
where k is the propagation vector, with modulus k = 2π/λ = ω/c, and φ is a

constant phase that can always be put equal to 0 by an appropriate choice of the
time origin. The wave described by (1.9) generates plane equal-phase surfaces (also
called wavefronts) that are perpendicular to k.
By substituting (1.9) into (1.1), one finds: ik · E = 0, that is: E ⊥ k. Analogously,
from (1.2) one derives B ⊥ k. In addition, Maxwell equations say that E ⊥ B,
√
that E and B have the same phase, and that B = εo μo E. Summarizing, the
electromagnetic plane wave is a transverse wave, in which electric and magnetic
field vectors are mutually orthogonal and lie in a plane that is perpendicular to the
propagation direction.
The Poynting vector of the plane wave is parallel to the propagation vector k. By
inserting (1.9) into (1.8), the following expression is found:
k
SP = cεo E o2 cos2 (ωt − k · r + φ). (1.10)
k
Optical fields oscillate at very high frequencies. For instance, at the wavelength
λ = 1 µm, the corresponding oscillation period is T = λ/c = 3.3×10−15 s = 3.3 fs.
In most cases it is more significant to consider time averaged quantities instead of
instantaneous values. To this end, it is useful to recall a general property of sinusoidal
functions: given two functions oscillating at the same frequency, a(t) = A cos(ωt +
φ A ) and b(t) = B cos(ωt + φ B ), the time average of the product is:
1
a(t)b(t) = AB cos(φ A − φ B ). (1.11)
2
Using (1.11), it is found that the time-average of the modulus of SP , which is the
intensity of the wave, is given by:
εo E o2
I =c . (1.12)
2
1.2 Electromagnetic Waves in Vacuum 5
As shown in (1.9), the electric field can be described by using complex exponen-
tials of the type E = Eo exp[−i(ωt − k · r + φ)] instead of real sinusoidal functions.
In this text the exponential notation will be often used to simplify the calculations. In
any case it should never be forgotten that only the real part of the complex exponential
corresponds to the real electric field.
1.3 Polarization of Light
The direction of polarization of an electromagnetic wave is the direction of the vector

E. Assume that a plane wave is propagating in the z direction. The electric field is
oriented in the x-y plane. Its x and y components are both sinusoidal functions of
time:
E x = E xo cos(ωt − kz), (1.13)
E y = E yo cos(ωt − kz + φ), (1.14)
where φ is the phase shift between the two components. Putting X = E x /E xo =

cos(ωt −kz) and Y = E y /E yo = cos(ωt −kz +φ) = cos(ωt −kz) cos φ −sin(ωt −
kz) sin φ, it is possible to write an implicit equation containing X and Y , but not the
spatio-temporal coordinates:

X cos φ − Y = 1 − X 2 sin φ. (1.15)
By squaring both sides of (1.15) one obtains:
X 2 + Y 2 − 2X Y cos φ = sin2 φ, (1.16)
hence:
E x2 E y2 Ex E y
2
+ 2
−2 cos φ = sin2 φ. (1.17)
E xo E yo E xo E yo
Equation (1.17) describes the trajectory of the endpoint of the vector E in the
plane x-y. The motion is periodic in time, with period T = 1/ν. In the general case,
the trajectory is an ellipse, as shown in Fig. 1.1, and the wave is said to be elliptically
polarized. In the particular case of φ = 0, the ellipse reduces to a straight line, i.e.,
the direction of the electric field vector does not change with time. The straight line
forms an angle α with the x axis, given by the relation tan(α) = E yo /E xo . Linear
polarization is also obtained when φ = π , in this case the polarization direction
forms an angle −α with the y axis.
In the case φ = ±π/2 and E xo = E yo , the endpoint of the vector E sketches out
a circle in one optical period, and so the wave is said to be circularly polarized. The
rotation of E can be clockwise, if φ = π/2, or anti-clockwise, if φ = −π/2.
Fig. 1.1 Polarization of electromagnetic waves for different values of the phase shift between E x
and E y
A generic polarization state can also be described as a superposition of two circular

polarizations, each having a different amplitude and phase. If the two amplitudes
are different the resulting polarization is elliptical. If the amplitudes are equal, the
superposition corresponds to a linear polarization, having direction controlled by the
phase shift between the two circular polarizations.
It is usually said that the light emitted by the Sun and by conventional sources is
unpolarized. What is meant here is that the direction of the electric field observed in
a given point is an essentially random function of time
1.4 Paraxial Approximation
Representing the field of the monochromatic wave as: E(r, t) = E(r)exp(−iωt),

and substituting this expression into the wave equation (1.7), allows the so-called
Helmholtz equation to be obtained:
1.4 Paraxial Approximation 7
(∇ 2 + k 2 )E(r) = 0 (1.18)
In real situations one never deals with ideal plane waves, but rather with collimated
beams, like those emitted by lasers. It is therefore useful to develop approximated
treatments applicable to light beams possessing a relatively small distribution of
wave vectors. Since the discussion will only concern the spatial properties of mono-
chromatic fields, the time-dependent term exp(−iωt) will be omitted from now on
in this chapter inside all the expressions involving the electric field.
Considering a light beam that travels along the z axis, and substituting the fol-
lowing expression for the electric field inside (1.18):
E(x, y, z) = U (x, y, z)exp(ikz), (1.19)
the Helmholtz equation becomes:
∂ 2U ∂ 2U ∂ 2U ∂U
+ + + 2ik = 0. (1.20)
∂x2 ∂ y2 ∂z 2 ∂z
It is now assumed that U is a slowly varying function of z, where slowly varying

means that the spatial scale over which U has significant variations is much bigger
than the wavelength λ, as qualitatively shown in Fig. 1.2. Mathematically, this is
equivalent to state that:
dU
kU. (1.21)
dz
A consequence of (1.21) is that the second derivative of U with respect to z is

much smaller than kdU/dz, and so it can be neglected inside (1.20), obtaining the
so-called paraxial equation:
∂ 2U ∂ 2U ∂U
+ + 2ik = 0. (1.22)
∂x2 ∂ y2 ∂z
Fig. 1.2 Paraxial approximation

In order to better clarify the meaning of the paraxial approximation, it is useful to

consider the simple case of a wave made by the superposition of three plane waves,
one having vector k directed along the z axis, the two others with k laying in the
plane x-z and forming angles ±α with the z axis. By assuming that the three plane
waves have the same amplitude Uo , U (x, y, z) can be written as:
U (x, y, z) = Uo {1 + exp[ik(cos α − 1)z][exp(ikx sin α) + exp(−ikx sin α)]}.
The first derivative with respect to z is:
dU
= ik(cos α − 1)(U − Uo ).
dz
If α is small, then the quantity cos α − 1 is also small with respect to 1. Therefore
it is demonstrated that:
dU

|k(U − Uo )| < |kU |.
dz
1.4.1 Spherical Waves
An important particular solution of the wave equation (1.7) is a wave with spherical
wavefronts, whose electric field, in scalar form, is expressed as:
Ao
E= exp(ik|r − ro |) (1.23)
|r − ro |
The origin of the spherical wave is at point ro having coordinates xo , yo , z o .

Putting the origin on the z axis means xo = yo = 0, and the distance |r − ro | is given
by:

|r − ro | = x 2 + y 2 + (z − z o )2

x 2 + y2
= (z − z o ) 1 + (1.24)
(z − z o )2
The paraxial approximation is applied to (1.24) by assuming that the fraction

appearing under the square root in the last term is small compared to 1. By expanding
the square root in a power series, and keeping only the first-order term, the following
expression is derived:

x 2 + y2
|r − ro | ≈ (z − z o ) 1 + . (1.25)
2(z − z o )2
Fig. 1.3 Convention for the sign of the radius of curvature of the spherical wave
To use (1.25) instead of (1.24) is analogous to approximating the spherical wave-

front with a paraboloid. The validity of (1.25) is clearly limited to situations in which
only that part of the wavefront for which the distance from the z axis is much less
than z − z o is considered.
Calling R(z) = z − z o the radius of curvature of the wavefront allows the complex
amplitude of the electric field associated to the spherical wave to be written as:

Ao x 2 + y2
U (x, y, z) = exp ik . (1.26)
R(z) 2R(z)
The radius of curvature can be positive or negative: the chosen convention (shown
in Fig. 1.3) being that positive R indicates a wavefront having convexity oriented
toward positive z, i.e. when the wave is diverging.
1.4.2 Gaussian Spherical Waves
As will be shown in Chap. 3, the beam emitted by a laser can usually be described as
a spherical wave presenting a Gaussian amplitude distribution in the x-y plane. The
electric field of this Gaussian spherical wave is:

x 2 + y2 k(x 2 + y 2 )
U (x, y, z) = A(z)exp − exp i
w2 2R

k(x + y )
2 2
= A(z)exp i , (1.27)
2q
where w is the distance from the z axis at which the field amplitude is reduced by a
factor 1/e. The parameter w can be called radius of the Gaussian wave.
The quantity q appearing in the right-hand side of (1.27) is called the complex
radius of curvature. It is defined by the expression:
1 1 λ
= +i (1.28)
q R π w2
By substituting (1.27) inside (1.22), one obtains:

k 2 dq
2 2ik q d A
− 1 x + y 2
+ + 1 A(z) = 0 (1.29)
q 2 dz q A dz
Equation (1.29) can be satisfied for all possible values of x and y provided that:
dq d A(z) A(z)
=1; =− . (1.30)
dz dz q(z)
By using the initial conditions q(z o ) = qo , A(z o ) = Ao , the following solutions

to (1.30) are found:
A(z) qo
q(z) = qo + z − z o ; = . (1.31)
Ao q(z)
It is therefore demonstrated that the Gaussian spherical wave is a solution of the

paraxial equation. This means that the wave does not change its functional form
under propagation. What happens is that the amplitude A(z) and the complex radius
of curvature q(z) change with z following simple algebraic laws. In particular the
evolution of q(z) is identical to that of the radius of curvature of the spherical wave.
The case of a Gaussian spherical wave having a planar wavefront at z = 0 is now
considered. If the radius of curvature of the wavefront is infinite for z = 0, then the
reciprocal of the complex radius of curvature at z = 0, as derived from (1.28), is:
1/qo = iλ/(π wo2 ), where wo is the beam radius at z = 0. The field distribution at
z = 0 is:
2
k(xo2 + yo2 ) x + y2
U (xo , yo , 0) = Ao exp i = Ao exp − o 2 o . (1.32)
2qo wo
Having assigned U (xo , yo , 0), the field distribution at the generic coordinate z
is immediately derived by using the two conditions of (1.31) that describe how the
amplitude and the complex radius of curvature are modified during the propagation
along the z axis. Note that the propagation problem is solved through the algebraic
relations (1.31) without making use of the differential equation (1.22). It is found:

qo k(x 2 + y 2 )
U (x, y, z) = Ao exp i , (1.33)
qo + z 2(qo + z)
which can also be written as:
wo k(x 2 + y 2 )
U (x, y, z) = Ao exp [−iψ(z)]exp i , (1.34)
w(z) 2q(z)
where
iπ wo2
q(z) = z + qo = z − , (1.35)
λ

2 2
λz z
w(z) = wo 1 + = wo 1 + , (1.36)
π wo2 zR
and λz z
ψ(z) = arctan = arctan . (1.37)
π wo2 zR
The quantity z R is called Rayleigh length, and is defined as follows:
π wo2
zR = . (1.38)
λ
The behavior of the beam radius w(z) is reported in Fig. 1.4. It is seen that the
Gaussian spherical wave has the minimum radius at the position (usually called
beam waist) in which its wavefront is planar. The Rayleigh length represents the
propagation distance at which the beam radius becomes larger than wo by a factor
√
2. The beam radius is slowly growing for z ≤ z R and tends to grow linearly as
z
zR.
Equation (1.34) indicates that the propagating Gaussian spherical wave acquires,
with respect to the plane wave, an additional phase delay of ψ(z) that, according to
(1.37), varies from −π/2 for z z R to π/2 for z
z R , as shown in Fig. 1.5.
By using (1.28) and (1.35) the z-dependence of the radius of curvature of the
wavefront is derived:
Fig. 1.4 Gaussian spherical

wave: behavior of the beam
radius, w(z), as a function of
the propagation distance

wave: behavior of the
additional phase delay, ψ(z),
as a function of the
propagation distance

wave: behavior of the radius
of curvature of the
wavefront, R(z), as a
function of the propagation
distance
1 π wo2 2 zR2
R(z) = z + =z+ . (1.39)
z λ z
The behavior of R(z) is illustrated in Fig. 1.6.

It is interesting to discuss the behavior of the Gaussian spherical wave at distances
from the waist plane much larger than the Rayleigh length. If z
z R , one finds
from (1.36):
λz
w(z) ≈ . (1.40)
π wo
Since the beam radius grows proportionally to distance, most of the propagating
power is concentrated inside a cone having the aperture:
λ
θo = . (1.41)
π wo
The angle θo is the divergence angle of the beam. Equation (1.41) expresses a
very general concept for wave propagation: the finite transversal size of the beam
is necessarily associated to an angular spread of the wave vector directions. Apart
from a numerical factor of the order of one, the divergence angle is given by the ratio
between wavelength and transversal size of the light beam. As it will be seen in the
following sections, the same result is found in the classical diffraction problem, in
which a plane wave crossing a screen with a circular aperture of radius D/2 generates
a beam with an angular spread ≈2λ/D.
To complete the discussion concerning the behavior at z
z R , (1.39) shows that
R(z) approaches the value of z, that is, it tends towards the value expected for a
spherical wave centered at z = 0.
The paraxial equation admits other beam-like solutions that, like the spherical
Gaussian wave, do not change their shape during free propagation. An interesting
family of such solutions is that of Hermite-Gauss spherical waves, whose amplitude,
in Cartesian coordinates, is the product of a Gaussian function times two Hermite
polynomials:
wo
U (x, y, z) = Ao exp [−i(l + m + 1)ψ(z)]
w(z)
k(x 2 + y 2 ) √ x √ y
exp i Hl 2 Hm 2 , (1.42)
2q(z) w(z) w(z)
where the Hermite polynomial of order n is defined as
dn
Hn (ξ ) = exp(ξ 2 )(−1)n exp(−ξ 2 ). (1.43)
dξ n
Applying (1.43) to generate, as an example, the polynomials of order 0, 1, 2, it is

found:
H0 (ξ ) = 1; H1 (ξ ) = 2ξ ; H2 (ξ ) = 4ξ 2 − 2. (1.44)
The Hermite-Gaussian beam of order (0,0) is the simple Gaussian beam. As will
be seen in Chap. 3, the Hermite-Gaussian beams are useful to describe the field
distribution inside laser cavities.
To summarize, the treatment developed in this section has demonstrated that
Gaussian spherical waves have propagation properties similar to those of spherical
waves in the geometrical optics approach. However one important difference is that
all the formulas developed in this section include the effects due to diffraction.
1.5 Diffraction Fresnel Approximation
The classical diffraction theory concerns the situation illustrated in the scheme of
Fig. 1.7. A monochromatic plane wave, with propagation vector k parallel to the z
axis is transmitted through an aperture Σ in an opaque screen laying on plane x-y. In
Fig. 1.7 Diffraction from an

aperture Σ in the plane x-y
order to exactly determine the spatial distribution of the electromagnetic field in the
semi-space beyond the screen, one should solve the wave equation with appropriate
boundary conditions. In most cases the exact calculations are too complicated, and
so approximate theories have to be developed. The simplest approach would be that
of geometrical optics: the electric field of the transmitted wave would be different
from zero only in the cylindrical region having axis parallel to z and the aperture Σ as
the basis. Since it is experimentally observed that the pattern of the transmitted wave
extends outside the simple shadow of the aperture, it is necessary to go beyond this
geometrical optics approach. The standard diffraction theory, put forward by Huy-
gens and successively elaborated by Fresnel and Kirchhoff, is an approximate theory
based on the following model. Each infinitesimal element of Σ generates a spherical
wave, having amplitude and phase determined by the incident field. The electric field
at an observation plane at distance z from the screen is given by the superposition of
all the spherical waves coming from the different infinitesimal elements composing
the aperture Σ. The phase of each spherical wave is fixed by the distance between
the considered surface element and the observation point. Note that the superposition
principle is direct consequence of the linearity of Maxwell equations.
The incident field is not necessarily a plane wave. It can be generally described by
a complex amplitude E i (xo , yo ), where xo , yo are the coordinates on the screen plane.
The diffracted field is given at point P(x, y, z) by the following surface integral:

i exp(ik|r − ro |)
E(r) = E i (ro ) cos θ d xo dyo , (1.45)
λ Σ |r − ro |
where the vectors r = (x, y, z) and ro represent, respectively, the position of P and
the position of the surface element d xo dyo . θ is the angle formed by the vector r − ro
with the z axis. Therefore:
z
cos θ = , (1.46)
(x − xo ) + (y − yo )2 + z 2
2
1.5 Diffraction Fresnel Approximation 15
and

(x − xo )2 + (y − yo )2
|r − ro | = (x − xo )2 + (y − yo )2 + z 2 = z 1+ .
z2
(1.47)
The imaginary unit i placed in front of the integral (1.45) indicates that the spher-
ical wave coming from the generic surface element is phase-shifted by π/2 with
respect to the incident wave. Provided that the diffracted field is evaluated at an
observation point satisfying the condition z
λ, the integral (1.45) represents a
good approximation to the exact solution.
The integral (1.45) can be extended to the whole plane x-y by inserting a transmis-
sion function τ (xo , yo ), which takes the value 1 inside the aperture Σ and 0 outside.
The introduction of τ (xo , yo ) allows for generalizing the treatment to partially trans-
mitting screens and/or phase screens, in which case τ becomes a complex quantity
with a modulus taking intermediate values between 0 and 1.
The generalized diffraction integral is written as:
∞ ∞
i exp(ik|r − ro |)
E(r) = d xo τ (xo , yo )E i (ro ) cos θ dyo . (1.48)
λ −∞ −∞ |r − ro |
Diffraction theory, as expressed in (1.48), constitutes an approximate propagation

theory of electromagnetic waves. In fact, once a field distribution is defined on the
plane z = 0 immediately beyond the screen plane, the integral (1.48) gives the field
distribution at any later generic plane z.
In many cases it is useful to consider the Fresnel approximation, which assumes
that the angle θ is small for all the considered points. Mathematically, this is equiva-
lent to the assumption that the fraction appearing inside the square root in the right-
hand member of (1.47) is small compared to 1. By truncating the series expansion
of (1.47) at the first order, the following expression is obtained:

(x − xo )2 + (y − yo )2
|r − ro | ≈ z 1 + . (1.49)
2z 2
Considering the fraction inside the integral (1.48), the approximation |r − ro | = z

can be used for the denominator, but not for the numerator. In general, in the case of
an exponential of the type exp[i(A + A1 )], the assumption that A1 is much smaller
than A is not in itself a sufficient condition to neglect A1 . Since the exponential is
a periodic function with period 2π , A1 can be neglected only if it is much smaller
than 2π .
By substituting (1.49) inside (1.48), and putting cos θ = 1, the diffraction integral
under the Fresnel approximation becomes:

i exp(ikz) ∞
E(r) = d xo
λz −∞
∞
ik
τ (xo , yo )E i (ro )exp [(x − xo )2 + (y − yo )2 ] dyo (1.50)
−∞ 2z
The integral (1.50) is a convolution of a complex Gaussian nucleus with the field
distribution present in the plane immediately after the screen.
The Fresnel approximation is valid provided that the first neglected term at the
exponent of the complex exponential is small compared to 2π :
k(x − xo )4
2π . (1.51)
4z 3
Considering, as an example, a circular aperture of diameter D, (1.51) is verified if:
D4
z3
. (1.52)
λ
However, it should be mentioned that the comparison of the Fresnel approximation
with exact numerical results shows a good agreement even in cases in which the
condition (1.52) is not fully satisfied.
The Fresnel approximation is now applied to describe the free propagation of a
Gaussian spherical wave. It is assumed that the wave has its waist on the plane z = 0.
This means that the radius of curvature of the wavefront is infinite at z = 0, and thus:
1/qo = iλ/(π wo2 ), where wo is the beam radius at z = 0. The field distribution at
z = 0 is:
k(xo2 + yo2 )
E i (xo , yo ) = Ao exp i . (1.53)
2qo
The field distribution at the generic plane z is derived by inserting (1.53) into
(1.50), and putting τ (xo , yo ) = 1. The diffraction integral is calculated by separating
the integration variables. The result is:
i exp(ikz)
E(x, y, z) = Ao J (x, z)J (y, z), (1.54)
λz
where:
∞ k(x − xo )2 kx 2
J (x, z) = exp i exp i o d xo . (1.55)
−∞ 2z 2qo
Recalling that:
∞ ∞
x2 (x − a)2 √
exp − 2 d x = exp − 2
d x = π w, (1.56)
−∞ w −∞ w
1.5 Diffraction Fresnel Approximation 17
where a is an arbitrary constant, the integral (1.55) is readily calculated by expressing

the integrand function as a displaced Gaussian function:
∞
ik qo + z 2 2x xo x2
J (x, z) = exp x − + d xo
−∞ 2 qo z o z z

ikx 2 ∞ ik(qo + z) qo x 2
= exp exp xo − d xo . (1.57)
2(qo + z) ∞ 2qo z qo + z
The final expression of E(x, y, z) is:

qo ikz k(x 2 + y 2 )
E(x, y, z) = Ao e exp i (1.58)
qo + z 2(qo + z)
This result, which exactly matches (1.33), is not surprising because the Fresnel
approximation of the diffraction integral is essentially equivalent to the paraxial
approximation.
1.6 Fraunhofer Diffraction
A further simplification to diffraction theory is introduced by the Fraunhofer approx-

imation, which consists of neglecting the terms quadratic in xo and yo in the exponent
inside (1.50). This simplification is valid when the terms to be neglected are small
in comparison to π . This requires:
(xo2 + yo2 )
z
. (1.59)
λ
For a circular aperture with diameter D, the maximum value of xo2 + yo2 is D 2 /4,
therefore the condition (1.59) becomes:
D2
z
. (1.60)
4λ
In most practical cases it may be very difficult to satisfy such a condition. If, for
instance, λ = 0.6 µm, and D = 5 cm, then (1.60) requires z
3 km. This example
explains why the Fraunhofer approximation is also called the far-field approximation.
As it will be described in the next chapter, a converging lens can be used to bring
the far field into the lens focal plane. By exploiting this possibility, the Fraunhofer
approximation can be utilized in various applications.
Fig. 1.8 Decomposition of an optical wave into plane waves
The integral (1.50), as written in the Fraunhofer approximation, becomes:

i exp(ikz) ik 2
E(x, y, z) = exp (x + y 2 )
λz 2z
∞ ∞
d xo τ (xo , yo )E i (xo , yo )exp[−i(k/z)(x xo + yyo )]dyo . (1.61)
−∞ −∞
Apart from a multiplicative factor, the field distribution E(x, y, z) given by (1.61)
is the two-dimensional Fourier transform of the distribution at z = 0. The two
variables that are conjugated to xo , yo are: f x = x/(λz) and f y =y/(λz). These are
the components of the spatial frequency f, having modulus f = f x2 + f y2 .
In order to grasp the physical meaning of the quantity f, it should be recalled that
the field τ E i present immediately after the plane xo -yo can always be described as a
superposition of plane waves. As shown in Fig. 1.8, each plane wave is characterized
by a different vector k. The direction of k is determined by assigning the two angles αx
and α y , formed by k with the axes x and y. Once given αx and α y , the angle αz between
k and z may be calculated from the relation: cos2 αx + cos2 α y + cos2 αz = 1. For
every pair of values of αx and α y , there corresponds a spatial frequency f with
components f x = λ−1 cos αx e f y = λ−1 cos α y . The modulus of f is:

1 − cos2 αz sin αz
f = = . (1.62)
λ λ
The term “spatial frequency” comes from the fact that f −1 is the spacing (spatial
period) of the projected wavefronts on the plane xo -yo , as shown in Fig. 1.8. If the
plane wave propagates perpendicularly to z, αz = π/2 and f −1 = λ. If k is parallel
to z, αz = 0, and f −1 becomes infinite.
Equation (1.62) indicates that the plane wave characterized by αx and α y is repre-
sented on the plane z in the far field by the point x = f x λz and y = f y λz. In the case
where the field incident on the screen is a plane wave directed along z, E(x, y, z) is
simply the Fourier transform of the transmission function τ (xo , yo ).
1.6 Fraunhofer Diffraction 19
1.6.1 Rectangular and Circular Apertures
As a first example of the application of the Fraunhofer approximation, a rectangular

aperture illuminated by a plane wave propagating along z is here considered. The
assumed transmission function is:

xo yo
τ (xo , yo ) = r ect r ect , (1.63)
Lx Ly
where r ect (x) is unitary inside the interval −0.5 ≤ x ≤ 0.5 and is 0 elsewhere.
The transform of (1.63) is: F{τ (xo , yo )} = L x L y sin c(π L x f x ) sin c(π L y f y ) where
sinc(x) = sinx/x. By substitution into (1.61), it is found:

i exp(ikz) ik 2 x Lx yLy
E(x, y, z) = Ao exp (x + y 2 ) L x L y sinc π sinc π ,
λz 2z λz λz
(1.64)
where Ao is the amplitude of the incident wave.

Figure 1.9 shows the image of the diffracted wave, whereas Fig. 1.10 illustrates the
transversal behavior of the optical intensity I (x, y, z), proportional to |E(x, y, z)|2 ,
along the x axis. While the wave incident on the aperture consists only of the prop-
agation vector parallel to z, the diffracted wave has a distribution in k vectors that
becomes wider as the size of the aperture reduces. In fact, the width of the main
diffraction lobe, evaluated along x as the distance between the first two zeroes, is:
Δx = λz/L x , or, in other terms, the angular aperture of the main diffraction lobe is
λ/(2L x ).
As a second example, consider a circular aperture having a transmission function:
r
o
τ (xo , yo ) = circ , (1.65)
R
Fig. 1.9 Fraunhofer

diffraction from a square
aperture: image of the
diffracted wave
Fig. 1.10 Fraunhofer

diffraction from a square
aperture: intensity versus x
where circ(ro /R) is equal to 1 inside the circle of radius R and 0 elsewhere. Using
(1.61), the diffracted field is found to be:

i exp(ikz) ik 2 J1 [r R/(λz)]
E(r ) = Ao exp r R2 , (1.66)
λz 2z r R/(λz)
where J1 is the order 1 Bessel function. The intensity distribution I (r ), proportional

to |E(r )|2 , is called Airy function. Figure 1.11 shows the image of the diffracted
wave, and Fig. 1.12 gives the behavior of I (r ). The radius corresponding to the first
zero is: 0.61 λz/R. Here again it is found that the angular spread of the diffracted
wave is of the order of the ratio between wavelength and size of the aperture.

diffraction from a circular
aperture: image of the
diffracted wave

diffraction from a circular
aperture: behavior of the
diffracted intensity I (r ) as a
function of the distance from
the z axis
1.6.2 Periodical Transmission Function
Consider a plane wave propagating along z that illuminates a square aperture of

side L that has a transmission function varying sinusoidally along x, with period
d = 1/ f o :
1 + m cos (2π f o xo ) x y
o o
τ (xo , yo ) = r ect r ect , (1.67)
2 L L
where the parameter m is the modulation depth.
The Fourier transform of the fraction that appears in the right-hand member of
(1.67) is:

1 + m cos 2π f o xo 1 m m
F δ( f x , f y ) + δ( f x + f o , f y ) + δ( f x − f o , f y ),
=
2 2 4 4
(1.68)
where δ(x) is the Dirac delta function, which is infinite at x = 0 and zero elsewhere.
Recalling that the Fourier transform of the product of two functions is the convolution
of the two transforms, the far-field diffraction pattern is given by:

i L 2 exp(ikz) ik 2 yL
E(x, y, z) = Ao exp (x + y 2 ) sinc π
2λz 2z λz

xL m L m L
sinc π + sinc π (x + f o λz) + sinc π (x − f o λz) .
λz 2 λz 2 λz
(1.69)
The behavior of the diffracted intensity as a function of x is reported in Fig. 1.13.

The three peaks correspond to three diffracted waves. The central peak, i.e., the zero-
th order wave, represents the fraction of incident wave that is still propagating along
Fig. 1.13 Diffracted

intensity from a periodical
transmission function
z, with a distribution of k vectors arising from the diffracting effect due to the finite
size of the aperture. The lateral peaks, symmetrically positioned around the central
one, represent waves with propagation vectors laying in the x-z plane and forming
the angles ±λ/d with the z axis. These two waves constitute the diffraction order +1
and −1, respectively. The spatial separation between order 0 and order ±1 is f o λz,
which means that the angular separation is ± f o λ. The two first-order diffraction
peaks are broadened, like the central peak, because of the finite size of the aperture.
At this point it is easy to understand what happens if the transmission function
τ (xo , yo ) is a periodic function with a square profile instead of a sinusoidal profile.
The square profile contains, with decreasing amplitude, all the harmonics. As a
consequence, diffracted waves with orders higher than 1 are generated.
An other interesting case is that of a phase screen, having a transmission function
that is a complex function with unitary modulus and sinusoidal phase:
x y
m sin (2π f o xo ) o o
τ (xo , yo ) = exp i r ect r ect . (1.70)
2 L L
Such a transfer function can be realized by using a transparent material with a

periodic modulation of index of refraction or of thickness along the coordinate xo .
Recalling that:
∞ m
m sin 2π f o xo
exp i = Jq exp(i2πq f o xo ), (1.71)
2 q=−∞
2
where Jq is the q-th order first-type Bessel function, and using the convolution
theorem, one finds:
i L 2 exp(ikz) ik yL
E(x, y, z) = Ao exp (x 2 + y 2 ) sin c π
2λz 2z λz
∞
m πL
Jq sin c (x − q f o λz) (1.72)
q=−∞
2 λz
Equation (1.72) shows that the sinusoidal phase profile presents, in principle, all
possible diffraction orders. The amplitude of the generic order q is determined by
the value of the corresponding Bessel function for the argument m/2. Note that the
transmitted beam (that is, the zero-th order) can be totally suppressed in the particular
case of m/2 coinciding with a zero of J0 .
As it will be seen in the next chapter, the diffraction from periodical transmission
profiles has important applications to spectral measurements performed by using
gratings. The Fraunhofer approximation allows calculating not only the geometrical
properties, but also the power partitioning over the various diffraction orders of the
grating.
Problems
1.1 Consider the plane electromagnetic wave whose electric field is given by the
expressions (in SI units) E x = 0, E y = 200 cos[12π × 1014 (t − z/c) + π/2], and
E z = 0. What are the frequency, wavelength, direction of motion, amplitude, initial
phase angle, and polarization of the wave?
1.2 Consider the plane electromagnetic wave in vacuum whose electric field is given
by the expression E = E o cos(ωt − kz). Assuming that the electric field amplitude
E o is 100 V/m, calculate the amplitude of the magnetic field Bo in units of Tesla and
the intensity I in units of W/m2 .
1.3 A plane electromagnetic wave propagating along the z axis has the √ following
electric field components: E x (t, z) = E o cos(ωt − kz), and E y (t, z) = 3E o cos
(ωt − kz). Determine the angle formed by its polarization direction with the x axis.
1.4 A Gaussian beam at wavelength λ = 0.63 µm has a radius at the beam waist
of wo = 0.5 mm. Calculate the beam radius w and the radius of curvature of the
wavefront at a distance of 5 m from the beam waist.
1.5 A plane electromagnetic wave has wavelength λ = 500 nm and a wave vector
k that lies in the x-z plane, forming an angle αx = 60◦ with the x axis. Determine
the spatial frequency of the phase pattern on the z = 0 plane.
1.6 A plane wave of wavelength λ = 500 nm illuminates a circular aperture having
diameter D = 100 µm. The diffraction pattern is observed on a screen placed at a
distance of 30 cm from the plane of the aperture. (i) check whether the condition for
Fraunhofer diffraction is satisfied; (ii) determine the position of the first zero of the
Airy function.
Chapter 2
Optical Components and Methods
Abstract In this chapter, after presenting Maxwell equations inside matter and
describing reflection and refraction processes, the interaction of optical waves with
simple optical components such as mirrors, prisms, and lenses is examined. The dif-
ferent methods for measuring the power spectrum of an optical beam are described
and compared in Sect. 2.3. In Sect. 2.4, after treating wave propagation in anisotropic
materials, the methods used for fixing or modifying the polarization state of an opti-
cal beam are presented. Finally, Sect. 2.5 introduces the important subject of optical
waveguides.
2.1 Electromagnetic Waves in Matter
Inside a medium Maxwell’s equations take the following form:
∇ ·D=ρ (2.1)
∇ ·B=0 (2.2)
∂D
∇ ×H=J+ (2.3)
∂t
∂B
∇ ×E=− , (2.4)
∂t
where D is the electric displacement vector (measured in units of C/m2 ). The quan-
tities ρ and J are, respectively, the electric charge density (C/m3 ) and the current
density (A/m2 ). Here the treatment is limited to dielectric media, in which there are
no free electric charges or currents, so that one can put J = ρ = 0.
The relations connecting D to E and B to H inside the medium become:
D = εo E + P (2.5)
B = μo (H + M). (2.6)

DOI 10.1007/978-3-319-20627-1_2
26 2 Optical Components and Methods
The new vectors P and M are, respectively, the unit-volume electric polar-
ization and the unit-volume magnetization inside the medium. In this book non-
ferromagnetic media are only considered, and so it is possible to simplify the treat-
ment by putting M = 0. The wave equation inside the medium is the following:
1 ∂ 2E ∂ 2P
∇2E = + μo . (2.7)
c2 ∂t 2 ∂t 2
In order to solve wave propagation problems, it is necessary to associate an equa-
tion relating P to E to (2.7). It is useful to recall what is the physical meaning of
P. The medium is a collection of positively charged nuclei and negatively-charged
bound electrons. Under the action of an electric field the electron clouds are slightly
displaced from their equilibrium position, so that a macroscopic electric dipole is
created. Limiting the discussion to the time-dependence, it is clear that the medium
response to the application of an electric field cannot be instantaneous. Therefore, a
convolution relation should be used:
t
P(r, t) = εo R(t − t )E(r, t )dt , (2.8)
−∞
where R(t − t ) is the response function of the medium.

The electric field E(r, t) associated with a generic light beam can always be
described as a weighted superposition of sinusoidal functions oscillating at different
frequencies, as expressed by the integral:
∞
1
E(r, t) = dωE(r, ω)exp(−iωt). (2.9)
2π −∞
The complex quantity E(r, ω) represents the weight of the component at angular
frequency ω. From a mathematical point of view, E(r, ω) is the Fourier transform
of E(r, t).
The inversion of (2.9) gives:
∞
E(r, ω) = dt E(r, t)exp(iωt). (2.10)
−∞
Recalling that the Fourier transform of the convolution of two functions is the
product of their Fourier transforms, the Fourier transform of P(r, t) is expressed by:
P(r, ω) = εo χ (ω)E(r, ω), (2.11)
where χ (ω), called electric susceptibility, is given by:

∞
χ (ω) = R(t)exp(iωt)dt. (2.12)
−∞
2.1 Electromagnetic Waves in Matter 27
By substituting (2.11) inside (2.5), it is found that the electric displacement

D(r, ω), generated by a monochromatic electric field oscillating at the frequency
ω, is proportional to E(r, ω):
D(r, ω) = εo (1 + χ )E(r, ω) = εo εr E(r, ω). (2.13)
The dimensionless constant εr is called relative dielectric constant or relative

electric permittivity. The wave propagation velocity inside the dielectric medium is:
1 c
u=√ = (2.14)
εo εr μo n
The quantity n, called index of refraction of the medium, is given by:

√
n= εr = 1 + χ. (2.15)
In the more general case, in which the relative magnetic permeability of the
√
material, μr , may be different from 1, it is found that n = εr μr .
Since the frequency of oscillation cannot change when the wave goes from one
medium to the other, the change of velocity is associated with a change of wavelength.
Specifically, if λ is the vacuum wavelength, then the wavelength inside a medium of
index of refraction n is: λ = λ/n. √
Since χ (ω) is frequency-dependent, also n(ω) = 1 + χ (ω) is frequency-
dependent. This phenomenon is called optical dispersion. As an example, over its
transparency window going from 0.35 to 2.3 µm, the index of refraction of the
borosilicate glass called BK7 decreases monotonically from 1.54 at 0.35 µm to
1.49 at 2.3 µm. This behavior is called normal dispersion. The dispersion is said to
be anomalous when the index of refraction increases as a function of wavelength.
As it will be seen in Chap. 6, the propagation of ultrashort light pulses is strongly
influenced by optical dispersion.
The electric field of the monochromatic plane wave inside the medium is still
given by (1.9). The only difference is that the modulus of the propagation vector is
now: ω ω ω
k= = 1 + χ = n, (2.16)
u c c
and (1.12) becomes:
εo n E o2
I =c . (2.17)
2
The propagation inside a medium with a real χ proceeds similarly to vacuum
propagation, the only difference being in the propagation velocity. However, if χ
is a complex quantity, then n is also complex, and can be written as: n = n +
in . Assuming that a monochromatic plane wave, traveling along z, enters into the
medium at z = 0, the electric field at the generic coordinate z is:
E = Eo exp{−i([ωt − (ω/c)(n + in )z]}. (2.18)
The intensity I (z), which is proportional to the square of the field modulus, is
given by:
I (z) = Io exp[−2(ω/c)n z] = Io exp(−αz), (2.19)
where Io is the intensity at z = 0. Equation (2.19) shows that, if n > 0, the wave
intensity decays exponentially during propagation with an attenuation coefficient,
α = 2(ω/c)n , proportional to the imaginary part of the refractive index. The atten-
uation may be due to absorption or to scattering.
2.2 Reflection and Refraction
In this section the laws of reflection and refraction are introduced and applied to
describe the behavior of different optical components, such as mirrors and lenses.
2.2.1 Dielectric Interface
A monochromatic plane wave is incident at the planar boundary between two dielec-
tric media, characterized by refractive indices n 1 and n 2 . According to the geometry
shown in Fig. 2.1, the wave is coming from medium 1, and its wave-vector ki forms
an angle θi with the normal to the boundary. The planar boundary is taken to coin-
cide with the x-y plane, and ki lies on the y-z plane. The plane containing ki and
the normal to the planar boundary is known as the incidence plane. In the case of
Fig. 2.1, y-z is the incidence plane.
Part of the wave will be reflected and part will be transmitted. Let θr and θt
be the reflection and transmission angles, respectively. By imposing the continuity
conditions at the boundary for the tangential components of the electric and magnetic
Fig. 2.1 Reflection and

refraction at a dielectric
boundary
2.2 Reflection and Refraction 29
fields, it is possible to calculate the direction and the complex amplitude of the
reflected and transmitted (also called refracted) waves.
It is useful to consider separately two different polarization states of the incident
wave: (a) a wave linearly polarized in a direction perpendicular to the incidence
plane, this is usually called σ case; (b) a wave linearly polarized in the incidence
plane, this is usually called π case. Any arbitrary polarization state of the incident
wave can always be described as a linear superposition of these two states.
• σ case.
Putting |ki | = ko n 1 , where ko = 2π/λ, the electric field of the incident wave is:
Ei = E i exp(−iωt + iko yn 1 sin θi + iko zn 1 cos θi )x, (2.20)
where x is the unit vector directed along x. The electric fields Er and Et of the
reflected and transmitted waves are written as:
Er = Er exp(−iωt + iko yn 1 sin θr − iko zn 1 cos θr )x, (2.21)
and
Et = E t exp(−iωt + iko yn 2 sin θt + iko zn 2 cos θt )x. (2.22)
The continuity condition for the tangential component of the electric field at z = 0
is:
E i exp(iko yn 1 sin θi )x + Er exp(iko yn 1 sin θr )x = E t exp(iko yn 2 sin θt )x. (2.23)
In order to satisfy (2.23) for any y, the reflection and transmission angles should
satisfy the relations:
θi = θr , (2.24)
and
n 1 sin θi = n 2 sin θt . (2.25)
Equation (2.25) is known as Snell’s law.

By using (2.24) and (2.25), (2.23) simply becomes
E i + Er = E t (2.26)
In order to derive Er and E t , a second equation is needed besides that of (2.26).

The continuity of the tangential component of the magnetic field is expressed as:
Hi cos θi − Hr cos θr = Ht cos θt (2.27)
Recalling that the electric

√ field amplitude is related to the magnetic field amplitude
by the expression H = εr εo /μo E, (2.27) can be written as:
n2
E i cos θi − Er cos θr = E t cos θt . (2.28)
n1
The solution of the system of (2.26) and (2.28) can be presented in terms of the
transmission and reflection coefficients, τσ and ρσ , respectively:
Et 2n 1 cos θi 2 sin θt cos θi

τσ = = = , (2.29)
Ei n 1 cos θi + n 2 cos θt sin (θi + θt )
Er n 1 cos θi − n 2 cos θt sin (θi − θt )

ρσ = = =− . (2.30)
Ei n 1 cos θi + n 2 cos θt sin (θi + θt )
If the incident beam is coming from medium 2, one can repeat the treatment
finding two new coefficients, τσ and ρσ , related to the previous ones as follows:
ρσ = −ρσ ; τσ τσ = 1 − ρσ2 (2.31)
• π case.
Following a treatment similar to that of the σ case, it is found that:
Et 2n 1 cos θi 2 sin θt cos θi

τπ = = = , (2.32)
Ei n 1 cos θt + n 2 cos θi sin (θi + θt ) cos (θi − θt )
Er n 1 cos θt − n 2 cos θi tan (θi − θt )

ρπ = = =− . (2.33)
Ei n 1 cos θt + n 2 cos θi tan (θi + θt )
If the beam arrives at the boundary from medium 2, it is found that the relations
between the coefficients τπ , ρπ and the coefficients τπ and ρπ are identical to those
of (2.31).
In the case of normal incidence, θi = θr = θt = 0, and ρσ coincides with ρπ :
n2 − n1
ρσ (θi = 0) = ρπ (θi = 0) = − . (2.34)
n2 + n1
If n 2 > n 1 , the field reflection coefficient is negative. This indicates that the phase
of Er is changed by π with respect to E i . If n 2 < n 1 , Er has the same phase as E i .
In order to derive how the incident power splits between transmission and reflec-
tion, it should be recalled that the power arriving on a generic surface Σ, is determined
as the time average of the flux of the Poynting vector through the surface. Projecting
the Poynting vector on the direction perpendicular to the surface, allows the following
expression for Pi to be given:
Pi = (1/2)cεo n 1 Σ|E i |2 cos θi . (2.35)

Similarly, for the transmitted and reflected power:
Pr = (1/2)cεo n 1 Σ|Er |2 cos θi Pt = (1/2)cεo n 2 Σ|E t |2 cos θt . (2.36)
Therefore the reflectance R and the transmittance T are given by:
Pr |Er |2
R= = = |ρ|2 (2.37)
Pi |E i |2
Pt n 2 cos θt
T = = |τ |2 . (2.38)
Pi n 1 cos θi
Note that R + T = 1, whereas the sum |ρ|2 + |τ |2 is, in general, = 1.

The behavior of R as a function of θi depends on whether n 1 is smaller or larger
than n 2 . Therefore the two cases are discussed separately.
• n1 < n2
Considering, as an example, an air/glass boundary with a ratio n 2 /n 1 = 1.5, the
behavior of Rπ e Rσ for a wave coming from air is illustrated in Fig. 2.2. At normal
incidence (θi = 0), using (2.34), it is found that Rπ (θi = 0) = Rσ (θi = 0) = 0.04.
The reflectance Rσ is monotonically increasing with θi reaching a value of 1 at
θi = π/2 = 90◦ . On the other hand, Rπ at first decreases till it vanishes at θi = θ B ,
and then increases, becoming equal to 1 at θi = π/2.
The angle θ B is derived by noting that the reflection coefficient ρπ vanishes if
the denominator of the fraction at right-hand side of (2.33) becomes infinite. This
happens if: θi + θt = π/2. From this condition, by using Snell’s law, the following
expression of θ B , called Brewster’s angle, is derived:
Fig. 2.2 Reflectance versus incidence angle for n 2 /n 1 = 1.5

n2
tan θ B = (2.39)
n1
As an example, if n 2 /n 1 = 1.5, θ B = 56◦ , and Rσ (θ B ) = 0.147.

√
In general, ρ is a complex quantity that can be written as ρ = R exp(iφ), where
φ is the phase shift of the reflected field with respect to the incident field. In the σ
case, (2.30) shows that φσ = π over the whole interval 0 ≤ θi ≤ π/2. In the π case,
(2.33) indicates that φπ is equal to π within the interval 0 ≤ θi < θ B , and vanishes
for θ B ≤ θi ≤ π/2.
• n1 > n2
In this case, Snell’s law indicates that θt is larger than θi . Therefore, as θi grows,
there will be an incidence angle, θc , smaller than π/2, at which θt becomes π/2. If
θi ≥ θc there is no transmitted beam, and so the power reflection coefficient is equal
to 1. The total reflection angle or limit angle, θc , is defined by the relation:
sin θc = n 2 /n 1 (2.40)
Considering the glass/air interface, n 1 /n 2 = 1.5, for which θc = 41.8◦ , the

behavior of Rπ and Rσ is illustrated in Fig. 2.3. As in the previous case, Rπ vanishes
at the Brewster angle, θ B = 33.7◦ .
The field reflection coefficient is a real quantity for θi ≤ θc . If θi > θc , cos θt
becomes imaginary:

n 2 n 21
cos θt = 1 − sin2 θt = 1 − 2 sin2 θi = i
1
sin2 θi − 1. (2.41)
n2 n 22
Fig. 2.3 Reflectance versus incidence angle for n 1 /n 2 = 1.5

By substituting (2.41) into the expressions for ρ, it is seen that the field reflection
coefficient becomes complex. In the σ case, the phase φσ of ρσ is 0 for θi ≤ θc , and
grows from 0 to π in the interval θc ≤ θi ≤ π/2 according to the expression:

φσ sin2 θi − sin2 θc
tan = . (2.42)
2 cos θi
In the π case, φπ vanishes in the interval 0 ≤ θi ≤ θ B , and is equal to π in

the interval θ B < θi ≤ θc , decreasing from π to 0 in the interval θc ≤ θi ≤ π/2
according to the expression:

φπ cosθi sin2 θc
tan = . (2.43)
2 sin2 θi − sin2 θc
2.2.2 Reflection from a Metallic Surface
The most commonly used mirrors are made by depositing a thin metallic layer on a
glass substrate. It is therefore interesting to discuss the case in which medium 2 in
Fig. 2.1 is a metal. Metals contain electrons that are free to move inside the crystal.
In a simplified approach, the relative dielectric constant of the metal is written as
the sum of two terms, one due to bound electrons, ε , and the other related to free
electrons: σ
εm = ε + i , (2.44)
εo ω
where σ is the frequency-dependent electric conductivity of the metal:
1 + iωτ
σ (ω) = σ (0) . (2.45)
1 + ω2 τ 2
The quantity τ is a relaxation time that depends on the nature of the metal. The
zero-frequency conductivity is given by:
N e2 τ
σ (0) = , (2.46)
me
where N is the density of free electrons, m e is the electron mass, and e is the electron
charge.
The free electron gas inside the metal has a characteristic frequency of oscillation,
called the plasma frequency, related to σ (0) by the expression:

σ (0) N e2
ωp = = (2.47)
ε εo τ ε εo m e
Electromagnetic radiation having frequency close to the plasma frequency is

strongly absorbed by the metal, and so the refractive index must be written as a
√
complex quantity, n m = εm = n + in . Limiting the discussion to normal inci-
dence (θi = 0), the reflectance of the air-metal interface, derived by using (2.34), is
given by:
|n m − 1|2 n 2 + n 2 + 1 − 2n
R= = . (2.48)
|n m + 1|2 n 2 + n 2 + 1 + 2n
Approximate expressions for n and n can be derived from (2.44) by assuming

that the contribution of bound electrons is negligible in comparison to that of free
electrons. Taking into account that, for a typical metal, τ ≈ 10−13 s and ω p ≈
−
1015 s 1 , three different frequency intervals can be distinguished:
• infrared: ω 1/τ ω p . The result is: R = 1.
• near infrared and visible: 1/τ ω ω p . It is found that: R = 1 − 2/(ω p τ ), the
reflectance is large, but lower than 100 %.
• ultraviolet: ω p ω. It is found: R = ω4p /(16ω6 τ 2 ) 1. In this case the
reflectance is very low.
Experimental values of n , n and R for three metals at three different wavelengths
are reported in Table 2.1. The data refer to thin layers deposited by evaporation on a
glass substrate.
Table 2.1 Complex index of refraction and normal-incidence reflectance for some metal surfaces
at three distinct wavelengths
Metal λ (nm) n n R
Silver 400 0.08 1.9 0.94
550 0.06 3.3 0.98
700 0.08 4.6 0.99
Aluminum 400 0.4 4.5 0.93
550 0.8 6 0.92
700 1.5 7 0.89
Gold 450 1.4 1.9 0.40
550 0.33 2.3 0.82
700 0.13 3.8 0.97
One problem with metallic mirrors is that the fraction of incident power that is not
reflected is absorbed. When reflecting high-intensity beams, such as those coming
from laser sources, the temperature rise due to this absorption may cause damage to
the metallic layer.
2.2.3 Anti-reflection Coating
The reflectance of a boundary separating two transparent media can be strongly

modified by inserting thin dielectric layers. The structure investigated in this section,
shown in Fig. 2.4, consists of a single layer of refractive index n 2 and thickness d
separating media 1 and 3 which have refractive indices n 1 and n 3 , respectively. It is
assumed that n 1 < n 2 < n 3 .
Considering a monochromatic plane wave perpendicularly incident on the bound-
ary 1–2, and calling E i the incoming electric field, the first contribution to the reflected
field is: Er 1 = ρ12 E i . The transmitted field E t1 = τ12 E i is partially back-reflected
by the interface 2–3, producing a second contribution to the reflected field given
by: Er 2 = exp(iΔ)ρ23 τ21 E t1 . This contribution is phase-shifted with respect to the
first one because the wave has traveled back and forth inside layer 2. The shift is:
Δ = 4π n 2 d/λ, where λ is the vacuum wavelength of the incident wave. Note that, in
this case, the coefficients ρ12 and ρ23 have the same sign, so that no contribution to
the phase shift between Er 1 and Er 2 is coming from the reflections at the interfaces.
Since the wave back-reflected from the surface 2–3 is partially reflected again
when it reaches the surface 2–1, the total reflected field should be calculated by
adding an infinite number of multiple reflections. Assuming that the reflectance of
the interfaces 1–2 and 2–3 is small, which also means that τ21 τ12 ≈ 1, the summation
can be truncated by taking only the first two terms:
Er = Er 1 + Er 2 = [ρ12 + exp(iΔ)ρ23 ]E i (2.49)
It is clear from (2.49) that Er vanishes through destructive interference if Er 2

and Er 1 have the same amplitude, but opposite sign. The amplitudes of the two
Fig. 2.4 Anti-reflection

coating
fields are the same if ρ12 = ρ23 . According to (2.34), this is equivalent to saying
√
that (n 2 /n 1 ) = (n 3 /n 2 ), or: n 2 = n 3 n 1 . This means that the intermediate layer
should have an index of refraction that is the geometric mean of n 1 and n 3 . The two
contributions have opposite sign if Δ = π , which implies d = λ/(4n 2 ). This means
that the layer thickness must be a quarter of the wavelength inside the medium.
The fact that the condition imposed on the layer thickness d is wavelength-
dependent indicates that there is a price to pay for eliminating reflection: the method
works exactly at only one specific wavelength. In other words this anti-reflection
coating is chromatic.
Instead of using the approximate expression given by (2.49), Er can be exactly
calculated by summing up an infinite number of multiple reflections or by writing the
boundary conditions at the two interfaces. The latter approach is the more convenient,
also because it can be more easily generalized to an arbitrary number of layers.
Consider first the interface 1–2; at left there are two waves, the incident wave
with field E i and the reflected wave with field Er , at right there is a wave leaving the
interface, with field E t and one arriving at the interface, with field Er . The boundary
conditions are:
E i + Er = E t + Er
E i − Er = (n 2 /n 1 )(E t + Er ). (2.50)
Taking into account that the wave crossing layer 2 acquires a phase shift of Δ/2,
the following boundary conditions are written for the interface 2–3:
exp(iΔ/2)E t + exp(−iΔ/2)Er = E t
exp(iΔ/2)E t + exp(−iΔ/2)Er = (n 3 /n 2 )E t . (2.51)
Equations (2.50) and (2.51) constitute a linear system of four equations in four
variables that can be exactly solved. The expression obtained for Er is:
τ12 τ21 exp(iΔ)

Er = ρ12 E i + ρ23 E i (2.52)
1 − ρ23 ρ21 exp(iΔ)
Recalling that τ12 τ21 = 1 − ρ122 , it is immediately apparent that the conditions for
a vanishing Er derived from (2.52) coincide with those imposed by the approximate
expression (2.49).
Table 2.2 gives the index of refraction and the transparency window of several
dielectric materials. The index of refraction of the thin film depends not only on
the wavelength, but also on the preparation method, and so the values of n listed
in Table 2.2 should be only considered as indicative. The deposition of layers with
precisely controlled thickness is usually performed by vacuum evaporation methods.
In order to have an idea of the required thickness, consider the anti-reflection coating
of an air-glass surface. Taking λ = 550 nm, n 1 = 1 (air), and n 3 = 1.52 (glass), the
Table 2.2 Properties of materials used for deposition of thin films

Material n Transparency window (µm)
Criolite N a3 Al F6 1.35 0.15–14
Magnesium fluoride Mg F2 1.38 0.12–8
Silicon oxide Si O2 1.46 0.17–8
Aluminum oxide Al2 O3 1.62 0.15–6
Silicon monoxide Si O 1.9 0.5–8
Zirconium oxide Zr O2 2.00 0.3–7
Cerium oxide CeO2 2.2 0.4–16
Titanium oxide T i O2 2.3 0.4–12
Zinc sulfide Z nS 2.32 0.4–14
Cadmium telluride CdT e 2.69 0.9–16
Silicon Si 3.5 1.1–10
Germanium Ge 4.05 1.5–16
Lead telluride PbT e 5.1 3.9–20
√
coating should be made with a material having n 2 = 1.52 = 1.23 and thickness
d = 550/(4n 2 ) = 112 nm. In practice, given the list of available materials in
Table 2.2, the best choice would be magnesium fluoride. A quarter-wave layer of
Mg F2 reduces the reflectance from 4 to 1.3 %.
By using more than one layer it is possible to cancel completely the reflectance
of the air-glass surface. For instance, the sequence of two quarter-wave layers √ with
refractive indices n 2 and n 2 , produces a vanishing reflectance if n 2 /n 2 = n 3 /n 1 .
Putting n 1 = 1 e n 3 = 1.52, one finds n 2 /n 2 = 1.23. Zero-reflectance is obtained
by choosing from Table 2.2 Zr O2 (n 2 = 2.00) and Al2 O3 (n 2 = 1.62).
Up to now normal incidence has been assumed. What happens if θi is different
from 0, as shown in Fig. 2.5? The phase difference between the first and the second
reflection becomes:
Fig. 2.5 Anti-reflection

coating: non-normal
incidence. The optical path
difference between kr 2 and
kr 1 is: 2n 2 AB − n 1 AH
2π n 2 d 2π n 1 d 4π n 2 d
Δ = 2 − tgθt sinθi = cosθt , (2.53)
λ cosθt λ λ
where θt , the propagation angle inside medium 2, is related to θi by Snell’s law.

It is important to note that, once d and n 2 are fixed, the wavelength at which the
reflectance vanishes, corresponding to Δ = π , depends on θi .
2.2.4 Multilayer Dielectric Mirror
By using a sequence of dielectric layers it is possible to obtain high reflectance at

any desired wavelength. An important advantage of multilayer dielectric mirrors,
as compared to metallic mirrors, is that there is no absorption. This means they can
stand high intensities of visible light without being damaged. An important difference
is that metallic mirrors have a broad-band reflectance, whereas the reflectance of
dielectric mirrors is narrow-banded.
The structure of a multilayer dielectric mirror is sketched in Fig. 2.6: between
medium 1 (air) and medium 2 (glass substrate) there is a sequence of alternating
layers ABAB..A with refractive indices n A > n B , and n A > n 2 .
Clearly, the maximum reflectance will be obtained when the fields reflected by all
the interfaces add up in phase at the first interface, so that constructive interference is
exploited. The field reflected at the interface 1- A is phase-shifted by π with respect
to the incident field because n A > n 1 . The field reflected at the first A–B interface
does not suffer any phase-shift because n A > n B , but, upon arriving at the interface
1-A, it will acquire the phase shift 4π n A d A /λ because of the propagation back and
forth within the layer thickness d A . This second reflection will be in phase with the
first one if the layer thickness corresponds to a quarter wavelength: d A = λ/(4n A ).
By analyzing all subsequent reflections in a similar way, it is immediately clear that
the maximum reflectance, Rmax , is obtained if all the layers A and B have a quarter
wavelength thickness. The number 2N + 1 of layers is always an odd number, in
order to ensure that the last reflection (occurring at the A-2 interface) is also in phase
with the previous ones.
Fig. 2.6 Multilayer

dielectric mirror
A detailed calculation can be developed by writing the boundary conditions at

each interface, and noting that each pair of boundary conditions corresponding to a
specific interface, such as (2.50) and (2.51), can be expressed in terms of a matrix
equation. Therefore each interface introduces a transmission matrix, and the solution
is always expressible as a cascade of matrices, independently from the number of
layers.
The maximum reflectance at normal incidence is found to be:
2
1 − (n A /n 1 )(n A /n 2 )(n A /n B )2N
Rmax = . (2.54)
1 + (n A /n 1 )(n A /n 2 )(n A /n B )2N
Note that an expression of the type

1 − a
,
1+a
can be transformed, by introducing the new variable b = ln a, into

1 − a
e−b/2 − eb/2
b
= = tanh .
1+a e−b/2 + eb/2 2
Therefore Rmax can be expressed as:
Rmax = tanh2 (N ξ ), (2.55)
where:
n
A 1
nA
n
A
ξ = ln + ln + ln (2.56)
nB 2N n1 n2
By increasing N , Rmax is increased and, at the same time, the high-reflectance

band becomes narrower and narrower. In principle, as N tends towards infinity,
Rmax tends to 1. The greater is the ratio of n A /n B , the quicker Rmax converges to its
maximum value. For example, alternating zinc sulfide (n A = 2.32) and magnesium
fluoride (n B = 1.38) layers, a reflectance of R = 0.99 can be reached using only 13
layers.
Similarly to the case of the anti-reflection coating, analogous calculation of the
reflectance of the multilayer mirror can be performed for any incidence angle, pro-
vided the propagation path inside each layer is appropriately expressed. The dielectric
multilayer mirror offers a great flexibility because it can be designed for any value of
reflectance at any wavelength and any incidence angle. In particular, beam splitters
having any arbitrary power partition between transmitted and reflected beam can be
made.
The idea that the optical properties of a periodic structure can be considerably var-
ied not by changing materials, but by simply acting on the geometry has gained more
and more importance in the last decade. Two-dimensional and three-dimensional
periodic structures made of dielectric or semiconductor materials, called photonic

crystals, are increasingly utilized in photonic devices. It is interesting to note that this
kind of phenomenon occurs also in Nature. For example, the brilliant colors of some
butterflies are due not to absorption, but to the presence of periodic dielectric layers
on their wings. A phenomenon typical of periodic structures is iridescence, due to the
fact that the wavelength of maximum reflectance depends on the incidence angle.
This pleasing visual effect, also called opalescence, can be observed with opals,
semi-precious stones consisting of a three-dimensional periodical arrangement of
sub-micrometric silica spheres.
2.2.5 Beam Splitter
As mentioned in the previous section, beam splitters are semi-reflecting mirrors that
divide an optical beam into two parts propagating in different directions. In order to
understand the behavior of some optical systems utilizing beam splitters, it is useful
to know the phase relation between transmitted and reflected fields.
Let τ = τo exp(iφτ ) and ρ = ρo exp(iφρ ) be, respectively, the transmission and
reflection coefficients for the electric field. The actual value taken by these coefficients
depends on many parameters, such as the beam-splitter structure, the incidence angle,
and the wavelength of the incident beam. On the other hand, as will be shown, the
phase difference φτ − φρ is expressed by a general law.
The scheme of Fig. 2.7 shows a beam-splitter symmetrically illuminated on both
sides by two identical light beams having the same phase, polarization, incidence
angle, and intensity Io . Calling I1 and I2 the intensities of the two output beams and
assuming no losses, it should be expected, for symmetry reason, that I1 = I2 = Io .
Since the field E 1 at output 1 is the sum of two contributions, one coming from
reflection and the other from transmission, E 1 = ρ E o + τ E o , the intensity at output
1 is given by: I1 = |ρ + τ |2 Io . The condition I1 = Io is satisfied if:
|ρ + τ |2 = [τ τ ∗ + ρρ ∗ + 2Re(τρ ∗ )] = 1. (2.57)
Fig. 2.7 Two identical light

beams impinging on the two
sides of a beam splitter
Since τ τ ∗ + ρρ ∗ = 1, the quantity Re(τρ ∗ ) = τo ρo cos(φτ − φρ ) must be nil,

therefore: π
φ τ − φρ = . (2.58)
2
Equation (2.58) is a general relation, valid for any lossless beam-splitter, regardless
of the structure of the mirror or the incidence angle. The only limitation is that the
beam-splitter should be symmetrical. It is easy to generalize to the non-symmetrical
case: calling φτ and φρ the phase shifts for a beam incident to the left of the mirror,
and φτ and φρ the phase shifts for a beam incident to the right of the mirror, instead
of (2.58), the following relation is found:
φτ + φτ − φρ − φρ = π. (2.59)
2.2.6 Total-Internal-Reflection Prism
The total-internal-reflection glass prism, shown in Fig. 2.8, behaves as a 100 %-

reflectance mirror if the incident beam undergoes total reflections at the two glass-air
interfaces. This happens if the incidence angles at the glass-air boundaries are larger
than the limit angle. In the case of the figure, θi = 45◦ , which is larger than θc for
a normal optical glass. The spurious reflection of about 4 % at the air-glass input
interface can be eliminated by an anti-reflection coating.
Whereas multilayer dielectric mirrors present high reflectance in a narrow wave-
length band, the total-internal-reflection mirror is a broadband component. The only
limitations could come from the decrease of the refractive index of the glass as the
wavelength is increased, which could move θc to a value larger than 45◦ .
An interesting property of the total-internal-reflection prism is that the reflected
beam is always propagating in the opposite direction with respect to the incident
beam, whatever is the angle of incidence, provided that the propagation vector of the
beam lies in a plane perpendicular to the prism axis (this is the plane of Fig. 2.8).
Such a property is not shared by normal mirrors. Instead of a prism, it is possible
to use a trihedral (see Fig. 2.9), known as the corner cube. With this, the beam is
retro-reflected after three internal reflections, and it is possible to obtain a back-
propagating reflected beam for any direction of incidence. In practical applications,
Fig. 2.8 Total internal

reflection prism
Fig. 2.9 Corner cube
Fig. 2.10 Cat’s eye
the back-reflectors placed on vehicles or on wayside posts have a structure imitating

an array of corner cubes.
This same property of the corner cube is also shared by the combination of a lens
plus a mirror placed in the focal plane of the lens, known as a cat’s eye (see Fig. 2.10).
2.2.7 Evanescent Wave
A linearly polarized beam impinges on the glass-air planar boundary with an inci-
dence angle θi ≥ θc . Assume that the boundary coincides with the x-y plane, that the
propagation vector k1 of the incident beam lies in the y-z plane, and that the beam
is polarized along x as shown in Fig. 2.11a. Using the subscript 1 for glass and 2 for
air, the electric fields of the incident and transmitted waves can be written as:
Ei = E i exp[−i(ωt − k1y y − k1z z)]x

Et = E t exp[−i(ωt − k2y y − k2z z)]x, (2.60)
where k1y = −ko n 1 sinθi and k1z = ko n 1 cosθi .

The continuity condition at the boundary z = 0 imposes k1y = k2y . Using the
relation:
k2 = ko n 2 = k2y 2 + k2z 2 , (2.61)
Fig. 2.11 a Geometry for

total internal reflection.
b Transmission of the
evanescent wave by a pair of
prisms

it is found: k2z = ko n 2 2 − n 1 2 sin2 θi . If θi > θc , the term under square root is
negative, so that k2z becomes imaginary. Putting n 2 2 − n 1 2 sin2 θi = iΓ n 2 , where
Γ is a real quantity, it is found:
Et = E t exp[−i(ωt + ko n 1 sinθi y)]exp(−Γ ko n 2 z)x. (2.62)
The condition of total internal reflection at the boundary between glass and air
does not imply that there is no electromagnetic field in the air beyond the glass.
Indeed Eq. (2.62) describes a wave, known as evanescent wave, running parallel to
the glass-air interface, with an amplitude that is an exponentially decreasing function
of the penetration distance z. The electromagnetic energy density inside medium 2,
being proportional to |E t |2 , is also exponentially decreasing with z, becoming smaller
by a factor e−1 at the distance:
1 λ
δ= = . (2.63)
2Γ ko n 2 4π n 1 2 sin2 θi − n 2 2
For instance, if n 1 = 1.5, n 2 = 1, θi = π/4, it follows that Γ = 0.354, therefore

δ = λ/4.44. Taking the central wavelength of the visible range, λ = 0.55 µm, it is
found that δ = 120 nm.
In order to derive the amplitude of the transmitted and reflected fields, the follow-
ing boundary conditions are written:
E i + Er = E t (2.64)
E i − Er = iγ E t , (2.65)
where γ = Γ n 2 /(n 1 cosθi ) is a real quantity. The solutions of (2.64) and (2.65) are:
2 1 − iγ
Et = Ei ; Er = Ei . (2.66)
1 + iγ 1 + iγ
Note that |E i | = |Er |, thus confirming total reflectance, even if |E t | is different

from 0.
By placing a second prism in the vicinity of the totally reflecting one, it is possible
to exploit a tunnel effect, by “catching” some part of the evanescent wave. Such a
technique, called frustrated total internal reflection, allows for the realization of a
partially transmitting mirror, with a transmittance controlled by the thickness of
the air-gap between the two prisms. Consider the configuration in Fig. 2.11b, where
there is an air-gap of thickness d between the two prisms. In order to write down the
boundary conditions at the two interfaces, it is necessary to assume that inside the
gap there are two evanescent waves decaying in opposite directions. The continuity
conditions at the first interface are:
E i + Er = E t + Er , (2.67)
and
E i − Er = iγ (E t − Er ), (2.68)
where E t and Er are the fields of the two evanescent waves, the first decaying as
a function of z and the second growing as a function of z. Similarly, the continuity
conditions at the second interface are:
exp[−d/(2δ)]E t + exp[d/(2δ)]Er = E t , (2.69)
and
exp[−d/(2δ)]E t − exp[d/(2δ)]Er = −(i/γ )E t . (2.70)
By eliminating E t and Er , E t and Er can be calculated. In particular, the trans-

mittance T is found to be:
|E t |2 exp(−d/δ)
T = = 2 2 , (2.71)
|E i | 2
1 + exp(−d/δ) /2 + A 1 − exp(−d/δ)
where A = (1 − γ 2 )2 /(16γ 2 ). It is seen from (2.71) that, for d tending to 0, T tends

to 1, as expected. If d δ, T is approximated by:
4exp(−d/δ)
T ≈ , (2.72)
1 + 4A
which shows an exponential decay of T as a function of d. The set of two prisms is

therefore equivalent to a variable-transmittance attenuator.
2.2.8 Thin Lens and Spherical Mirror
So far in this chapter only situations involving plane waves and plane boundaries have
been considered. In this section spherical boundaries are discussed, concentrating
the attention on the important optical component of the lens.
The lens, as shown in Fig. 2.12, is formed by joining two spherical caps made
of glass. The radii of curvature of the two surfaces, ρ1 and ρ2 , are conventionally
taken to be positive if the surfaces are convex. The thin-lens approximation consists
in assuming that the lens thickness is much smaller than the radii of curvature of the
two spherical surfaces. Neglecting reflections at the two surfaces, the effect of the
lens on the incident field E 1 is described by a transmission function τ f (x, y):
E 2 (x, y, z, t) = τ f (x, y)E 1 (x, y, z, t). (2.73)
This transmission function has a unitary modulus and a phase that is locally
dependent on the glass thickness:
τ f (x, y) = exp{iko [nΔ(x, y) + Δo − Δ(x, y)]}, (2.74)
where n is the glass refractive index, Δ(x, y) is the lens thickness at the transversal
position (x, y), and Δo = Δ(0, 0) is the thickness on the optical axis.
The phase shift is the sum of two contributions, one due to the path inside the
glass, ko nΔ(x, y), and the other due to the path in air, ko [Δo − Δ(x, y)]. The lens
thickness at the generic point (x, y) is:

Δ(x, y) = Δo + ρ1 2 − (x 2 + y 2 ) − ρ1 + ρ2 2 − (x 2 + y 2 ) − ρ2 (2.75)
Fig. 2.12 Thin lens made of two glass caps with radii of curvature ρ1 and ρ2
Equation (2.75) can be simplified by assuming that the lens diameter is much
smaller than the radii of curvature of the two surfaces, so that:

x 2 + y2 x 2 + y2
ρ1 2 − (x 2 + y2) = ρ1 1− ≈ ρ1 1 − , (2.76)
ρ1 2 2ρ1 2
and
x 2 + y2
ρ2 2 − (x 2 + y 2 ) ≈ ρ2 1 − . (2.77)
2ρ2 2
In the power expansion of the square-root terms, only the first-order terms in
(x 2 + y 2 ) are kept. This is equivalent to approximate the spherical surface with
a paraboloidal surface, similarly to the paraxial approach used in Chap. 1 for the
description of spherical waves.
Inserting (2.76) and (2.77) into (2.75):

x 2 + y2 1 1
Δ(x, y) = Δo − + . (2.78)
2 ρ1 ρ2
By substituting inside (2.74):

x 2 + y2
τ f (x, y) = exp(iko nΔo )exp −iko , (2.79)
2f
where f , called focal length, is defined by the relation:

1 1 1
= (n − 1) + (2.80)
f ρ1 ρ2
Note that f has a sign: f > 0 means converging lens, f < 0 diverging lens. Since
f depends on n, and, in turn, n changes with wavelength, the lens focuses different
wavelengths to different distances, that is, it suffers from a chromatic aberration.
The spherical mirror is described similarly to the lens, the main difference being
that it works in reflection instead of transmission. The mirror shown in Fig. 2.13
has center in O and radius of curvature ρ. Consider a beam parallel to the optical
axis, incident at point P with an incidence angle α. The reflected beam intersects the
optical axis at point F. The distance of F from O is: O F = O P/(2cosα). Since O F
depends on α, parallel beams impinging on the mirror in different points produce
reflected beams crossing the optical axis at different positions. However, if only
beams traveling close to the optical axis are considered (i.e. where cosα ≈ 1), it
is found that O F ≈ O P/2 = ρ/2. Under this approximation, which is equivalent
to treating the surface as paraboloidal instead of spherical, it can be said that the
spherical mirror brings all parallel beams to converge at a single point a distance ρ/2
from the mirror surface. In other words, the mirror behaves in reflection just as a lens
Fig. 2.13 Spherical mirror

with radius of curvature ρ
having a focal length f = ρ/2. Note that the focal length of the spherical mirror
is independent of the wavelength. All large astronomical telescopes use parabolic
mirrors instead of lenses.
2.2.9 Focusing Spherical Waves
A spherical wave traveling along the z axis impinges on a thin lens located at the
plane z = 0. If R1 is the radius of curvature of the wavefront, then, under the paraxial
approximation, the electric field E 1 of the wave at z = 0 is given by:

Ao x 2 + y2
E 1 (x, y, 0) = exp iko . (2.81)
R1 2R1
The field distribution transmitted through the lens, E 2 = τ f E 1 , calculated by

using (2.79) and (2.81), is again that of a spherical wave, with the modified radius
of curvature R2 , which is related to R1 by:
1 1 1
= − . (2.82)
R2 R1 f
If R1 = f , R2 becomes infinite, which means that a spherical wave originating

from the left focus is converted into a parallel beam (plane wave). If R1 > f , R2 < 0,
indicating that the diverging wave is converted into a converging wave. If R1 f ,
R2 ≈ f , that is, a plane wave is converted into a spherical wave converging at the
right focal plane. In practice, the wave will not converge to a point, but the beam
spot at focus will be described by an Airy function (see Sect. 1.6.1) because of the
diffraction effect introduced by the finite dimension of the lens-diameter.
The ratio between the lens radius and the focal length is called numerical aperture
(NA) of the lens. It describes the light-gathering capacity of the lens.
Relation (2.82) can also be interpreted in the following way: given a point-like
object P at distance R1 on the left, the lens creates a real image Q of P at distance
R2 on the right. This is the standard interpretation given by geometrical optics. If
P is on the optical axis, Q is also on the same axis. What happens if the point-like
object P is off-axis? Assigning to P the coordinates (xo , 0, R1 ), the electric field of
the spherical wave originated at P , evaluated at the lens plane z = 0, is now:
ik
Ao o
E 1 (x, y, 0) = exp [(x − xo ) + y ] .
2 2
(2.83)
R1 2R1
Using again the relation E 2 = τ f E 1 , after some algebra the following expression
is found for E 2 :

Ao exp(iφ) iko f
E 2 (x, y, 0) = exp (x + x o )2 + y 2 , (2.84)
R1 2R2 R1 − f
where φ depends on xo and on the lens parameters, but not on x and y. Equation (2.84)
shows that the image of P is at point Q having coordinates [− f xo /(R1 − f ), 0, R2 ].
It can be said that the lens creates an image QQ of the segment PP . If R1 > f , the real
image is upside-down, with a magnification f /(R1 − f ). This approach is therefore
equivalent to the geometrical-optics treatment. The topic of image formation has
important applications but will not be further discussed in this text.
Another case to be considered is that of a plane wave having a propagation direc-
tion slightly tilted with respect to the optical axis. To simplify the discussion it is
assumed that the wave vector k is in the plane x-z, forming an angle α with x. At
z = 0 the incident field is expressed as E 1 = Ao exp(iko αx), where sin α has been
approximated as α. After using E 2 = τ f E 1 again, and applying some algebra, it is
found that:

iko
E 2 (x, y, 0) = Ao exp(iφ)exp − (x − α f ) + y
2 2
(2.85)
2f
where φ = kα 2 f /2. Equation (2.85) describes a spherical wave that converges at

the point (α f, 0, f ) located on the right focal plane, as shown in Fig. 2.14.
Generalizing the treatment, it can be stated that a biunique correspondence is
established between the propagation direction of the incident wave and points on the
right focal plane.
2.2.10 Focusing Gaussian Spherical Waves
If the wave impinging on the lens is a Gaussian spherical wave with a complex radius
of curvature of the wavefront q1 , then it can be immediately shown, by using (2.73)
and (2.79), that the transmitted wave is still a Gaussian spherical wave. The new
Fig. 2.14 A plane wave

corresponds to a point at the
focal plane of a lens
complex radius of curvature q2 satisfies the relation:
1 1 1
= − . (2.86)
q2 q1 f
Since the beam radius is unchanged by crossing the lens, w1 = w2 , one finds that
(2.86) coincides with (2.82). As an example, if f > 0 and E 1 describes a diverging
Gaussian wave with R1 > f , the transmitted Gaussian wave has a negative radius
of curvature of the wavefront, R2 < 0, and is, therefore, converging. There is a basic
difference between this case and the one treated in the previous section: the theory
of Gaussian waves accounts intrinsically for diffraction effects, such that the wave
converges to a finite-size beam spot.
Once q2 at z = 0 is assigned, it is possible to obtain the distance z f at which the
focused beam takes the minimum radius, and also the minimum radius w f , by using
relations 1.35 and 1.36. The results of the calculation are:
|R2 | w1
zf = , wf = . (2.87)
1 + [λ|R2 |/(π w1 )]
2 2
1 + [π w12 /(λ|R2 |)]2
Assuming that R1 f , which means |R2 | ≈ f , and π w12 λ f , the relations of

(2.87) simplify as follows:
λf
z f ≈ f, w f ≈ (2.88)
π w1
It is useful to calculate the Rayleigh length at focus:

2
π w2f λ f
zfR = ≈ . (2.89)
λ π w1
The length 2z f R defines the field depth, that is, the interval of distance over which
the beam remains focused. As an example, putting f = 20 cm, w1 = 2 mm, λ =
0.5 µm (note that π w12 /(λ f ) is 1), one finds that w f = 16 µm and z f R = 1.6 mm.
From a mathematical point of view, the Gaussian spherical wave has an infinite
extension on the transversal plane. In reality, practical optical components have a
limited size, and so the tails of the Gaussian distribution are lost. The empirical
criterion for applying the theory of Gaussian waves is if the lost power is sufficiently
small. Considering a lens (or a mirror) with diameter d, the power fraction intercepted
by the lens is given by:
d/2 2
0 2πr exp(− 2r
w2
)dr d2
Tf = ∞ = 1 − exp − 2 (2.90)
2r 2
0 2πr exp(− w2 )dr
2w
As an example, if d = 2w, the intercepted power fraction is 86 %, if d = 3w,

T f = 0.99. Therefore, if d > 3w, it is reasonable to assume the applicability of the
Gaussian wave theory.
2.2.11 Matrix Optics
Within the geometrical optics approximation, a useful approach is that of associating

a 2 × 2 matrix, called ABC D matrix, with each optical component.
The ray present at the coordinate z is characterized by two parameters, the distance
r (z) from the optical axis, and the angle its direction forms with the optical axis, φ(z).
Within the paraxial approximation, φ(z) is simply the derivative of r with respect to
z. The ray is represented by a one-column matrix:

r
R= . (2.91)
φ
Every optical component is represented by a 2 × 2 matrix of the type:

A B
M= , (2.92)
C D
connecting the output vector Ru to the input vector Ri according to:
Ru = MRi . (2.93)
Using refraction laws and the properties of lenses and mirrors discussed in the
preceding sections, the ABC D matrices of the components shown in Fig. 2.15 can
be written as follows.
Fig. 2.15 Different optical

components that are
described by ABCD
matrices: a generic ray,
b free propagation, c planar
boundary between two
different media, d lens,
e spherical boundary
between two media,
f spherical mirror
• Free propagation
1d
M= (2.94)
01
• Lens
1 0
M= −1 (2.95)
f 1
• Planar boundary between two media

1 0
M= (2.96)
0 nn 21
• Spherical boundary between two media

1 0
M= n 1 −n 2 1 n 1 (2.97)
n2 ρ n2
• Spherical mirror
1 0
M= −2 (2.98)
ρ 1
Note that the determinant of a generic ABC D matrix is given by:
AD − BC = n 1 /n 2 . (2.99)
If the two media are the same, the determinant is equal to 1, as expected if the
optical element does not introduce losses.
An optical system made by a cascade of various elements is represented by a matrix
that is the ordered product of all the individual matrices describing each element.
The order is important because the matrix product is generally non-commutative.
2.3 Fourier Optics
As discussed in Sect. 1.6, the Fraunhofer approximation of the diffraction integral is

only valid in the far field, at distances from the screen that are much larger than the
typical size of optical instruments. Therefore, it is of great practical importance to
find a method of transporting the far-field behavior to short distances. In this section
it will be shown that the insertion of a lens can solve the problem.
Consider a thin lens having a focal distance f and positioned at z = 0, that is
illuminated from the left by a wave propagating along the z axis. By assuming that
the field distribution in the left focal plane of the lens is known, E i (xo , yo , − f ),
the aim of the calculation is to derive the field distribution at the right focal plane,
E f (x f , y f , f ). The calculation is performed according to the following scheme:
(a) the wave is propagated from the plane z = − f to the plane z = 0 by using
Fresnel diffraction, as expressed by (1.50); (b) at z = 0 the field distribution is mul-
tiplied by the lens transmission function τ f (x, y); (c) the obtained field distribution
is propagated to the plane z = f by using again Fresnel diffraction.
Step (a) of the calculation gives:

i exp(ik f ) ∞
E(x, y, 0) = d xo
λf −∞
∞
ik
E i (xo , yo , − f )exp [(x − xo ) + (y − yo ) ] dyo .
2 2
(2.100)
−∞ 2f
Steps (b) and (c) require calculating the integral:

i exp(ik f ) ∞
E f (x f , y f , f ) = dx
λf −∞
∞ ik
E(x, y, 0)τ f (x, y)exp [(x − x f )2 + (y − y f )2 ] dy. (2.101)
−∞ 2f
2.3 Fourier Optics 53
By inserting (2.100), and the expression of τ f given by (2.79), into (2.101), one
obtains:
∞ ∞
exp[2ik f + ik(n − 1)Δo ] ∞
E f (x f , y f , f ) = − d x o dy E (x
o i o o , y , − f ) dx
λ2 f 2 −∞ −∞ −∞
∞
ik
dy exp [(x − x f )2 + (y − y f )2 + (x − xo )2 + (y − yo )2 − x 2 − y 2 ] . (2.102)
−∞ 2f
The d xd y integral is solved by writing the integrand function as the exponential of

a square term. Noting that: (x − x f )2 +(y − y f )2 +(x − xo )2 +(y − yo )2 − x 2 − y 2 =
(x − xo − x f )2 + (y − yo − y f )2 − 2(xo x f + yo y f ), it is finally found:
exp[2ik f + ik(n − 1)Δo ]

E f (x f , y f , f ) =
λf
∞ ∞
ik
d xo dyo E i (xo , yo , − f )exp (xo x f + yo y f ) . (2.103)
−∞ −∞ f
Equation (2.103) represents a very important result: it states that the field dis-
tribution at the right-hand focal plane of the lens is the two-dimensional Fourier
transform of the distribution at the left-hand plane. Such a result is consistent with
the consideration, developed at the end of Sect. 2.2.9, that a biunique correspondence
exists between the propagation direction of the incident wave and points on the right
focal plane. In fact to perform the spatial Fourier transform of the field distribution
is the same as describing the field by a superposition of plane waves.
Equation (2.103) has a fundamental role in many optical processing techniques.
For instance, it suggests that spatial filtering of optical signals can be performed
by placing an appropriate filter at the right-hand focal plane, and reconstructing
the filtered signal with a second confocal lens. An example is shown in Fig. 2.16,
where a pinhole is used as a low-pass filter, i.e. a filter transmitting only low spatial
frequencies.
Fig. 2.16 Scheme of spatial filtering of an optical beam

2.4 Spectral Analysis
The electric field E(t) associated with a generic light beam can always be described,
as discussed in Sect. 2.1, as a weighted superposition of sinusoidal functions oscil-
lating at different frequencies. The Fourier transform of E(t) is the complex quantity
E(ω) (given by (2.10)) that represents the amplitude and the phase of the field com-
ponent at angular frequency ω.
The power spectrum of the light beam, S(ω), is defined as:
εo |E(ω)|2
S(ω) = c (2.104)
2
It should be noted that S(ω) does not contain any information on the phase of E(ω).
This means that the knowledge of S(ω) is not sufficient, in general, for reconstructing
the temporal behavior of E(t).
The usual method to measure the power spectrum of an optical signal is that of
collimating a sample beam and sending it through a dispersive system in order to
angularly separate the different frequency components. This method was introduced
by Newton, who first observed the spectrum of solar light through a glass prism.
Instead of using the dispersion properties of glass, angular separation can be gen-
erated by diffraction gratings, i.e. structures with periodic transmission functions.
Almost all modern commercial spectrometers are based on gratings. As will be seen,
there is another method, consisting of the use of a narrow-band tunable filter. This
method is conceptually similar to that used to measure the power spectrum of electri-
cal signals. In the optical case, the narrow-band filter is a Fabry-Perot interferometer.
An important parameter for assessing the performance of a spectrometer is the
frequency resolution δν, defined as the minimum frequency separation that can be
resolved by the spectrometer. Calling νo the central frequency of the signal to be
analyzed, the resolving power of the instrument at that frequency is calculated as:
νo λo
Pr = = , (2.105)
δν δλ
where λo = c/νo and δλ is the minimum wavelength separation that can be resolved.
When writing the right-hand member of (2.105), the assumption δν νo has been
made. In fact, by differentiating νλ = c around νo , it is obtained: λo dν − νo dλ = 0,
hence (2.105).
2.4.1 Dispersive Prism
Optical components fabricated from dispersive materials refract light of different

wavelengths by different angles. Consider a collimated light beam that impinges
on a glass prism at an incidence angle θi , as shown in Fig. 2.17. Since the angle
2.4 Spectral Analysis 55
Fig. 2.17 Dispersive prism
of refraction depends on the refractive index, which is wavelength dependent, the

direction of the emergent beam is also wavelength dependent.
The deflection angle θ , calculated from the refraction law, is given by:

θ = θi − α + arcsin sin α n 2 − sin2 θi − sinθi cosα , (2.106)
where α is the prism angle and n the refractive index of the glass.
What’s the minimum wavelength separation δλ that can be resolved? Clearly, this
depends on the minimum angular resolution. Even assuming that the incident beam
is monochromatic and perfectly collimated, the output beam will present an angular
spread due to diffraction. If L is the minimum size between prism size and incident
beam diameter, and λo is the wavelength of the incident beam, the angular spread
will be δθ ≈ λo /L. In order to resolve two different wavelengths, the difference in
their deflection angles should be larger than δθ . Therefore, the wavelength resolution
is given by:
−1
dθ
δλ = δθ, (2.107)
dλ
where dθ/dλ is calculated by assigning the dependence of n on λ. Taking typical

values for the parameters appearing inside (2.106), α = θi = π/4, n = 1.5, it is
found that dθ /dλ ≈ dn/dλ. Therefore, the resolving power of the dispersive prism
can be written as:
λo dn
Pr = ≈L (2.108)
δλ dλ
If dn/dλ = (10 µm)−1 , and L = 1 cm, then Pr = 103 .
2.4.2 Transmission Grating
The diffraction grating is an optical component that periodically modulates (in the
transversal direction) the amplitude or phase of an incident wave. The treatment
here is limited to linear gratings in which the diffracting elements consist of parallel
lines ruled on glass or metal. These gratings may be used either in transmission or
in reflection. The most common transmission grating is made of a periodic array of

narrow slits on an opaque screen. The slit width is smaller than the wavelength, so
that the diffracted field has a wide angular spread.
Assume that the grating plane coincides with the x-y plane, and that the slits are
parallel to the x axis and have spacing d. The wave vector of the incident plane wave
lies in the y-z plane and forms an angle θi with the z axis, as shown in Fig. 2.18. The
spatial distribution of the field in the half-space beyond the grating is determined by
the interference among the diffracted waves coming from all the slits. Note that in
the approximation of a far-field projection the interfering beams can be considered
as parallel. Diffractive interference will occur only in those directions satisfying the
condition:
qλ = d(sin θq − sin θi ), (2.109)
where q = 0, ±1, ±2, . . .. Equation (2.109) defines a discrete set of directions in

which a diffracted beam can be observed.
In particular, the beam corresponding to the order of q = 0 is simply the trans-
mitted beam: sin θ0 = sin θi .
Equation (2.109) says that the directions of the diffraction orders (except for the
order 0) are wavelength-dependent. Therefore the grating is a dispersive structure.
As such it can be used, like the dispersive prism, to analyze the power spectrum of
an optical signal.
Similarly to the case of the prism, the limit to the wavelength resolution comes
from the diffraction effect due to the finite size of the grating. At θi = 0 the first-order
diffracted field, as derived from the theory of Fraunhofer diffraction, is expressed by
(1.69). This equation shows that, in the far-field plane at distance z from the grating
plane, the peak corresponding to the diffracted field is transversally displaced by
Fig. 2.18 Transmission

grating. The path difference
between the two diffracted
beams is: AC − B H =
d sin θq − d sin θi
λz/d and has a width λz/L. This means that the resolving power of the grating is:
λz/d L
Pr = = = N, (2.110)
λz/L d
where L is the size of the grating, and N the number of illuminated slits. Note that N
coincides with L/d only if the collimated beam fully covers the grating surface. By
introducing the number of slits per unit length, f o = 1/d, N = f o L. Taking, e.g.,
f o = 103 mm−1 and L = 1 cm, the resolving power is 104 , one order of magnitude
larger than that of the dispersive prism. As shown by (2.109), the angular separation
between two wavelengths grows by increasing the diffraction order: for the order q,
the resolving power becomes Pr = q N . However, the fraction of diffracted optical
power in a given order decreases with the value of q.
A problem arising from the use of a grating in spectral analysis is an ambiguity
that can appear if the optical signal has a very broad wavelength spectrum. Indeed,
considering (2.109), it can be seen that the first-order diffraction of the wavelength
λ1 appears in the same direction as the second-order diffraction of the wavelength
λ1 /2. This means that there is an upper limit to the spectral width that can be unam-
biguously analyzed by the grating. For very broad spectra it is more convenient to
use a dispersing prism.
Instead of an amplitude grating, a phase grating can be used: this is discussed in
Sect. 1.6.2. Most of the properties of the amplitude grating are also shared by the
phase grating.
The description of the diffraction properties of gratings assumes that the diffracted
field is observed in the far field. Since, in order to have a good resolving power,
the grating size must be of the order of centimeters, far field means a distance of
kilometers, as discussed in Sect. 1.6. However, by exploiting the properties of lenses
(or spherical mirrors) discussed in Sect. 2.3, the far field can be brought to the focal
plane of a lens placed beyond the grating: this is the approach adopted by all practical
spectrometers.
2.4.3 Reflection Grating
Reflection gratings are made by creating a linear alternation of reflecting and opaque
zones on a surface, as shown in Fig. 2.19.
In the case of the reflection grating, the diffractive effects are observed in the
backward direction. The directions of constructive interference are given by:
qλ = d(sin θi + sin θq ), (2.111)
where q is 0, ±1, ±2, . . . and θi is the incidence angle. Inside (2.111) diffraction
angles are positive if the corresponding directions stay in the upper half-plane with
Fig. 2.19 Reflection

grating. The black zones are
opaque, and the grey zones
are reflecting
respect to the grating plane normal. The zeroth-order beam propagates at the angle
θ0 = −θi , corresponding to specular reflection, that is independent from λ.
In a transmission grating the diffraction lobe due to a single slit presents maximum
intensity in correspondence with the direction of incidence. As a consequence, the
largest fraction of diffracted power goes to order 0. This is useless from the point
of view of spectral analysis because the zeroth order is not dispersive. Similarly, in
the case of a reflection grating the largest fraction of diffracted power would appear
in specular reflection. Diffraction gratings can be optimized such that most of the
power goes into a certain diffraction order, q0 = 0, leading to a high diffraction
efficiency for that order. This is indeed possible by using a so-called blazed grating
(see Fig. 2.20), in which the reflecting elements are tilted by an angle θg with respect
to the grating plane, obtaining a sawtooth-like profile.
The specular reflection from the tilted elements appears in a direction forming
an angle θr = −(θi − θg ) + θg = (2θg − θi ) with the normal to the grating plane.
It is intuitive to expect that the maximum of the diffraction lobe associated to a
single reflecting element occurs in the direction of specular reflection relative to that
segment. By imposing that such direction coincides with that of the diffraction order
q0 , θqo = θr , the following relation is found:
θi + θqo
θg = . (2.112)
2
By substituting (2.112) inside (2.111):
qo λ = d[sin θi + sin (2θg − θi )]. (2.113)

Fig. 2.20 Blazed reflection grating
Once λ and θi are fixed, (2.113) prescribes which is the value of θg to be chosen
in order to maximize the power diffracted at order qo .
There are two particularly interesting cases. The first corresponds to the Littrow
configuration, in which the incident beam arrives normally to the tilted elements,
θi = θg , so that (2.113) becomes:
qo λ = 2d sin θg . (2.114)
In this case θqo = θi , which means that the order qo propagates in the opposite
direction with respect to the incident beam. The grating behaves as a wavelength-
selective mirror, because it reflects only the wavelength that satisfies (2.114). As
an example, consider a grating with 1200 lines per millimeter used in the Littrow
configuration at the first order (qo = 1) for λ = 600 nm. Equation (2.114) gives
θg ≈ 21◦ . The back-reflected wavelength can be tuned by slightly rotating the grating,
in order to change the incidence angle.
The second interesting case is that of normal incidence. If θi = 0, θg = −θqo /2.
The condition of constructive interference is: qo λ = d sin (2θg ). This is the config-
uration utilized in standard spectrometers.
2.4.4 Fabry-Perot Interferometer
The Fabry-Perot interferometer is a frequency filter, made by two parallel mirrors,

as shown in Fig. 2.21.
Fig. 2.21 Fabry-Perot interferometer
Considering a monochromatic plane wave at normal incidence, the aim is to

calculate the transmittance of the interferometer, TFP , as a function of the frequency
of the wave. If R1 and R2 are the reflectances of the two mirrors, one might naively
guess that TFP is given by the product (1 − R1 )(1 − R2 ). This is wrong, because
the field of the transmitted wave needs to be calculated by taking into account the
interference process arising by the superposition of the multiple reflections of the
waves bouncing back and forth between the two mirrors.
The treatment is analogous to the one performed for the anti-reflection coating.
By introducing the field reflection and transmission coefficients of the two mirrors,
ρ1 , τ1 , ρ2 , τ2 , and calling L the distance between the two mirrors, the electric field
of the transmitted wave, E t , is given by a converging geometric series:
τ1 τ2 eiΔ/2
E t = E i τ1 τ2 eiΔ/2 [1 + ρ1 ρ2 eiΔ + (ρ1 ρ2 eiΔ )2 + · · · ] = E i , (2.115)
1 − ρ1 ρ2 eiΔ
where E i is the electric field of the incident wave, and Δ is the phase shift introduced
by the round-trip propagation between the two mirrors. If λ is the wavelength of
the incident wave and n√the refractive index√ of the medium that fills the cavity, Δ =
4π n L/λ. Putting ρ1 = R1 eiφ1 , ρ2 = R2 eiφ2 , |τ1 |2 = 1− R1 , and |τ2 |2 = 1− R2 ,
it is found:
|E t |2 1 + R1 R2 − R1 − R2
TF P = = √ , (2.116)
|E i |2 1 + R1 R2 − 2 R1 R2 cosδ
where δ = Δ + φ1 + φ2 . The Fabry-Perot interferometer is usually made of two

identical mirrors. Putting R1 = R2 = R, (2.116) becomes:
(1 − R)2
TF P = . (2.117)
1 + R 2 − 2Rcosδ
As shown in Fig. 2.22, TFP is a periodic function of the phase delay δ, and takes
the maximum value, Tmax = 1, when cosδ = 1, that is δ = 2qπ , q being a positive
integer. TFP is minimum when cosδ = −1, that is, δ = (2q + 1)π . The minimum
value is: Tmin = [(1 − R)/(1 + R)]2 . For instance, if R = 0.99, Tmin = 0.25 × 10−4 .
The ratio between Tmax and Tmin is the contrast factor F:
Fig. 2.22 Transmittance of the Fabry-Perot interferometer as a function of the round-trip phase
shift δ, which is linearly related to the frequency of the light wave
Tmax 4R
F= =1+ . (2.118)
Tmin (1 − R)2
Since δ is linearly dependent on the frequency ν of the incident wave, the abscissa
scale in Fig. 2.22 can be considered as a frequency scale. As such the figure shows
that the interferometer is a periodic filter in frequency-space. The frequencies corre-
sponding to maximum transmission (resonance frequencies) are given by:
c φ1 + φ2
νq = (q − ). (2.119)
2n L 2π
The frequency distance between two consecutive transmission maxima, called
free spectral range, is:
c
Δν = νq+1 − νq = . (2.120)
2n L
An important parameter characterizing the performance of the interferometer is
the width at half-height, W , of the transmission peak, expressed in units of radians.
R is close to 1, the approximate expression of W derived from (2.117) is: W
If √
4/ F 2(1 − R). In frequency units, the peak width, δν, is derived by noting that
δν/Δν = W/(2π ). Therefore:
1− R
δν = Δν (2.121)
π
As an example, if R = 0.99, δν ≈ Δν/300. Taking L = 1 cm and n = 1, one
finds Δν = 15 GHz and δν = 50 MHz.
Equation (2.121) seems to suggest that, once fixed Δν, the resolving power of the
interferometer could be made arbitrarily large by choosing a value of R very close
to 1. However, when 1 − R is made very small, there are effects due to diffraction or
misalignment losses that become important. In practice, it is difficult to obtain values
of the ratio Δν/δν (usually called finesse of the Fabry-Perot interferometer) larger
than 300. The minimum achievable frequency resolution is of the order of 1 MHz.
The finding that a sequence of two highly reflecting mirrors could give rise to
a complete transmission of the beam power, when cosδ = 1, may appear rather
counter-intuitive. It is important to take into account that TFP is the result of a

steady-state calculation. If it is assumed that the interferometer is illuminated by
an intensity step-function, starting at t = 0, it is clear that the initial transmittance
is very low. If the incident frequency corresponds to a resonance, what happens is
that optical power is progressively accumulated inside the interferometer, and the
output intensity progressively grows, by approaching asymptotically the value of the
input intensity. As an example, if R = 0.99 and the incident power is 1 W, the output
power of 1 W will be reached only when the power bouncing back and forth inside the
cavity attains the value of 100 W. It can be shown that the time constant controlling
such a transient is the reciprocal of the frequency resolution δν. In other terms, the
larger is the resolving power of the instrument, the longer is its time-response.
In order to use the Fabry-Perot interferometer for spectral analysis, one should be
able to continuously shift the transmission peaks by changing the optical path n L, so
that a scan of the frequency interval between two consecutive resonant frequencies
can be performed. What is the change in L, δL, required to scan a given free spectral
range? Assuming that the interferometer is initially tuned to νq , the condition to

satisfy is that the new resonance νq−1 corresponding to the modified length L + δL
should coincide with νq :
c c
νq = q = νq−1 = (q − 1) (2.122)
2L 2(L − δL)
From (2.122) it is found:
L c λq
δL = = = . (2.123)
q 2νq 2
Therefore in order to scan a free spectral range it is sufficient to change the mirror
distance by half wavelength.
In practical cases, the scan is performed by attaching one of the mirrors to a
piezoceramic, and varying the ceramic thickness by the application of a slow voltage
ramp. The mirror displacement is proportional to the applied voltage, and therefore
changes linearly with time. In such a measurement, the time scale is transformed
into a frequency scale, and so the observed temporal dependence of the transmitted
power coincides with the spectral distribution of the incoming optical signal.
As an alternative approach, scanning of the resonant frequency can also be per-
formed by keeping L fixed and changing the refractive index of the medium inside
the Fabry-Perot cavity. This can be done by putting the interferometer inside a sealed
gas cell and sweeping the pressure.
When L is larger than a few centimeters, spherical mirrors are used instead of
plane mirrors. The advantage of this approach is that the alignment tolerance is less
critical, and diffraction losses are lower.
To avoid ambiguity in spectral analysis, the spectral width of the incoming beam
should be narrower than the free spectral range. This means that the Fabry-Perot
interferometer is useful only for narrow-band signals.
2.5 Waves in Anisotropic Media 63
2.5 Waves in Anisotropic Media
Manipulation of light polarization is a key issue in several applications. Main goals

include controlled modification of the polarization state of a beam or selective trans-
mission of that portion of beam polarized in a particular linear direction. To this end,
components based on materials that exhibit an anisotropy in transmission or refrac-
tion are usually exploited. All media constituted by randomly positioned atoms (or
molecules), like gases, liquids, and amorphous solids (glasses) are isotropic, that is,
their macroscopic properties are independent of the orientation with respect to an
applied field. In contrast, ordered media (crystals) or partially ordered media (liquid
crystals) may be anisotropic. A consequence of this anisotropy is that the velocity
of an optical wave inside the medium will depend on its propagation direction and
polarization state. As will be seen, anisotropic materials can be used to change the
polarization state of an optical beam.
Inside an anisotropic medium the relation between the electric field and the electric
displacement is described by a tensor:

Di (r, ω) = εo εi j E j (r, ω). (2.124)
The indices i, j can take values 1, 2, 3, corresponding to the axes x, y, z. Equa-

tion (2.124) describes a system of three linear equations connecting the Cartesian
components of E to those of D. The electric permittivity tensor, εi j , is expressed by
a 3 × 3 matrix. It can be demonstrated that such a matrix is symmetric, which means
that εi j = ε ji . Symmetric matrices can be made diagonal by an appropriate choice of
the coordinate system. With reference to the diagonal axes, usually called principal
axes, the off-diagonal elements of the electric permittivity matrix vanish:
⎛ ⎞
ε11 0 0
⎝ 0 ε22 0 ⎠ (2.125)
0 0 ε33
From now on, unless otherwise specified, it will be assumed that the reference
system is the system of principal axes.
In the general case in which E has an arbitrary direction inside the anisotropic
medium, the vector D will not be parallel to E. Considering a monochromatic plane
wave with wave vector k, propagating inside an anisotropic medium, Maxwell’s
equations indicate that D and H are both perpendicular to k, whereas E is perpen-
dicular to H, but not, in general, to k. However, if E is directed along one of the
principal axes, E and D are parallel, but the relative permittivity will take different
values depending on the specific principal axis. Note that the polarization direction
inside an anisotropic medium is the direction of D, not of E.
It is useful to distinguish three situations:

√
• isotropic medium, ε11 = ε22 = ε33 , the refractive index n = ε11 is a scalar
quantity. The wave propagation velocity is the same for all directions and all
polarizations.
• uniaxial medium, ε11 = ε22 = ε33 , it is possible to define an ordinary refractive
√ √ √
index n o = ε11 = ε22 and an extraordinary refractive index n e = ε33 . The
z axis is called optical axis.
• biaxial medium, ε11 = ε22 = ε33 , the medium is characterized by three distinct
refractive indices.
In order to describe wave propagation in a uniaxial medium, it is better to start with
a simple situation, namely that of a wave propagating in a direction perpendicular to
the optical axis. Assuming that k is parallel to the x axis, D lies in the y-z plane. If the
√
wave is linearly polarized along y, then its velocity will be c/n o , where n o = ε22 .
Analogously, if the wave is polarized along z, then the propagation velocity will be
√
c/n e , where n e = ε33 . In both cases, the wave polarization does not change during
propagation. What happens now if the linear polarization direction is not coincident
with a principal axis? By exploiting the linearity of Maxwell equations, the plane
wave can be described as a linear combination of two waves, one polarized along y
and the other polarized along z. The two waves travel with different velocities, and so
a phase delay arises. When recombined, the resulting wave is elliptically polarized.
In the general case, in which k is not directed along a principal axis, the situation is
more complicated. It would be useful to discover if it is also the case that there are two
“special” linear polarization directions possessing the property that the polarization
remains unchanged during propagation. Luckily, straightforward theory shows not
only that those special directions exist, but also provides the means of determining
them. Once the directions are found, the propagation is treated by describing D as
the sum of two principal polarization modes.
The principal directions are found by using a geometric construction based on the
index ellipsoid that is described by the equation:
x2 y2 z2
+ + = 1. (2.126)
n12 n22 n32
Consider the ellipse generated by intersecting the ellipsoid with the plane perpen-
dicular to k and containing the origin of the coordinate system, as shown in Fig. 2.23.
The two principal modes have polarization directions given by the ellipse axes. In
addition, if n a and n b are the lengths of the ellipse semi-axes, then the propagation
velocities of the two modes are c/n a and c/n b , respectively.
In the case of a uniaxial medium, by taking the symmetry axis parallel to z, and
putting n 1 = n 2 = n o , and n 3 = n e , (2.126) becomes:
x 2 + y2 z2
+ = 1. (2.127)
no2 ne 2
Fig. 2.23 Index ellipsoid
From now on, the discussion is restricted to uniaxial media. It can be assumed
without loss of generality that k is in the plane x-z, forming an angle θ with the z
axis. The plane that is perpendicular to k and passes through the origin has equation:
xsinθ + zcosθ = 0. The intersection of this plane with the index ellipsoid gives an
ellipse that has one axis coincident with y, and the other one, placed in the x-z plane,
forming an angle θ with z. The length of the first semi-axis is n o . The length of the
second semi-axis, called n θ , satisfies the equation:
1 cos2 θ sin2 θ
= + (2.128)
n 2θ no2 ne 2
A generic plane wave with wave vector k can be described as the superposition of
an ordinary wave and an extraordinary wave. The former, polarized perpendicularly
to the plane defined by k and z, travels at velocity c/n o . The latter, polarized in the
plane defined by k and z, travels at velocity c/n θ . If θ = 0, then n θ = n o , that
is, for a wave propagating along the optical axis the medium behaves as though it
were isotropic. If θ = 90◦ , the extraordinary wave is polarized along z, and has a
propagation velocity c/n e .
For the ordinary wave, D has no components along z, therefore D is parallel to E,
and the Poynting vector is parallel to k. For the extraordinary wave, if 0 < θ < 90◦ ,
D has nonzero components along both z and x. Therefore D is not parallel to E. The
consequence is that the Poynting vector forms with k a non-zero angle, ψ, called
walk-off angle. ψ is given by the relation:

sin(2θ )n 2θ 1 1
tan ψ = − 2 (2.129)
2 ne 2 no
If the propagating wave is a collimated beam, the effect of walk-off is to generate

a spatial separation between ordinary and extraordinary beams.
Up to now the treatment has concerned the propagation inside the crystal. It is
also important to see what happens when a wave coming from an isotropic medium,
impinges on the crystal surface. The discussion is here limited to the case of a uniaxial
crystal having the optical axis in the incidence plane. An incident wave that is linearly
polarized perpendicularly to the incidence plane (σ case) propagates in the crystal
as an ordinary wave, whereas an incident π wave will behave as an extraordinary
wave. Since the refractive index is different in the two cases, the refraction angles
calculated by Snell’s law will also be different. In the case of arbitrary polarization,
an object seen through the uniaxial crystal will present a double image, hence the
name birefringent crystal.
The refractive indices of a few birefringent crystals are given in Table 2.3. Bire-
fringence is said to be positive if n e > n o , as in the case of quartz crystal, and negative
if n e < n o , as in the case of calcite crystal.
2.5.1 Polarizers and Birefringent Plates
Many applications require a well-defined polarization state. Conventional optical

sources produce light having a polarization state that changes randomly with time
(unpolarized light). Some lasers emit polarized light. Birefringent crystals are used
to generate polarized light or to modify the state of polarization of a light beam.
The simplest polarizer is made from a dichroic material. This is an anisotropic
material having, for instance, a large absorption coefficient for light polarized along
the optical axis, and a small absorption for light polarized perpendicularly to the
Table 2.3 Refractive indices of some birefringent crystals

Crystal no ne λ (nm)
Calcite, CaCO3 1.658 1.486 589
Quartz, SiO2 1.543 1.552 589
Rutile, TiO2 2.616 2.903 589
Corundum, Al2 O3 1.768 1.760 589
Lithium niobate, 2.304 2.212 589
LiNbO3
Barium borate, 1.678 1.553 589
BaB2 O4
Fig. 2.24 Polaroid polarizer.

The polymer chains are
parallel to y, and so only the
linear polarization parallel to
x is transmitted
optical axis. The most common dichroic material, Polaroid, is composed of a parallel
arrangement of linear chains of conjugated polymers (see Fig. 2.24).
The optical axis coincides with the chain direction. The polymer presents high
electric conductivity along the chain and low conductivity across the chain. In a
grossly simplified description it behaves as a conductor if the applied electric field is
parallel to the optical axis, and as an insulator if the electric field is perpendicular to
the optical axis. In practice the component polarized perpendicularly to the chains
also suffers a non-negligible attenuation. Polarizers made of Polaroid are cheap,
but the selectivity is not very large (a typical value of the extinction ratio, defined
as the transmittance ratio between unwanted and wanted polarization state, could
be ≈1/500). Since absorption heats the material, high-intensity beams damage the
polarizer.
High-selectivity polarizers, having an extinction ratio about 105 , consist of a pair
of prisms made of a birefringent crystal, typically calcite. One of the most utilized
polarizers is the Glan–Thompson polarizer, shown in Fig. 2.25, which is made of two
calcite prisms cemented together by their long faces. The cement has a refractive
index intermediate between the ordinary and the extraordinary indices of calcite.
The optical axes of the calcite crystals are parallel and aligned perpendicular to the
incidence plane, that is, the plane of the figure. Light of arbitrary polarization is split
into two rays, experiencing different refractive indices; the π -polarized ordinary ray is
totally internally reflected from the calcite-cement interface, leaving the σ -polarized
extraordinary ray to be transmitted. The prism can therefore be used as a polarizing
beam splitter. Traditionally a resin called Canada balsam was used as the cement in
assembling these prisms, but this has largely been replaced by synthetic polymers.
Fig. 2.25 Glan-Thompson

prism
Fig. 2.26 Birefringent plate
The Glan-Thompson polarizer operates over a very wide wavelength range, from
350 nm to 2.3 µm. This type of polarizer can handle much higher optical powers
than the dichroic polarizers.
Thin plates made of a birefringent crystal are used to change the polarization state
of a light beam. The plate faces are parallel to the x-z plane, and the light beam
travels in the y direction, perpendicular to the optical axis z, as shown in Fig. 2.26.
If the incident beam is linearly polarized in the x direction (ordinary wave), then
the propagation across the plate does not modify the polarization state, but simply
introduces a phase shift Δφo = 2π n o d/λ, where d is the plate thickness. Similarly,
no modification of the polarization state occurs for the beam polarized along z,
which is phase-shifted by: Δφe = 2π n e d/λ. An input wave linearly polarized in an
arbitrary direction can be described as the sum of an ordinary and an extraordinary
wave. The two waves are initially in phase, but acquire, on crossing the plate, a phase
difference Δφ = 2π(n e − n o )d/λ. In general, the output polarization is elliptical.
An important particular case is a birefringent plate that gives a phase difference of
|Δφ| = π . This is known as a half-wave plate. When linearly-polarized light is
incident upon a half-wave plate, the emergent light is also linearly-polarized. If the
polarization direction of the incident light makes an angle α with the optical axis, then
the polarization direction of the emergent beam makes an angle −α with the same
axis, therefore the direction of polarization has been rotated through an angle 2α, as
shown in Fig. 2.27. A plate for which |Δφ| = π/2 is called a quarter-wave plate.
When linearly-polarized light is incident upon a quarter-wave plate, the emergent
light is in general elliptically polarized. If, however, α = π/4, the emergent light is
circularly polarized.
2.5.2 Jones Matrices
The propagation of polarized light beams trough anisotropic optical elements is

conveniently described by using the Jones formalism, in which the electric field of
the light wave is represented by a two-component vector and the optical element by
a 2 × 2 transfer matrix.
Fig. 2.27 Rotation of the

polarization direction in a
half-wave plate
In general, the electric field of a monochromatic plane wave propagating along z

can be described as the sum of two components:
E(z, t) = E x exp[−i(ωt − kz + φx )]x + E y exp[−i(ωt − kz + φ y )]y, (2.130)
where x and y are unit vectors, directed along x and y, respectively. Equation (2.130)
can be written as:
E(z, t) = E o exp[−i(ωt − kz + φx )]V, (2.131)

where E o = E x 2 + E y 2 , and V is the Jones vector represented by the column
matrix:
cosα
, (2.132)
eiφ sin α
where α = arctan(E y /E x ) and φ = φ y − φx .

The adopted sign convention considers the angles measured counter-clockwise as
being positive.
The vector (2.132) corresponds to a generic elliptical polarization. If the wave is
linearly polarized, φ = 0. The following Jones vectors describe a linearly polarized
field with a polarization direction forming the angle α with the x axis, or directed
along the x axis, or along the y axis, respectively:

cosα 1 0
; ; . (2.133)
sinα 0 1
For a circularly polarized wave, α = π/4 and φ = ±π/2. Therefore the Jones
vectors associated with clockwise and counter-clockwise circular polarization are
respectively given by:
1 1 1 1
√ ; √ . (2.134)
2 i 2 −i
Two waves represented by the vectors V1 and V2 have orthogonal polarization

states if the scalar product of the two vectors is zero. As an example, the linear
polarization states along x and y are mutually orthogonal. An arbitrary polarization

state can always be described as the sum of two orthogonal polarization states.
An optical component modifying the polarization state of an incident beam is
described by a transfer matrix T. If Vi represents the incident wave, then the Jones
vector of the emergent wave is given by: Vu = TVi . If the beam goes through a
sequence of components with matrices T1 , . . . , Tn , then the emergent wave is given
by: Vu = Tn . . . T1 Vi . Note that, in general, the T matrices do not commute.
The polarizers having optical axis along x and y are represented, respectively, by
the matrices Tpx e Tpy :

10 00
Tpx = ; Tpy = . (2.135)
00 01
The birefringent plate having optical axis directed along x introduces a phase
shift Δφ between the x and y components of the electric field, without modifying
the amplitude ratio E x /E y . It is represented by:

1 0
Tr = . (2.136)
0 exp(iΔφ)
The phase shift is given by Δφ = 2π(n e − n o )d/λ, where d is the plate thickness.
The half-wave plate corresponds to Δφ = π , and the quarter-wave plate to Δφ =
π/2.
The polarization rotation matrix,

cosθ −sinθ
R(θ ) = , (2.137)
sinθ cosθ
represents an optical component that rotates the polarization direction. If the incident
wave is linearly polarized at an angle θ1 , the emergent wave is linearly polarized at
the angle θ2 = θ1 + θ .
It is important to take into account that Jones vectors and matrices change when
the coordinate system is changed. If a field is described by the vector V, once the
coordinate system is rotated by the angle θ , the same field is described by the new
vector V :
V = R(−θ )V. (2.138)
A polarizer having its optical axis that forms an angle θ with the x axis is repre-
sented by the matrix:

cos2 θ sinθ cosθ
Tpθ = R(θ )Tpx R(−θ ) = . (2.139)
sinθ cosθ sin2 θ
Similarly, the matrix describing an optical retarder (birefringent plate) having the
optical axis rotated by θ with respect to x is written as:
Trθ = R(θ )Tr R(−θ )

1 + [exp(iΔφ) − 1]sin2 θ [1 − exp(iΔφ)]sinθ cosθ
= . (2.140)
[1 − exp(iΔφ)]sinθ cosθ exp(iΔφ) − [exp(iΔφ) − 1]sin2 θ
2.5.3 Rotatory Power
Certain materials present a particular type of birefringence: the two fundamental

modes of propagation are circularly rather than linearly polarized waves, with right-
and left-circular polarizations traveling at different velocities, c/n + and c/n − . This
property, known as optical activity, is typical of materials having an intrinsically
helical structure at the molecular or crystalline level.
As mentioned in Sect. 1.3, a linearly polarized wave can be decomposed into the
sum of two circularly polarized waves having the same amplitude. By using Jones
matrices, one can write:

cosθ 1 1 1 1
= exp(−iθ ) + exp(iθ ) . (2.141)
sinθ 2 i 2 −i
If L is the thickness of the optically active medium, the phase delay between the
two circularly polarized waves is given by: φ = 2π L(n + −n − )/λ. The recombination
of the two emergent waves produces a linearly polarized beam (see Fig. 2.28) with a
polarization direction rotated with respect to that of the incident beam:

1 1 1 1
exp(−iθ ) + exp[i(θ − φ)]
2 i 2 −i

cos(θ − φ/2)
= exp(−iφ/2) . (2.142)
sin(θ − φ/2)
The rotation per unit length, called rotatory power ρ, is:
φ π(n + − n − )
ρ= = . (2.143)
2L λ
Fig. 2.28 Rotatory power

A typical optically active crystal is quartz. For propagation along the optical axis,
the rotatory power is 4 rad/cm at λ = 589 nm. Also disordered media containing
chiral molecules, as, for instance, sugar molecules, can show optical activity. For
an aqueous solution containing 0.1 g/cm3 of sugar, ρ = 10−2 rad/cm. The sugar
concentration can therefore be determined by measuring the rotatory power.
From a formal point of view, the optical activity can be explained only assuming
the presence of anti-symmetric off-diagonal terms inside the εi j matrix. Such a
presence would contradict the assertion that the matrix εi j is symmetric. However, it
can be shown that such a symmetry property applies only for spatially homogeneous
fields, that is, infinite wavelength fields. If λ is finite, the local value of D depends,
in general, on E and its spatial derivatives. At first order:
∂Ej

Di = εo εi j E j + iγi jm . (2.144)
∂ xm
jm
whereas εi j is symmetrical, the third-order tensor γi jm is anti-symmetrical with

respect to the indices i j: γi jm = −γ jim . If there is no absorption, then γi jm is
real. In the case of an incident plane wave with wave vector k, it is found that
∂ E j /∂ xm = i E j km . Hence:

Di = εo εi j E j + iγi jm E j km . (2.145)
jm
D is related to E by the new electric permittivity tensor:

εi j = εi j + i γi jm km , (2.146)
m
which is the sum of a symmetric and an anti-symmetric term, the latter vanishing if
the modulus of k vanishes, that is, if the wavelength becomes infinite.
For an isotropic medium, (2.145) becomes:
Di = εo [εr E i + iγ (E × k)i ], (2.147)
where (E × k)i denotes the i-th component of the cross product, and γ is a scalar
quantity. Assuming k parallel to z, the D components along x and y, as derived from
(2.147), are:
Dx = εo [εr E x + iγ k E y ] (2.148)
D y = εo [εr E y − iγ k E x ]. (2.149)
These two equations show that a linear polarization cannot propagate unchanged
inside an optically active medium, due to the fact that E x generates D y , and viceversa.
Considering that the electric field of a right- (left-) circularly polarized beam is written
as: E ± = E x ± i E y , by using (2.148) and (2.149) the quantities D ± = Dx ± i D y

may be written as:
D + = εo [εr + γ k]E + (2.150)
D − = εo [εr − γ k]E − (2.151)
These two equations show that circular polarization indeed represents a propaga-
tion mode inside the optically active
√ medium. The refractive indices seen by the two
circular polarizations are: n ± = εr ± γ k. Assuming that the second term under
square root is much smaller than the first one, putting k = nω/c and εr = n 2 ,
one obtains the approximate relation: n ± = n[1 ± γ k/(2)], and, consequently,
ρ = π γ k/(nλ).
2.5.4 Faraday Effect
There are various means to change the optical properties of a medium by applying
an external field. In this section the magneto-optic effect known as Faraday effect is
described. It consists in the induction of rotatory power through the application of
a static magnetic field. The effect is longitudinal, since the applied magnetic field
B is parallel to the propagation direction of the linearly-polarized optical wave. The
polarization direction is rotated by an angle θ = C V B L, where L is the propagation
length and C V is the Verdet constant. The sign of θ changes with the sign of B, but it
does not change by reversing the propagation direction. Therefore the rotation effect
is doubled if the beam travels back and forth inside the medium.
In the standard configuration of the Faraday rotator the medium has a cylindrical
shape and is placed inside a solenoid. The longitudinal magnetic field is created by
an electric current flowing into the solenoid.
In the presence of the magnetic field, the relation between D and E can be written,
similarly to (2.144), as:
D = εo [εr E + iγ B (B × E)], (2.152)
where γ B is called magneto-gyration coefficient. The rotatory power is given by:
π γB B
ρ = CV B = − . (2.153)
nλ
The big difference between a passive rotator, such as a quartz plate, and the Faraday
rotator is made clear by considering the situation in which the wave transmitted in
one direction is reflected back and retransmitted in the opposite direction. In the
case of the passive rotator there is no net polarization rotation because the rotation
generated in the forward path is canceled during the backward path, while in the
case of the Faraday rotator the effect is doubled. The Faraday rotator is said to be a
nonreciprocal device.
The Faraday effect can be found in gases, liquids, and solids. Since its microscopic
origin is connected to the Zeeman effect, the Verdet constant becomes larger when the
wavelength of the optical wave gets close to absorption lines. The most commonly
utilized materials are doped glasses and crystals, like YIG (yttrium iron garnet), TGG
(terbium gallium garnet) and TbAlG (terbium aluminum garnet). As an example, the
Verdet constant of TGG is: C V = −134 rad/(T · m) at λ = 633 nm, and becomes
−40 rad/(T · m) at λ = 1064 nm.
There are some ferromagnetic materials (containing Fe, Ni, Co) that exhibit large
values of C V , but they also have a very large absorption coefficient, so the Faraday
rotation can only be observed in reflection. They have an important application in
optical memories, as will be mentioned in Chap. 8.
2.5.5 Optical Isolator
An optical isolator is a device that transmits light in only one direction, thereby
preventing reflected light from returning back to the source. Isolators are utilized in
many laser applications, since back-reflected beams often have deleterious effects
on the operating conditions of optical devices.
A Faraday isolator may be constructed by placing a Faraday rotator between two
polarizers, as shown in Fig. 2.29. The optical axis of polarizer 1 is along x, the Faraday
cell rotates clockwise the polarization direction by π/4, the optical axis of polarizer
2 makes a π/4 angle with the x axis, so that the forward traveling wave is fully
transmitted by the device. If a portion of the transmitted wave is back-reflected by a
Fig. 2.29 Faraday isolator

subsequent optical component, it travels inside the isolator in the backward direction.
Assuming that reflection does not alter the polarization state, the reflected beam is
fully transmitted by polarizer 2, and its polarization direction suffers a clockwise
rotation of π/4 by the Faraday cell. The backward beam becomes y-polarized, and
therefore is blocked by polarizer 1.
The behavior of the isolator can be mathematically described by Jones matrices.
Let Vi and Vu be the Jones vectors of the incident and emergent waves, respec-
tively, and call T1 , T2 , T3 , the matrices representing polarizer 1, Faraday rotator, and
polarizer 2, respectively. It is important to recall that all vectors and matrices must
be expressed with reference to the same x-y coordinate system. The output field is
related to the input field by the following expression:
Vu = T3 T2 T1 Vi
√ √ √
=
1/2 1/2 √2/2 −√ 2/2 10 1
= √
2/2
. (2.154)
1/2 1/2 2/2 2/2 00 0 2/2
As expected, the emergent wave is linearly polarized at π/4 with respect to the
x axis. The back-reflected beam crosses the sequence of devices in opposite order,
and is completely blocked by the isolator:

0
Vu = T1 T2 T3 Vu = . (2.155)
0
Note that T3 seems to have no role in the proposed scheme. In practice, however,
it may happen that back-reflection modifies the polarization state. In such a case the
role of polarizer 2 becomes important.
Fig. 2.30 Faraday isolator including a half-wave plate

The sequence of a Faraday rotator plus a half-wave plate constitutes another

nonreciprocal device, as schematically depicted in Fig. 2.30. If the former rotates
the polarization direction by π/4, and the latter by −π/4, a linearly polarized beam
does not change its polarization state by crossing the device. On the contrary, for a
beam coming from the opposite direction, the two rotations have the same sign, so
that its polarization direction is rotated by π/2. Such a device can be used to make
uni-directional the wave propagation in a circular path.
2.6 Waveguides
Waveguides are strips of dielectric material embedded in another dielectric material

of lower index of refraction. Starting from the advent of optical communications, the
development of integrated optical systems has become more and more important. In
such systems the light propagates in a confined regime through optical waveguides,
instead of free space, without suffering from the effect of diffraction.
A light beam traveling in a direction that is only slightly tilted with respect to the
guide axis is totally reflected at the boundary between the two media, provided the
incidence angle is larger than the limit angle. So that the guide can completely trap
a light beam in a diffraction-less propagation: the optical power should be mainly
confined inside the guide, with only an evanescent tail extending outside.
Many aspects of guided propagation can be understood by studying the simple
case of a planar dielectric waveguide, consisting of a slab of thickness d and refractive
index n 1 , usually called core, surrounded by a cladding of lower refractive index n 2 ,
as shown in Fig. 2.31. In a planar waveguide, confinement is guaranteed only in one
dimension. Light rays making angles θ with the z-axis, in the x-z plane, undergo
total internal reflection if cosθ ≥ n 2 /n 1 . The electromagnetic theory predicts the
existence of waveguide modes, that is, transverse field distributions that are invariant
in the propagation. Here, the discussion is limited to waves linearly polarized along
x, that is, to transverse electric-field modes (TE modes). The propagation properties
are derived by writing the Helmholtz equation inside the two media,
Fig. 2.31 Scheme of a

planar waveguide
2.6 Waveguides 77
(∇ 2 + (n 1 ko )2 )E(r) = 0 (2.156)
(∇ 2 + (n 2 ko )2 )E(r) = 0, (2.157)
and imposing the boundary conditions.

By assuming that the wave propagates along z with a propagation constant β, the
electric field can be generally written as:
E(x, z, t) = A(x)exp[i(βz − ωt)] (2.158)
By substitution of (2.158) into (2.156), the equation relative to medium 1 (the

core) becomes:
d2 A
+ [(n 1 ko )2 − β 2 ]A(x) = 0. (2.159)
dx2
Similarly, inside the cladding:
d2 A
+ [(n 2 ko )2 − β 2 ]A(x) = 0 (2.160)
dx2
Since the propagation constant of the guided wave must be intermediate between
n 2 ko and n 1 ko , the solution of (2.159) is a periodic function, A1 cos(γ1 x), and that
of (2.160) is an exponential function, A2 exp(−γ2 x), where the real quantities γ1 and
γ2 are given, respectively, by:

γ1 = n 1 2 ko 2 − β 2 , (2.161)
and
γ2 = β 2 − n 2 2 k o 2 (2.162)
By imposing the continuity of the field amplitude and its derivative at x = d/2,
it is found:
A1 cos(γ1 d/2) = A2 exp(−γ2 d/2), (2.163)
and
A1 γ1 sin(γ1 d/2) = A2 γ2 exp(−γ2 d/2). (2.164)
Non-zero field amplitudes A1 and A2 exist only if the following condition is

satisfied: γ2
tan(γ1 d/2) = . (2.165)
γ1
This is a transcendental equation in the variable β. A graphic solution can be

found by plotting the right- and left-hand sides as functions of β.
What has been calculated is the propagation constant of the fundamental mode. It
can be shown that the waveguide can support a discrete sequence of modes, whose
propagation constants are derived by generalizing (2.165) as follows:
γ2
tan(γ1 d/2 − mπ/2) = , (2.166)
γ1
where m = 0, 1, . . ..
Whereas a solution always exists for the fundamental mode, a value of the propa-
gation constant β1 of the m = 1 mode satisfying the condition n 2 ko ≤ β1 ≤ n 1 ko is
found only if the mode frequency ν is larger than the cut-off frequency νc given by:
c
νc = , (2.167)
2d n 1 2 − n 2 2
By increasing the mode order m, the cut-off frequency increases. Therefore, by

increasing the frequency, the guide accepts more and more propagation modes.
According to (2.167), to be single-mode the waveguide must be thin, and the refrac-
tive index difference n 1 − n 2 needs to be small. As it will be seen in the following
chapters, there are various applications requiring single-mode propagation.
Concerning the field distribution of the waveguide modes, it should be noted that,
by increasing m, the field penetrates deeper into the cladding. Correspondingly, the
propagation constant βm decreases with m.
Most of the considerations developed for the propagation in slab waveguides also
apply for the case of channel waveguides, in which the confinement of the light beam
is guaranteed on two dimensions. A very important example of channel waveguide
is the optical fiber (Fig. 2.32a), which is a waveguide with a cylindrical symme-
try that will be discussed in greater details in Chap. 6. Figure 2.32b–d illustrate the
typical geometries of different channel waveguides fabricated on planar substrates,
which constitute the basic structures of several integrated optical devices, like optical
modulators (Chap. 4) and semiconductor lasers or amplifiers (Chap. 5). In the recent
years, the evolution of communication systems has driven the development of opti-
cal circuits mainly based on silicon deposited on silicon oxide (silicon on insulator,
SOI) in which several electronic and photonic functions are integrated on the same
Fig. 2.32 Scheme of the

most commonly used optical
waveguide geometries:
a optical fiber, b embedded
waveguide, c rib waveguide,
d ridge waveguide. The
darker regions represent the
cores in which the light is
confined
2.6 Waveguides 79
chip. Miniaturization is a key issue for this application, as it will be discussed in

Sect. 8.1.3, and waveguides with cross-sections of hundreds of nanometers and very
high refractive contrast are needed. This provides high confinement of the light that
can be bent in curved waveguides, thus making feasible the fabrication of complex
networks of waveguides on a very small area.
Problems
2.1 Given that the wavelength of a light signal in vacuum is 600 nm, what will it be
in a glass block having a refractive index n = 1.50?
2.2 A glass block having an index of 1.55 is coated with a layer of magnesium
fluoride of index 1.32. For light traveling in the glass, what is the total reflection
angle at the interface?
2.3 Yellow light from a sodium lamp (λ = 589 nm) traverses a tank of benzene (of
index 1.50), which is 30-m long, in a time τ1 . If it takes time τ2 for the light to pass
through the same tank when filled with carbon disulfide (of index 1.63), determine
the value of τ2 − τ1 .
2.4 Light having a vacuum wavelength of 600 nm, traveling in a glass (n = 1.48)
block, is incident at 45◦ on a glass-air interface. It is then totally internally reflected.
Determine the distance into the air at which the amplitude of the evanescent wave
has dropped to a value of 1/e of its maximum value at the interface.
2.5 A beam of light in air strikes the surface of a transparent material having an
index of refraction of 1.5 at an angle with the normal of 40◦ . The incident light has
component E-field amplitudes parallel and perpendicular to the plane-of-incidence of
10 and 20 V/m, respectively. Determine the corresponding reflected field amplitudes.
2.6 The complex index of refraction of silver at λ = 532 nm is n + in = 0.14 −

3.05i. Calculate: (i) the reflectance of an air-silver surface at normal incidence; (ii)
the transmittance of a 100 µm silver plate.
2.7 A Gaussian beam at wavelength λ = 0.63 µm having a spot size at the beam
waist of wo = 0.5 mm is to be focused to a spot size of 30 µm by a lens positioned
at a distance of 0.5 m from the beam waist. What focal length should the lens have?
2.8 A Gaussian beam with λ = 800 nm and wo = 0.1 mm at z = 0 is propagating

along the z axis. A plano-convex lens, made of a spherical glass cap having a radius
of curvature of 4 cm, is placed at z 1 = 20 cm. The refractive index of the glass is
n = 1.8. Determine: (i) at which coordinate z 2 the beam is focused; (ii) the beam
spot size at z 2 ; (iii) the lens diameter to be chosen in order to collect 99 % of the
beam power.
2.9 A blazed reflection grating with pitch d = 1200 nm and blazing angle θg =
20◦ is used in the Littrow configuration. Determine which incident wavelength is
diffracted at second order in the backward direction.
2.10 A Fabry-Perot interferometer made by two identical mirrors at distance d =

1 mm is illuminated by a light beam containing two wavelengths λ1 = 1064 nm and
λ2 = λ1 + Δλ, where Δλ = 0.05 nm. Determine which is the minimum reflectance
of the two mirrors required to obtain a resolving power Pr ≥ λ1 /Δλ.
2.11 Consider a light beam at λ = 589 nm traveling along the z-axis and linearly
polarized along the x-axis. The light beam crosses, in succession, a quarter-wave
plate and a polarizer. The optical axis of the quarter-wave plate is in the x-y plane,
forming an angle π/4 with the x-axis. The optical axis of the polarizer is parallel
to the y-axis. Write the Jones matrix of the quarter-wave plate in the x-y reference
frame, and calculate the fraction of the incident power that is transmitted by the
polarizer.
2.12 Consider the following sequence of components: a polarizer P1 , a Faraday

rotator that rotates counterclockwise the linear polarization direction of the incoming
beam by 45◦ , a calcite half-wave plate, and a second polarizer P2 . Assume that
the incoming beam travels along y and is linearly polarized along x, that the two
polarizers have optical axes parallel to x, and that the optical axis of the calcite plate
lies in the x-z plane. (i) Calculate which angle the optical axis of the plate should
form with the x axis in order to ensure a unitary transmission. (ii) Write the Jones
matrix of the half-wave plate in the x-z reference frame.
Chapter 3
The Laser
Abstract All the light sources utilized before the invention of the laser are based
on spontaneous emission processes from excited atoms. Since each atom emits in a
random direction with a random phase, conventional light sources are intrinsically
chaotic. The laser is a completely different light source, because it is based on stim-
ulated emission instead of spontaneous emission, which means that atoms emit in a
cooperative way, generating a monochromatic and collimated light beam. The funda-
mental ingredient for making a laser is the optical amplifier. A mechanism selecting
wavelength and direction is provided by the positive feedback that transforms the
amplifier into an oscillator. In this chapter, after shortly mentioning conventional
light sources, especially thermal sources, and after giving a brief historical introduc-
tion to the laser, the general scheme of the oscillator is first discussed. Successively,
by using an elementary treatment of the interaction between light and atoms, the
conditions under which a set of atoms can become an optical amplifier are derived.
After discussing the general laser scheme, the steady-state and pulsed behavior is
presented, together with a concise description of a few solid-state and gas lasers. The
chapter ends with a comparison between the laser and the conventional light source.
3.1 Conventional Light Sources
Conventional light sources, like flames, incandescent lamps, and electric discharge
lamps, operate by converting an input energy (chemical, thermal, and electrical,
respectively) to light. The light is emitted over the full solid angle, usually with
a very broad frequency spectrum, except for the case of low gas-pressure electric
discharge lamps. The main application of these sources is to provide illumination,
in substitution to the Sun.
It is of historical importance to discuss the emission of incandescent sources,
such as the Sun (which converts nuclear energy from fusion reactions into light), or
a common household tungsten-filament lamp. The study of the emission spectrum
from these incandescent sources and of its dependence on the absolute temperature T
has had a profound impact on modern physics. In the theoretical treatment proposed
during the nineteenth century, an ideal body consisting of a closed cavity kept in

DOI 10.1007/978-3-319-20627-1_3
82 3 The Laser
thermal equilibrium at temperature T was considered, under the assumption that the
inner walls of the cavity could emit radiation at all possible frequencies. By using a
general thermodynamic argument, Kirchhoff demonstrated that the spectrum of the
“blackbody” radiation from such a cavity depends only on T , and is independent
of the nature of the material or the shape of the cavity. Defining ρ(ν, T )dν as the
density of electromagnetic energy inside the cavity in the frequency interval between
ν and ν + dν, a very strong inconsistency between predictions based on Maxwell
equations and experimental observations of ρ(ν, T ) was discovered. By making the
revolutionary assumption that the energy of the dipole oscillating at frequency ν can
only take discrete values that are multiples of the fundamental quantum hν, Planck
was able to provide the solution to this problem in 1901, obtaining the following
expression:
8π hν 3 1
ρ(ν, T ) = . (3.1)
c 3 exp(hν/k B T ) − 1
Besides the Boltzmann constant of k B = 1.38 × 10−23 J · K−1 , (3.1) contains the
new universal constant of h, called Planck constant, having value 6.55 × 10−34 J · s.
By making the derivative of (3.1), one finds that the frequency νm corresponding
to the maximum value of ρ(ν, T ) is proportional to T , following the approximate
relation: νm ≈ 5k B T /h.
Introducing the peak wavelength λm = c/νm , expressing λm in meters e T in
Kelvin, the following relation, usually called Wien’s law, is obtained:
λm T = 0.288 × 10−2 m · K. (3.2)
Figure 3.1 shows the blackbody emission spectra of the Sun (5500 K) and a
tungsten-filament lamp (1500 K). In the first case the peak of the emission is in
the visible range, whereas in the second case a large portion of the emitted radiation
is in the infrared, indicating that the incandescent lamp is a low-efficiency source of
visible light.
Fig. 3.1 Blackbody emission spectra, as calculated for the Sun (T = 5500 K) and for a tungsten-
filament lamp (T = 1500 K). The two curves are normalized to the same maximum value
3.1 Conventional Light Sources 83
The electric-discharge lamp, also known as the fluorescent lamp, uses electron
collisions to generate the atomic excitation for light emission. Notwithstanding the
fact that the working temperature can be much lower, this approach provides a higher
conversion efficiency of electrical power to visible light power than incandescent
sources.
3.2 Origins of the Laser
Whereas traditional light sources consist of a collection of independently emitting

atoms, the laser is a source in which individual emitters have a correlated phase and
also well defined emission direction.
In an important paper published in 1916 Einstein considered a blackbody cavity
in thermal equilibrium containing a set of two-level atoms, and showed that the
balance between absorption and emission processes becomes compatible with the
blackbody law only if one assumes the existence of a previously unknown process,
called stimulated emission. At that time, it was not noted that this new process can
result in optical amplification. It took a further four decades before proposals aimed
at using stimulated emission to amplify optical beams were put forward.
The first amplifier based on stimulated emission, called maser (acronym of
“Microwave Amplification by Stimulated Emission of Radiation”), used ammonia
molecules in a high magnetic field, and had emission in the microwave frequency
range, at 23.87 GHz. The experiment was performed at the Columbia University,
New York, by Gordon, Zeiger and Townes in 1954. They sent a beam of excited
ammonia molecules into a resonant cavity, obtaining a self-sustaining emission. A
maser amplifier had an important role in the first demonstration of long-distance
communications using the reflection of a microwave signal from an orbiting satellite
in space. In this experiment, performed by a Bell Labs team led by Pierce in 1960,
a maser was used as a very sensitive low-noise device to amplify the faint signals
coming from the satellite.
The fundamental step for the birth of Quantum Electronics (this was the initial
name given to this new field) was the invention in 1960 of the first oscillator emitting
visible light, the ruby laser, created by Maiman, a researcher working at the Hughes
Research Laboratories in California. Such a realization had been prepared by the
theoretical studies of Townes in the United States, and Basov and Prokhorov in
Russia, studies that were awarded the Nobel Prize in 1964. The acronym “laser”
comes from maser, with the substitution of “light” for “microwave”. The laser is
indeed an oscillator, even if the acronym laser (Light Amplification by Stimulated
Emission of Radiation) seems to describe an amplifier.
Starting from 1960, Photonics (this is the name now more frequently used instead
of Quantum Electronics) had a very rapid development. In a few years, laser action
was demonstrated in a wide variety of solid, liquid, and gaseous materials. A turning
point was the invention in 1962 of the semiconductor laser; a very small device having
high efficiency and long operating lifetime, which can be fabricated using the same
84 3 The Laser
technology developed for integrated electronic circuits and devices. Normally every
type of laser emits one (or more) well-defined wavelengths that are characteristic
of the material used for amplification. A very important property of semiconductor
lasers is that they can be designed to work at a pre-selected wavelength, by tuning
the chemical composition of the active layer.
3.3 Properties of Oscillators
An oscillator is a device that generates a sinusoidal signal with a stable amplitude and
fixed frequency. The capability of producing oscillating electric fields at higher and
higher frequencies has played a basic role in the development of telecommunication
systems, of computers, and, in general, of all electronic instrumentation. The laser
is a device that extends the working principle of the electronic oscillator to opti-
cal frequencies. The analogy with the electronic device is conceptually important,
because it explains at first glance the different emission properties of a laser with
respect to traditional light sources. At the same time, the analogy immediately sug-
gests that lasers can play a very important role in the information and communication
technology.
In this section the general properties of oscillators are discussed, showing that an
oscillating output signal can be generated by a scheme involving the combination of
amplification with positive feedback. The discussion is developed generally, without
specifying the exact nature of the signal. The description of the optical amplifier and
the corresponding oscillator is deferred to the following sections.
In the block scheme presented in Fig. 3.2, the input and output signals Si e Su
may represent a displacement in the case of the mechanical oscillator, a voltage for
the electronic oscillator or an electric field for a microwave or optical oscillator. The
input signal Si goes through an amplifier with gain G. The fraction β of the amplifier
output is fed back and added to Si . Therefore, in the stationary regime, the output
signal Su is given by:
Su = G Si + βG 2 Si + β 2 G 3 Si + · · · (3.3)
Fig. 3.2 General scheme of the oscillator: amplifier with positive feedback
3.3 Properties of Oscillators 85
By assuming that |βG| < 1, the geometric series represented by (3.3) is conver-
gent. One obtains:
G
Su = Si . (3.4)
1 − βG
The quantity βG is called loop gain. The transfer function of the amplifier with
feedback, defined as the ratio between output and input, is given by:
Su G
T = = . (3.5)
Si 1 − βG
The signal Si is taken as a sinusoidal function of time with angular frequency ω.

Generally speaking, both G and β are complex functions of ω. In order to make the
treatment more explicit, assume that β is independent of ω and that the frequency
dependence of G has a resonance at ωo :
Go
G(ω) = (3.6)
1 + i(ω − ωo )τ
The power amplification |G(ω)|2 is:
|G o |2
|G(ω)|2 = (3.7)
1 + (ω − ωo )2 τ 2
The function (3.7) is a Lorentzian (from Lorentz), and represents the effective
behavior of absorption (or gain) as a function of the angular frequency in most
practical cases. The quantity τ −1 is the half-width at half-height of the Lorentzian
function.
Through substitution of (3.6) inside (3.5), one finds the transfer function of the
system with feedback:
Go G fo
T (ω) = = , (3.8)
1 + i(ω − ωo )τ − βG o 1 + i(ω − ωo )τ f
where G f o = G o /(1 − βG o ), and τ f = τ/(1 − βG o ). It is evident that positive

feedback increases the peak gain, and, at the same time, decreases the frequency
band, such that the product between frequency band and peak gain is left unchanged.
Figure 3.3 shows the behavior of the power transfer function, |T (ω)|2 , for increasing
values of the loop gain, within the interval 0 < βG o < 1.
For βG o tending to 1, G fo tends to infinity and the bandwidth Γ f = 2τ −1 f tends
to 0. The condition βG o = 1 is the threshold condition for the operation of the
oscillator: in fact, if G f o becomes infinite, the system can provide a finite output
signal even if the input is infinitesimal.
What kind of output signal is generated by such an oscillator? Since the bandwidth
becomes infinitely narrow around ωo , the output signal must be a sinusoidal oscilla-
tion at frequency ωo . In the general case, in which β also depends on ω, the angular
86 3 The Laser
Fig. 3.3 Behavior of the

frequency response of the
amplifier with positive
feedback for various values
of the loop gain
frequency corresponding to maximum loop gain can be different from ωo . If the loop
gain is larger than 1, the oscillation is triggered by the noise always present inside
a real-world loop: from amongst all the noise frequency components, the system
preferentially amplifies the component with angular frequency ωo .
The treatment developed thus far would indicate that the output signal Su is
continuously growing when the gain is above the threshold value G oth = β −1 .
This is clearly impossible, because it would violate energy conservation. Since the
supply power of the amplifier has a finite value, the output power must also have a
finite upper bound.
The steady-state value of Su can be derived using a treatment that includes nonlin-
ear effects. It is a general property of an amplifier that the gain must decrease when
the amplitude of the output signal becomes large. This effect is called saturation.
As shown qualitatively in Fig. 3.4, the gain G o must be a decreasing function of Su ,
going to 0 when Su becomes large.
The working operation of an oscillator is as follows: initially the oscillator is
designed to have a small-signal gain larger than β −1 . As long as Su grows through
Fig. 3.4 Saturation of the

amplifier: behavior of gain as
a function of the output
signal Su . The working point
of the oscillator is given by
the condition βG o (Sus ) = 1
3.3 Properties of Oscillators 87
successive amplifications within the loop, the gain G o (Su ) decreases: the stationary
state is reached when Su attains the saturation value Sus , in correspondence of which
the saturated loop gain is equal to 1. By looking at Fig. 3.4 it is evident that, if
Su < Sus the loop gain is >1 so that Su is growing toward the stationary value,
whereas, if Su > Sus , the loop gain becomes <1 so that Su is now decreasing
toward the stationary value. The conclusion is that the stable point of the oscillator
corresponds to the situation in which the saturated loop gain is equal to 1.
3.4 Emission and Absorption of Light
The general discussion of the preceding section shows that the essential ingredient
for building an optical oscillator is the optical amplifier. In this section an elementary
description of the interactions between atoms and light is presented, as a preliminary
step toward the treatment of the optical amplifier.
An atom is made of a nucleus with a positive electric charge at the center and a
surrounding electron cloud of negative electric charge. Viewing the atom as a micro-
scopic planetary system, its internal energy can be defined as the sum of the poten-
tial energy associated to the spatial distribution of electric charges plus the kinetic
energy of the electrons that are undergoing a rotational motion around the nucleus.
The configuration of the atom can be determined, at least in principle, by solving
the Schrödinger equation for the system. Quantum mechanics demonstrates that the
internal energy of the atom can take only discrete values, here called E 0 , E 1 , E 2 . . .
in increasing order, where E 0 is the energy of the equilibrium configuration (ground
state), and E j , with j > 0, is the energy of the j-th excited state. In an intuitive
model the excited state corresponds to one (or more) electrons having moved from
the equilibrium orbit to an outer, higher energy, orbit. The atom may make a transition
from its initial state to an upper excited state by absorbing a photon, or it may decay
from an excited state to a lower-energy state by emitting a photon, but all transitions
must satisfy the energy conservation relation:
Ei − Ek
νik = , (3.9)
h
where E i > E k . The atom in the state k can move to the state i by absorbing a
photon, only if it is illuminated by light having frequency νik . Analogously the atom
decaying from state i to state k emits a quantum of light at frequency νik . Each type of
atom has different energy levels, and so it has a different discrete set of characteristic
frequencies νik . It should be mentioned that not all atomic transitions are radiative,
i.e., are compatible with the absorption or emission of light. There are quantum
mechanical selection rules (not discussed here) that determine which transitions are
radiative.
It is useful to have in mind the order of magnitude of the energy difference
for atomic transitions associated to visible light. Considering, as an example, light
88 3 The Laser
Fig. 3.5 Absorption and emission processes between levels 0 and 1
with a wavelength of λ = 0.5 µm, and frequency ν = c/λ = 6 × 1014 Hz, the
corresponding energy difference, calculated from (3.9), is 2.5 eV.
Consider a box containing Nt identical atoms per unit volume, illuminated by a
monochromatic wave with intensity I and frequency ν close to ν10 . If the intensity
is divided by the photon energy, it gives the photon flux Φ = I /(hν), which is the
number of photons crossing the unit area in the unit time.
A schematic of the different absorption and emission processes is illustrated in
Fig. 3.5. Some atoms will absorb a photon moving from the ground level 0 to level
1. The number of absorption processes per unit time and unit volume must be pro-
portional to the number of available photons and to the number of atoms on level 0.
Every absorption reduces by 1 the number of atoms on level 0, therefore:
d N0
− = σ01 (ν)Φ N0 , (3.10)
dt
where N0 (t) is the number of atoms per unit volume on level 0 at time t, called the
population of level 0 at time t. The minus sign in front of the derivative inside (3.10)
indicates that absorption processes decrease the population of level 0.
The proportionality constant σ01 (ν), called absorption cross-section, represents
the probability of absorption of a photon by the atom, and has dimensions of an area.
Since the energy of the excited atomic levels is always defined with some spread,
σ01 (ν) is a function of ν that is peaked at ν10 , but has a non-zero broadening. It is
typically described by a Lorentzian function.
The excited atom goes back to the ground state by spontaneously emitting a pho-
ton. The spontaneous emission is spatially isotropic, and has a spectral distribution
identical to that of the absorption cross-section. It is important to point out that the
radiation spontaneously emitted by a given atom has no memory of the direction,
frequency or phase of the exciting radiation, and that it is uncorrelated with the
emission of other nearby atoms.
By assuming that at t = 0 there are some excited atoms on level 1, it is found
that the intensity of spontaneous emission decays exponentially with time. The time
constant τ10 , which is characteristic of the considered transition, is called lifetime
−1
of the level 1. The reciprocal A10 = τ10 represents the decay probability from 1
3.4 Emission and Absorption of Light 89
to 0 per unit time. Considering that every spontaneous emission reduces the num-
ber N1 by 1, and that the number of spontaneous emission processes per unit time
and unit volume is given by A10 N1 , the decrease of the population of level 1 is
described by:
d N1
− = A10 N1 (3.11)
dt
In 1917 Einstein demonstrated that the presence of quasi-resonant radiation at a
frequency close to ν10 stimulates the decay of the excited atom from state 1 to state 0.
In analogy with the absorption process, the number of stimulated emission
processes per unit time and unit volume can be written as:
d N1
− = σ10 (ν)Φ N1 , (3.12)
dt
where σ10 (ν) is the stimulated emission cross-section, which, as shown in the next
section, coincides with the absorption cross-section.
If the atoms are at an absolute temperature of T , then the ratio between N1 and
N0 is given by the Boltzmann factor:

N1 hν10
= exp − . (3.13)
N0 kB T
As an example, for a transition in the red light region (hν10 = 2 eV) at room
temperature (T = 300 K), it is found: N1 /N0 = e−80 ≈ 10−35 . This means that the
probability of finding an excited atom is negligible, for a transition in the visible at
room temperature.
If all the atoms are in the lower energy state (i.e., the level 0), a quasi-resonant light
beam initially generates only absorption processes. If some fraction of the atoms are
in level 0, and the rest are in level 1, then a competition arises between absorption and
stimulated emission processes. If N0 > N1 , absorption prevails. However, if N0 <
N1 , then the number of generated photons overcomes the number of absorptions.
The condition of N0 < N1 is called the population inversion condition.
The important aspect of the stimulated emission process is that, in contrast to the
case of spontaneous emission, the emitted radiation has the same direction, polar-
ization, frequency and phase of the incident radiation. As such, a light beam having
a frequency close to ν10 experiences amplification after passing through a volume
containing a collection of atoms that are prepared in a state of population inversion.
In order to build an optical amplifier, it is necessary to find a means of creating and
maintaining a population inversion.
What happens if a set of two-level atoms is illuminated with a photon flux Φ at
frequency ν10 ? By combining (3.10), (3.11), and (3.12), and taking into account the
90 3 The Laser
conservation condition that N0 + N1 = Nt , the following dynamic equation for the

population N1 (t) is obtained:
d N1
= σ01 (ν)Φ(Nt − 2N1 ) − A10 N1 . (3.14)
dt
The first term on right-hand side of (3.14) describes the balance between absorp-
tion and stimulated emission processes, the second term takes into account spon-
taneous emission. Assuming that the photon flux is suddenly activated at t = 0
and remains constant for t > 0, and setting the initial condition of N1 (0) = 0, the
integration of (3.14) gives:
σ01 Φ Nt
N1 (t) = {1 − exp[−(A10 + 2σ01 Φ)t]} (3.15)
A10 + 2σ01 Φ
The behavior of N1 (t) is shown in Fig. 3.6. The stabilization to the steady-state
response is governed by the time constant: τ = (A10 + 2σ01 Φ)−1 . The steady-state
values of N1 (t) and N0 (t), derived for t → ∞, are:
Nt 1 + A10 /(σ01 Φ)
N1s = , N0s = Nt . (3.16)
2 + A10 /(σ01 Φ) 2 + A10 /(σ01 Φ)
In presence of the flux Φ the set of atoms becomes less absorbing, because the
population of the ground state is reduced, and simultaneously there are stimulated
emission processes partially compensating absorption. In any case, N1s is always less
than N0s , but both stationary values tend to Nt /2 as Φ tends to infinity. Consequently,
when Φ → ∞, the material becomes completely transparent, because the number
Fig. 3.6 Behavior of N1 (t) in in presence of a step-function photon flux

3.4 Emission and Absorption of Light 91
of stimulated emission processes compensates for the absorption processes. The first
of relations (3.16) can be written as:
Nt 1
N1s = (3.17)
2 1 + Φs /Φ
where Φs = A10 /(2σ10 ) is the saturation flux for the considered transition.
3.4.1 Einstein Treatment
In his 1917 article, Einstein not only introduced the new concept of stimulated emis-
sion, but also derived in a simple way the relations connecting emission and absorp-
tion processes. In this section the main steps of the treatment are reproduced.
Einstein considered a set of atoms inside a blackbody cavity at temperature T,
observing that, at thermal equilibrium, the number of atoms per unit time making
the transition from level 0 to level 1 must coincide with the number of atoms making
the opposite transition. The radiation thermally generated inside the cavity is not
at all monochromatic, but it presents a frequency spread comparable or even larger
than the bandwidth Δν10 of the absorption cross-section of the considered atomic
transition. In such cases, in order to express the number of absorption processes,
one should use, instead of (3.10) that refers to a monochromatic photon flux, the
following integral:

d N0
= −N0 σ01 (ν )φ(ν )dν , (3.18)
dt
where φ(ν ) is now the incident flux per unit frequency. The total incident flux is
given by the integral:

Φ = φ(ν )dν (3.19)
Of course, (3.18) reduces to (3.10) if the incident flux is monochromatic: φ(ν ) =

Φδ(ν − ν), where δ(ν − ν) is Dirac delta function. In contrast, if the bandwidth of
φ(ν ) is much larger than Δν10 , then (3.18) can be approximated as:
d N0
= −c−1 B01 φ(ν10 )N0 , (3.20)
dt
where
B01 = c σ01 (ν )dν (3.21)
is the absorption coefficient (expressed in units: m3 s−2 ).

92 3 The Laser
Balancing spontaneous emission and absorption processes, and introducing the

energy density ρ(ν) = (hν)φ(ν)/c instead of the photon flux, one finds:
hν10 A10 N1 = ρ(ν10 )B01 N0 (3.22)
Taking into account that N0 /N1 is given by (3.13), (3.22) gives a dependence of
ρ(ν) on temperature that, rather surprisingly, is different from the expression (3.1). At
this point, Einstein made the hypothesis that a third process must be occurring, which
he called stimulated emission. By expressing the number of stimulated emission
processes per unit time with a relation similar to that used for absorption processes,
(3.12) transforms into:
d N1
= −c−1 B10 φ(ν)N1 (3.23)
dt
where B10 is the stimulated emission coefficient:

B10 = c σ10 (ν )dν (3.24)
The balance between absorption and emission, including now also stimulated
emission, becomes:
hν10 A10 N1 + ρ(ν10 )B10 N1 = ρ(ν10 )B01 N0 . (3.25)
From (3.25) the following expression is derived:

hν10 A10
ρ(ν10 ) = . (3.26)
B01 (N0 /N1 ) − B10
By taking into account (3.13), and comparing (3.26) with (3.1), the following
relations are obtained:
A10 8π ν10 2
B10 = B01 , = (3.27)
B01 c3
Note that the equality B10 = B01 implies that σ10 = σ01 .
Relations (3.27) show symmetry between absorption and stimulated emission
processes, and demonstrate that the spontaneous emission coefficient A10 is propor-
tional to B01 . The quantum-mechanical calculation, not discussed here, shows that
B01 is proportional to ν10 . As a consequence, A10 is proportional to the cube of the
frequency: this means that the spontaneous decay time τ10 = (A10 )−1 is much longer
for infrared transitions with respect to visible transitions, and becomes very short for
UV or soft X-ray transitions. As will be seen, very short decay times of the excited
level make it much more difficult to create and maintain a population inversion for
a given transition, and so it is comparatively difficult to build optical amplifiers and
lasers in the UV or soft X-ray region.
3.5 Optical Amplification 93
3.5 Optical Amplification
In this section it is described how to obtain optical amplification. It is clear from the
previous sections that it is impossible to obtain a steady-state population inversion on
the transition 0–1 using an approach dealing only with levels 0 and 1. The excitation
process must involve a third level.
Three-level system. The discussion starts by considering a three-level system, like
the one shown in Fig. 3.7. The scheme is the same as that used in the first laser; the
ruby laser. The aim is to obtain a population inversion between the ground level 0
and the excited level 1. In the jargon of Photonics the process producing population
inversion is called pumping process, and the amplifying medium is called active
medium. The idea is to pump with light resonant with the transition between level 0
and level 2. In presence of a photon flux Φ p (ν) at ν02 , absorption processes exciting
atoms to level 2 occur. The spontaneous decay from level 2 to level 1 populates
level 1. Intuitively, one could predict that population inversion can be obtained if the
spontaneous decay process from 2 to 1 is quicker than the process from 1 to 0. In
a regime of dynamic equilibrium, this does indeed create a “bottleneck” that favors
the growth of N1 .
The quantitative treatment of the three-level scheme is based on the so-called
rate equations. These describe the population dynamics by taking into account all
possible emission and absorption processes among the three levels:
d N1
= A21 N2 − A10 N1 (3.28)
dt
d N2
= W02 p (N0 − N2 ) − (A20 + A21 )N2 , (3.29)
dt
where A10 , A20 and A21 are spontaneous decay probabilities per unit time, and W02 p
is the probability per unit time of a transition from
level 0 to level 2 under the action
of the pump. Using (3.18), one finds: W02 p = σ02 (ν )φ p (ν )dν .
Fig. 3.7 Three- and four-level schemes. The vertical direction indicates increasing energy of the
atomic levels
94 3 The Laser
Since the sum of the three populations coincides with the total number of atoms
per unit volume Nt :
N0 + N1 + N2 = Nt , (3.30)
N0 can be eliminated from (3.29), so that a system of two equations with two unknown
functions is obtained. If W02 p is constant, then the system can be easily solved. Here
the discussion is limited to the steady-state solutions, obtained by putting all the
time-derivatives equal to 0.
The steady-state population inversion is given by:
A10 W02 p +A20 +A21

1− A21 W02 p
Ns = N1s − N0s = Nt . (3.31)
A10 2W02 p +A20 +A21
1+ A21 W02 p
In order to obtain amplification, Ns should be positive, which means that the sign
of the fraction appearing at right-hand side of (3.31) must be positive. For very large
values of the pump parameter, W02 p A20 + A21 , it is a necessary condition that
A21 > A10 , in agreement with the intuitive considerations mentioned above. It is
interesting to note that, if A21 A10 and simultaneously W02 p A20 + A21 , then
(3.31) reduces to Ns = Nt . This means that all the atoms have been excited to level
1.
The condition Ns = 0 gives the threshold value of the pump parameter:
A20 + A21
W02 pth = A10 . (3.32)
A21 − A10
To avoid excessive values of W02 pth , systems in which A21 is much larger than
A10 are usually chosen. The first laser (i.e., the ruby laser) has the simple three-level
scheme, with A21 many orders of magnitude larger than A10 .
Four-level system. Optical amplification can be also achieved by using a four-level
scheme, illustrated in Fig. 3.7. In this scheme, the pump excites the atom from the
ground level to level 3; level 2 is then populated by the spontaneous decay from 3 to
2, and optical amplification occurs between 2 and 1, at a frequency ν21 . Intuitively
it can be expected that population inversion between 2 and 1 can be achieved, if the
spontaneous decay time from 3 to 2 is short, from 2 to 1 is long, and from 1 to 0 is
short.
There is an important difference between the three- and four-level schemes. In
the three-level scheme, the amplifier exploits the transition 1 to 0, in which the lower
level is the ground level. As a consequence, in order to reach threshold, which means
zero population inversion, at least half of the atoms must be excited. In the four-level
scheme the amplifier exploits a transition between two excited states. This means
that, since level 1 is initially empty, a positive population inversion is technically
achieved when even only one atom is brought to level 2. As a result, the four-level
3.5 Optical Amplification 95
scheme requires much lower pump power with respect to the three-level scheme for
lasing, and so most practical lasers operate using a four-level scheme.
The dynamics of the four-level system is described by the following rate equations:
d N3
= W03 p (N0 − N3 ) − (A30 + A31 + A32 )N3 (3.33)
dt
d N2
= A32 N3 − (A20 + A21 )N2 (3.34)
dt
d N1
= A31 N3 + A21 N2 − A10 N1 , (3.35)
dt

where W03 p = σ03 (ν )φ p (ν )dν .
The system (3.33)–(3.35) is completed by:
N0 + N1 + N2 + N3 = Nt . (3.36)
Again, the discussion is limited to the steady-state solutions, by putting all time
derivatives equal to 0. The steady-state population inversion is found to be:
(1 − C1 )W03 p
Ns = N2s − N1s = Nt , (3.37)
C2 + C3 W03 p
where
A31 A21 + A31 A20 + A32 A21
C1 = (3.38)
A10 A32
(A21 + A20 )(A30 + A31 + A32 )

C2 = (3.39)
A32
A31 A21 + A31 A20 + A32 A10 + 2 A10 A21 + 2 A10 A20
C3 = . (3.40)
A10 A32
As expected from intuitive considerations, the threshold pump parameter W03 pth is
equal to 0. The steady-state population inversion Ns becomes positive when C1 < 1.
In many practical cases A20 A21 and A31 A32 , so that the condition Ns > 0
simply reduces to: A21 < A10 .
Since the rate equations constitute a set of linear first-order differential equations,
one can determine analytically the complete time-dependent solutions, once the ini-
tial conditions are given. The important feature of these solutions is that they give
the response time of the atomic system τr . The atomic system can follow variations
of the pump power only if they happen on a time-scale that is not shorter than τr .
Optical amplifier. Consider the propagation of a monochromatic light beam, at
frequency ν close to ν21 , inside an amplifying medium made of four-level atoms.
96 3 The Laser
Assuming that the beam propagates along the z axis and that the amplifier extends
from z = 0 to z = L, the infinitesimal variation of the photon flux Φ over dz at z is
given by the following balance between stimulated emission and absorption:
dΦ = σ (ν)(N2 − N1 )Φdz. (3.41)
For simplicity of notation σ appears instead of σ12 .

In the small-signal regime the photon flux can be considered sufficiently weak
so as not to appreciably modify the populations of atomic levels. If N2 − N1 is
independent of z, then (3.41) becomes a linear equation, and its integration gives:
Φ(z) = Φo e g(ν)z , (3.42)
where Φo is the input flux at z = 0, and g(ν) = σ (ν)(N2 − N1 ) is the gain per unit
length. The total gain, calculated as the ratio output to input, is:
G = e g(ν)L (3.43)
Note that (3.41) does not contain the contribution of spontaneous emission,
because spontaneous emission is spread over all directions and all frequencies within
the transition bandwidth. As such, it can be assumed that very few photons are spon-
taneously emitted along z at frequency ν. Spontaneous emission is an intrinsic noise
source for the optical amplifier. Concerning noise, the difference between the elec-
tronic amplifier and the optical amplifier is that for an electronic amplifier hν is
smaller than k B T (i.e. the noise is essentially thermal), while for the optical ampli-
fier hν is much larger than k B T , so that thermal noise is negligible.
3.6 Scheme and Characteristics of the Laser
As discussed in Sect. 3.3, an oscillator is an amplifier inserted in a positive feedback

scheme. Therefore, in order to make a laser (i.e., an optical oscillator), one should
create a feedback loop in which some part of the amplifier output is returned to
the input. Two typical schemes are shown in Fig. 3.8. In the ring cavity the optical
beam follows a closed loop, propagating either clockwise or counter-clockwise. In
the two-mirror scheme, the two flat or curved mirrors are set up facing each other,
so that an optical wave can bounce back and forth between the two mirrors. Such
a cavity is known as Fabry-Perot cavity, because it has the same structure of the
Fabry-Perot interferometer treated in Sect. 2.4.4. Compared to the ring cavity, the
Fabry-Perot cavity is simpler, more compact, and has the advantage that, in a single
round-trip, the laser beam crosses twice the amplifier. Usually the amplifier has a
cylindrical geometry and the laser beam propagates along the axis of the cylinder.
One of the mirrors is partially transmitting, so that a fraction of the optical power
traveling inside the cavity is coupled out as the laser emission.
3.6 Scheme and Characteristics of the Laser 97
Fig. 3.8 Laser configurations. Mirrors 1 and 3 are totally reflecting, the laser output comes from
mirror 2 that is partially transmitting
Because of the finite transverse dimensions of the active medium and mirrors,
a light beam that travels even slightly off-axis will leave the cavity region after a
given number of passes between the two mirrors. Therefore, in order to be amplified
by many passes, the laser beam must travel very closely along the cavity axis. This
results in a highly directional laser output.
The laser threshold condition corresponds to the situation in which the loop gain is
equal to 1. Introducing an idealized signal, E(t), into the cavity, and then determining
how the signal changes after a single one-loop pass, allows the loop gain to be
deduced.
The calculation is performed by considering a Fabry-Perot cavity with a distance
L between mirrors, and an electric-field reflectivity of ρ1 and ρ2 for mirror-1 and
mirror-2, respectively. The amplifier is made of a material with refractive index n,
has a length l, and provides a gain G for the optical intensity (or photon flux). In
order to fully understand the behavior of the laser cavity it is necessary to consider
the light beam as a wave, possessing amplitude and phase, instead of a stream of
photons. Taking into account that the intensity is proportional to the modulus-square
of the field, and that a wave of angular frequency ω, traveling for a distance z in
a medium with index of refraction n, undergoes a phase shift of (ω/c)nz, then the
effect of the amplifier
√ is that of multiplying the electric field of the input wave by a
complex factor G exp[i(ω/c)nl]. By considering also the free propagation in air
over the distance L − l and the reflections on the two mirrors, the electric field E (t)
of the wave that has completed a round trip is given by:
E (t) = E(t)ρ1 ρ2 G exp[i(2ω/c)(L − l + nl)]. (3.44)
The threshold condition, that is the unitary loop-gain condition, is obtained by

putting E = E:
ρ1 ρ2 Gexp[i(2ω/c)(L − l + nl)] = 1. (3.45)
98 3 The Laser
Since the loop gain is a complex quantity, the threshold condition breaks down
into two conditions:
• The modulus of the loop gain must be equal to 1.
• The phase of the loop gain must be equal to 2qπ , where q is a positive integer.
In general, ρ1 and ρ2 are complex quantities, because, as described in Sect. 2.2,

the reflected field differs in both amplitude
√ and phase with√respect to the incident
field. Therefore, one may have: ρ1 = R1 exp(iφ1 ), ρ2 = R2 exp(iφ2 ), where R1
are R2 are the reflection coefficients for the optical power. Given that G = exp(gl),
the unity modulus condition determines the threshold value gth of the gain per unit
length inside the amplifier:

R1 R2 exp(gth l) = 1. (3.46)
The phase condition defines a discrete set of oscillation frequencies, νq , usually

called longitudinal modes. Taking L = L − l + nl as the optical path between
mirror 1 and mirror 2, the following expression is derived from (3.45):
c φ1 + φ2
νq = (q − ). (3.47)
2L 2π
This coincides with the relation (2.119), derived in Sect. 2.4.4 for the resonance
frequencies of the Fabry-Perot interferometer.
The frequency difference between two adjacent modes, Δν = νq+1 − νq , is:
c
Δν = . (3.48)
2L
As it would be expected, this relation reproduces (2.120). Note that 2L is the

optical path corresponding to one round trip inside the laser cavity, therefore 2L /c
is the time taken by light to make one round trip. Consider a numerical example: if
L = 7.5 cm, 2L /c = 0.5 ns, and Δν = 2 GHz.
Figure 3.9 qualitatively shows g(ν), which is proportional to the stimulated-
emission cross-section σ21 , and so is a function centered at ν21 , with a half-height
width Δν21 . The figure refers to a laser operating above threshold, because the peak
value of g(ν) is larger than gth . The laser has two longitudinal modes, corresponding
to two auto-frequencies, running in the cavity. In general, the laser will operate only
if at least one of the auto-frequencies νq belongs to the frequency interval over which
the above-threshold condition is satisfied. This will certainly happen if Δν is smaller
than Δν21 .
3.7 Rate Equations 99
Fig. 3.9 Gain per unit length as a function of frequency. A set of auto-frequencies is also shown
3.7 Rate Equations
The general discussion of Sect. 3.3 indicates that the steady-state behavior of the
oscillator above threshold is determined by the gain-saturation effects. In this section
a simple nonlinear laser model is discussed, in which saturation effects are included.
The model considers an amplifier working on a four-level scheme, inserted into
a Fabry-Perot cavity and operating on a single longitudinal mode. The treatment is
greatly simplified by assuming that the dynamic behavior of the laser is determined
by only two time-dependent quantities, the total number of photons inside the cavity,
Q(t), and the population of the upper laser level N2 (t). The implicit assumption is
that level 3 has a very fast spontaneous decay toward level 2, and level 1 is rapidly
decaying toward level 0, so that N1 is close to 0 and N2 can be taken to coincide with
the population inversion. The nonlinear equations expressing the balance between
gain and losses are the following:
dQ Q
= V B Q N2 − (3.49)
dt τc
d N2 N2
= W − B Q N2 − , (3.50)
dt τ
where B is proportional to the stimulated emission coefficient, V is the volume
occupied by the laser mode inside the active medium and W is the number of atoms
brought to level 2 by the pump per unit time and volume. W is proportional to the
pump power Pp . The time constant τ is the spontaneous decay time of level 2, and
τc represents the decay time of the photon number inside the cavity in absence of
pump. In general, τc , sometimes called photon lifetime of the laser cavity, depends
100 3 The Laser
on all cavity losses, including, besides the useful loss of the semi-transparent output
mirror, also losses due to diffraction, to misalignments and to defects of the optical
components. If the only important loss is due to the output mirror having reflectance
R, then τc is expressed as:
2L
τc = . (3.51)
c(1 − R)
The two rate equations have a very simple meaning. Equation (3.49) indicates
that the total number of photons inside the cavity increases because of stimulated
emission, but decreases because of losses. Equation (3.50) says that the number of
atoms on the upper level N2 increases because of the pumping process, but decreases
because of the emission processes.
The steady-state solutions, Q s and N2s , are obtained by putting the time-
derivatives equal to 0. In general, nonlinear systems may admit more than one steady-
state solution. Indeed, in the case of (3.49) and (3.50), it is found that two solutions
exist:
Qs = 0
N2s = W τ, (3.52)
and
Q s = V τc (W − Wth )
N2s = (V Bτc )−1 , (3.53)
where
Wth = (V Bτ τc )−1 . (3.54)
By performing a stability analysis on the pump parameter W , it is possible to

determine which of the solutions is stable and therefore has a real physical basis.
The solution (3.52) is stable, provided that W ≤ Wth . In this situation, the value of
the pump parameter is too small to compensate for cavity losses. The loop gain is
then smaller than 1 and so the laser is below threshold. Note that below threshold the
laser cavity contains the photons due to spontaneous emission processes, and even
if Q s is small, it will not be exactly zero. The mathematical result of Q s = 0 is due
to the fact that the contributions of spontaneous emission do not appear in the above
simplified rate equations.
As the parameter W increases, there is an exchange of stability between the two
solutions. The solution (3.53) is stable, provided that W > Wth , i.e., if the loop gain
is larger than 1, meaning that the laser is above threshold.
The solution (3.53) also indicates that, when the laser is above threshold, the
population of the upper level does not grow as W is increased, but remains clamped
to the threshold value: this is consistent with the fact that in the steady-state regime
the saturated loop-gain should be equal to 1. Below threshold the pump power is
3.7 Rate Equations 101
Fig. 3.10 Laser output

power as a function of pump
power. Ppth is the threshold
pump power
completely used to populate the upper level of the atomic transition until the loop
gain inside the cavity approaches 1. Above threshold the fraction of pump power that
exceeds the threshold value is converted into laser power.
With the assumption that cavity losses are due only to the transmission through
the output mirror, the laser output power is given by:
Qs
Ps = hν . (3.55)
τc
The behavior of Ps as a function of the pump power is presented in Fig. 3.10.

An important parameter is the laser efficiency, defined as the ratio between output
optical power and pump power:
Ps
η= . (3.56)
Pp
For the first laser, i.e., the ruby laser, η was of the order of 10−3 . As it will be seen,
there are lasers for which η can approach 0.5.
As a final comment to this section, it should be emphasized that the treatment has
implicitly assumed that the active medium is made of a dilute set of atoms. In other
words, the atoms are sufficiently far apart one from the other so that the energy level
structure of the isolated atom is not modified. Most existing lasers are consistent
with this assumption, but there are exceptions, the most important of which are
semiconductor lasers. In any case, even if some change in the structure of the rate
equations needs to be introduced, the main conclusions drawn from the discussion
of (3.49) and (3.50) remains largely valid for all types of lasers.
3.8 The Laser Cavity
The two finite-size mirrors of the Fabry-Perot laser constitute an open cavity that
is intrinsically lossy, because of diffraction. A general property of optical cavities
is that there is a discrete number of possible field configurations, known as cavity
modes. In this section the discussion is limited to Fabry-Perot cavities made with
102 3 The Laser
Fig. 3.11 Laser cavity made by two spherical mirrors at distance L
spherical mirrors. A very remarkable result of the treatment is that, within the paraxial
approximation, the cavity modes are Gaussian spherical waves.
The cavity, as shown in Fig. 3.11, is made by two spherical mirrors at distance
L. The radii of curvature of the mirrors, called ρ1 and ρ2 , are considered positive if
the mirror concavity is directed inwards in the cavity. Gaussian spherical waves are
possible candidates as cavity modes because their shape is invariant in free propaga-
tion and in reflection from a spherical mirror, as discussed in Sect. 1.4.2. The general
condition ensuring that the wave reflected by a spherical mirror retraces back exactly
the incident wave is that the radius of curvature of the wavefront of the incident wave
coincides with the radius of curvature of the spherical mirror surface. If such a con-
dition can be satisfied for both mirrors, the Gaussian spherical wave is then reflected
back and forth between the two mirrors without a change in the transverse profile.
By assuming that the beam waist inside the cavity is at distance z m from mirror 1,
and recalling that the z-dependence of the radius of curvature of the wavefront is
given by (1.39), the two conditions are:
zR2 zR2
zm + = ρ1 ; L − z m + = ρ2 (3.57)
zm L − zm
The two unknown quantities inside (3.57) are z m and z R , or z m and wo , if one
recalls that wo is related to z R by (1.38). Introducing the two parameters g1 =
1 − L/ρ1 and g2 = 1 − L/ρ2 , the solutions of (3.57) are expressed as follows:
√
g2 (1 − g1 ) g1 g2 (1 − g1 g2 )
zm = L ; zR = L. (3.58)
g1 + g2 − 2g1 g2 g1 + g2 − 2g1 g2
Noting that, for a symmetrical cavity, g1 = g2 , one finds from (3.58) that z m =
L/2, which means that the waist of the Gaussian spherical wave is in the center of
the cavity.
3.8 The Laser Cavity 103
The beam radii w1 and w2 at the two mirrors are given by:

Lλ g2 Lλ g1
w1 2 = ; w2 2 = . (3.59)
π g1 (1 − g1 g2 ) π g2 (1 − g1 g2 )
Up to this point, only the spatial dependence of the mode was considered. The
second step is that of determining the resonance frequencies, by using the requirement
that the complete round-trip phase delay be a multiple of 2π . The calculation must
take into account that the Gaussian spherical wave acquires during propagation an
additional phase delay ψ(z), given by (1.37), with respect to the plane wave. The
resonance condition reads:

2π νq
2 L − ψ(z m ) + ψ(L − z m ) + (φ1 + φ2 ) = 2qπ, (3.60)
c
where q is a positive integer, and φ1 , φ2 are the phase shifts due to the reflections
from the mirrors. By combining (3.58) and (3.60), it is finally found:

√
φ 1 + φ2 arccos g1 g2 c
νq = q − + (3.61)
2π π 2L
If compared with (3.47), (3.61) shows that the Gaussian wave approach produces a
rigid shift of the resonance frequencies found with the plane wave treatment, keeping
unchanged the frequency difference between adjacent resonances.
Equation (3.61) gives real resonance frequencies only if the condition
0 ≤ g1 g2 ≤ 1 (3.62)
is satisfied.
The condition (3.62) can be conveniently displayed in the g1 –g2 plane, as shown
in Fig. 3.12. The shaded areas in the diagram of Fig. 3.12, in which the condition
(3.62) is fulfilled, correspond to stable cavities, and the clear areas, in which (3.62) is
violated, correspond to unstable cavities. Unstable cavities are very lossy, therefore
lasers normally use stable cavities.
It is instructive to discuss a numerical example. Considering a stable symmetric
cavity with g1√g2 = 0.5, the beam waist of the Gaussian wave, calculated from (3.58),
is: wo ≈ 1.1 Lλ/π. Therefore, the Gaussian wave emitted √ by such a laser cavity
has a divergence angle given by: θo = λ/(π wo ) ≈ 0.9 λ/π L. Taking, for instance,
λ = 500 nm and L = 10 cm, it is found: wo ≈ 0.14 mm and θo ≈ 1.14 × 10−3 rad.
It should be recalled that the full theory based on the paraxial approximation
predicts the existence of generalized Gaussian modes presenting a more complicated
transverse field distribution, as the Hermite-Gauss modes described in Sect. 1.4.2.
The presence of the higher-order transverse modes in the laser output is considered in
almost all cases a nuisance. It is a general fact that the higher-order transverse modes
extend farther out in the transverse direction and have more of their energy at greater
104 3 The Laser
Fig. 3.12 A g1 –g2 stability diagram for a general spherical laser cavity
distances from the axis than does the Gaussian mode. Suppose the transverse size
of a cavity is defined by a limiting aperture of diameter 2a, which can be the mirror
(or the laser rod) diameter or even an iris placed inside the cavity for mode-control
purpose. All the mode losses will increase as the aperture radius a is reduced toward
the mode size w. This loss will be larger for the higher-order modes because of their
larger transverse spread. The single-mode condition can be expressed by stating that
the quantity N F = a 2 /(λL), known as the Fresnel number of the cavity, should have
a value around 1. In the common terminology, a laser operating on a single transverse
mode, i.e., emitting a Gaussian wave, is said to work on the TEM00 mode, where
TEM stays for “transverse electromagnetic”.
3.9 Solid-State and Gas Lasers
It has been shown in the previous sections that it is possible to obtain optical ampli-
fication by exciting specific energy levels of atoms. Ions, which are atoms that have
lost or gained one or more electrons, can be also candidates for laser action.
Molecules have an internal energy due not only to their electronic structure, but
also to their vibrational and rotational motion. As predicted by quantum mechanics,
molecules present discrete vibrational and rotational energy levels, and can make
transitions among these levels by absorbing or emitting electromagnetic radiation.
While lasers based on electronic transitions emit in the visible or near infrared,
molecular lasers typically emit in the middle infrared, because the energy jumps
associated with vibrational and rotational transitions are of the order of 0.1 and
0.01 eV, respectively.
3.9 Solid-State and Gas Lasers 105
Fig. 3.13 Ruby laser:

simplified scheme of the
energy levels of the ion Cr 3+
There are also lasers utilizing transitions that are both electronic and vibrational
(termed “vibronic”). These lasers are interesting because they can be tuned to emit
within a rather wide band of wavelengths.
In this section, only solid-state and gas lasers are considered. The descriptions of
semiconductor lasers and fiber-optic lasers are deferred to Chaps. 5 and 6, respec-
tively.
Solid-state lasers. The term solid-state laser indicates a laser using a dielectric crystal
(or a glass) doped with a small quantity of impurities as the active medium. The first
laser ever made (i.e., the ruby laser in 1960) was a solid-state laser. Ruby is a piece
of Al2 O3 crystal, in which a small fraction of Al 3+ ions are replaced by chromium
ions. In the ruby crystals used as active media the percentage of Cr2 O3 is typically
0.05 % by weight. The pure Al2 O3 crystal, called corundum, by itself is perfectly
transparent in the visible. It appears red when doped with chromium, because Cr 3+
ions have two absorption bands in the visible range, one in the green and the other
in the violet, as shown in the simplified scheme of Fig. 3.13.
The ruby laser is a three-level laser: level 0 is the ground level, level 1 is the
first excited level, level 2 is represented by the two absorption bands. The pump
power comes from a lamp, emitting light on a broad spectrum that includes the green
and violet wavelengths required to excite the chromium ions to level 2. Level 1 is
populated by spontaneous decay processes from level 2. The lifetime of level 2 is
of the order of picoseconds, while that of level 1 is of the order of milliseconds.
As shown in Sect. 3.5, it is possible to create population inversion between 1 and 0
under such conditions. The fact that level 2 is made by two relatively broad bands
instead of a sharp energy level is very helpful, because it ensures a larger absorption
of the broad-band light emitted by the lamp. The energy difference between excited
and ground level is E 1 − E 0 = 1.79 eV, corresponding to red light with wavelength
λ = 0.6943 µm.
The active medium is a cylindrical monocrystal with a length of a few centimeters
and diameter of a few millimeters. In the first laser the two end faces of the ruby
bar were made flat and parallel, and were coated with a thin silver layer, in order
to create cavity mirrors. The pump light was provided by a helical xenon flashtube
surrounding the active rod. The ruby laser has a high threshold power, a low efficiency
(typically, η ≈ 0.1 %), and exhibits large fluctuations in output power. Consequently,
although it is of historical and illustrative importance, the ruby laser is no longer used
in modern applications.
106 3 The Laser
Following the example of the ruby laser, many different solid-state lasers have
been developed, mainly by choosing active ions that belong to the transition metals
(in particular, Cr, Ti), or rare earths (in particular, Nd, Er, Yb). A common feature
of such ions is that they have excited levels with very long decay times, termed
metastable levels. These metastable levels make it easy to build up the population
inversion needed for optical amplification.
By doping the Al2 O3 crystal (which is also known as sapphire) with titanium
instead of chromium, a titanium-sapphire laser material can be produced. When
pumped by green light, this laser is tunable over a very broad wavelength range,
from 700 to 1000 nm, by exploiting a vibronic transition.
The most important solid-state laser is the neodymium laser, in which the host
medium is often a YAG (yttrium aluminium garnet, or Y2 Al5 O12 ) crystal, in which
some of the Y 3+ ions are replaced by N d 3+ ions. Besides YAG, other host media,
such as yttrium vanadate (Y V O4 ), can be used.
The simplified scheme of the energy levels of the ion N d 3+ in YAG is shown in
Fig. 3.14. The main difference with respect to the ruby laser is that the laser transition
involves two excited levels, that is, one deals with a four-level scheme. The two
absorption bands that play the role of level 3 are centered on the wavelengths 0.73
and 0.81 µm. These bands are coupled by a fast non-radiative decay (A32 ≈ 107 s−1 )
to level 2, which is the upper level of the laser transition. The spontaneous decay
time from 2 to 1 is of the order of hundreds of microseconds, much longer than
the spontaneous decay time from 1 to 0. The laser emission is in the infrared, at
λ = 1.064 µm. The structure of the Nd:YAG laser is similar to that of the ruby laser.
The Nd:YAG laser has a much lower threshold power and a much larger efficiency
(η around 1–3 %) when compared to the ruby laser. While the ruby laser can only
work as a pulsed laser, the Nd laser can operate both as a continuous wave (CW)
laser with output power up to hundreds of watts, and a pulsed laser with a pulse
repetition rate of 10–100 Hz. The half-height gain bandwidth is around 200 GHz,
and so the laser operates simultaneously on many longitudinal modes, when it has a
Fabry-Perot cavity with a length of a few centimeters.
In the early times of solid-state lasers, the pump light was provided by electric
discharge lamps. The modern trend is to use, instead of lamps, another laser, the
semiconductor laser. In the case of the Nd laser, the optimum pump source is the
Fig. 3.14 Neodymium laser:

simplified scheme of the
energy levels of the ion
N d 3+
3.9 Solid-State and Gas Lasers 107
AlGaAs semiconductor laser, which emits at 0.81 µm, and is in resonance with one
of the absorption lines of the N d 3+ ion. The important advantage in such a scheme
is that all the pump power can be used to create the population inversion, while less
than 50 % of the lamp pump power has a wavelength overlap that drives the excitation
from level 0 to level 3. Another important advantage is that the collimated light from
the semiconductor pump laser can be easily directed into the target laser medium.
This is not the case for discharge lamps, because they emit over the whole solid
angle, making it difficult to fully harness all the pump power available. Typically a
Nd laser pumped by a semiconductor laser has an output power of about 30 % the
input pump power.
The Nd laser is useful in a great many applications, including materials processing
(drilling, welding, marking of industrial products, etc.), optical radar, and biomedical
applications. It can also be used to create green laser light, through a process known
as second-harmonic generation, as it will be discussed in Chap. 7.
A useful property of rare earth ions (Nd, Er, Yb, etc.) is that they can provide
optical amplification even when randomly dispersed in a transparent glass matrix,
instead of a highly ordered crystal. A particularly important exploitation of this
property is represented by the development of fiber-optic amplifiers and oscillators,
as will be discussed in Chap. 6.
Gas lasers. Shortly after the invention of the ruby laser, it was shown that it is also
possible to have population inversion in a gaseous medium. Gas lasers are almost
always pumped using an electric discharge inside the active medium. The gas is
typically contained inside a capillary tube. If the voltage applied across the tube is
sufficiently large, a fraction of the gas atoms (or molecules) becomes ionized, and the
gas, which is an insulator at rest, becomes a conductor of electric current. The free
electrons, accelerated by the external electric field, collide with the atoms, generating
a great variety of excited states. By choosing the gas mixture and pressure correctly,
along with the value of the electric current density, laser oscillation can be achieved
in many gaseous media.
The first gas laser, which also was the first CW laser (Bell Telephone Laboratories,
1960), was the He-Ne laser, made using a transition between two excited levels of the
neon atom, emitting infrared radiation at 1.15 µm. The capillary tube is filled with a
mixture of helium and neon atoms, typically in the ratio 10 : 1, at a pressure around
one hundredth of an atmosphere. The helium atoms excited by electronic collisions
selectively transfer their excitation energy to neon atoms, generating population
inversion on certain neon transitions. Besides operating at 1.15 µm, the He-Ne laser
can emit at 0.633 and 3.39 µm. The most interesting wavelength for applications is
the red emission. Commercial He-Ne lasers typically have a cavity length between
10 and 100 cm, and an output power between 1 and 50 mW. The output beam has a
very stable pointing direction and a very small angular spread—the divergence angle
is typically 10−4 rad. The laser can be frequency-stabilized, in order to provide a
monochromatic beam with a linewidth of the order of 1 kHz. During a few decades
following its invention, the He-Ne laser was the laser of choice for those applications
requiring low power and high beam-stability, such as bar-code reading, telemetry,
108 3 The Laser
velocimetry, and alignment in civil and mechanical engineering. In most of these

applications the He-Ne laser has now been replaced by the smaller and cheaper
semiconductor laser.
An important gas laser is the carbon dioxide (CO2 ) laser, emitting in the middle
infrared at the wavelength λ = 10.6 µm, corresponding to a transition between
two excited vibrational levels of the CO2 molecule. The active gas mixture, made
of carbon dioxide, nitrogen and helium, is pumped by an electric discharge, which
can be longitudinal (electric current flowing along the axis of the optical cavity) or
transversal. The gas mixture is usually flowed along the laser tube in order to refresh
the active medium. The structure of the CO2 laser depends on the desired output
power. The most common configuration utilizes a longitudinal electric discharge
and a slow longitudinal gas flow. Cooling the lateral walls of the laser tube with a
water circulation system provides the necessary heat dissipation. With a laser tube
1 m long, filled at a pressure of 2.5 × 103 Pa, a continuous output power of the order
of 100 W can be produced. By replacing the slow gas flow with a supersonic flow
(which also acts as an effective means of heat removal) the output power can be
increased to 1 kW. A even higher output power of up to 20 kW can be reached by
increasing the pressure to ≈ 1.3 × 104 Pa and using a transversal electric discharge
and transversal gas flow. In some cases, a radio-frequency electric discharge (at 30–
50 MHz) is used instead of a continuous discharge. This has the advantage that no
electrodes inside the laser tube are needed. The CO2 laser has a very high efficiency,
typically η ≈ 20 %. This is an extremely important property for a high-power laser
used in industrial applications. Indeed the main applications of the CO2 laser are in
the area of materials processing. Surgical applications have also been tested, using
the laser as a lancet to exploit the fact that biological tissues strongly absorb 10.6-µm
radiation.
Another useful class of gas lasers is the excimer laser, which emits powerful
pulses of ultraviolet light. In this laser a pulsed electric discharge creates excited
bi-atomic molecules formed by a noble gas element (argon or krypton) and a halo-
gen element (fluorine or chlorine). The molecule is formed because the chemical
behavior of an excited noble gas atom is similar to that of an alkali atom. The excited
molecules dissociate by emitting ultraviolet light. For instance, the argon-fluorine
(Ar-F) excimer laser can emit 10–100 ns pulses at 193 nm, with a repetition rate of
hundreds of H z and an average output power of several watts. The excimer laser has
very important applications in the fields of photolithography for microelectronics
and ophthalmology for the myopia treatment.
3.10 Pulsed Lasers
Some of the most important applications of lasers involve pulsed operation. There
are different methods of obtaining laser pulses. One method is to pulse the pump
driving the laser emission, as it is the case for excimer lasers. Another possibility is
to use an external switch that transmits the output of a CW laser for short periods of
time. Naturally, the minimum pulse duration that can be obtained with this approach
3.10 Pulsed Lasers 109
cannot be shorter than the switch response time, and the peak power of the pulse
cannot exceed the steady-state power of the CW laser. These limitations can be
overcome, as will be shown, by inserting an internal switch or modulator directly
into the laser cavity.
3.10.1 Q-Switching
The technique of “Q-switching” is used to generate short and powerful light pulses
from lasers that have an upper-level spontaneous decay time τ much longer than the
photon decay time of the laser cavity, i.e., τ τc .
The principle is as follows: an optical switch is inserted into the cavity, before one
of the mirrors. When the pump is switched on and the switch is closed introducing
extra losses, the population inversion can build up to a very high value, but does not
reach the oscillator threshold. When the population inversion reaches its peak value,
the switch is suddenly opened, and the loop gain of the cavity is almost instantly
boosted well above the oscillation threshold. This results in an extremely rapid build-
up of the photon number and a subsequent depletion of the population inversion by
stimulated emission. This process converts most of the pump energy stored by the
excited atoms into the emission of a short and powerful laser pulse.
The name Q-switching refers to the quality factor Q of the laser cavity, which is
a figure of merit quantifying the cavity losses, and is defined as:
angular frequency × energy stored

Q= (3.63)
power dissipated.
Applying this general definition to the laser case, it is found that Q = ω21 τc . The
process above described changes suddenly the state of the laser cavity from large
losses (short τc , that is, low Q) to small losses (long τc , that is, high Q), hence the
name Q-switching.
The transient behavior following the switch-on operation at time t = 0 can be
derived by solving the two time-dependent rate Eqs. (3.49) and (3.50), with appro-
priate initial conditions. The initial number of excited atoms, N2 (0), is calculated
from (3.50) under the assumptions that (i) there are no photons in the cavity and (ii)
the time derivative is equal to 0. This gives: N2 (0) = W τ . If the initial number of
photons is actually set to zero, then the system cannot dynamically evolve. However,
in practice there are always some initial photons present in the cavity, due to sponta-
neous emission. A typical transient, obtained by choosing Q(0) = 1, is sketched in
Fig. 3.15. The upper curve shows that the pumping process makes the population of
the upper level to grow much larger than the threshold value when the optical switch
is closed. The cavity Q is switched on at t = 0. Then the number of photons begins
to grow, leading to a pulse with a characteristic half-height duration, τ p , of the order
of τc . As a result of the growth in photon number, the population of the upper state
will decrease from its initial value to a very low final value.
110 3 The Laser
Fig. 3.15 Time evolution of the population of the upper laser level (upper diagram) and of the
laser output power (lower diagram) in a Q-switched laser. The Q-switching operation is performed
at t = 0
In order to derive the order of magnitude of the peak power, it can be assumed
that the total number of photons inside the pulse coincides with the total number of
excited atoms at t = 0, V N2 (0), leading to a peak output power estimate of:
V Wτ
Ppeak = hν . (3.64)
τc
Recalling that the steady-state output power Ps is given by (3.55), and assuming,
for simplicity, that W Wth , the ratio of peak power to steady-state power in the
emission is:
Ppeak τ
= . (3.65)
Ps τc
Equation (3.65) clearly shows that the Q-switching operation is only advantageous
for active media having a long spontaneous decay time of their excited levels, such
as solid-state and CO2 lasers. If one takes τ = 300 µs and τc = 10 ns for the case
of a Nd-YAG laser, then the power ratio can be up to 3 × 104 . A laser designed for
an average output power of 1 W can have a Q-switching pulse (sometimes called a
“giant pulse” in the literature) with a peak power of 30 kW.
The optical switch used in Q-switching can be realized exploiting the electro-optic
or acousto-optic effects later described in Chap. 4. Another interesting possibility,
frequently utilized, is the inclusion of a saturable absorber in the optical cavity

for “passive” Q-switching. The absorber can be modeled as the two-level system
described in Sect. 3.4, with an intensity-dependent absorption. The absorption loss
is high in the early stages of the oscillation build-up, because the slowly increasing
intensity is not large enough to saturate the absorption. However, once the intensity
in the cavity exceeds a certain level (the saturation level, see (3.17)), the loss quickly
decreases, and the resulting effect is similar to that of a sudden opening of an optical
switch.
3.10.2 Mode-Locking
Q-switching produces pulses with a duration ranging from a few nanoseconds to

hundreds of nanoseconds. Another technique, known as “mode-locking”, is capable
of producing pulses on the femtosecond and picosecond scale. Whereas Q-switching
can work equally well with lasers in single-mode or multi-mode operation, mode-
locking requires a large number of longitudinal modes in the cavity.
Consider a CW laser oscillating on N longitudinal modes. The frequency differ-
ence between adjacent modes is Δν. For a Fabry-Perot cavity, Δν is given by (3.48).
The electric field associated to the laser beam is written as:

N
E(t) = E k exp[−i(ωk t + φk )], (3.66)
k=1
where the amplitudes E k are real and time-independent. The laser intensity, which
is proportional to |E(t)|2 , is the sum of the intensities of the N modes plus the sum
of N (N − 1) interference terms, each having zero time-average. If the phases φk are
random and mutually uncorrelated, then the interference terms tend to cancel each
other, such that:

N
I (t) = Ik + ΔI (t), (3.67)
k=1
where ΔI (t) is a small randomly fluctuating term.

The mode-locking operation consists in introducing a fixed relation among the
phases of different modes, expressed by:
φk − φk−1 = φo , (3.68)
where φo is an arbitrary constant. To simplify the algebra, it is assumed that N is an

odd number and that the frequency of the central mode coincides with the central
frequency ν21 of the gain band. Since N is odd, it can be written as: N = 2n + 1.
112 3 The Laser
By calling m the integer number that takes all values in the interval −n ≤ m ≤ n,
assuming equal amplitudes, E m = E o , and using (3.68), (3.66) then becomes:

n
E(t) = E o exp{−i[(ωo + mΔω)t + mφo ]} = A(t)E o exp(−iωo t), (3.69)
m=−n
where

n
A(t) = exp[−im(Δωt + φo )]. (3.70)
m=−n
It is convenient to introduce the new temporal variable t , connected to t by the

relation Δωt + φo = Δωt . To use t instead of t is equivalent to shifting the time
axis by the quantity φo /Δω. The summation in the right-hand member of (3.70) is a
geometric progression with common ratio exp(iΔωt ). It is found:

n
sin[(2n + 1)Δωt /2]
A(t ) = exp(−imΔωt ) = . (3.71)
−n
sin(Δωt /2)
The time-dependent intensity is derived from (3.71) by recalling that I (t ) is

proportional to E o 2 A2 (t ). As an example, the function I (t ) is plotted in Fig. 3.16,
for N = 11 and Δν = 100 MHz. The intensity profile is a periodic sequence of
pulses, separated by a period of:
2π 1
T = = . (3.72)
Δω Δν
Fig. 3.16 Temporal behavior of the output intensity of a mode-locked laser in the case of an equal-
amplitude mode spectrum. In this numerical example 11 modes with a frequency separation of
100 MHz are considered
The peak intensity, I p , is obtained by calculating A2 (t ) in the limit of t tending

to 0. The result is:
I p = N 2 Io = N I L , (3.73)
where Io is the intensity of one mode and I L = N Io is the average laser intensity.
It is important to note that the pulsed behavior is the consequence of an inter-
ference effect, As such, mode-locking does not change the average power of the
emission, it simply re-distributes the power non-uniformly in time. Therefore, the
energy contained in a single pulse must be equal to that of the integrated power
emitted by the CW laser over a period T . Consequently, the pulse duration τ p can
be determined from the following equality:
T
τ p I p = T N IL τp = . (3.74)
N
Recalling that T is the reciprocal of Δν and that N = Δν21 /Δν, the very important
result is found that τ p = (Δν21 )−1 , regardless of the cavity length. As an example,
the N d-Y AG laser has a typical gain bandwidth of 200 GHz, and so the minimum
pulse duration that can be obtained from mode-locked operation is 5 ps. For a Nd-
glass laser, the bandwidth is one order-of-magnitude broader than that of a N d-Y AG
laser, and so its minimum pulse duration is of the order of 500 fs. The pulse duration
can be reduced by another order of magnitude by mode-locking the Ti-sapphire laser.
The Fourier transform of the emitted field, E(ω), can be described as the product
of an infinite comb of Dirac delta functions with a rectangle function:
∞

ω−ω
o
E(ω) = E o r ect 2 δ(ω − ωk ) (3.75)
Δω12
k→−∞
where r ect (x) is a function which takes value 1 if −0.5 ≤ x ≤ 0.5 and 0 if |x| > 0.5.
Recalling that the inverse Fourier transform of the product of two functions of ω is
the convolution of the two anti-transforms, and that the Fourier anti-transform of r ect
is the sinc function, it can be verified that the anti-transform of (3.75) is precisely
given by (3.69). The interesting aspect of this Fourier approach is that it allows for
a general treatment to the more realistic case of unequal mode amplitudes. Indeed
the typical situation would be that in which the amplitudes of E k are distributed
according to a bell-shaped function, with maximum at k = 0. By using a Gaussian
distribution, E(ω) can be expressed as:
∞

(ωo − ωk )2
E(ω) = E o exp δ(ω − ωk ). (3.76)
k→−∞
Δω12 2
It is apparent that the anti-transform of (3.76) gives an E(t) that consists in a

periodic sequence of Gaussian pulses, each having a duration inversely proportional
114 3 The Laser
to the bandwidth. It should be noted that the secondary maxima that are present in the
intensity plot of Fig. 3.16 disappear when a Gaussian distribution of mode amplitudes
is used instead of the rectangle function.
Mode-locking can be achieved in various ways. One method is to internally mod-
ulate the losses of the laser at a frequency which is equal to the intermode frequency
spacing Δν. With this approach the laser behavior can be intuitively understood as
a self-adjusting process: if losses are minimal at times to , to + T , etc., in order to
maximize the efficiency the laser will prefer to emit pulses at to , to + T , etc., instead
of operating at constant intensity. The loss modulation can be obtained by inserting
into the laser cavity an electro-optic or acousto-optic modulator.
The description of the N -mode laser as a collection of N independent oscillators
is a rather rough approximation. In practice, even without any external intervention,
the modes are interacting because they are driven by emission from the same active
atoms. Generally speaking, the existence of interactions among the modes tends
to favor the process of mode-locking. What prevents the laser from spontaneously
operating in a mode-locked regime is the fact that real longitudinal modes are not
exactly equally spaced in frequency. The optical path appearing in (3.47) contains the
index of refraction of the active medium, which is a frequency-dependent quantity. As
a consequence, (3.47) should be viewed as an implicit expression for the resonance
frequencies of the laser cavity, producing a spacing that is slightly different for each
pair of adjacent modes. Only if the modulation depth exceeds a threshold value, the
amplitude (or phase) modulation forces the modes to become equally spaced and
locks the mode phases.
Mode-locking can also be achieved by passive methods. A frequently used
approach is to insert a saturable absorber in the cavity. Initially, the multimode laser
has random intensity fluctuations, but the saturable absorber selectively transmits
the fluctuations of high intensity, while absorbing those of low intensity. After many
passes through the cavity, only a single high-intensity pulse is formed. In this scheme,
it is essential to use a saturable absorber with a recovery time much shorter than the
period T . Semiconductor saturable-absorber mirrors (SESAMs) can also provide
passive mode-locking. Furthermore, as it will be explained in Chap. 7, spontaneous
mode-locking can be achieved in solid-state lasers by exploiting the nonlinear optical
behavior of the active medium. This method is called Kerr-lens mode locking.
Ultrashort pulses of high peak power can be obtained by simultaneously operat-
ing in the same laser cavity mode-locking and Q-switching. In such a situation the
obtained mode-locked train has an envelope that follows the time evolution of the
Q-switching pulse. As an example, Fig. 3.17 illustrates the result of an experiment
in which a passively mode-locked diode-pumped N d-Y V O4 laser is Q-switched
by an acousto-optic switch. The output train is made up of about 20 mode-locked
pulses, separated by a period of 5.6 ns corresponding to the cavity round trip time.
The measured width of the pulses was about 9 ps.
3.11 Properties of Laser Light 115
Fig. 3.17 Output intensity of a simultaneously mode-locked and Q-switched N d-Y V O4 laser
3.11 Properties of Laser Light
Laser light differs from the light emitted by conventional sources in many aspects.
The most relevant properties of laser emission are reviewed below.
3.11.1 Directionality
Laser radiation is characterized by a high degree of directionality. While traditional

light sources emit in all directions and with a wide frequency range, the laser (as an
oscillator) has a built-in selection mechanism for frequency and direction. The origin
of the directionality can be intuitively understood, if one considers that only rays
traveling back-and forth between the cavity mirrors, along the cavity axis, are strongly
amplified by the active medium. As discussed in Sect. 3.8, the spatial properties of
laser beams can be described in terms of Gaussian waves. If the cavity supports
several transverse modes, then the fundamental mode TEM00 , can be selected by
placing an aperture within the cavity. The TEM00 laser emission has an angular
spread θo typically in the range 10−4 − 10−3 radians.
The beam can be further collimated by a telescope consisting of a pair of two
converging lenses sharing a common focus. If the focal length of the first lens is
f 1 and that of the second lens is f 2 > f 1 , then the beam diameter is increased by
the factor f 2 / f 1 , and, consequently, the beam divergence is decreased by the same
factor.
116 3 The Laser
In many applications, such as writing and reading optical memories, a laser beam
is focused to a very small spot size. Using a lens with focal length f , the spot size
in the focal plane is approximated by w f ≈ θo f . Taking θo = 10−4 and f = 1 cm,
it is found that w f = 1 µm. This shows that the small divergence of a laser beam
enables the concentration of the output power to an area having the size of the optical
wavelength. This is useful for many applications. In the case of materials processing,
the intensity at focus can be easily large enough to strongly heat the material, and
locally produce melting or even vaporization.
3.11.2 Monochromaticity
There are applications, such as optical communications, in which it is essential to

have a single-frequency optical source. This requires suppression not only of higher-
order transverse modes, but also of all-but-one longitudinal modes. In almost all
cases, the bandwidth of the frequency-dependent gain profile Δν21 is much larger
than the frequency spacing between adjacent longitudinal modes, Δν, and so the laser
has a natural tendency to operate with multiple modes. Single-mode behavior can
be enforced by using compound cavities or, more effectively, by using a frequency-
selective feedback mechanism. An example of this approach will be given in Chap. 5,
where single-mode semiconductor lasers will be discussed.
The emission frequency of a single-mode laser, ν L , always has fluctuations orig-
inated by external disturbances. Since ν L is linearly dependent on the optical path
L between the two laser mirrors, the frequency stability of the laser depends on
the stability of L . As shown in Sect. 2.4.4, to shift by Δν the mode frequency, it is
sufficient to change the mirror separation by half wavelength. This means that small
variations in the mirror separation can induce large changes of the laser emission
wavelength. Therefore, in order to minimize the wavelength fluctuations of the single-
mode laser, the cavity length must be kept fixed by rigid construction, with isolation
from mechanical vibrations, and also with temperature control. In a stable single-
mode laser the linewidth δν L of the emitted radiation can be of the order of 1 MHz,
and even lower values of 1 kHz can be achieved using active frequency-stabilization
methods.
A useful feature of some types of lasers is the possibility to tune the emission
wavelength without changing the active medium, but simply acting on an optical
component inside the cavity.
It should be noted that, in principle at least, one can filter both the direction and
frequency of light emitted by a conventional source, in order to achieve a divergence
and bandwidth comparable to a laser source. However, the resulting beam would
be not only extremely weak, but also characterized by large amplitude and phase
fluctuations.
3.11 Properties of Laser Light 117
3.11.3 Spectrum of Laser Pulses
The light emitted by conventional optical sources contains many frequency com-
ponents. Each component has a random phase, uncorrelated with that of the other
components. In the laser case, it is possible to have a multi-frequency emission in
which the phases of different frequency components are mutually correlated, as it has
been shown in Sect. 3.10.2. This is very important because it allows for the possibil-
ity of optical signals having a pre-determined time-dependent (and space-dependent)
amplitude and phase, analogous to what can be done in the range of radio-frequencies
or microwaves with an electronic function generator.
In this section the relation between pulse duration and spectral width is derived
by considering a specific pulse shape, that of the time-dependent field amplitude
described by a Gaussian function. Assuming that the intensity peak is at t = 0, the
electric field can be expressed as:

t2
E(t) = Ao exp − exp(−iωo t), (3.77)
2τi 2
where ωo is the central frequency, and τi is the time delay at which the intensity is
reduced by a factor 1/e with respect to the peak value. The pulse duration τ p , defined
as the width at half-height of the intensity peak, is related to τi as follows:
√
τ p = 2 ln2τi = 1.67τi . (3.78)
Recalling the definite integral:

∞ √
π
exp(−a 2 t 2 )dt = , (3.79)
0 2a
the Fourier transform of (3.77) can be immediately calculated, obtaining:

√ τi 2 (ω − ωo )2
E(ω) = 2π τi Ao exp − . (3.80)
2
Note that E(ω) is a real function, which means that all the frequency components
have the same phase. This type of optical pulse is known as a “transform-limited”
pulse. Equation (3.80) shows that the power spectrum of the Gaussian pulse, S p (ω) ∝
|E(ω)|2 , is also a Gaussian function. The Gaussian pulse, together with its power
spectrum, is illustrated in Fig. 3.18. The width at half-height of the power spectrum,
Δν p , satisfies the relation:
2 ln2
Δν p τ p = = 0.441. (3.81)
π
118 3 The Laser
Fig. 3.18 The Gaussian pulse and its power spectrum
It is a general property of transform-limited light pulses that the bandwidth-

duration product is a number of the order of 1. The precise value of the product
depends on the shape of the pulse. In the more general case of pulses which are
not transform-limited, (3.81) gives only the lower bound of the bandwidth-duration
product. To consider an extreme situation, a short pulse of solar light can be created
by using a fast switch on a Sun beam. This pulse would have the spectrum of Sun
light, regardless of its duration, so that the bandwidth-duration product would be
larger than 0.441 by many orders of magnitude.
Problems
3.1 What is the momentum of a 1018 -Hz X-ray photon?

3.2 If 0.5 % of the light intensity incident into a medium is absorbed per millimeter,
what fraction is transmitted if the medium is 10 cm long? Calculate the absorption
coefficient α.
3.3 Calculate the photon decay time τc for a ring laser cavity having length 30 cm
and mirror reflectances R1 = 0.95 and R2 = 0.85.
3.4 An Nd-YAG laser having a Fabry-Perot cavity of length 30 cm and an active
medium of length 5 cm is operated in the mode-locking regime. Assuming that the
index of refraction of the active medium is n = 2 and that the gain bandwidth is
200 GHz, calculate the repetition period T and the duration τ p of the emitted pulses.
3.5 Considering that the gain bandwidth of a He-Ne laser is 1.5 GHz, which is the
maximum Fabry-Perot cavity length that ensures single longitudinal mode operation?
3.6 Calculate the mirror reflectance required to sustain oscillation in a symmetric
Fabry-Perot laser cavity, assuming that the active medium has a small-signal gain
coefficient of 1 m−1 and is 0.1 m long.
Problems 119
3.7 Consider a Fabry-Perot cavity Nd-YAG laser, in which the Nd-YAG crystal has
length L = 2 cm and mirror 1 is totally reflecting. Assuming that the population
inversion inside the active medium is N2 − N1 = 1 × 1018 cm−3 , the stimulated
emission cross-section is σ21 = 3.5 × 10−19 cm2 , the loop gain is equal to 2, deter-
mine the reflectance of mirror 2.
Chapter 4
Modulators
Abstract The application of an external field can modify the optical properties of a
material. By controlling the strength and frequency of the external field, the ampli-
tude, phase, or polarization of the optical beam can be controlled and modulated.
The most important modulation methods are those using electric fields, but both
acoustic fields and magnetic fields can be exploited. This chapter describes linear
and quadratic electro-optic effects, and explains the most common modulation and
switching schemes. Modulation methods based on liquid crystals are also briefly
discussed. The final part of the chapter is devoted to the acousto-optic effect and its
applications to amplitude modulation and to beam deflection.
4.1 Linear Electro-Optic Effect
The electro-optic effect is the change in the index of refraction due to an applied
electric field. If the refractive index change is proportional to the applied field, then
the effect is known as Pockels effect. If the index change is quadratically-dependent
on the applied field, then the effect is known as Kerr effect. As it will be seen, the
Pockels effect can exist only in crystalline materials, and is absent in amorphous
materials, such as glasses and fluids. In contrast, the Kerr effect is present in all
classes of materials.
In general, crystals are anisotropic media, in which D and P are related to the field
E by a second-rank tensor, as was seen in Chap. 2. In particular:

3
Pi (ω) = εo χij Ej (ω), (4.1)
j=1
where the indices i, j can take the values 1, 2, 3. As usual, 1 = x, 2 = y, 3 = z. The

electric susceptibility tensor χij is connected to εij by the relations εii = 1 + χii , and,
for i = j, εij = χij . The two tensors have the same diagonal system of coordinates.
The electro-optic effect modifies the relation between P and E, by introducing
additional terms dependent on an external field E . In the case of the Pockels effect,
the additional terms are linear with respect to the external field components:
DOI 10.1007/978-3-319-20627-1_4
122 4 Modulators

3
Pi (ω) = εo (χij + χijk Ek )Ej (ω). (4.2)
j,k=1
Equation (4.2) is nonlinear because it contains terms proportional to the product

of two electric fields. A general description of nonlinear interactions will be given in
Chap. 7. In the case under discussion here, it is important to keep in mind that E is
the electric field associated with an optical wave, while E is a field generated by an
electronic device. Thus the oscillation frequency of E is several orders of magnitude
larger than that of E . The external field acts as a parameter modifying the electric
susceptibility of the medium, and, consequently, its refractive index.
In a centrosymmetric medium all the components of χijk vanish. In fact, the
presence of non-zero components would induce the existence of terms cubic in
the electric field inside the electromagnetic energy density. Cubic terms change sign
when the variable changes sign, but, in a centrosymmetric medium, the energy density
cannot depend on the sign of the electric field.
The tensor χijk has 33 = 27 components, but, since it is symmetric with respect
to the exchange of i with j, only 18 components are independent. For all crystal
classes, symmetry considerations can be used to predict which components of χijk
are independent and non-zero.
In the scientific literature the name electro-optic tensor designs rijk , which is
connected to χijk by the relation:
χijk = −(1 + χii )(1 + χjj )rijk (4.3)
A single index n can be introduced to replace the ij pair, using the following con-
vention: n = 1, 2, 3, 4, 5, 6 corresponds to ij = 11, 22, 33, 23, 13, 12, respectively.
For instance, in this contracted notation, r123 becomes r63 and r333 becomes r33 . The
values of the refractive indices and of the non-zero electro-optic coefficients of some
typical crystals are reported in Table 4.1.
Table 4.1 Properties of some electro-optic crystals at λ = 633 nm

Crystal Class Indices of refraction r (10−12 m/V)
LiNbO3 3m no = 2.272, ne = 2.187 r33 = 30.8, r13 = 8.6
r22 = 3.4, r51 = 28
KNbO3 mm2 n1 = 2.279, n2 = 2.329 r33 = 25, r13 = 10
n3 = 2.167
LiTaO3 3m no = 2.175, ne = 2.180 r33 = 33, r13 = 8
KDP 42 m no = 1.51, ne = 1.47 r41 = r52 = 8.6, r63 = 10.6
ADP 42 m no = 1.52, ne = 1.48 r41 = r52 = 28, r63 = 8.5
Lithium niobate (LiNbO3 ), potassium niobate (KNbO3 ), lithium tantalate (LiTaO3 ), potassium di-
hydrogenated phosphate (KH2 PO4 , usually designed as KDP), ammonium di-hydrogenated phos-
phate (NH4 H2 PO4 , usually designed as ADP)
4.1 Linear Electro-Optic Effect 123
Crystals that have a symmetry supporting the Pockels effect also exhibit piezo-
electric effects. This means that they expand and contract under the application of
an external electric field. Piezoelectric effects are usually negligible for frequencies
of the applied field larger than some MHz.
Up to frequencies of the order of 1 THz, the induced polarization contains contri-
butions from the displacements of both ions and electrons. Above 1 THz, there are
only electronic contributions and coefficients are proportionally smaller. The values
reported in Table 4.1 refer to the frequency range between 1 MHz and 1 THz.
From a general point of view the wave propagation inside the electro-optic crystal
should be treated by inserting (4.2) into the wave equation (2.7). Since, as it has been
seen in Chap. 2, the propagation characteristics are fully described by means of the
index ellipsoid, the effect of the electric field is expressed most conveniently in terms
of changes of the index ellipsoid. In presence of E , (2.126) is modified as follows:
1 1 1
+ a1 x 2
+ + a2 y 2
+ + a3 z2
n12 n22 n32 (4.4)
+ 2a4 yz + 2a5 xz + 2a6 xy = 1,
where the coefficients an are given by:

3
an = rnk Ek (4.5)
k=1
As an example, considering KDP, which has only three non-zero components of

rnk , (4.4) becomes:
x 2 + y2 z2
+ + 2r41 E1 yz + 2r52 E2 xz + 2r63 E3 yz = 1 (4.6)
no2 ne2
The effect of the additional terms is that of modifying the direction of the dielectric
axes and the length of the semi-axes. The modulator configurations described in the
following sections will deal with rather simple situations in which E is directed
along the optical axis z and the light beam propagates along a principal axis.
4.1.1 Phase Modulation
The phase of an optical wave can be electro-optically modulated. For this type of
application, the optical wave is linearly polarized in a direction that is parallel to
one of the principal axes of the crystal in presence of the external field, so that
its polarization state does not change during propagation. In order to illustrate the
method two specific situations are below discussed in some detail.
124 4 Modulators
Modulation based on r33 . Lithium niobate is a crystal often used in electro-optic

modulation. It has a high electro-optic coefficient r33 , as shown in Table 4.1. The
configuration adopted to access the r33 term is having the light beam traveling along
x, linearly polarized along the optical axis z (extraordinary wave), and the external
electric field E parallel to z. In this configuration the extraordinary wave propagates
without changing its state of polarization, whether E is present or not. Using (4.2),
it is found that the only non-zero component of P is that which lies along z:
P3 = εo (χ33 + χ333 E3 )E3 = εo (χ33 − ne4 r33 E3 )E3 (4.7)
√
Equation (4.7) indicates that the extraordinary index of refraction, ne = 1 + χ33 ,
is modified by the presence of E , and becomes:

ne4 r33 E3
ne = 1 + χ33 − ne4 r33 E3 = 1 + χ33 1 −
1 + χ33 (4.8)
r33 ne3 E3
≈ ne − ,
2
where, taking into account that ne4 r33 E3 is usually much smaller than 1 + χ33 , the
approximate expression is obtained by a truncated series expansion of the square
root term.
Calling L the crystal length, the phase delay caused by the Pockels effect is given
by:
L r33 ne3 E3 L

Δφ = 2π(ne − ne ) = −2π . (4.9)
λ 2 λ
If L is a few centimeters and λ is around 1 µm, then the ratio of L/λ is of the order
of 104 . This means that a refractive index variation of just 10−4 can give a phase
delay Δφ of the order of π radians.
A phase modulator can be realized by using a lithium niobate crystal cut as
a parallelepiped along the principal axes. A voltage difference V across the crystal
thickness d induces an electric field E = V /d directed along z. The wave propagates
without changing polarization, and acquires a phase delay Δφ. Taking Eo cos(ωo t) as
the electric field of the incident wave, and assuming, for simplicity, that the applied
voltage oscillates with an angular frequency ωm , such that V (t) = Vo sin(ωm t), the
electric field of the wave that travels for a length L inside the crystal is:
2π r33 ne3

E(t) = Eo cos ωo t − ne − Vo sin(ωm t) L . (4.10)
λ 2d
Equation (4.10) describes a wave at frequency ωo , phase modulated at the fre-
quency ωm .
Modulation based on r63 . If an external field directed along the optical axis is
applied to a KDP-like crystal, then the P components, as derived from (4.2), are
given by:
P1 = εo (χ11 E1 + χ123 E3 E2 )

P2 = εo (χ11 E2 + χ123 E3 E1 ) (4.11)
P3 = εo χ33 E3 ,
where it has been taken into account that E1 = E2 = 0 and χ123 = χ213 . The
effective electric susceptibility is therefore described by the matrix:
⎛ ⎞
χ11 χ123 E3 0
⎝χ123 E χ11 0 ⎠. (4.12)
3
0 0 χ33
The presence of E3 generates off-diagonal terms which couple P1 with E2 and P2 with
E1 . In order to describe wave propagation inside the crystal the matrix (4.12) must be
diagonalized, by introducing a new set of principal axes x , y , z , and new diagonal
terms χ11 , χ , χ . Noting that the relation between P and E is not modified by the
22 33 3 3
presence of E3 , it can be shown that z must coincide with z, the plane x -y coincides
with x-y, and χ33 = χ . Since the expressions for P and P are symmetrical with
33 1 2
respect to the exchange of E1 with E2 , the angle between x and x must be 45◦ . The
result of the diagonalization is:

χ11 = χ11 − χ123 E3
(4.13)
χ22 = χ11 + χ123 E3 .
Equation (4.13) show that the rotation symmetry around the z axis is lost, because
= χ . The new refractive indices are:
χ11 22
2
n1 = 1 + χ11 ≈ n 1 + no r63 E3
o
2
(4.14)
no r63 E3
2
n2 = 1 + χ22 ≈ no 1 − ,
2
where χ123 has been substituted by −no4 r63 , and the approximation χ123 E3 1+χ11
has been made.
The scheme of a phase modulator exploiting the properties of the KDP crystal is
shown in Fig. 4.1. The crystal is a parallelepiped cut along the axes x , y , z, with E
directed along z, the optical wave travels along z and is linearly polarized in the x
direction. Since the external voltage is applied along the propagation direction, this
configuration requires transparent electrodes, which can be realized by depositing
thin semiconductor layers on the crystal faces. The incident wave propagates without
changing polarization, and acquires a phase delay determined by the value of n1 .
126 4 Modulators
Fig. 4.1 Phase modulator based on r63 . The incident optical wave propagates along the optical
axis, z, and is linearly polarized along x . The external field is applied along z
Taking Eo cos(ωo t) as the electric field of the incident wave, and assuming again
a sinusoidal applied voltage, the electric field of the wave traveling for a length L
inside the crystal is:
2π r63 no3

Ex (t) = Eo cos ωo t − no + Vo sin(ωm t) L . (4.15)
λ 2L
Equation (4.15) has the same form as (4.10), such that the two schemes are con-
ceptually equivalent. In practice, the first scheme can work with much lower applied
voltages, provided the aspect ratio d/L is made very small. A more detailed compar-
ison will be made after treating the amplitude modulation.
4.1.2 Amplitude Modulation
Phase modulation can be converted into amplitude modulation by using an interfer-

ometric method or by exploiting a birefringence variation.
Interferometric modulator. The scheme of an amplitude modulator based on a
Mach-Zehnder interferometer is shown in Fig. 4.2. M1 and M2 are totally reflecting
mirrors, the beam splitter D1 divides the input beam in two identical parts, and the
Fig. 4.2 Amplitude

modulator based on the
Mach-Zehnder
interferometer
beam splitter D2 recombines the two beams at the two outputs A and B. Two identical
lithium niobate crystals are inserted in the upper and lower branches of the interfer-
ometer. The optical beams propagate through the crystals in a direction perpendicular
to the optical axis and are linearly polarized along the optical axis. Qualitatively the
behavior is as follows. If the upper and lower optical paths are identical, then there
is constructive interference at output A and destructive interference at output B. If
all the components are lossless, then the optical intensity at A, IA , coincides with the
input intensity, Io , and IB is zero. By applying an electric field to the upper crystal,
the index of refraction is changed, such that the wave traveling in the upper branch
will acquire an additional phase delay. As a consequence, the interferometer will be
unbalanced, and some part of the input intensity will appear at output B.
The electric field of the wave at the output A (or B), EA (EB ), is the sum of the
contributions coming from the upper and lower branch of the interferometer. Taking
Eo cos(ωo t) as the electric field of the incident wave, and considering all the phase
delays experienced by the beams propagating in the upper and lower branch of the
interferometer, the electric fields EA and EB of the two output beams are expressed
as:
Eo
EA = [cos(ωo t − φ1 ) + cos(ωo t − φ2 )]
2 (4.16)
Eo
EB = [cos(ωo t − φ3 ) + cos(ωo t − φ4 )]
2
Calling φτ and φρ the phase delays in transmission and reflection due to the beam
splitters D1 and D2 , and φM the phase delay due to reflection at M1 and M2 , the phase
delays of (4.16) are expressed as:
φ1 = φτ + 2π Lup /λ + φρ + φM
φ2 = φρ + 2π Ldown /λ + φτ + φM
(4.17)
φ3 = φτ + 2π Lup /λ + φτ + φM
φ4 = φρ + 2π Ldown /λ + φρ + φM .
The upper and lower optical paths are given by: Lup = ne L+Lo , Ldown = ne L+Lo ,
respectively, where L is the length of the two crystals and Lo is the path in air. Taking
into account that φτ and φρ are related by (2.58), it is found that: φ1 − φ2 = Δφ,
and φ3 − φ4 = Δφ − π . Δφ is the phase delay caused by the Pockels effect, and is
given by (4.9).
The intensity IA (IB ) at output A(B) is proportional to the square of EA (EB ). Recall-
ing that:
α+β α−β
cosα + cosβ = 2cos cos
2 2
it is found that:
φ − φ Δφ
1 2
IA = Io cos2 = Io cos2 , (4.18)
2 2
128 4 Modulators
and φ − φ Δφ
3 4
IB = Io cos2 = Io sin2 . (4.19)
2 2
Equations (4.18) and (4.19) show that, if Δφ = ±π , then the whole input intensity
is transferred to output B. The applied voltage needed for a phase shift of π , known
as the half-wave voltage, is given by:
λd
Vπ = . (4.20)
r33 Lne 3
Since the transmittance of the Mach-Zehnder interferometer, observed at out-

put B, ranges from 0 to 1 when the applied voltage is changed from 0 to Vπ , the
interferometer can be used as an electro-optic switch.
If the applied voltage is a continuously varying signal, for instance the sinusoidal
function V (t) = Vm sin(ωm t), then the device behaves as an amplitude modulator.
However, the output intensity is not linearly dependent on the applied voltage. In order
to obtain a linear relation, at least in the small signal regime, the working region should
be shifted around the inflexion point of the curve IB /Io versus V , as shown in Fig. 4.3.
This can be achieved by adding a DC bias voltage Vπ/2 = Vπ /2, which introduces a
fixed phase shift π/2. The total applied voltage becomes: V = Vπ/2 + Vm sin(ωm t).
By using (4.19), the ratio IB /Io takes the expression:
IB π π Vm sin(ωm t)
= sin2 + . (4.21)
Io 4 2Vπ
Fig. 4.3 Analog modulation based on the Mach-Zehnder interferometer

Fig. 4.4 Equivalent electric

circuit of the crystal with
applied voltage
If Vm /Vπ 1, (4.21) can be approximated as:
IB 1 π V sin(ω t)
1
m m π Vm sin(ωm t)
= 1 + 2sin ≈ 1+ , (4.22)
Io 2 2Vπ 2 Vπ
obtaining a linear relation between applied voltage and output intensity. The ampli-
tude modulation used in optical communications operates around the inflexion point
of the curve IB versus V .
The intrinsic response time of the Pockels effect can be smaller than 10−11 s. The
limits to the speed of the device come from electrical capacitive effects: the crystal
behaves as a capacitor with capacity C, charged by a voltage generator with internal
resistance Rs (see Fig. 4.4). Therefore, if the modulation frequency ωm is larger than
(Rs C)−1 , then most of the applied voltage will be dropped across Rs instead of across
the crystal. Considering the problem in the time domain, the response time is given
by Rs C. Recalling that C = εo εr A/d, where A is the surface area, a quick calculation
shows that the typical response time is of the order of 1 ns.
A method to reduce the response time is to put a coil with inductance Lo and a
resistor Ro in parallel to the crystal, such that ωm 2 = (Lo C)−1 and Ro Rs . At
resonance, the impedance of the parallel Ro Lo C is equal to Ro , and so the applied
voltage is dropped almost completely across the crystal. The price to pay in this
configuration is that the device can operate only in a band of modulation frequencies
centered at ωm with a width Δω = (Ro C)−1 .
The modulation induced by the applied field is effective only if the transit time
of the optical wave through the crystal is smaller than the modulation period Tm =
2π/ωm . Indeed, instead of using (4.9), Δφ should be more generally calculated as:

2π r33 ne 3 L ωr33 ne 2 t+τd
Δφ = − E3 dz = − E3 (t )dt , (4.23)
2λ 0 2 t
where τd = Lne /c is the transit time through the crystal, and E3 (t ) is the instan-
taneously applied field. In the right-hand member of (4.23), the spatial integra-
tion has been substituted by a temporal integration by putting dt = ne dz/c. If
E3 = Eo sin(ωm t), Δφ is given by:
130 4 Modulators
Fig. 4.5 Integrated

Mach-Zehnder modulator

ωr33 ne 2 τd Eo sin(ωm τd )
Δφ = − . (4.24)
2 ωm τd
In the limit ωm τd → 0, (4.24) coincides with (4.9). Assuming that the upper limit
to the modulation frequency is determined by the condition ωm τd = π/2, it is found
that the maximum value is: νmax = c/(4Lne ). By using a lithium niobate crystal long
1 cm, νmax = 3 GHz.
The lithium niobate modulators utilized inside optical communication systems are
integrated devices, in which the light beams travel inside embedded strip waveguides
fabricated by in-diffusing in the lithium niobate substrate a material (usually tita-
nium) that increases the extraordinary refractive index. A scheme of the integrated
modulator is shown in Fig. 4.5. The width of the waveguide is very small, typically
d = 5 µm, so that the half-wave voltage is reduced to about 1 V, as calculated from
(4.20) by assuming a waveguide length of a few centimeters. Light is coupled into
and out of the modulator by using optical fibers. The output capacitance is strongly
reduced in comparison with the bulk device, and so the response time is much faster.
Transit-time effects can be, in principle, eliminated by using the electrodes as a trans-
mission line in which the velocity of the traveling electrical signal matches that of
the light beam. Modulation speeds of several tens of GHz can be achieved with this
approach.
Modulation by birefringence variation. In order to perform an amplitude modu-
lation based on r63 , E is directed along z, and the optical wave travels along z, with
a linear polarization along x, as shown in Fig. 4.6. Since in this case the wave is
not polarized along a principal axis, its propagation has to be treated by separately
considering the two components of the electric field that are parallel to x and y and
travel inside the crystal with different velocities. If the electric field of the incident
wave is Ex = Eo cos(ωo t − kz), then the two components of the wave exiting the
crystal are:
Eo Eo
Ex = √ cos(ωo t − φx ) ; Ey = √ cos(ωo t − φy ). (4.25)
2 2
Fig. 4.6 Amplitude modulator based on r63 . The input wave is linearly polarized along x. The axis
of the output polarizer is along y
The relative phase delay is:
2π no3 r63 V
Δφ = φx − φy = , (4.26)
λ
where V = LE3 is the applied voltage. The output wave is elliptically polarized,
that is, a modulation of the polarization is produced. In order to transform the device
into an amplitude modulator, it is sufficient to add a polarizer with the optical axis
parallel to y, as shown in Fig. 4.6. The electric field of the wave transmitted by the
polarizer is the sum of the projections of Ex and Ey on the y axis:
Ey = (Eo /2)[cos(ωo t − φy ) − cos(ωo t − φx )]

(4.27)
= Eo sin(ωo t − φo )sin(Δφ/2),
where φo = (φx + φy )/2. The ratio between output and input intensity is:
Iu Δφ πV
= sin2 = sin2 , (4.28)
Io 2 2Vπ
where
λ
Vπ = (4.29)
2r63 no3
is the half-wave voltage for this configuration.

Equation (4.28) is identical to (4.19), therefore the two amplitude modulators are
equivalent in principle. However, by comparing (4.29) with (4.20), a very important
difference appears: if the electric field E is applied transversally to the direction
of wave propagation, then Vπ becomes proportional to the aspect ratio d/L. As a
consequence, if a small aspect ratio is used for the transverse modulation, as in the
case of the integrated modulator of Fig. 4.5, then Vπ can be of the order of 1 V. In
contrast, the longitudinal modulation scheme of Fig. 4.6, taking also into account that
132 4 Modulators
the electro-optic coefficient of KDP is much smaller than that of lithium niobate, has
a half-wave voltage of several kilovolts.
It is instructive to discuss the amplitude modulator of Fig. 4.6 by using the for-
malism of Jones matrices described in Sect. 2.5.2. Defining Vo as the Jones vector of
the input wave, Vu the vector of the output wave, T1 , T2 , T3 , the matrices describing
the input polarizer, the electro-optic crystal, and the output polarizer, respectively,
then Vu is given by:
Vu = T3 T2 T1 Vo

00 exp(−iΔφ) + 1/2 exp(−iΔφ) − 1/2 10 1
01 exp(−iΔφ) − 1 /2 exp(−iΔφ) + 1 /2 00 0 (4.30)

0
= .
exp(−iΔφ) − 1 /2
The ratio between output and input intensity is simply given by:

exp(−iΔφ) − 1 2
Iu
= = 1 − cos(Δφ) = sin2 Δφ , (4.31)
Io 2 2 2
which coincides with (4.28).

In order to operate the modulator in a linear region, a quarter-wave plate, having
x and y as principal axes, can be inserted after the crystal. Using the x and y axes
as the reference system, the matrix describing the plate is:

exp(−iπ/2) + 1/2 exp(−iπ/2) − 1/2
T2 = . (4.32)
exp(−iπ/2) − 1 /2 exp(−iπ/2) + 1 /2
The output signal becomes:

0
Vu = T3 T2 T2 T1 Vi = , (4.33)
exp(−i(Δφ + π/2) − 1 /2
and the ratio Iu /Io is given by:

Iu 1 − cos(Δφ + π/2) Δφ π
= = sin2 + . (4.34)
Io 2 2 4
The insertion of the quarter-wave plate has the effect of shifting the operating
point of the modulator to the inflexion point of the curve Iu /Io versus V . Note that in
the case of the interferometric amplitude modulator described in the previous section,
the same effect was obtained by applying a bias voltage equal to Vπ/2 (see Fig. 4.3).
A transverse modulation scheme could also be applied to the KDP modulator, as
shown in Fig. 4.7, where E is parallel to z, and the light beam propagates in the y
direction, with a linear polarization at 45◦ with z. The phase delay Δφ between the
x and z components is given by:
Fig. 4.7 Amplitude modulator based on r63 in a transverse scheme. The external electric field is
applied along z and the light beam propagates along y
2π L n3 r63 V
Δφ = φx − φz = no − ne + o , (4.35)
λ 2d
where d is the crystal thickness. Equation (4.35) shows that Δφ is non-zero even in
absence of E . This is obvious, because the light beam is not propagating along the
optical axis. However, the presence of the voltage-independent term complicates the
utilization of the transverse configuration.
The electro-optic methods described in this section are also utilized for the gen-
eration of ultrashort laser pulses. In the case of a Q-switched laser, a fast optical
switch can be realized by inserting into the cavity an electro-optic shutter that blocks
the laser beam when voltage is applied to the crystal and is suddenly opened by
removing the voltage. Electro-optic modulators can be used to drive a laser in the
mode-locking regime by intra-cavity modulation at a frequency ωm ≈ 2π Δν, where
Δν is the frequency difference between two adjacent longitudinal modes of the laser
cavity.
4.2 Quadratic Electro-Optic Effect
Centrosymmetric materials may exhibit a quadratic electro-optic effect, i.e. a refrac-

tive index change proportional to the square of the applied electric field:
Δn = nK E 2 , (4.36)
where nK , measured in m2 /V 2 , is the Kerr constant. It should be mentioned that

many authors define differently the Kerr constant, by writing Δn = λnK E 2 ,

where λ is the wavelength of the modulated light beam and the constant nK is mea-
sured in m/V2 .
134 4 Modulators
In general terms, the relation between P and E takes the following form:
⎛ ⎞

Pi = εo ⎝ χij Ej + χijkl Ek El Ej ⎠ (4.37)
j j,k,l
The tensor χijkl has 34 = 81 components. In an isotropic material, only the

components of the type χiiii or χiikk are non-zero.
Because of the Kerr effect, an isotropic material becomes uniaxial, taking the
direction of the applied√electric field as the optical axis. Assume that χ is the electric
susceptibility and n = 1 + χ the refractive index of the isotropic material, and that
E is parallel to z. For a wave traveling along y and linearly polarized along z the
extraordinary index of refraction is:
e 2
ne = n + nK E (4.38)
where n2e is proportional to the component χ3333 . For the wave traveling along y, but
polarized along x, the ordinary refractive index is:
o 2
no = n + nK E , (4.39)
where n2o is proportional to χ1133 . Therefore the two Kerr constants are generally
different.
In solid media the Kerr effect is mainly connected with a displacement of electron
clouds, whereas in fluid media important contributions may come from the orientation
of anisotropic molecules. The electro-optic effects of electronic origin are usually
small, but have very short response times. In contrast, orientational Kerr effects can
be large, but have correspondingly long response times.
In absence of an applied field a liquid consisting of elongated molecules behaves
as an isotropic medium because the molecular orientations are completely random.
When an electric field is applied, the molecules tend to orient parallel to each other,
and the medium becomes uniaxial. If E is not very large, then the first-order theory
predicts that the change in refractive index is proportional to E 2 , and that nK o =
−nK e /2. For large E , the molecules become all parallel to the field, and the medium
birefringence attains a saturation value.

In principle the Kerr effect can be utilized for electro-optic modulation by using
schemes similar to those of Pockels modulators. In practice, the most important
applications involve liquid crystals, and are discussed in the next section.
A very interesting aspect of the Kerr effect is that, being quadratic, it can also be
effective when the applied field is oscillating at very large frequencies, such as the
optical frequencies. This is known as the optical Kerr effect, and will be discussed
in Chap. 7.
4.2 Quadratic Electro-Optic Effect 135
4.2.1 Liquid Crystal Modulators
A particularly large quadratic electro-optic effect is shown by liquid crystals, with

very important applications in displays or in other situations not requiring fast
response times. In normal cases a solid material, when heated to its melting point,
becomes a liquid, and undergoes a phase transition from an ordered state to a dis-
ordered state. Liquid crystals are a particular class of materials that present one (or
more) intermediate states between the crystalline solid and the isotropic liquid. They
are fluid-like, yet the arrangement of molecules exhibits structural order. A charac-
teristic of liquid crystals is the rod-like shape of their molecules. Within the liquid
crystal family, there are three distinct molecular structures: nematic, cholesteric,
and smectic, as shown in Fig. 4.8. Only the first two structures are of importance in
modulators and displays. In the nematic phase, the elongated molecules are aligned
parallel to each other, but they are free to move relative to each other so that the
phase has liquid properties. In the cholesteric phase, the material is made up from
a large number of planes each having a nematic-like structure, but with each plane
showing a progressive change in the molecular axis direction from the one below.
The molecular axis directions thus display a helical twist through the material.
The nematic temperature range of a single compound is usually limited to a few
tens of degrees. For instance, the compound known as 5CB is crystalline below 24 ◦ C,
is nematic between 24◦ and 35.3 ◦ C, and becomes a liquid above 35.3 ◦ C. For display
applications a wide nematic range is highly desirable. To widen the nematic range,
mixtures of several components are commonly used.
Liquid crystals are optically anisotropic: the nematic phase is birefringent and the
cholesteric phase presents rotatory power. A light beam traveling across a nematic
liquid crystal layer in a direction perpendicular to the molecular orientation behaves
as an extraordinary wave if linearly polarized along the molecular axis. If an applied
electric field forces the molecules to align parallel to the propagation direction of
the light beam, then the wave becomes an ordinary wave. Therefore the effect of
the electric field is that to change the phase delay experienced by the light beam,
from 2π ne L/λ to 2π no L/λ, where L is the layer thickness. Since the liquid crystal
birefringence can be quite large, typically ne − no ≈ 0.1, a difference in phase delay
equal to π can be obtained with a layer thickness of just a few microns. It is therefore
Fig. 4.8 Liquid crystals

136 4 Modulators
Fig. 4.9 Optical switch based on a twisted nematic liquid crystal cell. The switch is open if there
is no applied voltage. The light beam is blocked if the liquid crystal molecules are forced to align
parallel to z by a sufficiently large voltage
possible to make an electro-optic switch or modulator using a very thin liquid crystal
film.
When a nematic liquid crystal material comes into contact with a solid surface,
the molecules can become aligned either perpendicular to the surface (homeotropic
ordering) or parallel to the surface (homogeneous ordering), depending on the treat-
ment of the surface. In the case of homogeneous ordering, the orientation direction of
the molecules is selected by rubbing the surface along a particular direction before it
comes into contact with the liquid crystal. The molecules will take up an orientation
parallel to the direction of rubbing.
A typical example of electro-optic effects in liquid crystals is given by the cell
using a “twisted nematic liquid crystal”, as shown in Fig. 4.9. The liquid crystal
layer is closed inside a glass cell, in which opposing walls are treated to produce
a homogeneous ordering with orientations at right angles to each other. Thus the
molecular axis undergoes a gradual rotation across the cell, from 0◦ at one wall to
90◦ at the opposite wall. The twisted nematic phase obtained in this way has the same
rotatory power of a cholesteric phase. When a linearly polarized light beam travels
across the cell its polarization direction undergoes a 90◦ rotation. By applying a strong
enough electric field perpendicularly to the cell walls, the molecules are forced to
align homeotropically. In this state the light beam will travel across the cell without
changing its polarization. By inserting the cell between two crossed polarizers, an
electro-optic switch can then be constructed. With no applied voltage, the light beam
is transmitted through the second polarizer. When voltage is applied the direction
of polarization of the light traversing the cell is not rotated and hence cannot pass
4.2 Quadratic Electro-Optic Effect 137
through the second polarizer. The device operates with an applied voltage of a few
volts, but has long response times, since the electro-optic effect is based on molecular
rotations. The typical application is in the field of displays, as it will be discussed in
Chap. 8.
4.3 Acousto-Optic Effect
The acousto-optic effect is the change in the refractive index of a medium due to
the mechanical strains associated with the passage of an acoustic wave inside the
medium. The refractive index changes are caused by the photoelastic effect which
occurs in all materials under a mechanical stress.
Consider the case of a monochromatic plane wave of frequency νi and wave vector
ki , traveling in a medium in which an acoustic wave of frequency νs and wave vector
ks is present. Qualitatively the interaction between the light wave and the acoustic
wave can be described in the following way. The acoustic wave induces a periodic
variation of the refractive index, which constitutes a three-dimensional phase grating,
as shown in Fig. 4.10. The periodicity is given by the acoustic wavelength λs = us /νs ,
where us is the sound velocity. It is assumed that ks is parallel to the z axis, and that
ki lies in the y-z plane, forming an angle θ with the y-axis, as shown in Fig. 4.11.
Considering the acoustic wavefronts as partially reflecting surfaces, a diffracted beam
can be formed only if there is constructive interference among the reflections coming
from all the reflecting surfaces, similarly to the case of the multilayer mirror treated
in Sect. 2.2.4. It is clear from this approach that the diffracted beam, if it exists, will
have a wave vector kd forming an angle θ with the y-axis.
The condition for constructive interference is found to be:
ks λi
sinθ = = , (4.40)
2ki 2nλs
Fig. 4.10 An acoustic wave

induces a traveling periodic
variation of the refractive
index
138 4 Modulators
Fig. 4.11 Acousto-optic

diffraction
where n is the index of refraction of the medium in which the waves are propagating.
Once λi and ks are fixed, a diffracted beam is found only if the direction of the
incident beam satisfies (4.40).
In practice, the diffraction angle is small, as shown by the following example.
Consider a visible light beam at λi = 600 nm, and a sound wave of frequency
νs = 100 MHz. Assuming a sound velocity inside the medium of us = 3 × 103 m/s,
one finds λs = us /νs = 30 µm. If n = 1.5, (4.40) gives θ = 0.4◦ . This means that
the diffracted beam forms an angle α = 2θ = 0.8◦ with the incident beam.
Since the acoustic wavefronts are traveling at velocity us , one should expect that
the frequency of the diffracted beam be Doppler shifted with respect to that of the
incident beam by the amount:
(kd − ki ) · us
ΔνD = = νs . (4.41)
2π
The mathematical treatment of the interaction between optical wave and acoustic
wave, both taken as plane monochromatic waves, is sketched-out in the following
section. A very important result is that the frequency νd and the wave vector kd of
the diffracted beam must satisfy the following conditions:
νd = νi ± νs , (4.42)
and
kd = ki ± ks . (4.43)
Equation (4.42) is essentially a confirmation of (4.41). Equation (4.43) is graph-

ically expressed by the triangles in Fig. 4.12. The two equations can be interpreted
as describing a collision between two quasi-particles, the incident photon, with
energy hνi and momentum (h/2π )ki , and the quantum of acoustic energy, known
as “phonon”, with energy hνs and momentum (h/2π )ks . The choice of the sign
inside (4.42) and (4.43) depends on the direction of the energy and momentum
exchange between optical and acoustic waves. The plus sign indicates that the phonon
4.3 Acousto-Optic Effect 139
Fig. 4.12 Conservation of wave vectors in two cases. Upper part: a phonon is absorbed. Lower
part: a phonon is generated
is absorbed, and its energy and momentum are added to the diffracted photon. Vicev-
ersa, the minus sign applies when a phonon is generated. In the situation illustrated
in Fig. 4.11, the acoustic wave transfers energy and momentum to the optical wave,
and so the plus sign is chosen. Of course, if the acoustic wave is a stationary wave,
then both signs are present at the same time. In the following examples, it will be
assumed that the plus sign applies.
Since the frequency of the acoustic wave is many orders of magnitude smaller than
the optical frequency, νd is very close to νi . Therefore it is a very good approximation
to assume that the modulus of kd is equal to that of ki . As a consequence, the
triangles of Fig. 4.12 can be seen as isosceles, so that it is immediately verified that
sin θ = ks /(2ki ), as it has been already anticipated by the relation (4.40).
4.3.1 Acousto-Optic Modulation
The treatment of the acousto-optic interaction develops similarly to that of the electro-
optic effect. The procedure is here outlined without detailed calculations, only report-
ing the final result.
The acoustic wave modifies the relation between E and P by introducing a strain-
dependent term. The propagation of the light wave in the presence of the acoustic
140 4 Modulators
field is described by inserting the modified relation into the wave equation, and by
assuming that a single diffracted beam is generated, as shown in Fig. 4.11.
The electric field of the incident monochromatic plane wave is given by:
Ei = (Eio /2)exp[−i(ωi t − ki · r)] + c.c.
The diffracted beam is also taken to be a monochromatic plane wave. Its electric
field is:
Ed = (Edo /2)exp[−i(ωd t − kd · r)] + c.c.
The general definition of the component Skl of the strain tensor is:

1 ∂uk ∂ul
Skl = + ,
2 ∂xl ∂xk
where uk (r) (k = 1, 2, 3) is the Cartesian component of the displacement u(r) of

the point r inside the strained medium. The acoustic wave generates a strain wave:
1 o
Skl (r, t) = S exp[−i(ωs t − ks · r)] + c.c. (4.44)
2 kl
In presence of the strain wave, the relation between electric field and electric
polarization becomes:

3

Pi = εo χij + fijkl Skl Ej , (4.45)
jkl=1
where fijkl is the acousto-optic tensor. It is usually more convenient to express the
equations in terms of the photoelastic tensor, pijkl , related to fijkl by:
fijkl = −(1 + χii )(1 + χjj )pijkl .
Limiting the treatment to isotropic media, the relation between the acousto-optic
and the photoelastic parameters simplifies to: f = −n4 p. As a consequence, instead
of (4.45), the following expression can be used:
P = εo [χ − n4 pS(r, t)]E, (4.46)
where S(r, t) = So exp[−i(ωs t − ks · r)] is now the amplitude of the strain wave.
Calculations are made in the stationary regime, such that the field amplitudes Eio and
Edo are taken as time-independent quantities. As the incident wave is propagating
inside the medium, its amplitude Eio decreases because some of the incident power is
transferred to the diffracted beam. Consequently, the amplitude of Edo grows from its
initial value of zero. In principle, the amplitude of the acoustic wave is also changed
by the interaction. However, the variation of acoustic power is negligible in all the
cases of practical interest.
By imposing the initial conditions of Eio (0) = Eo and Edo (0) = 0, the treatment
gives the following dependence of the field amplitudes on the propagation distance y:
Eio (y) = Eo cos(hycosθ )

(4.47)
Edo (y) = iEo sin(hycosθ ),
where:
π pn3
h= So . (4.48)
2λi
Taking L as the path-length inside the medium, the final result is that the ratio
between diffracted and incident intensity is given by:
3
Id π n pSo L
= sin2 (4.49)
Io 2λi
The oscillating behavior of the diffracted optical intensity versus L is explained as

follows. The incident wave transfers power to the diffracted wave until it is completely
extinguished. However, in turn, the diffracted wave generates a secondary diffracted
wave that exactly coincides with the incident wave. The result is that there is a
periodic exchange of power between the two waves during propagation, with the
sum of the two powers remaining constant in an ideal lossless material.
By using the amplitude So of the acoustic wave as a modulating signal, acousto-
optic modulation can be achieved. Since the experimental control parameter is the
intensity of the acoustic wave, Is , it is better to express the diffracted intensity as a
function of Is . Recalling that:

2Is
So = , (4.50)
ρm us 3
where ρm is the medium density, and further defining a figure of merit, M, that
summarizes the relevant material properties:
n6 p2
M= , (4.51)
ρm us 3
Equation (4.49) becomes:

Id πL
= sin2 √ MIs (4.52)
Io 2 λi
142 4 Modulators
As an example, if water is chosen as the acousto-optic material, the appropriate

material values are: n = 1.33, p = 0.31, us = 1500 m/s, ρm = 1000 kg/m3 .
The figure of merit is then calculated as: M = 1.58 × 10−13 m2 /W. Considering an
acoustic power of 1 W over an area of 1 mm × 1 mm, and taking L = 1 mm with
λi = 0.6 µm, it is found that Id /Io = 0.97. This example shows that comparatively
strong optical effects may be obtained using moderate acoustic power and short
propagation lengths.
One of the most commonly used materials in commercial acousto-optic mod-
ulators is tellurium dioxide, TeO2 , which has a wide transparency window (400–
4000 nm) and M = 34 × 10−15 m2 /W.
In order to easily separate the diffracted beam from the incident beam the diffrac-
tion angle θ should not be too small. In any case it should be somewhat larger than
the divergence angle of the optical beam. Since θ is proportional to the frequency
of the acoustic wave, as shown by (4.40), this criterion fixes the lower limit to the
acoustic wave frequency.
An acousto-optic cell can act as an optical switch for the Q-switching operation.
In the presence of an acoustic wave, a significant portion of the optical beam traveling
inside the laser cavity is diffracted out of the cavity, increasing the cavity losses and
keeping the laser below threshold. Once the acoustic wave is shut off by suddenly
putting the voltage across the piezoelectric transducer to zero, laser action can start
and the giant pulse is generated. Acoustic cells are also used in mode-locked lasers:
instead of having an acoustic traveling wave, an acoustic stationary wave is produced.
If the amplitude of the stationary wave oscillates at the frequency Δν = c/(2L ),
then the cell behaves as an amplitude modulator that forces the laser to operate in
the mode-locking regime.
4.3.2 Acousto-Optic Deflection
The acousto-optic effect can be used to generate a deflected beam in a controlled way,
without recurring to moving mirrors or prisms. Deflectors are useful in a variety of
applications, such as writing and reading operations with laser beams. Furthermore,
since the deflection angle depends on λi , the acousto-optic cell behaves similarly to
a dispersive prism, and can be used as a frequency filter.
As shown in the preceding section, the angle between the diffracted beam and the
incident beam is α = 2θ , where θ is given by (4.40). If θ is small, then sinθ
θ ,
hence:
λi λi
α = 2θ = = νs (4.53)
nλs nus
Equation (4.53) suggests that a scanning of the deflection angle can be achieved
by simply making a sweep of the acoustic frequency. If the sweep covers an interval
Δνs , then the corresponding variation of the deflection angle is:
λi
Δα = Δνs . (4.54)
nus
The important figure here is the number of distinguishable angular positions, N,

determined by the ratio between Δα and the angular divergence of the laser beam.
θo = λi /(π w), w being the radius of the Gaussian beam. The ratio is:
Δα
N= = Δνs τ, (4.55)
θo
where τ = w/us is the transit time of the acoustic wave across the light beam. As
an example, if us = 1500 m/s (water), w = 5 mm, νs is swept from 80 to 125 MHz
(Δνs = 45 MHz), one finds: N = 150.
There is an important conceptual point that should be clarified. If both ki and
the direction of ks are kept fixed, then the condition (4.43), which graphically cor-
responds to the closure of the triangle in Fig. 4.12, does not give any degree of
freedom: diffraction is obtained at one acoustic frequency only. This would imply
that one would not get any diffracted beam during a sweep of the acoustic frequency,
except at that instant in which the acoustic wave vector satisfies the equality (4.43).
How can this problem be overcome? The key to the solution lies in the observa-
tion that the acoustic wave is not truly a plane wave. It is transversally limited to a
finite cross-section, and this results in a spread in the direction of ks , as depicted in
Fig. 4.13. Therefore, as νs is changed, it may still be possible to close the triangle
of wave vectors by exploiting a component of the acoustic wave that has a different
direction.
Fig. 4.13 Acousto-optic deflection

144 4 Modulators
Problems
4.1 Consider the electro-optic amplitude modulator of Fig. 4.2, using a lithium nio-
bate crystal. The operating wavelength is λ = 1530 nm, the electro-optic crystal is
lithium niobate, the crystal length is L = 6 cm, the crystal thickness is d = 40 µm,
and the output intensity IB is equal to zero when no voltage is applied to the crystal. (i)
Calculate the value of the applied voltage Vπ required to obtain 100 % transmission
at input B; (ii) Calculate the upper limit to the modulation frequency, as determined
by the condition that the transit time of the light beam through the crystal be one
fifth of the modulation period.
4.2 Consider the electro-optic amplitude modulator of Fig. 4.2, using a lithium tanta-
late crystal. The operating wavelength is λ = 1550 nm, the crystal length is L = 4 cm,
the crystal thickness is d = 40 µm, and the output intensity IB is equal to zero when
no voltage is applied to the crystal. Assume that the applied voltage V is the sum
of a bias voltage Vo plus a signal voltage V1 . Calculate: (i) the value of Vo required
to bring the modulator in the linear regime; (ii) the upper limit to V1 , subject to the
condition that the maximum output intensity does not exceed the limit of 75 % of
the input intensity.
4.3 A light beam at λ = 1550 nm, linearly polarized along x, propagates along the
optical axis (the z axis) of a KDP crystal. A voltage V is applied to the crystal along
the z axis. After crossing the crystal the light beam goes through a sequence of a
quarter-wave plate, having the optical axis in the x-z plane forming an angle of 45◦
with the x axis, and of a polarizer, with optical axis along y. Calculate the value
of V required for a 100 % transmission of the incident beam through the sequence
crystal-plate-polarizer.
4.4 An amplitude modulator based on the electro-optic Kerr effect can be made by
putting the Kerr cell in between two crossed polarizers, as in Fig. 4.6. The external
electric field E is applied transversally, and the incident light is linearly polarized at
45◦ to the electric field. Calculate the half-wave voltage Vπ/2 by assuming that the
wavelength of light is 600 nm, the path-length of the light beam inside the Kerr cell
is 5 cm, the distance between the two capacitor plates is 2 mm, and the electro-optic
material is liquid nitrobenzene (nK e − no = 2.6 × 10−18 m2 /V2 ).
K
4.5 A light beam at λ = 1500 nm goes through an acousto-optic cell made of flint
glass (us = 3 km/s, n = 1.95) in which an acoustic wave of frequency 200 MHz
is propagating. Calculate: (i) the wavelength difference between the diffracted wave
and the incident wave; (ii) the deflection angle of the diffracted wave.
4.6 A light beam at λ = 650 nm goes through an acousto-optic cell made of water
(n = 1.33, p = 0.31, us = 1500 m/s, ρ = 1000 kg/m3 ) having a square cross-
section. Calculate: (i) the frequency of the acoustic wave producing a deflection angle
of 2◦ ; (ii) the acoustic power required to transfer all the input power to the diffracted
beam.
Chapter 5
Semiconductor Devices
Abstract The most important family of photonic devices is based on semiconductor

materials. Semiconductor lasers have high efficiency and long lifetime, are fabricated
by the same epitaxial growth techniques used in silicon microelectronics, can be
designed to emit at almost any wavelength in the visible and near infrared range.
In this chapter, after an introduction to the optical properties of semiconductors,
the double-heterojunction structures used for semiconductor lasers are described,
along with the main characteristics of laser emission. The subsequent sections deal
with semiconductor amplifiers and light emitting diodes. The last part of the chapter
concerns photodetectors and electro-absorption modulators.
5.1 Optical Properties of Semiconductors
The lasers described in Chap. 3 are based on transitions among energy levels of
single atoms (or ions or molecules). In the case of semiconductor lasers the situation
is different because the involved energy levels belong to the whole crystal, and not
to a single atom.
In this section, a very simplified qualitative picture of the optical properties of
semiconductors is given, with the aim of offering an intuitive description of the
main processes involved in the interactions between a semiconductor crystal and
an optical wave. From a phenomenological point of view semiconductors are solids
whose electrical conductivity is typically intermediate between that of a metal and
that of an insulator. Their conductivity can significantly increase by increasing the
temperature or doping the material with impurities.
The atoms in a crystal lattice are strongly interacting. As a result, instead of
individual energy levels, energy bands belonging to the whole crystal are formed. In
the absence of external excitations, the bands are either fully occupied by electrons or
are totally unoccupied (except for the special case of metals). The highest-energy fully
occupied band is known as the valence band, while the lowest-energy unoccupied
band is called the conduction band. These two bands are separated by a forbidden
energy band, with bandgap energy E g . Values of E g for various semiconductors are
given in Table 5.1.

DOI 10.1007/978-3-319-20627-1_5
146 5 Semiconductor Devices
Table 5.1 Properties of some semiconductors

Semiconductor E g (eV) λg (µm) Type of gap
Ge 0.66 1.88 I
Si 1.11 1.15 I
AlP 2.45 0.52 I
AlAs 2.16 0.57 I
AlSb 1.58 0.75 I
GaP 2.26 0.55 I
GaAs 1.45 0.85 D
GaSb 0.73 1.70 D
GaN 3.45 0.36 D
InN 2.00 0.62 D
InP 1.35 0.92 D
InAs 0.36 3.50 D
InSb 0.17 7.30 D
In the fourth column I and D indicate, respectively, indirect bandgap and direct bandgap
At an absolute temperature of zero, an ideal semiconductor crystal has zero electric

conductivity because there are no free charge carriers. An external energy source can
excite an electron, causing a transition from the valence band to the conduction band.
The transition leaves a hole in the valence band. Both electrons in the conduction
band and holes in the valence band are mobile charge carriers contributing to the
electric conductivity of the crystal.
The absorption of a photon with an energy of hν ≥ E g excites an electron to the
conduction band. By analogy with the case of an isolated atom, one could expect
that the excited electron decays spontaneously to the valence band by emitting a
photon. The situation inside the crystal is, however, more complicated, because the
electron must be treated as a wave-like particle, with a wave-vector k and momentum
p = hk/(2π ). Near the bottom of the conduction band, the energy versus momentum
relation is approximately parabolic:
h2k2
E = Ec + , (5.1)
8π 2 m c
where m c , called effective mass of the electron, is different from the free electron
mass m e because of the effect of the interactions with the ions of the crystal lattice. A
similar energy-momentum relation can be written for holes near the top of the valence
band. The illustration in Fig. 5.1 qualitatively shows the band structures of two semi-
conductors, Si and GaAs. In the case of Si, the minimum of the conduction band and
the maximum of the valence band do not occur at the same k value. A semiconduc-
tor that has such a band characteristic is called an indirect-bandgap semiconductor.
5.1 Optical Properties of Semiconductors 147
Fig. 5.1 Transitions in direct- and indirect-gap semiconductors
In contrast, a semiconductor like GaAs is a direct-bandgap semiconductor, because

the minimum of the conduction band and the maximum of the valence band occur
at the same k value. Table 5.1 indicates that both silicon (Si) and germanium (Ge)
are indirect-bandgap materials, whereas several semiconductors made by combining
a trivalent with a pentavalent atom, known as III–V semiconductors, have a direct
bandgap. By absorbing a photon, the electron takes up not only the energy, but also
the momentum of the photon. The order of magnitude of electron wave-vectors is
k ≈ 2π/a, where a ≈ 10−8 cm is the crystal lattice constant. The modulus of the
photon wave-vector is 2π/λ. In the case of visible or near infrared light, λ is about
four orders of magnitude larger than a, so that it can be said that the electron momen-
tum is practically unchanged by the photon absorption process. Therefore, as shown
in Fig. 5.1, the transition from the valence band to the conduction band is vertical in
the E, k plane.
As a general rule the excited electron will always go to occupy the lowest energy
level in the conduction band. In correspondence, the hole will go to the highest
available level in the valence band. To ensure momentum conservation, the decay of
the excited electron to the valence band (also called electron-hole recombination) can
be radiative (that is, a photon is emitted) only if the bottom of the conduction band
corresponds to the top of the valence band, i.e. if they have the same wave-number.
The maximum wavelength of light that can be absorbed (and emitted) by a given
semiconductor is:
hc
λg = . (5.2)
Eg
Values of λg are reported in Table 5.1. In the case of indirect-bandgap, the excited
electron quickly relaxes to the bottom of the conduction band, by making an intra-
band transition in which energy and momentum are exchanged with the crystal
lattice. At that point the electron-hole recombination is non-radiative, and involves
the creation of a phonon.
In an undoped semiconductor (usually called an “intrinsic” semiconductor), at
temperature T , there are always some electron-hole pairs, due to thermal excitation.
The electron concentration n i is equal to the hole concentration pi . Both concentra-
tions are small at room temperature, because k B T E g .
A large concentration of conduction electrons can be created by doping the semi-

conductor with “donor” atoms having excess valence electrons with respect to the
semiconductor. As an example, silicon, which is a tetravalent element, becomes an
n-type semiconductor when a small amount of pentavalent atoms, such as phos-
phorus (P) or arsenic (As), are added to it. The impurity atom needs only four of
its outer-shell electrons to bind with the neighboring Si atoms. This leaves its fifth
electron only loosely bound, and free to move into the conduction band, after a
room-temperature thermal excitation. In a similar way, n-type gallium arsenide can
be created by doping the GaAs with an element from the sixth column of the peri-
odic table, such as selenium (Se), which substitutes for an As atom in the lattice. To
form p-type semiconductors with large hole populations, the crystal is doped with
“acceptor” elements, having a deficiency of valence electrons with respect to the
semiconductor.
By itself, a doped semiconductor is not better than the intrinsic one, from the
point of view of light emission. In fact it can be shown that the product of the
electron and hole concentrations, n and p, is constant for a given semiconductor at
fixed temperature, that is, np = n i pi . The application of this law to an n-doped
semiconductor shows that, when n is above n i , p becomes smaller than pi . Since the
number of recombination processes is proportional to the product np, the number of
emitted photons is independent of the doping of the semiconductor.
5.2 Semiconductor Lasers
Efficient light emission from a direct-bandgap semiconductor requires the presence

of a large population of both type of carriers in the same crystal region. Such a
condition can be realized in the depletion region of a p-n junction. Semiconductor
lasers, which are currently the most important class of lasers, are based on forward-
biased junctions. For this reason they are also known as diode lasers.
5.2.1 Homojunction Laser
In an n-type semiconductor at low temperature electrons occupy all the energy states
of the valence band plus the lower energy states of the conduction band up to a
maximum level, E Fn , known as the Fermi energy. In a similar way, inside a p-type
semiconductor electrons occupy the energy levels up to a maximum level, E F p ,
known as the Fermi energy of the p-type semiconductor, which is located inside
the valence band. When a p-type and an n-type regions are brought into contact,
electrons diffuse from the n-region into the p-region, and holes diffuse in the opposite
direction. After a short transient period, in which recombination processes occur in a
narrow region on both sides of the junction, a steady-state is reached when this region,
known as the depletion-layer, is completely depleted of mobile charge carriers. A
5.2 Semiconductor Lasers 149
Fig. 5.2 Electron potential energy in an unpolarized (left) and directly polarized (right) p-n junction
double-layer of fixed charges, positive ions on the n-side and negative ions on the
p-side, creates an electric field that blocks further diffusion of majority carriers. A
positive potential difference between n-region and p-region is established, as shown
in Fig. 5.2. At steady-state the two Fermi energies are equalized, E Fn = E F p , in
analogy to what happens in hydrostatics when a liquid fills to the same level in a set
of communicating vases, regardless of their different shapes and volumes.
By applying a positive potential difference V between the p- and the n-regions,
an electric current i flows across the junction. The current-voltage characteristic of
the junction follows the law:

V
i = i s exp −1 , (5.3)
Vo
where i s is the inverse saturation current, and Vo = k B T /e. The inverse current is
small, but non-zero, because minority carriers (electrons in the p region and holes
in the n region) are always present at non-zero temperatures. The p-n junction acts
as a diode. Diodes are unidirectional devices that are used in electronics for voltage
rectification, logic circuits, and many other applications.
When the junction is forward-biased, that is V ≥ 0, the flow of majority carriers
across the junction produces large concentrations of both electrons and holes in the
depletion region, causing a high number of recombination processes to occur. If the
diode is fabricated from an indirect-bandgap semiconductor, like Si, then the energy
from the non-radiative recombination is transferred to the crystal lattice as vibra-
tional energy. If, however, the junction is made with a direct-bandgap material, then
the recombination is radiative, and the depletion region becomes a light emitter in a
wavelength range close to λg . At low current density, mainly spontaneous recombi-
nation processes occur. However, at large enough carrier concentrations, stimulated
electron-hole recombination processes become important, making optical amplifi-
cation possible.
As shown in Fig. 5.2, the forward bias makes the two Fermi levels unequal, E Fn
becomes larger than E F p . The difference ΔE F (J ) = E Fn − E F p grows as a function
of the current density J flowing across the junction. It is possible to have optical gain
Fig. 5.3 Structure of the homojunction laser
only if ΔE F (J ) becomes larger than E g . Precisely, light emission occurs in the

frequency band determined by the condition:
ΔE F (J ) ≥ hν ≥ E g . (5.4)
Equation (5.4) can be seen as the condition for population inversion in the current-
injected semiconductor.
The first semiconductor laser was operated in 1962 using a Ga As p-n junction.
Since a junction in which the same material is used for both the p and the n sides
is called a homojunction, this laser is known as a homojunction laser. The light
emission was in the near infrared, at a wavelength close to λg = 0.85 µm. The basic
structure of the homojunction laser is shown in Fig. 5.3. The electric current flows
along the x axis. The active region is the depletion layer with a length of 0.2–0.5 mm
and a width, in the y direction, of 0.1–0.2 mm. The thickness of the active layer along
x is very small, typically around 1 µm.
The feedback needed for laser action inside the Fabry-Perot cavity is provided
by the reflections from the two parallel end faces, which are prepared by cleavage
along a crystal plane. Since the refractive index of a semiconductor is fairly large,
the refractive-index mismatch at the air-semiconductor interface gives a sufficiently
high reflectance without the need for reflective coatings of the end facets. As an
example, the refractive index of GaAs is n = 3.6, so that the reflectance from (2.34)
is R ≈ 30 %.
The homojunction laser was operated continuously at the liquid nitrogen tem-
perature (T = 77 K), with a threshold current density Jth ≈ 1 kA/cm2 . When T
is increased, more and more electrons can occupy energy states above Fermi lev-
els. This has the effect of reducing the optical gain. It can be shown that Jth grows
exponentially with T , according to the law:

T
Jth (T ) = Jth (To ) exp −1 . (5.5)
To
At room temperature, Jth becomes of the order of 100 kA/cm2 . For such large
current densities, the junction is very quickly damaged, so that only pulsed opera-
tion is possible. A drawback of the homojunction laser is that the size of the laser
beam in the x direction is much larger than the thickness of the active region. As a
consequence, a large fraction of the beam power travels inside the p and n regions,
where it is absorbed. Another drawback is that the charge carriers injected inside
the depletion region quickly drift outside this region, so that only a small fraction is
available for stimulated electron-hole recombination.
5.2.2 Double-Heterojunction Structures
The performance of the semiconductor laser was greatly improved by using a double
heterojunction, instead of the homojunction described in the preceding section. The
double-heterojunction laser has an active medium sandwiched between p and n
materials that differ from the active material. This laser was invented around 1970
by Alferov and Kroemer, who were awarded the Nobel prize for physics in the year
2000. As shown in Fig. 5.4, the active region p-GaAs is sandwiched between the
regions p-Ga1−x Al x As and n-Ga1−x Al x As, where x is the fraction of aluminum
atoms. Typically, x = 0.3. Such a structure has many positive features, as illustrated
in Fig. 5.5. The bandgap of the active layer (E g1 = 1.5 eV for GaAs) is smaller
than that of the cladding layers (E g2 = 1.8 eV for Ga0.7 Al0.3 As). Therefore, the
“wings” of the laser beam are not absorbed by the cladding regions. As a general
rule for III–V semiconductors, a change in composition that leads to an increase in
bandgap also results in a reduced refractive index. In fact, the refractive index of
Ga0.7 Al0.3 As is 3.4, smaller than that of GaAs, so that an optical waveguide effect
exists for the laser beam. An additional benefit of the double-heterojunction is that
the difference between the bandgaps of the two junctions creates an energy barrier
that increases the carrier concentration in the active region. By increasing the carrier
concentration, the optical gain increases.
By using the double-heterojunction structure, the room-temperature threshold cur-
rent density is reduced by two orders of magnitude with respect to the homojunction
structure, reaching a value Jth ≈ 0.5 kA/cm2 .
In order to form a double-heterostructure without introducing material strain, a
very important condition must be satisfied, namely that the lattice period of the active
layer must closely match that of the cladding layers. A mismatch gives rise to the
Fig. 5.4 Double-

heterojunction in
GaAs
Fig. 5.5 Double heterojunction: a behavior of the energy bands; b transverse profile of the laser
beam; c refractive index profile
Fig. 5.6 Lattice constants of different semiconductors. Taken from Opensource Handbook of
Nanoscience and Nanotechnology
formation of interface defects known as misfit dislocations. These defects act as

centers for electron-hole non-radiative recombination that reduce the optical gain.
The lattice constants of different semiconductors are given in Fig. 5.6, together with
the corresponding bandgap values. Since the radii of Ga and Al ions are almost the
Fig. 5.7 The

double-heterojunction laser:
stripe-geometry
configuration
same, the lattice periods of GaAs and AlAs coincide, and thus lattice-matching for
the GaAs/GaAlAs structure is very good. In the case of the quaternary compound
Ga1−x I n x As1−y Py , which is used for optical-communications lasers, the alloy can
be lattice-matched to cladding layers made of InP by choosing a specific y/x ratio.
It is found that the optimal value is y = 2.2 x.
Stripe geometry. Double-heterojunction lasers often use the so-called stripe-geometry
configuration shown in Fig. 5.7. By introducing an insulating oxide layer, current
from the positive electrode is constrained to flow in a narrow stripe. By reducing
the cross-section, the total current is correspondingly reduced for a given current
density. Another method to reduce the cross-section is that of creating a refractive
index profile that produces a confinement along the y-direction, as is the case for
both the “channeled substrate laser”, and “buried heterostructure laser” structures.
The size of the active region along y can be as small as a few µm, with a threshold
current, i th , around 10 mA and applied voltage of 1.5–2 V.
Distributed feedback. As it has been discussed in Sect. 3.6, a Fabry-Perot laser
can operate with more than one longitudinal mode. The actual number of oscillating
modes is determined by the ratio between the frequency band Δν L of the amplifier and
the distance Δν between two adjacent longitudinal modes. In several applications,
among which optical communications is the most important, emission at a single
frequency is necessary. Among the different methods to realize a single-mode semi-
conductor laser, the distributed feedback geometry is the most widely used. Figure 5.8
shows the scheme of a distributed feedback laser, also known as DFB laser, which
consists of an active medium in which a periodic thickness variation is introduced in
one of the cladding layers. The periodic corrugation provides optical feedback for the
laser oscillation, and the end facets are coated with antireflection coatings. The laser
behaves as an optical amplifier in which the forward and the backward beam have a
frequency-selective coupling. These two counter-propagating beams are efficiently
coupled to each other only if their wavelength is such that
λ = 2n eff Λ (5.6)
Fig. 5.8 Distributed feedback laser
where Λ is the pitch of the periodic corrugation and n eff is an effective refractive
index. The exact value of n eff depends on specific details of the structure. The fab-
rication of a high-quality uniform grating is rather difficult, because the pitch has
submicron periodicity. For example, in the case of a 1550-nm InGaAsP laser with
n eff = 3.4, one finds that Λ = 0.23 µm.
Surface-emitting structure. The area of semiconductor lasers is in continuous evo-
lution, and many different types of structures have been tested. The semiconductor
lasers discussed so far are edge-emitting, i.e. they generate an optical beam prop-
agating parallel to the junction plane. However, in some cases, it is interesting to
have laser emission normal to the junction plane. Such lasers are known as vertical-
cavity surface-emitting lasers, or VCSELs. Compared to the edge-emitting laser,
the VCSEL has a much shorter length of active medium, and a much broader emit-
ting area. Since the single-pass gain is low, high-reflectivity mirrors must be used
to avoid high threshold currents. As shown in Fig. 5.9, such mirrors are made by
alternating several layers of two different semiconductors, following the multilayer
scheme described in Sect. 2.2.4. Due to the short length of the laser cavity, the spac-
Fig. 5.9 Structure of a vertical-cavity surface-emitting laser

Fig. 5.10 Active layer of a semiconductor laser made of a multiple quantum well, with a composi-
tion given by I n 0.5 Ga0.5 P/I n 0.5 Ga0.5−x Al x P. The plot represents the behavior of the Al fraction,
x, across the junction
ing of Δν between adjacent longitudinal modes is very large. Thus, if the frequency
of one mode coincides with the peak reflectivity of the multilayer mirrors, the two
adjacent modes fall outside the high-reflectivity band of the mirrors, and single-
longitudinal-mode operation is achieved. By limiting the diameter of the emission
area to 5–10 µm, a very low threshold current is obtained (i th ≈ 1 mA). Correspond-
ingly, the output power is also small, typically of the order of 1 mW. However, it is
possible to build large-area (300 µm diameter) VCSELs with output power of watts.
Multiple-quantum-well. A very interesting concept recently emerged in many dis-
tinct scientific and technological areas is the use of artificially tailored structures at
the nanoscale. It is beyond the scope of this text to explain the properties of nanostruc-
tured materials, but the important aspect is that electrons confined in nanostructures
have discrete energy levels, instead of continuous, with the energy of those levels
fixed by the geometry of the structure, following the rules of quantum mechanics. As
an example, Fig. 5.10 shows the nanostructured active layer of a semiconductor laser,
in which the active region consists of a number of alternating layers of narrow and
wide bandgap materials. The thickness of the narrow-bandgap layers (the “quantum
wells”) and that of the wide-bandgap barriers are both about 5 nm. A laser of this
type is known as a “multiple quantum well” (MQW) laser. MQW lasers are realized
in both edge-emitting and surface-emitting configurations. In general, they have a
very low threshold current density and a high efficiency. They also have a very stable
emission wavelength, not influenced by changes in electric current or temperature.
5.2.3 Emission Properties
Laser emission has been obtained from many different III–V semiconductors.
Table 5.2 lists the most common semiconductor lasers. The trivalent atom can be
gallium or aluminum or indium. The pentavalent atom can be arsenic or phosphorus
or antimony (Sb) or nitrogen (N ). By using combinations of two trivalent atoms
and/or two pentavalent ones, the emission wavelength can be tuned, in some cases,
Table 5.2 Operating wavelength range of semiconductor lasers

Compound λ (µm) Compound λ (µm)
I n x Ga1−x N 0.37–0.50 Ga0.42 I n 0.58 As0.9 P0.1 1.55
Ga1−x Al x As 0.62–0.85 I nGa As Sb 1.7–4.4
Ga0.5 I n 0.5 P 0.67 PbCd S 2–4
Ga0.8 I n 0.2 As 0.98 PbSSe 4.2–8
Ga1−x I n x As1−y Py 1.10–1.65 PbSnSe 8–32
Ga0.27 I n 0.73 As0.58 P0.42 1.31 Quantum cascade 2–70
over a very broad range. This is a distinctive property of semiconductor lasers: solid-
state and gas lasers operate at discrete wavelengths fixed by the position of atomic or
ionic energy levels (with a few exceptions), but semiconductor lasers can be designed
to operate at essentially any wavelength in the visible and near infrared.
The wavelength range 0.62–0.85 µm is covered by devices using Ga1−x Al x As.
Since the lattice constant of Ga1−x Al x As does not depend on the fraction x of Al
atoms (see Fig. 5.6), heterojunctions containing any fraction of Al can be grown
epitaxially. A limit to using large values of x comes from the fact that the compound
acquires an indirect bandgap when x ≥ 0.7. Ga1−x Al x As lasers emitting at 0.81 µm
have an important application for pumping Nd-YAG lasers, as described in Sect. 3.9.
Optical-communication lasers, emitting around 1.31 or 1.55 µm are based on the
quaternary compound Ga1−x I n x As1−y Py grown on top of InP. By changing the
fractions x and y, it would be possible to cover a wide range of bandgap energies, from
0.4 to 2.2 eV. In practice, because of the mismatch in atomic sizes between gallium
and indium, and also between arsenic and phosphorus, only the range 0.7–1.4 eV
(λ = 0.88–1.76 µm) can be usefully covered. The Ga0.8 I n 0.2 As laser, emitting at
0.98 µm, has an important role as the pump of the fiber-optic amplifier described
in the following chapter. III-V semiconductors using nitrogen as the pentavalent
atom exhibit a large bandgap, and so can emit blue and violet light. In particular,
the wavelength range 0.37–0.50 µm is covered by the I n x Ga1−x N laser, which is
useful for a variety of applications, such as writing and reading Blu-ray optical discs.
By using II–V I semiconductors, lasers operating in the mid- and far-infrared
range of wavelengths can be made. Sources in this range of wavelengths are use-
ful for such applications as chemical sensing and trace-gas analysis. Since the laser
efficiency strongly decreases when the photon energy becomes comparable to k B T ,
all these devices work at cryogenic temperatures. The same wavelength range can
be covered much more efficiently by a new family of semiconductor lasers, termed
quantum cascade lasers (QCL), which make use of MQW structures in III–V semi-
conductors. It goes beyond the scope of this text to describe the physical mecha-
nism of QCLs, which is completely different from that of diode lasers. It suffices to
mention that the QCL can be designed to emit at any wavelength in the mid- and
far-infrared, through appropriate choice of the nanostructure. The QCL can operate
at room temperature as a continuous source or as a source of picosecond pulses.
Fig. 5.11 Output power

versus input current for a
double-heterojunction laser,
at three different
temperatures
A double-heterojunction laser has an internal efficiency, defined as the ratio

between the number of radiative recombinations and the number of injected electron-
hole pairs, that can be close to 100 %. The overall efficiency of the laser, defined
as the ratio between emitted optical power and injected electric power, often termed
“wall-plug efficiency”, can be as large as 50 %.
Experimental curves of output power versus input current, at three different tem-
peratures, are shown in Fig. 5.11. It is found that the threshold current grows with
increasing the temperature, as predicted by (5.5) with a best-fit parameter To ≈ 160 K.
The output beam of a semiconductor laser is usually strongly diverging, because
of the smallness of the transverse size of the active region. Typically, the beam
observed at the output facet of a GaAs laser has an elliptical cross-section with a size
of 1 µm in the direction orthogonal to the junction (x direction), and of 10 µm in the
y direction. The far-field angular spread due to diffraction is about 50◦ along x and
5◦ along y. By using a cylindrical lens with two different focal lengths along x and
y, the beam can be collimated to obtain a circular spot having a divergence angle of
just a few milliradians.
The spectrum of the laser emission has a peak wavelength λ p that depends on pump
current and temperature. The wavelength interval Δλ L over which laser oscillation
can occur grows as the pump current increases, as described by (5.4). Some typical
values of Δλ L are reported in Table 5.3. If Δλ L grows and, at the same time, the
highest emission wavelength remains fixed to hc/E g , then the central wavelength
Table 5.3 Spectral properties of three semiconductor lasers

λ = 850 nm λ = 1300 nm λ = 1550 nm
Δλ L in nm 2.0 5.0 7.0
Δλ in nm 0.27 0.6 0.9
dλ p /dT in nm/◦ C 0.22 0.5 0.73
dλq /dT in nm/◦ C 0.06 0.12 0.18
of the emission spectrum, λ p , decreases with increasing pump current, at fixed

temperature.
Since E g is a decreasing function of T , λ p grows with T . Some typical values
of the derivative dλ p /dT are reported in Table 5.3. Clearly, in order to stabilize the
emission wavelength, precise temperature control is needed.
Table 5.3 also shows typical values for the wavelength separation between two
adjacent longitudinal modes, which, according to Sect. 3.6, is expressed as Δλ =
λq − λq−1 ≈ λ p 2 /(2L ). The optical path L is very short in the semiconductor laser,
so that the modes are rather widely spaced. However the bandwidth Δλ L is also large.
As a consequence, the laser usually oscillates on several modes. Using the data in
Table 5.3, the number of oscillating modes, as calculated from N = Δλ L /Δλ, is 7–8.
Since the index of refraction of semiconductors increases with T , the wavelength of
the generic qth mode also increases with T . Values of dλq /dT are reported in the
last row of the same table.
The CW output power of a semiconductor laser can only be several hundreds of
milliwatts, at most. By fabricating a two-dimensional array of lasers on a single sub-
strate, and then coupling all the outputs into a single optical fiber, a monochromatic
optical power of several tens of watts can be achieved. This scheme is often used to
pump solid-state and fiber lasers.
An important feature of a semiconductor laser is its very long lifetime: as an
example, a test conducted on a set of GaAlAs lasers, operating at 70 ◦ C with an
output power of 5 mW, has measured a lifetime of 30,000 h (3.5 years of continuous
operation). GaInAsP lasers have been found to have even longer lifetimes.
By pulsing the pump, a semiconductor laser can be made to emit short pulses, for
use, as an example, in portable optical radars. It can also be intensity modulated by
controlling the pump current. The modulating signal (typically 40–60 mA peak-to-
peak) is superposed to the direct bias current. The maximum frequency of modulation
depends on how quickly the carrier density in the active region can be varied. Typical
values are of the order of a few GHz, but, in specially designed devices, modulation
frequencies of 20–30 GHz have been demonstrated.
Diode lasers can also generate ultrashort pulses, using either active or passive
mode-locking. Typically, pulse-widths of between 0.5 and 5 ps with repetition rates
of between 1 and 100 GHz can be obtained. For repetition rates below roughly
10 GHz, an external cavity setup is generally required. Particularly interesting are the
vertical-external-cavity surface-emitting lasers (VECSELs), which have a vertical-
cavity surface-emitting structure with one of the two high-reflecting mirrors external
to the diode structure. Typical cavity lengths are from a few millimeters up to tens of
centimeters. The gain chip generally contains a highly reflective bottom section to
reflect the laser and pump light, an active semiconductor gain section, and an anti-
reflective top layer. By passively mode-locking optically pumped VECSELs, it is
possible to obtain much higher pulse energies than those generated by edge-emitting
structures.
5.3 Semiconductor Amplifiers 159
5.3 Semiconductor Amplifiers
A semiconductor optical amplifier, usually known as SOA, is realized by eliminating

the feedback elements in a semiconductor laser, usually by putting an anti-reflection
coating on the two facets of the active region. Of course, the operating wavelengths
of SOAs are the same as those of the corresponding lasers, listed in Table 5.2.
Typically, at λ = 1.55 µm, the gain per unit length g is of the order of 105 m−1
and the gain bandwidth is about 10 THz. Since g is somewhat dependent on the
polarization state of the input signal, a simple SOA does not generally preserve the
input polarization. However, it is possible to implement more complex amplifying
schemes that are polarization-preserving.
SOAs are attractive for many applications because they are small, have a large
optical gain, and can be driven by a low-voltage supply. However, the small size of
the active region makes the input coupling rather critical, while the large gain can
give rise to signal distortions because of saturation effects.
The most important applications of SOAs are in the optical-communication area,
as will be discussed in Chap. 8.
5.4 Electroluminescent Diodes
The electroluminescent diodes fabricated from semiconductors are known as light

emitting diodes (LEDs). An LED is a solid-state light source using the spontaneous
emission generated by radiative electron-hole recombinations in the active region of
the diode under forward bias. Unlike a laser, an LED does not have an optical cavity
or a threshold current. It starts emitting light as soon as a forward bias voltage is
applied to the junction, and emits incoherent and unpolarized spontaneous photons.
As in a laser, the emission wavelength is determined by the bandgap energy of the
used semiconductor. An other important electroluminescent diode is the organic
light-emitting diode, known as OLED, which is based on organic compounds.
5.4.1 LEDs
The first visible LED, emitting at 710 nm, was demonstrated by Holonyak in
1962 using a GaAsP diode. After this realization the research concerning electro-
luminescent materials was extended to many different III–V semiconductors. An
important step was made in 1994 with the invention of the blue LED. Akasaki,
Amano, and Nakamura were awarded the Nobel prize in Physics in the year 2014
for this invention. The light emitted by an LED has a wavelength λg = hc/E g , and
a bandwidth Δλ given by:
Fig. 5.12 LED emission in

the three fundamental colors
2k B T λg 2
Δλ = (5.7)
hc
As an example, an LED emitting 500-nm light at room temperature, has an emis-
sion bandwidth predicted by (5.7) of about 10 nm. Commercially available LEDs
cover the spectral range from the near ultraviolet to the near infrared, with wave-
lengths ranging from about 370 nm to 1.65 µm. As shown in Fig. 5.12, red-orange
LEDs use Ga Asx P1−x , and blue-green LEDs use I n 1−x Gax N . The emission wave-
length is tuned by changing the composition, which modifies the parameter of
x. Some LEDs even use indirect-bandgap semiconductors doped with impurities
although these offer lower efficiency.
The structure of a surface-emitting LED is shown in Fig. 5.13. Spontaneous emis-
sion radiates in all directions, but the downward-emitted light is absorbed, and the
extraction of the upward-emitted light is possible only over a limited solid angle.
The limitation to extraction efficiency is due to reflection from the semiconductor-air
interface. The spontaneous photons that reach this interface at an angle of incidence
larger than the limit angle θc are totally internally reflected. If the refractive index
of the semiconductor is 3.4, then this critical angle is just 17◦ , and so the extrac-
tion efficiency is very low. Various techniques can be used to increase the extraction
efficiency: the reflectance of the semiconductor-air interface can be reduced by an
anti-reflection coating, and the downward-emitted light can be recuperated by using
a transparent substrate associated to a highly-reflecting bottom surface.
An important parameter that characterizes a light source is the ratio between
emitted optical power and injected current. For a visible LED, this ratio is around
0.5 W/A. The overall LED efficiency, η, can be as large as 25 %. As a comparison,
the efficiency of an incandescent lamp is about 1 %. High efficiency also implies
less heat generation and less damage. The lifetime of an LED can reach 50,000 h of
continuous operation.
LEDs are currently used in a large variety of applications, some of which are
described in Chap. 8. When an LED replaces traditional lamps, then white light
emission is usually required. White light can be obtained by mixing the emissions
5.4 Electroluminescent Diodes 161
Fig. 5.13 Structure of a GaAsP LED. A layer of Te-doped n–Ga As0.6 P0.4 is grown on the substrate
n–Ga As. The double heterojunction is formed by superposing a layer of Zn-doped p–Ga As0.6 P0.4 .
The LED emits in the upward direction, while the downward emitted light is absorbed by the
substrate
of three LEDs operating at the three fundamental colors (red-green-blue, RGB).

However, in practice, it is often more convenient to make use of a single blue-emitting
LED coated with phosphor molecules. The phosphors absorb the blue emission, and
spontaneously re-emit light of a longer wavelength. By mixing different phosphors
with different colors, white light emission is achieved. The design and fabrication of
this white light source is both simpler and cheaper than a complex RGB system.
5.4.2 OLEDs
In an organic light-emitting diode (OLED), the electroluminescent material is an

organic compound, which is electrically conductive and behaves as an organic semi-
conductor. Two thin films of organic semiconductor, forming an organic heterostruc-
ture, are sandwiched between two electrodes on a transparent substrate. Electronic
charges are transported and injected into the heterostructure from the electrodes:
electrons from the cathode, and holes from the anode. In analogy to the case of semi-
conductor devices, the radiative recombination of electron and hole generates light.
The wavelength of this emission depends on the band gap of the organic material.
OLEDs can be used to produce light of a very wide range of wavelengths by modi-
fying the precise structure of the organic material. OLEDs are fabricated from small
organic molecules or conjugated polymer chains, by using standard techniques of
thin film deposition. These are simpler and less expensive than the epitaxial growth
used for fabricating LEDs. The potential applications of OLEDs include thin, low-
cost displays with a low driving voltage, wide viewing angle, and high contrast and
color gamut. Polymer LEDs have the added benefit of offering the possibility for
printable and flexible displays. OLEDs are used to make self-luminous displays for
portable electronic devices such as cellular phones, digital cameras, and as a source of
backlighting for liquid-crystal displays. A recent development is the White Organic
Light-Emitting Diode (WOLED), in which red, green, and blue OLED materials are
sandwiched together. WOLEDs have potential for use in television technology and
architectural lighting.
5.5 Photodetectors
A photodetector is a device that measures optical power by absorbing energy from

an optical beam. All photodetectors are square-law detectors that respond to the
intensity, rather than the field amplitude, of an optical signal. Commonly used pho-
todetectors are based on either thermal effects or electric effects.
Thermal detectors convert the absorbed energy into heat. In a bolometer, light is
absorbed by a metallic or semiconducting resistor, or by a termistor. The tempera-
ture variation induces a change in the electric resistance of the absorbing material,
which is measured, for instance, by a Wheatstone bridge. Highly sensitive mod-
ern bolometers use superconducting absorbers operating at cryogenic temperatures.
Another possibility is to use a thermocouple, which is typically made by a ring with
two bimetallic junctions. If one junction is kept at fixed temperature, and the other
is heated by light absorption, then an electric current flows inside the ring, because
of an effect known as thermoelectric effect or Seebeck effect. The intensity of the
electric current is proportional to the absorbed optical power. Thermocouples are
self powered and require no external form of excitation. In general, the response of
thermal detectors does not depend on the wavelength of the incident radiation. How-
ever, they have a slower response time and lower sensitivity, compared to electric
detectors.
More important then thermal detectors are the detectors that use electric effects,
which include photoelectric detectors, photoconductors, and photodiodes. They can
be made to be extremely sensitive, and can have response times capable of following
very fast optical signals.
5.5.1 Photoelectric Detectors
The photoelectric effect is the emission of an electron when a material absorbs energy
from a light beam. The emitted electrons are called photoelectrons. The effect was
discovered by Hertz in 1887 and explained by Einstein in 1905, for which he was
5.5 Photodetectors 163
awarded the Nobel Prize in 1921. The photoelectron is generated only if the photon
energy hν satisfies the condition:
hν ≥ E i , (5.8)
where E i is the ionization energy of the electron inside the material. In the case of
solids, E i is usually known as the “work function”. The excess energy, hν − E i ,
is transformed in kinetic energy of the photoelectron. The material can be a gas,
a liquid, or a solid. Considering isolated atoms in the ground state, low ionization
energies are found for elements belonging in the first column of the periodic table,
i.e. the alkali metals. This is not surprising because these elements have only one
electron in their outer electron shell. The cesium atom has the lowest binding energy
of E i = 3.8 eV. The highest binding energy is 24.5 eV for helium, a noble gas.
Equation (5.8) states that a photoelectron can only be generated if the wavelength of
the photon is below a threshold value given by:
hc
λth = . (5.9)
Ei
For the cesium atom, λth = 327 nm, which is in the ultraviolet. While no isolated
atom emits photoelectrons under excitation from visible light, the situation is more
favorable for solids. The work function of solid cesium is 1.94 eV, corresponding to
a threshold wavelength of λth = 640 nm. Even lower values of the work function are
found for cesium compounds, such as cesium antimonide. The threshold wavelength
can reach a value of about 1.1 µm, extending the applicability of photoelectric
detectors to the near infrared.
Photoelectric cell. A photoelectric cell, also called a vacuum photodiode, can be
made by collecting the photoelectrons emitted by the photosensitive surface of a
negative electrode (photocathode) at a positive electrode (anode). The quantum effi-
ciency of the photosensitive surface layer, η, defined as the ratio between the number
of photoelectrons and the number of absorbed photons, can vary between 10−4 and
10−1 , depending on the incident wavelength and the nature of the surface layer. The
two electrodes are placed in a vacuum tube with a transparent window. An electric
current flows in a closed circuit configuration under illumination of the photocathode.
The photocurrent i ph is obtained by multiplying the electron charge by the number
of photoelectrons collected by the anode per second:
ηe P
i ph = , (5.10)
hν
where P is the incident optical power. The sensitivity of the photoelectric cell is
expressed by the ratio i ph /P (Ampère/Watt). As an example, if hν = 2 eV (red
light) and η = 0.1, then the sensitivity is 50 mA/W.
Fig. 5.14 Structure of a photomultiplier tube
Photomultiplier tube. The sensitivity of a photoelectric detector can be improved

by many orders of magnitude by multiplying the number of emitted electrons via the
process of secondary emission. This occurs when the accelerated electrons impact
intermediate electrodes, called dynodes, maintained at ever-increasing potential-
differences. The resulting device, known as a photomultiplier tube, is schematically
illustrated in Fig. 5.14. The photocathode contains combinations of materials such
as cesium, rubidium and antimony, specially selected to provide a low ionization
energy. The number of dynodes varies between 10 and 14. Their shape and posi-
tion are designed in order to guide the electrons along trajectories which maximize
the collection efficiency. Typically the potential difference between two successive
dynodes is around 100 V and the supply current is around 1 mA. A photoelectron
emitted by the photocathode generates m secondary electrons when it impacts the
first dynode. Each secondary electron, when collected by the second dynode, further
produces secondary electrons. If there are n dynodes, then the number of electrons
arriving at the anode is m n . Assuming m = 4 and n = 10, the gain of a photomulti-
plier tube is: G = m n = 410 ≈ 106 . With 14 dynodes, the gain can approach 108 . If
the quantum efficiency is the same, then the sensitivity of a photomultiplier tube is
larger than that of a photoelectric cell by a factor G.
The photoelectric effect has a very short intrinsic response time, comparable to
the oscillation period of the optical field. The response time of the device is limited
by two effects: (i) there is a spread in the transit time of the electrons through the
multiplication structure, such that the G electrons produced by a single photoelectron
do not simultaneously reach the anode, but arrive with a spread of a few nanoseconds;
(ii) the device has an equivalent capacitance, C, between the anode and the ground
that gives a time constant of R L C, because of the non-zero load resistance R L . If
C = 20 pF and R L = 50 , then the time constant is 1 ns.
A photomultiplier tube with a gain G = 108 produces an easily measurable output
pulse, in response to the absorption of a single photon. Indeed G electrons bunched in
a time interval τ give a current pulse with a peak value Ge/τ . Therefore the voltage
pulse measured across the load resistance has a peak value Ge R L /τ . If G = 108 ,
τ = 1 ns, R L = 50 , then the peak voltage is 0.8 V. When observing the single-
photoelectron pulses, very large amplitude variations are found from one pulse to
the other, because m is not a fixed number, but a statistical variable with a Poisson
probability distribution.
When the optical power incident on the photocathode is so low that the number
of photoelectrons per unit time is less than τ −1 , the electric signal at the photomulti-
plier output appears as a discrete sequence of pulses. The experimental observation
of this discreteness has a great conceptual relevance, because it demonstrates that
electromagnetic energy is absorbed by matter in quantum steps.
5.5.2 Semiconductor Photodetectors
Semiconductor photodetectors are based on the generation of an electron-hole pair

caused by the absorption of a photon. Sometimes this effect is called “internal pho-
toelectric effect”, whereas the electron extraction process described in the preceding
section is called external photoelectric effect. The photogeneration of electron-hole
pairs can be exploited for photodetection purposes in different ways.
Photoconductors. The conductivity of a semiconductor that has electron and hole
concentrations of n and p, respectively, is:
σ = e(μe n + μh p), (5.11)
where μe and μh are the electron and hole mobilities, respectively. When a semi-
conductor is illuminated with light of a sufficient photon energy, carriers in excess
of the equilibrium concentrations are generated. As a consequence, the conductiv-
ity increases by an amount that is proportional to the number of photogenerated
electron-hole pairs. Similarly to the photoelectric effect described in the previous
section, the photoconductivity has a threshold photon energy, and a corresponding
threshold wavelength, λth . For an intrinsic semiconductor, λth is determined by the
condition that the photon energy should be larger than the bandgap energy. Accord-
ing to the wavelength to be detected, different semiconductors, such as germanium,
silicon, lead sulfide, indium antimonide can be used. In the case of photoconductive
detectors operating at mid-infrared, low temperature operation is necessary, in order
to avoid that the thermal excitation of electron-hole pairs obscures the photoexcita-
tion. Instead of using small-bandgap semiconductors, the detection of mid-infrared
radiation is more frequently performed using doped photoconductors. As an example,
p-type germanium doped with Z n, cooled at 4 K, can detect mid-infrared radiation
up to wavelengths of about 40 µm.
Photodiodes. A very important category of photodetectors is represented by pho-
todiodes. A photodiode is a type of detector capable of converting light into either
current or voltage, depending upon the mode of operation. As discussed in Sect. 5.1,
a large internal electric field exists in the depletion layer of a p-n junction. If an
electron-hole pair is generated by photon absorption within this layer, then the two
charge carriers rapidly drift in opposite directions, the electron toward the n region
Fig. 5.15 Current-voltage characteristic of a photodiode. Under illumination the characteristic

shifts downward by an amount proportional to the light intensity
and the hole toward the p region. If the device is left in an open circuit configuration,
then an external potential appears between the p and n regions. This is known as
the photovoltaic effect. The usual way to operate the detector is to apply a reverse
voltage across the junction. The current-voltage characteristic expressed by (5.3), as
modified by photon absorption, becomes:
V
i = i s exp − 1 − i ph , (5.12)
Vo
where i ph , the photogenerated inverse current, is given by (5.10). As shown in

Fig. 5.15, the effect of illumination is that of vertically translating the current-voltage
characteristic. The internal resistance of a photodiode is large, and thus it behaves
essentially as a current generator.
Many photodiodes use a structure in which an intrinsic region is interposed
between the n and p regions. These photodiodes, known as p-i-n photodiodes exhibit
a larger sensitive volume, and faster response time, with respect to p-n photodiodes.
Furthermore the p-i-n structure has a width of the depletion region that is practically
independent of the applied voltage.
For the detection of visible and near-infrared radiation silicon photodiodes are
the most common choice, and can have a large efficiency: η = 0.4–0.7. For the
wavelength interval 1.0–1.6 µm, InGaAs photodiodes are mostly used.
Avalanche photodiodes. The current-voltage characteristics shown in Fig. 5.15 illus-
trate a feature that is not described by (5.12). When the reverse bias voltage is suf-
ficiently large, a sudden increase in reverse current is observed. What happens in
this region is that carriers crossing the depletion layer acquire sufficient energy to
ionize lattice atoms by collision. The carriers produced in this way may, in turn, gen-
erate further ionizing collisions in a cascade. This mechanism, known as “avalanche

breakdown”, can be destructive if the reverse current is allowed to become too large.
An avalanche photodiode (APD) is a photodiode designed to operate under very
high reverse bias, in order to exploit the internal amplification caused by the avalanche
breakdown. The applied voltage is kept slightly below the breakdown voltage, so that
the APD still behaves as a linear device, giving an output signal proportional to the
incident optical power. The actual value of the breakdown voltage depends on the
semiconductor material, device structure and temperature. It is typically around 50 V
for InGaAs/InP APDs.
Some particular types of APDs can reach such a high gain that they become
single-photon detectors, a kind of solid-state photomultiplier tube, but requiring a
supply voltage more than one order of magnitude lower than a real photomultiplier
tube. A possible drawback is that the gain G is rather sensitive to temperature and
aging.
5.5.3 CCD Image Sensors
It is possible to convert an optical image into a digital sequence of electrical signals

using an array of photodetectors. The invention of the charge-coupled-device (CCD)
structure represented a real revolution for image recording and transmitting. Boyle
and Smith were awarded the Nobel prize in Physics in the year 2009 for this invention.
The photosensitive surface is divided into a two-dimensional array of detectors.
The single detector, known as “pixel”, a term coming from the contraction of “picture
element”, is a metal-oxide-semiconductor (MOS) capacitor. In visible CCD cameras
the semiconductor is doped-silicon. A positive voltage applied between metal and
silicon traps the photogenerated electrons at the silicon surface. The number of
trapped electrons is proportional to the integral of the optical power incident on the
pixel during the measurement interval. The electric charge relative to each pixel is
amplified, sequentially read-out, and then recorded in digital form. Two-dimensional
arrays typically consist of 1024×1024 pixels. The pixel size is of the order of 10 µm.
Colored images can also be digitally converted by simultaneously recording images
in the three fundamental colors. Infrared CCD cameras, based on III–V or II–VI
semiconductors are also commercially available.
5.6 Electro-Absorption Modulators
An alternative to the electro-optic modulators described in Chap. 4 are the electro-

absorption modulators based on semiconductors. The idea is to exploit the change
in bandgap energy induced by an external electric field, an effect known as the
Franz-Keldish effect. To ensure that the semiconductor is transparent to the optical
signal, the wavelength λ of the signal to be modulated must be slightly larger than
Fig. 5.16 Qualitative

description of
electro-absorption;
modification of the
wavelength dependence of
the absorption coefficient
(proportional to the
photocurrent) at increasing
applied fields
λg , so that the photon energy is less than E g . The applied electric field induces a
decrease of E g , which is equivalent to an increase of λg , as shown in Fig. 5.16. If
the new value of E g is smaller or equal to the photon energy, then the modulated
semiconductor absorbs the light beam. Therefore a time-dependent electric field
induces an amplitude modulation of the optical field. Analogously, electro-absorption
switching can be realized.
Electro-absorption is different from the electro-optic effects described in Chap. 4,
because the effect of the applied field is to modify the imaginary part of the refractive
index, instead of the real part. The change of E g is rather small: typically, a field of
about 107 V/m is needed to induce a shift of bandgap energy of 20 meV.
An interesting aspect is that it is possible to design a monolithic structure in
which the electro-absorption modulator is integrated with a laser, to act as an external
modulator, or inside the laser cavity to act as an intra-cavity modulator for active
mode-locking.
The electro-absorption modulators used in optical communications (at wave-
lengths around 1.55 µm) are based on the quaternary compounds InGaAsP or
InGaAlAs grown on an InP substrate. These modulators often have the form of a
waveguide, with the electric field applied in a direction perpendicular to the propa-
gation direction of the light beam. Multiple quantum well structures can be used to
obtain high modulation speed and large extinction ratios, with low drive voltages.
As an example, with a waveguided structure of length 1.5 mm and drive voltage of
11 V, an extinction ratio of 30 (about 15 dB) and a maximum modulation frequency
of 20 GHz can be obtained.
Problems
5.1 Consider a semiconductor laser made of a Fabry-Perot cavity in which the

optical feedback is only due to the refractive-index mismatch at the air-semiconductor
interface. Assuming that the cavity length is 1 mm and that the refractive index of
Problems 169
the active medium is 3.5, calculate: (i) the reflectance of the interfaces; (ii) the small-
signal gain coefficient g (cm−1 ) at threshold; (iii) the frequency separation between
two adjacent longitudinal modes of the cavity.
5.2 Consider a Gax I n 1−x N semiconductor laser. Assuming that the bandgap energy
E g is a linear function of the gallium fraction x and using the data of Table 5.1,
calculate the value of x to be chosen for obtaining an emission wavelength of 420 nm.
5.3 A current is injected into a GaInAsP semiconductor laser of bandgap energy

E g = 0.92 eV and refractive index n = 3.5 such that the difference in Fermi levels
E f n − E f v (see Eq. (5.4)) is 0.95 eV. If the laser cavity has length L = 300 µm,
determine the maximum number of longitudinal modes that can oscillate.
5.4 Calculate the bandwidth Δλ of the light emitted by a blue Ga0.45 I n 0.55 N LED
at room temperature, by assuming that the bandgap energy E g of the semiconductor
is a linear function of the gallium fraction x and using the data of Table 5.1.
5.5 A photodiode with an efficiency of η = 0.1 is illuminated with 10 mW of

radiation at 750 nm. Calculate the voltage output across a load resistance of 50 .
5.6 A ten-dynode photomultiplier tube with an efficiency of η = 0.1 is illuminated

with 1 µW of radiation at 750 nm. Assuming that each incident electron generates
4 secondary electrons at all dynodes, calculate the voltage output across a load
resistance of 50 .
Chapter 6
Optical Fibers
Abstract This chapter describes the waveguiding properties of optical fibers, with
emphasis on the single-mode fibers used in optical communications. After examining
modal behavior and attenuation processes, the dispersive propagation of ultrashort
pulses is treated in some detail. A section is dedicated to the description of several
different types of fibers and their properties, followed by a presentation of some fiber
components, such as fiber couplers and fiber mirrors, which play an important role in
the setup of fiber sensors and active fiber devices. The last two sections specifically
deal with the active devices, such as fiber amplifiers and fiber lasers.
6.1 Properties of Optical Fibers
As discussed in Sect. 2.6, optical waveguides can confine an optical beam indefi-
nitely, avoiding diffraction effects. An optical fiber is a cylindrical waveguide of one
transparent material, clad by a second transparent material. It has important applica-
tions in many areas, primarily optical communications. The basic structure in shown
in Fig. 6.1. Both regions are typically made of glass, with the core glass having a
higher refractive index than the cladding. Within the frame of geometrical optics the
confinement effect can be explained by considering a ray which passes through the
center of the guide and is incident on the core-cladding interface at an angle larger
than the limit angle. The ray is then able to travel along the guide by means of a
sequence of total internal reflections.
The glasses used in optical fibers are made predominantly by silica, Si O2 , with
the addition of small amounts of other oxides, such as germania (GeO2 ) and alumina
(Al2 O3 ). Doping silica with germania or alumina raises the refractive index, and so
a silica-based optical fiber could be made from a germania-doped silica core with a
pure silica cladding. Silica is an ideal material because it can be purified to a high
degree to become extremely transparent.
The fabrication of optical fibers is a two-stage process. In the first stage, a vapor-
deposition method is used to make a cylindrical preform with the desired refractive

DOI 10.1007/978-3-319-20627-1_6
172 6 Optical Fibers
Fig. 6.1 The basic structure of an optical fiber
index profile and the relative core-cladding dimensions. The preform is a glass rod
with a diameter between 1 and 10 cm and roughly 1 m length. Glass fibers are
fabricated by pulling from the preform in a fiber-drawing tower. The fiber drawing
process, as illustrated in Fig. 6.2, begins by feeding the glass preform into the drawing
furnace. The drawing furnace softens the end of the preform to the melting point.
When the preform is heated close to the melting point, a thin fiber can be pulled
out of the bottom of the preform. During the pulling process, the fiber diameter is
held constant by automatically adjusting the pulling speed with a feedback system.
Fibers are pulled at a rate of 10–20 m/s. The pulling process preserves the transverse
refractive index profile of the preform, by simply shrinking the diameter from a
few centimeters to typically 125 µm. The fiber from a single preform can be many
kilometers long. Before the fiber is wound up, it usually receives a polymer coating
for mechanical and chemical protection.
Fig. 6.2 Fiber drawing

process
6.1 Properties of Optical Fibers 173
6.1.1 Numerical Aperture
A ray that is incident on the fiber front-face at an angle θi is refracted at an angle θt

given by the Snell’s law of (2.25):
n o sinθi = n 1 sinθt (6.1)
where n o is the refractive index of the medium in which the fiber is immersed, and n 1
is the refractive index of the core glass. The transmitted beam hits the core-cladding
interface at an incidence angle π/2 − θt , and is totally reflected if this angle is larger
than the limit angle of arcsin(n 2 /n 1 ), where n 2 is the refractive index of the cladding
glass. The incident ray is trapped inside the fiber if θi ≤ θmax , where the maximum
acceptance angle θmax is determined by the condition:
π/2 − θt = arcsin(n 2 /n 1 ) (6.2)
By using (6.1) and (6.2) it is found that θmax is given by:

n12 − n22
sinθmax = (6.3)
no
The quantity n o sinθmax is known as the “numerical aperture” of the fiber, usually
designated as N A, the same symbol used for the numerical aperture of the lens.
Although the two definitions are somewhat different, the physical meaning is the
same: the numerical aperture is a measure of the light-gathering capacity of the
optical component. As an example, taking n 1 = 1.48, n 2 = 1.46, n o = 1, one finds
N A = 0.24. In this case the maximum acceptance angle is θmax = 14◦ .
From now on it will be always assumed that n o = 1. Sometimes it is also useful
to introduce the parameter Δ, defined as:
n12 − n22 n1 − n2
Δ= ≈ (6.4)
2n 1 2 n1
6.1.2 Modal Properties
The optical fiber described in the preceding section is known as a “step-index fiber”,
because the refractive index profile presents an abrupt change at the core-cladding
interface. It is also possible to guide a light beam using a refractive index profile
that decreases smoothly inside the core from the center to the cladding. These fibers
are known as “graded-index” fibers. Their refractive index profile can be generally
expressed in the form:

n(r ) = n 2 + (n 1 − n 2 ) 1 − (r/a)γ , (6.5)
where a is the core radius and r is the radial distance. The validity of (6.5) is limited
to the interval 0 ≤ r ≤ a. If the exponent γ tends to infinity, (6.5) reproduces a
step-index profile. When γ = 2, there is a parabolic profile.
The propagation of optical waves along fibers must be treated, following the gen-
eral scheme outlined in Sect. 2.6, by using the Helmholtz equation and the boundary
conditions at the core-cladding interface. Since the geometry is two-dimensional,
the modes are characterized by two integer numbers, m and l. Using the cylindrical
coordinates z, r , and θ , the electric field of a generic mode, written in scalar form, is
given by:
Elm (r, θ, z, t) = ABlm (r, θ )exp[i(βlm z − ωt)], (6.6)
where βlm is the propagation constant and Blm (r, θ ) is the transverse mode profile,
which is independent of the propagation distance z.
The main results of the theoretical treatment are summarized as follows. The
propagation properties are described by a set of so-called “linearly polarized” (L Plm )
modes, where an L Plm mode has m field amplitude maxima along the radial coordi-
nate and 2l maxima along the circumference. The lowest order mode, i.e. the L P01
mode, has a bell-shaped field amplitude distribution that is well approximated by a
Gaussian function:

r2
B01 (r, θ ) = Bo exp − 2 , (6.7)
w
where w is known as the mode field radius. It has been seen in Sect. 1.4.2 that freely
propagating Gaussian waves have z-dependent parameters. In contrast, the Gaussian
wave propagating inside the fiber has constant radius and a fixed planar wavefront.
While there is always a solution for the fundamental mode L P01 , the existence
of higher-order modes depends on the fiber parameters. It is useful to introduce a
dimensionless parameter, V , known as the normalized frequency, defined as:
2πa 2 2πa
V = n1 − n22 = N A. (6.8)
λ λ
It can be shown that, if V is smaller than a critical value, Vc , then the fiber can
support only the fundamental mode. In the case of a “step-index” fiber, the critical
value is: Vc = 2.405. In the more general case of a profile following the law (6.5),
Vc is given by:

2
Vc = 2.405 1 + . (6.9)
γ
As an example, for γ = 2, Vc = 3.512. Above Vc the second mode, L P11 , can

also propagate. As long as V increases, more and more higher-order modes can
propagate. When V 1, the number of modes, N , that can propagate is given by
6.1 Properties of Optical Fibers 175
Fig. 6.3 Intensity of the

fundamental mode plotted as
a function of the ratio
between the distance from
the axis and the core radius,
at different values of the
normalized frequency V
V2
N= (6.10)
2
At fixed fiber parameters, the condition V = Vc defines a cut-off wavelength:
2πa 2 2πa
λc = n1 − n22 = N A. (6.11)
Vc Vc
At wavelengths λ ≥ λc the fiber behaves as a single-mode waveguide.

The intensity profile of the fundamental mode is shown in Fig. 6.3 for three values
of V . The optical power is mainly traveling inside the core, but there is also an
exponential tail (the evanescent wave described in Sect. 2.2.7) extending inside the
cladding. If V is larger, then the confinement inside the core is more effective because
the n 1 – n 2 difference is larger. The dependence of the beam radius w on V is described
by the empirical law:
w = a(0.65 + 1.62 V −1.5 + 2.88 V −6 ) (6.12)
Single-mode step-index fibers are designed by keeping the value of V close to

2. Smaller values of V are avoided because the field amplitude extends increasingly
into the cladding as V decreases. If the field amplitude is not negligible at the edge
of the cladding, then the power losses may become significant.
The single-mode condition imposes small core radius values. As an example, a
single-mode fiber, operating at λ = 1550 nm, and designed to have V = 2, can be
made by taking a/λ = 2.5 and N A = 0.12. The core radius is only 3.9 µm. The
cut-off wavelength of such a fiber is 1290 nm.
The propagation velocity of the fundamental mode is expected to be intermediate

between the values c/n 1 and c/n 2 , because the wave extends over both core and
cladding. Its propagation constant can be written as:
2π
β01 = ne f f , (6.13)
λ
where the effective refractive index n e f f has values in the interval n 1 ≥ n e f f ≥ n 2 .
Of course, every mode has its own effective refractive index. As a general rule, if
the indices l and m are increased, then the transverse distribution is more extended
into the cladding, and so the effective refractive index is expected to decrease.
6.2 Attenuation
The beam propagation inside the optical fiber is unavoidably lossy, so that the optical
power decreases with the propagation distance z. Assuming linear losses, the optical
power decays following an exponential law:
P(z) = Po exp(−αz), (6.14)
where Po is is the power launched into the fiber and α is the attenuation coefficient,
already defined in Sect. 2.1. The unit commonly used to characterize loss in fibers is
decibel per kilometer (dB/km). Denoting by αo the attenuation coefficient measured
in dB/km, it is found:

10 Po
αo = log10 . (6.15)
z P(z)
If α is expressed in units of km−1 , then the numerical relation between the two
coefficients is:
αo = 10 α log10 e = 4.34 α (6.16)
From (6.15) it is found that the propagation distance at which the incident power
is halved, z 1/2 , is given by:
10 log10 2 3
z 1/2 = ≈ . (6.17)
αo αo
As an example, if αo = 0.15 dB/km, then z 1/2 = 20 km.

Fiber losses are due to two effects, light scattering and absorption. Scattering
arises because there are submicroscopic random inhomogeneities in density and
composition inside the glass, which give rise to local fluctuations in the refractive
index. The inhomogeneities originate from thermal fluctuations in the melt of the
6.2 Attenuation 177
silica that are then frozen in, as the glass cools. Light scattering from submicroscopic
scatterers, known as Rayleigh scattering, gives a wavelength-dependent attenuation
coefficient that scales as λ−4 .
Silica glasses have an absorption band in the ultraviolet, due to electronic tran-
sitions, and absorption bands in the mid-infrared, due to vibrational transitions. At
wavelengths greater than about 1.6 µm, absorption losses become predominant with
respect to scattering losses. The combination of scattering and absorption losses gives
rise to the attenuation curve shown in Fig. 6.4, which has a minimum of ≈0.15 dB/km
at about 1.55 µm. Indeed, long-distance optical communication systems use semi-
conductor lasers emitting a 1.55 µm beam, as the carrier that is modulated and
transmitted through the optical fiber.
The attenuation curve of Fig. 6.4 also shows a secondary minimum around 1.3 µm,
due to the presence of an absorption peak at about 1.4 µm. This peak, due to vibra-
tional transitions of the O–H bond, occurs because of small quantities of water trapped
inside the glass during the fiber fabrication process. This absorption peak no longer
exists in recent fibers, because of improvements in the purification process. However,
several networks using fibers with water impurities are still installed and in use.
Historically, the first optical-communication experiments, operating at short dis-
tances, were performed at 850 nm. In optical-communication jargon, the wavelength
region around 850 nm was called the first attenuation window. The region around
1.3 µm is called the second attenuation window, and the one around 1.55 µm the
third attenuation window.
All the glasses made by different combinations of oxides offer similar infrared
absorption bands, so that the attenuation behavior shown in Fig. 6.4 is not much
changed by varying the oxide composition. It is, however, possible to shift the vibra-
tional absorption to longer wavelengths by using fluoride-based glasses instead of
oxides. The so-called “ZBLAN” glasses, made from a mixture of zirconium, barium,
lanthanum, aluminum, and sodium fluorides, offer attenuation minima around 3 µm.
Fig. 6.4 Attenuation coefficient of a silica fiber versus wavelength

The minimum value of the attenuation coefficient is one order of magnitude lower
than for silica glasses.
In addition to intrinsic losses, there are also fiber losses due to the departure
from the ideal straight line configuration. In a bent fiber, the incidence angles at the
core-cladding interface change, and so the condition of total internal reflection is no
longer satisfied, meaning that the optical power is only partially trapped inside the
fiber. In general, bending losses are larger for those modes that extend more into
the cladding. Typically bending losses become important when bend radii are of the
order of millimeters.
6.3 Dispersion
Optical dispersion, as defined in Sect. 2.1, is related to the dependence of the refractive
index of a material on the wavelength of the propagating beam. As discussed in
Sect. 6.1.2, a mode propagating in an optical fiber has an effective refractive index,
which is a type of weighted average between n 1 and n 2 . Since both materials are
dispersive, one might think that the dispersion of n eff is determined by that of the
core and cladding materials, to a first approximation. The situation, however, is more
complex, because the parameter V is changed by changing the wavelength. Therefore,
the transverse field distribution is modified. As a consequence the relative weighting
of n 1 and n 2 varies with λ. The conclusion is that the single-mode dispersion is
a combination of both a material effect and a waveguide effect. The waveguide
dispersion is very sensitive to the shape of the radial profile of the refractive index,
such that it may significantly change on going from a step-index to a graded-index
fiber.
An intermodal dispersion occurs in multimode fibers, due to the differences in
effective refractive indices of the various modes. Even if the wavelength is the same,
the different modes have different propagation velocities. A single light pulse entering
an N -mode fiber is divided into N pulses, each traveling with a different velocity. At
the output a broadened pulse is observed, or, for a longer propagation, a sequence of
N separated pulses.
6.3.1 Dispersive Propagation of Ultrashort Pulses
In this section, the discussion is focused on the propagation of short optical pulses in
single-mode fibers. Since a pulse can be viewed as the superposition of many waves
with different wavelengths, the fact that each wavelength travels with a different
velocity clearly modifies the time-dependent pulse-shape. As it has been shown in
Sect. 3.11.3, the shorter the pulse is in time, the wider its optical spectrum. Therefore,
one should expect that dispersion effects will be largest for the shortest pulses.
6.3 Dispersion 179
In order to describe pulse propagation it is useful to express the propagation

constant β(ω) = (ω/c)n e f f (ω) as a power series around the central pulse frequency
ωo :
∞

d k β (ω − ωo )k
β(ω) = , (6.18)
dωk k!
k=0
where the derivatives d k β/dωk are evaluated at ωo .

Assuming that the pulse spectral width, Δω, is much smaller than ωo , the power
expansion can be truncated at the quadratic term:
1
β(ω) = βo + β1 (ω − ωo ) + β2 (ω − ωo )2 , (6.19)
2
where βo = β(ωo ),
dβ
β1 = , (6.20)
dω
and
d 2β
β2 = . (6.21)
dω2
The electric field of the input pulse, entering the single-mode fiber at z = 0, is
given by
E(t, 0) = A(t, 0)exp(−iωo t). (6.22)
For simplicity, the term B01 (r ), which describes the transverse distribution of the
field amplitude and is invariant during the propagation, has been omitted.
The mathematical treatment of dispersive propagation becomes simpler if just one
specific pulse shape is considered, such as the Gaussian shape described by (3.77).
This has a Fourier transform given by (3.80).
In the frequency domain, the effect of propagation may be regarded as a frequency-
dependent phase delay, given by φ(ω, z) = β(ω)z. In this way, the Fourier transform
of the field at z is simply given by:
E(ω, z) = E(ω, 0)exp[iβ(ω)z] (6.23)
By substituting (6.19) and (3.80) inside (6.23) one obtains:

√ (ω − ωo )2
E(ω, z) = 2π τi Ao exp(iβo z)exp −(τi 2 − iβ2 z) + iβ1 z(ω − ωo ) .
2
(6.24)
By using (2.9), it is possible to return to the time domain, in order to recover the
expression for the time-dependent electric field at the generic coordinate z:
∞
1
E(t, z) = E(ω, z)exp(−iωt)dω
2π −∞

Ao (t − β1 z)2 (1 + iβ2 z/τi 2 )
= exp[−i(ωo t − βo z)]exp −
1 − iβ2 z/τi2 2τi 2 (1 + β2 2 z 2 /τi 4 )
(6.25)
It is useful to separately highlight the behavior of the field amplitude and phase
by writing the electric field as:
E(t, z) = A(t, z)exp{−i[ωo t − βo z + φ(t, z)]}, (6.26)
where:

Ao (t − β1 z)2
A(t, z) = exp − (6.27)
(1 + β2 2 z 2 /τi 4 )1/4 2τi 2 (1 + β2 2 z 2 /τi 4 )
and
(t − β1 z)2 zβ2 /τi 2 arctan(β2 z/τi 2 )
φ(t, z) = − . (6.28)
2τi 2 (1 + β2 z 2 /τi 4 )
2 2
Equation (6.27) shows that the time-dependent pulse amplitude remains Gaussian
during the dispersive propagation. The pulse intensity, I (z, t), proportional to the
modulus square of (6.27), reads:

Io (t − β1 z)2
I (t, z) = exp − , (6.29)
1 + (z/L D )2 τi 2 (1 + z 2 /L D 2 )
where Io is the peak intensity of the input pulse. Equation (6.29) contains the disper-
sion length L D , defined as:
τi 2
LD = (6.30)
|β2 |
According to (6.29), the intensity observed at the coordinate z is maximum at

time t p = β1 z. This means that the pulse peak is propagating with a velocity, known
as the “group velocity”, which is given by:
u g = β1 −1 . (6.31)
6.3 Dispersion 181
The quantity β1 can be written as:
1 dn e f f n g
β1 = ne f f + ω = , (6.32)
c dω c
where n g is called group refractive index. As is the case for most transparent materials,
the refractive index of silica glass, n gl , is an increasing function of ω. Although n e f f
does not coincide with n gl , its frequency-dependence is similar, so that the derivative
dn e f f /dω is positive. Therefore it can be quite generally said that the group velocity
is smaller than the phase velocity c/n e f f , according to (6.32).
Recalling that the pulse duration τ p is related to τi by (3.81), the law describing
the change of pulse duration upon propagation can be derived from (6.29):

z2
τ p (z) = τ p (0) 1 + (6.33)
L D2
Equation (6.33) shows that L D represents the propagation distance above which
there is a significant increase of the pulse duration. While the pulse spectrum remains
unchanged during propagation, the pulse duration grows with z. As a consequence
the bandwidth-duration product is no longer given by (3.81), but becomes:

z2
Δν p τ p (z) = 0.441 1 + (6.34)
L D2
Since a lossless propagation has been considered, the pulse energy must be con-
served. Indeed (6.29) shows that the intensity decreases in such a way as to keep the
product between peak intensity and pulse duration constant during propagation.
The quantity β2 is known as the “group velocity dispersion” (GVD) parameter, and
is usually measured in units of ps2 /km. At the optical-communication wavelength,
λ = 1.55 µm, it is found that β2 = −20 ps2 /km for a typical single-mode fiber. It is
instructive to numerically investigate the effect of pulse duration. For a 20-ps pulse,
(6.30) gives L D = 20 km. After
√ a propagation length of 20 km, the pulse duration
is increased by a factor of 2. For a 1-ps pulse, L D is much shorter, only 50 m.
Here, the pulse duration is increased by a factor of 400, after propagating 20 km.
This clearly shows that the stretching effect due to dispersion plays a very important
role for ultrashort pulses.
In the area of optical communications, it is customary to use the dispersion para-
meter D, instead of β2 . This is defined as:
dβ1 2π cβ2
D= =− 2 (6.35)
dλ λ
D is measured in units of ps/(nm km). It represents the group delay per wavelength
(in nm) and propagation length (in km). At λ = 1.55 µm, it is found that D =
17 ps/(nm km) for a typical single-mode fiber. The so-called “normal” group-velocity
Fig. 6.5 Material and

waveguide dispersion in a
single-mode silica fiber with
a parabolic index profile
dispersion regime corresponds to the situation in which D is negative (and β2 is

positive). “Anomalous” dispersion corresponds to a positive D (and negative β2 ).
Figure 6.5 shows the behavior of D as a function of λ for a single-mode silica
fiber with a parabolic index profile. The behavior is mainly determined by the mate-
rial dispersion, with only a slight modification introduced by waveguide dispersion.
The zero-dispersion wavelength is about 1.3 µm. The fiber dispersion is normal at
wavelengths less than 1.3 µm, and anomalous for longer wavelengths.
6.4 Fiber Types
In this section a review of the main types of fibers available is given. A first distinction
must be made between single-mode fibers and multimode fibers. As discussed in
Sect. 6.1.2, single-mode fibers have V ≈ 2 and a core diameter of ≈8 µm, for
transmission at 1.55 µm. Multimode fibers have a large V and core diameter typically
around 50 µm. The numerical value of the cladding diameter is not critical as long
as it is large enough to confine the fiber modes entirely. The cladding diameter of
single-mode fibers is usually standardized to 125 µm. Multimode fibers have a larger
N A, which makes it much easier to couple optical signals into the fibers. They are
also cheaper and can carry a higher optical power. Intermodal dispersion limits their
utility for pulse transmission to very short distances.
Low-dispersion fibers. Optical communications require low-dispersion fibers. There
are available fibers in which the zero-dispersion wavelength is shifted from 1.3 µm
to about 1.55 µm, by using the w-shaped profile shown in Fig. 6.6. In this type of
fiber, known as “dispersion shifted” (DS) fiber, the waveguide dispersion almost
exactly compensates the material dispersion at 1.55 µm, as shown in Fig. 6.7. Other
fibers, known as “dispersion-flattened” fibers, are designed to have a small non-zero
dispersion over a wide wavelength interval around 1.55 µm, as shown in Fig. 6.7. In
the area of optical communications, as it will be seen in Chap. 8, it is also important
6.4 Fiber Types 183
Fig. 6.6 Refractive

index profile in a
“dispersion-shifted” fiber
Fig. 6.7 Group velocity

dispersion versus λ for a
“dispersion-shifted” and a
“dispersion flattened” fiber
to have fibers with large normal dispersion (D ≈ −60 ps nm−1 km−1 ), available for
use in dispersion compensation. A short stretch of these fibers (known as “dispersion
compensating” DIS-CO) can compensate for the dispersion accumulated over a long
propagation span in an SMR fiber.
Polarization-maintaining fibers. An aspect of waveguiding that has not yet been
touched upon in this chapter is that of signal polarization. The fundamental mode
L P01 is a linearly polarized mode, but its polarization direction is completely arbitrary
as long as the fiber has perfect cylindrical symmetry about its axis. In fact, real fibers
invariably have some slight local asymmetry, either intrinsic or due to their environ-
ment, which creates a small birefringence varying randomly along their length. As
a result, the initial polarization is lost after some tens of meters, so that it becomes
impossible to predict the polarization state of the output signal. In the case of pulse
propagation, the random changes in the polarization can induce a pulse broadening.
This dispersion effect, known as “polarization-mode dispersion” (PMD), is usually
smaller than the GVD effect. For some applications it is desirable that fibers transmit
light without changing the polarization state. Such fibers are called “polarization-
maintaining” fibers. One possibility is to break the cylindrical symmetry of the fiber
by using an elliptical core. An alternative scheme is that of fabricating a preform that
Fig. 6.8 Polarization-maintaining fibers
Fig. 6.9 Microstructured fiber. Courtesy of NKT Photonics
contains two glass rods, inserted on the opposite sides of the fiber core, as shown
in Fig. 6.8. These rods are both made of a glass that is different from the cladding,
and so mechanical stress due to their different thermal expansion coefficients arises
inside the fiber structure during the cooling process. Because of stress-induced bire-
fringence, the fiber becomes strongly birefringent, so that a linear polarization state
parallel to one of the principal axes is transmitted without a change in polarization.
This type of fiber is sometime called “panda” fiber.
Microstructured fibers. A new family of fibers, called microstructured fibers, con-
tain cylindrical air holes running parallel to the fiber axis, as shown in Fig. 6.9. If
the holes are organized in a regular periodic pattern, then the fiber is also called
photonic-crystal fiber. These fibers are usually made of pure silica glass. Both the
hole diameter and the minimum distance between holes are of the order of the wave-
length of light. Microstructured fibers have higher losses and are more expensive than
standard fibers. However, they are interesting for various applications, because their
more complicated structure gives new degrees of freedom, offering performances
not accessible using standard fibers. Once λ is fixed, there is a very limited flexibility
regarding the choice of core diameter of standard single-mode fibers. In contrast,
microstructured fibers can be designed for single mode operation with a large mode-
area (mode diameter larger than 10 µm) or a very small mode-area (mode diameter
around 1 µm). Large mode-areas are interesting for transmitting high optical pow-
ers, whereas small mode-areas produce intensities high enough to enhance nonlinear
effects, such as the optical Kerr effect and other effects that will be described in the
following chapter. The dispersion properties of microstructured fibers can also be
6.4 Fiber Types 185
drastically different from those of standard fibers. In particular, the zero-dispersion

wavelength can be positioned anywhere in the visible or near infrared.
Plastic optical fibers. Fibers can also be made from polymers, by using, for instance
a core of polymethyl methacrylate (PMMA, n = 1.495) and a cladding of fluoroalkyl
methacrylate (n = 1.402). These fibers have a transparency window in the visible,
between 500 and 700 nm, have large attenuation coefficients (50–100 dB/km) and
also a large NA. They can be useful for sensors, for short-distance multimode data-
transmission, and for architectural lighting.
6.5 Fiber-Optic Components
Several optical components can be made using fibers. They are useful for the manipu-
lation of optical signals that are already traveling inside fibers, so that all-fiber setups
can be realized. In this section attention is focused on two components, fiber couplers
and fiber mirrors.
Couplers. Fiber couplers can couple light from one or more input fibers to one or
more output fibers. Several fibers can be thermally fused, such that their cores come
into contact over a length of some centimeters, as shown in Fig. 6.10. When the
evanescent field of one fiber reaches the core of the other fiber, a fraction, F, of the
power traveling in the first fiber is coupled into the other. An optical signal coming
from fiber 1 can be progressively transferred to fiber 2. The power fraction that is
coupled into fiber 2 depends on the length of the coupler, L, according to the law:
F = sin2 (AL), (6.36)
where the parameter A is an increasing function of λ. By varying L, fiber beam-

splitters that have an arbitrary power ratio between the two output fibers can be
designed. Fiber couplers introduce very low losses, typically 0.2–0.5 dB. Since the
power transfer depends not only on the coupling length, but also on wavelength, it
is possible to design couplers, which, for instance, entirely transfer a λ1 signal from
fiber 1 to fiber 2, while leaving a λ2 signal in fiber 1. Such dichroic couplers are used
in fiber amplifiers to combine the signal input and the pump wave, as it will be seen
in Sect. 6.6. Their insertion loss may be very small (below 1 dB) for both inputs.
Fig. 6.10 Scheme of a fiber

coupler
In optical-communication systems wavelength-sensitive couplers are used as mul-

tiplexers to combine several input signals with different wavelengths, or to separate
different wavelengths.
Fiber Bragg gratings. An other type of fiber component is the wavelength-selective
fiber mirror, which is conceptually similar to the multilayer mirrors discussed in
Sect. 2.2.4. The idea is to create a periodic change of refractive index within the core.
Such a phase grating, known as “fiber Bragg grating”, has maximum reflectivity
when the phase delay corresponding to the propagation along one period is equal
to π . Therefore the wavelength of maximum reflectivity, λo , known as the Bragg
wavelength, is given by the relation
λo = 2mnΛ, (6.37)
where m is an integer, Λ is the periodicity of the grating and n is the refractive index
of the core.
The maximum reflectivity is calculated using the same relations (2.55) and (2.56)
given in Sect. 2.2.4 for multilayer mirrors. By putting n 1 = n 2 = n B = n, and
n A = n + Δn inside (2.56), Rmax can be expressed as:

Δn
Rmax = tanh N2
, (6.38)
n
where N is the number of periods. The length, L, of the grating is typically of the
order of a few millimeters, and so N = L/Λ is of the order of 104 . Even if Δn is
small, a maximum reflectivity very close to 100 % can be obtained. As an example,
taking Δn = 2 × 10−4 , N = 4 × 104 , n = 1.5, it is found: Rmax = 99.99 %. The
reflection bandwidth is typically narrower than 1 nm.
Fiber Bragg gratings are fabricated by exploiting the photosensitivity of germania-
doped silica glasses to ultraviolet radiation. Radiation at a wavelength around 240 nm,
usually coming from an excimer laser, induces a permanent change of refractive index
in the core. The increase in refractive index due to UV illumination, Δn, is small,
typically about 10−4 . However, this can be enhanced by forcing molecular hydrogen
into the fiber core through a high-pressure low-temperature diffusion process. There
are various methods to obtain a spatially periodic illumination of the core. The method
shown in Fig. 6.11 uses a periodic phase mask, creating interference fringes with the
required periodicity.
Fig. 6.11 Fabrication of a

fiber Bragg grating
6.5 Fiber-Optic Components 187
Fiber Bragg gratings are used as mirrors for fiber lasers, as will be seen in Sect. 6.7.
They also have many applications in optical communication systems and in fiber
sensors, as will be seen in Chap. 8. A recent interesting development is that of “long
period gratings”, in which the modulation period is larger than the wavelength. Inside
these gratings there is a wavelength-selective transfer of optical power from core to
cladding. The power displaced into the cladding is ultimately dissipated, and so the
grating selectively suppresses certain wavelengths, acting as a filter.
Fiber gratings with a linear variation in the grating period (termed chirped grat-
ings) are particularly useful for dispersion compensation. In these gratings, the
Bragg wavelength varies with position, thereby broadening the reflected spectrum.
A chirped grating has the property of adding dispersion, because different wave-
lengths are reflected from the grating subject to different delays. In this way, it is
possible to achieve a very large group delay dispersion, sufficient for compensating
the dispersion of a long span of transmission fiber, in a short length of fiber grating.
6.6 Fiber Amplifiers
An optical fiber can become the active medium for optical amplification, if the fiber
core is doped with rare-earth ions, such as N d 3+ , Pr 3+ , Er 3+ , and Y b3+ . A strong
motivation for the development of efficient fiber optical amplifiers came from the
area of optical communications. In the case of long-distance links the attenuation
due to fiber losses becomes large, and so it is necessary to periodically restore the
power level of the traveling optical signals. Fiber amplifiers offer the best physical
compatibility with fiber transmission systems.
Amplifiers based on erbium-doped fibers, which exhibit gain over a range of
wavelengths centered on 1550 nm, were developed in the late 1980s. An important
factor for the success of the erbium-doped fiber amplifier (EDFA) was the discov-
ery that relatively high concentrations of erbium could be incorporated without ion
clustering into the fiber core by co-doping with alumina.
The energy level scheme for the Er 3+ ion in silica glass is presented in Fig. 6.12.
By exploiting the long lifetime (about 12 ms) of the first excited level, optical gain
can be obtained for the transition between this excited level and the ground state.
The system then behaves like a three-level system, similar to that of the ruby laser
described in Sect. 3.9. Because the ions are incorporated into a disordered solid
matrix the energy levels are considerably broadened, and so the gain is spread over
a wavelength bandwidth of about 40 nm. The peak gain is obtained at 1530 nm, as
shown in Fig. 6.13.
Like the solid-state lasers described in Sect. 3.9, fiber amplifiers are optically
pumped. Because of the constraints of the fiber geometry, they are longitudinally
pumped, so that both pump and signal utilize the waveguiding effect of the fiber.
This ensures an excellent overlap of the two waves over the entire length of the
device, and allows the pump power to be completely absorbed and utilized.
Fig. 6.12 Simplified scheme

of the energy levels of the
Er 3+ ion in a silica fiber
Fig. 6.13 Optical gain

versus wavelength for a
typical erbium-doped fiber
amplifier
Optical pumping is carried out with semiconductor lasers, emitting at 980 nm or

1480 nm. In the former case, the ion excited at level 2 rapidly decays non-radiatively
to the lower level of the band 1, so that a population inversion can be established
between this lower level and level 0. In the latter case, the ion is excited to a level
that is already inside band 1. By a rapid intraband non-radiative decay, the excitation
is subsequently transferred to the lower level of the band 1.
A typical scheme of an erbium-doped fiber amplifier is shown in Fig. 6.14. The
pump and signal beams are coupled into the fiber by using a dichroic fiber coupler
of the type described in the previous section. An isolator is inserted after the doped
fiber, in order to avoid back-reflections, which could have the undesired effect of
transforming the amplifier into an oscillator. The residual pump power, possibly
present at the output of the amplifier, is eliminated through a wavelength filter or a
dichroic coupler. The pump light can also be injected in the opposite direction from
the signal, or in both directions. The latter configuration provides a more uniform
distribution of pump power along the length of the amplifier.
6.6 Fiber Amplifiers 189
Fig. 6.14 Scheme of an EDFA
Once the type of doped fiber and the pump power are chosen, the parameter to be
fixed is that of the length of the doped fiber, L. If the gain medium is too long, part of it
will not be pumped because the pump power will be totally absorbed before it reaches
the far end of the fiber. In a three-level system, such as an EDFA, the unexcited ions
that remain in the ground state absorb signal photons, so that the un-pumped regions
of the fiber have a detrimental effect on the amplifier performance. Considering that
too short an amplifier also does not use the pump power very efficiently, it is clear
that an optimum length for the fiber amplifier exists.
Typical parameters for an optical-communications amplifier are the following.
The concentration of erbium atoms is Nt ≈ 2 × 1024 m−3 , the pump power Pp =
10 − 20 mW, the stimulated emission cross-section σ = 8 × 10−25 m−2 , the
amplifier length L = 10–20 m. The power gain can be as large as 104 (40 dB), with
a conversion efficiency of 30–40 %.
The efficiency of the amplifier can be improved by using doubly-doped fibers
containing both Er 3+ and Y b3+ ions. In these fibers the pump power excites the
ytterbium ions, which have an absorption cross-section larger than that of erbium
ions. In a second step, there is a spontaneous transfer of excitation energy from
ytterbium to erbium, so that a population inversion is built up in the erbium ions.
This type of co-doping is mainly used for high-power amplifiers, with output power
in the range of watts.
The main factor that limited the development of high-power amplifiers in early
years was that the pump beam emitted by arrays of semiconductor lasers is spatially
multimodal. The coupling efficiency of these multimodal beams inside the small core
of the single-mode amplifying fiber is actually rather low. The problem is solved
by using fibers with a double cladding, known as “cladding-pumped” fibers. The
structure of these fibers is shown in Fig. 6.15. The signal to be amplified is confined
within the core, whereas the pump is confined within the undoped inner cladding.
Since the diameter of the inner cladding is much larger than that of the core, coupling
of pump power into the fiber is much easier. The pump power is continuously fed from
inner cladding to core, where it is absorbed, ensuring a high-efficiency operation.
The efficiency of power transfer from inner cladding to core is improved by using
an elliptical core.
In comparison with other optical-gain media, rare-earth doped-fibers have several
very useful characteristics. The combination of broad gain bandwidth, polarization-
independent gain, long lifetime of the excited level, and relatively large saturation
Fig. 6.15 Structure of a cladding-pumped optical fiber
power ensures full transparency to all modulation formats. The EDFA offers the possi-
bility of simultaneous amplification of multiple signals at different wavelengths (like
the multiple channels in a wavelength-division-multiplexing optical-communication
system). There is also a low sensitivity of the EDFA gain to pump fluctuations
occurring on a time-scale shorter than the lifetime of the excited level τ10 . Another
important advantage is that the fiber geometry, with its large surface-to-volume ratio,
has a good efficiency for heat dissipation.
Using schemes similar to those of EDFAs, fiber amplifiers operating at various
near-infrared wavelengths can be made by doping silica fibers with N d (amplifi-
cation at 1.06 µ or 1.34 µm), or Y b (amplification in the range 1.05–1.12 µm).
Optical amplifiers using Pr -doped ZBLAN fibers have also been tested for opera-
tion in the wavelength interval 1.28–1.34 µm, where some of the existing optical-
communication systems operate.
A completely different family of fiber amplifiers is represented by Raman ampli-
fiers, which make use of the stimulated Raman scattering process. They will be
described in Chap. 7.
6.7 Fiber Lasers
It is clear from the general discussion of Sect. 3.3 that the fiber amplifier can be
transformed into a laser by providing an optical feedback. The initial interest in fiber
lasers came from the optical-communications area, because it was thought possible
to use them as sources of high-repetition-rate pulse trains. Although the interest for
such an application quickly vanished, fiber lasers have reached a high degree of
reliability for use in many different areas.
Fiber lasers can operate both CW and pulsed. In comparison with solid-state
lasers, they have several advantages.
• When using a single-mode fiber, the output beam has a very smooth Gaussian
profile, therefore it can be focused to a very small focal spot, at variance with the
behavior of high-power semiconductor or N d-Y AG lasers.
• The very good coupling of pump power into the fiber gives high-efficiency oper-
ation.
6.7 Fiber Lasers 191
Fig. 6.16 Fiber laser schemes: a Ring cavity, b Fabry-Perot cavity
• The high surface-to-volume ratio of fiber structure provides an effective heat dis-
sipation mechanism, so cooling problems are greatly simplified.
CW fiber lasers are based on Fabry-Perot cavities (see Fig. 6.16b) with mirrors
made by fiber Bragg gratings. They use cladding pumped fibers in order to achieve
high efficiency. A particularly interesting fiber laser is the ytterbium laser. The Y b3+
ion behaves like a four-level system, is pumped at 915 nm and emits in the 1000–
1100 nm wavelength region. The emission wavelength is fixed by the peak reflectance
of the FBG mirrors. The Yb-fiber laser can provide high-power outputs (up to some
kilowatts). Because of its efficiency, compactness, and simplicity of cooling require-
ments, it is often used in industrial processes, such as marking, welding and precision
cutting.
Pulsed fiber lasers can be built to operate in the Q-switching regime, generating
kilohertz-frequency trains of nanosecond pulses, or in the mode-locking regime.
Mode-locked fiber lasers are usually based on ring cavities (see Fig. 6.16a). The
pump beam is injected into the doped fiber by using a dichroic fiber coupler. Another
fiber coupler is then used to extract the output beam. The mode-locking behavior
is induced by inserting a modulator or a saturable absorber inside the ring. These
lasers can largely be based on telecom components, which have been developed for
reliable long-term operation and have a moderate cost. Mode-locked erbium-doped or
ytterbium-doped fiber lasers can generate trains of picosecond pulses with repetition
rates up to 20 GHz. The very large gain bandwidth achievable in an ytterbium-doped
fiber is capable of supporting pulses shorter than 20 fs.
Problems
6.1 A step-index fiber has a numerical aperture of 0.15, a core refractive index of
1.47 and a core diameter of 20 µm. Calculate (i) the maximum acceptance angle of
the fiber, (ii) the refractive index of the cladding, (iii) the approximate number of
modes with a wavelength of 800 nm that the fiber can carry.
6.2 An optical power of 10 mW is launched into an optical fiber of length 500 m.

If the power emerging from the other end is 8 mW, calculate the fiber attenuation αo
in dB/km.
6.3 An erbium-doped fiber amplifier has length L = 11 m. If the stimulated-
emission cross-section at λ = 1530 nm is 1.2 × 10−20 cm2 and the population
inversion is N2 − N1 = 7 × 1017 cm−3 , calculate the small-signal gain in dB units.
6.4 A light beam at λ = 1550 nm is coupled into a step-index optical fiber having
numerical aperture N A = 0.12 and core radius a = 4.5 µm. (i) specify whether the
fiber behavior will be monomodal or multimodal; (ii) calculate the parameter w of
the fundamental mode.
6.5 A transform-limited Gaussian laser pulse having duration τ p = 4 ps and angular
frequency ωo is coupled into a single-mode optical fiber. The fiber has an attenuation
coefficient αo = 0.15 dB/km and a frequency-dependent propagation constant β(ω)
following the law:
β(ω) = βo + β1 (ω − ωo ) + (1/2)β2 (ω − ωo )2 (6.39)
where β1 = 4.90 × 10−9 s/m, β2 = −1 × 10−27 s2 /m. By assuming that the fiber
length is L = 50 km, calculate: (i) the time taken by the peak of the pulse to propagate
along the fiber; (ii) the pulse duration τ p (L) at the fiber output; (iii) the ratio r between
the output and input energy of the laser pulse; (iv) the ratio r = PL /P0 , where P0
and PL are the input and output peak power of the laser pulse, respectively.
6.6 Consider a step-index optical fiber with a core radius a = 4.5 µm and core
refractive index n 1 = 1.494. Assuming that the normalized frequency is V = 2.15
at λ = 1550 nm, calculate: (i) the cladding refractive index; (ii) the width w of the
fundamental mode; (iii) the wavelength at which the fiber ceases to be single-mode.
6.7 An erbium-doped fiber amplifier has length L = 10 m and a stimulated emission
cross section σ = 1.2 × 10−20 cm2 at λ = 1550 nm. Calculate the value of the
population inversion N2 − N1 required to attain a small signal gain of 20 dB.
6.8 Calculate pitch and length of a fiber Bragg grating having a reflectivity peak
Rmax = 0.995 at λ = 1500 nm, assuming that the core average refractive index is
n = 1.5 and that Δn = 2 × 10−4 .
6.9 Calculate the effective refractive index n e f f at λ = 1550 nm for a single-mode
optical fiber having a core radius a = 5 µm, core refractive index n 1 = 1.484,
cladding refractive index n 2 = 1.480, by assuming that n e f f = n 1 f 1 + n 2 f 2 ,
where f 1 and f 2 are the power fractions traveling inside the core and the cladding,
respectively. The profile of the single mode is given by the Gaussian function
exp[(x 2 + y 2 )/w2 ]. The value of n e f f should be given with 4 digits.
6.10 A transform-limited Gaussian laser pulse having duration τ p = 1 ps is cou-
pled into a single-mode optical fiber with a dispersion parameter β2 = 10 ps2 /km.
Calculate at which propagation distance the pulse duration is doubled.
Chapter 7
Nonlinear Optics
Abstract The field of nonlinear optics ranges from fundamental studies of the inter-
action of intense laser light with matter to applications such as frequency conversion,
non-resonant amplification, all-optical switching. The treatment developed in this
chapter restricts the attention to those topics that are of interest for applications, and
limits the notational complications to a minimum. Second-harmonic generation is
described in some detail in Sect. 7.2, because it is considered a case study for illus-
trating the mathematical approach and the experimental methods that are generally
utilized in nonlinear optics. Section 7.3 completes the description of second-order
effects with a concise treatment of parametric amplifiers and oscillators and of sum-
frequency generation. Section 7.4 is devoted to a review of the main phenomena
arising from the optical Kerr effect. The subject of Sect. 7.5 is stimulated Raman
scattering, with particular emphasis on applications based on optical-fiber devices.
Stimulated Brillouin scattering is described in Sect. 7.6.
7.1 Introduction
Nonlinear optics is the field of optics that considers phenomena occurring when the
response of a medium to an applied electromagnetic field is nonlinear with respect
to the amplitude of that field.
The proportionality relation between electric field and induced electric polariza-
tion that was introduced in Sect. 2.1 implies that a monochromatic field oscillating at
ω generates a polarization P also oscillating at ω. As a consequence, the propagation
of a wave across a transparent medium can change the spectrum by attenuating or
suppressing some frequency component, but cannot generate new frequencies.
Nonlinear optical effects become appreciable when the displacement of charges
from their equilibrium position inside the material becomes so large that the induced
dipole ceases to be proportional to the field amplitude E. This happens if the incident
field amplitude is not negligible in comparison with the internal field E i that holds the
charges together in the material. A typical value for E i is 1011 V/m, a figure attainable
with laser fields, whereas the electric-field amplitude of conventional optical sources
can reach at most 103 V/m. Nonlinear optics was born in the early 1960s immediately
after the first intense laser fields became available.
DOI 10.1007/978-3-319-20627-1_7
194 7 Nonlinear Optics
The standard classification of nonlinear optical effects is based on the expansion

of the electric polarization P in a series of powers of E:
P = εo (χ (1) E + χ (2) E 2 + χ (3) E 3 + · · · ). (7.1)
Equation (7.1) can be considered as a generalization of the linear relation

expressed by (2.11). For sake of simplicity, the notation adopted in (7.1) neglects
dispersion and uses scalar quantities.
If propagation in a medium exhibiting a non-zero χ (2) is considered, an input
field oscillating at ω, E(t) ∝ exp(−iωt), generates a polarization component ∝
χ (2) exp(−2iωt), oscillating at 2ω. Once the expression of the electric polarization
is inserted into the wave equation (2.7), the 2ω component becomes the source
of a second-harmonic field. Analogously, if the input field contains two distinct
frequencies ω1 and ω2 , then it is found that the quadratic term generates the sum
frequency ω1 + ω2 , and the difference frequency ω1 − ω2 . A similar discussion could
be made for the cubic term. Besides optical frequency conversion, another feature
of nonlinear interactions is that they can involve field-dependent modifications of
certain material properties. This second feature opens the way to phenomena in
which one beam can be used to control the amplitude and/or the phase of a second
beam. To be more specific, through the optical Kerr effect a strong pump beam can
change the index of refraction of the medium in which a signal beam propagates, so
that a fully optical switch can be realized. In principle, nonlinear phenomena appear
as the natural candidates for the realization of integrated all-optical devices, such as
gates and logical circuits, to be used in various applications related to information
and communication technology and optical signal processing.
7.2 Second-Harmonic Generation
Consider a monochromatic plane wave having angular frequency ω and wave vector
kω propagating along the y axis inside a second-order nonlinear medium character-
ized by a nonlinear susceptibility χ (2) . During propagation a second-harmonic wave
will be generated, so that the total electric field of the waves traveling along y is:
1
E(y, t) = {E ω (y) exp[−i(ωt − kω y] + c.c.]}
2
1
+ {E 2ω (y) exp[−i(2ωt − k2ω y] + c.c.]}, (7.2)
2
where kω = (ω/c)n ω and k2ω = (2ω/c)n 2ω . Here n ω and n 2ω are, respectively,
the refractive indices experienced by the input beam, called sometimes the pump
beam, and by the second-harmonic beam during the propagation along the nonlinear
medium.
7.2 Second-Harmonic Generation 195
Note that the field described by (7.2) is a real quantity, because of the presence of
the complex conjugate terms. In the previous chapters many calculations have been
carried out by representing the real electric field in terms of a complex function,
whose real part is the field itself. Such an approach is justified only if the real field
and its complex representation satisfy the same propagation equations. This is true
as long as the equations are linear in the fields. The propagation equations treated in
this chapter contain nonlinear terms, and so it is mandatory to deal with real fields.
The equations describing the y-dependence of the amplitudes E ω and E 2ω are
derived as follows. First, by neglecting the cubic term inside (7.1), and inserting
(7.2) into (7.1), an expression for P(y, t) is found. This expression will contain
terms oscillating at ω, 2ω, 3ω, 4ω, and also time-independent terms. Considering
only those terms that oscillate at ω and 2ω, P(y, t) is given by:
εo χ (1) εo χ (1)
P(y, t) = E ω e−i(ωt−kω y) + c.c.] + E 2ω e−i(2ωt−k2ω y) + c.c.]
2 2
εo χ (2) 2 −2i(ωt−kω y) εo χ (2)
+ Eω e + c.c.] + E 2ω E ω∗ e−i(ωt−k2ω y+kω y) + c.c.]
4 4
(7.3)
As a second step, (7.2) and (7.3) are substituted inside the wave equation (2.7).
By separately considering all the terms oscillating respectively as exp(−iωt) and
exp(−2iωt), and further assuming that the terms containing the second-order deriv-
ative with respect to y can be neglected, the following two equations are obtained:
d E ω (y) ωχ (2)
=i E 2ω E ω∗ exp(iΔky), (7.4)
dy 2cn ω
d E 2ω (y) ωχ (2) 2
=i E exp(−iΔky), (7.5)
dy 2cn 2ω ω
where Δk = k2ω − 2kω . Taking the entrance face into the medium coincident with
the plane y = 0, the initial conditions are:
E ω (0) = 0; E 2ω (0) = 0. (7.6)
Recalling that the beam intensity is related to E ω by:

εo n ω c
Iω (y) = |E ω (y)|2 , (7.7)
2
and noting that:

d Iω (y) εo n ω c d E ω∗ ∗ d Eω
= Eω + Eω , (7.8)
dy 2 dy dy
the following relation is obtained from (7.4) and (7.5):
d Iω d I2ω
+ =0 (7.9)
dy dy
Equation (7.9) indicates that, during propagation into the lossless nonlinear
medium, there is a conservation of the overall optical power: the nonlinear medium
acts only as a catalyst, and so the second-harmonic generation (SHG) process simply
consists of an exchange of power between the two fields.
It is not possible to find an exact analytical solution to the system of (7.4) and
(7.5). However, an approximate solution is immediately found by assuming that the
amplitude of the generated second-harmonic wave is so weak that the amplitude
of the fundamental wave is negligibly reduced by the SHG process. If E ω (y) is
considered a constant, coinciding with E ω (0), then the solution of (7.5) is:
iωχ (2) 2 1 − e−iΔky

E 2ω (y) = E ω (0) (7.10)
cn 2ω iΔk
The second-harmonic (SH) intensity is given by:
2(ωχ (2) )2 4
2 Δky
I2ω (y) = Iω2 (0)y 2 sin . (7.11)
c2 n 2ω n 2ω (Δk)2 2
Equation (7.11) shows that I2ω (y) is an oscillating function of y, at fixed Δk. This
means that there is a continuous exchange of power between pump and SH beam
during propagation. The first maximum of I2ω (y) occurs at the propagation distance
L c = π/Δk, known as the coherence length of second-harmonic generation. If the
propagation distance y exceeds L c , the SH power starts to decrease because the newly
generated SH contributions have an opposite phase with respect to those generated
for y < L c . It is important to note that the amplitude of the spatial oscillation gets
bigger and bigger as Δk decreases. Of course, if the SH power becomes large, the
approximation of no-pump-depletion is no longer applicable.
The conversion efficiency, η2ω , defined as the ratio between SH intensity at the
output and pump intensity at the input, is expressed as:
I2ω (L) 2(ωχ (2) )2

2 Δk L
η2ω = = Iω (0)L 2 2 sinc , (7.12)
Iω (0) c n 2ω n 2ω 2
where L is the propagation length, and sinc(x) = sinx/x. The efficiency η2ω is
maximum if
Δk = k2ω − 2kω = 0, (7.13)

Fig. 7.1 Efficiency of the

second-harmonic generation
process plotted as a function
of the phase mismatch
as shown in Fig. 7.1. This condition, known as the “phase-matching” condition,

implies that
n 2ω = n ω , (7.14)
i.e. the SH wave should travel at the same phase velocity as the pump wave, so that
all the contributions generated during propagation can add in phase.
In general terms, time-invariance and translational invariance (which means
monochromatic plane waves interacting in an infinitely extended crystal) impose
energy and momentum conservation among the interacting waves. Using the photon
concept, energy conservation is satisfied when two pump photons are converted into
one SH photon, and momentum conservation requires that the momentum of the gen-
erated SH photon is the sum of the momenta of the two annihilated pump photons.
Therefore the phase-matching condition can be seen as a momentum conservation
condition.
By assuming Δk = 0, (7.4) and (7.5) can be exactly solved, obtaining:
I2ω (L) = Iω (0) tanh2 (L/L S H ), (7.15)
and
Iω (L) = Iω (0)sech2 (L/L S H ), (7.16)
where the length L S H is defined as

√
λ n ω n 2ω
LSH = . (7.17)
π χ (2) E ω (0)
It is easy to verify that I2ω (L) + Iω (L) = Iω (0) for any L. Recalling that the
asymptotic limit of the hyperbolic tangent for L going to infinity is equal to 1,
Fig. 7.2 The second-

harmonic intensity as a
function of the propagation
distance. At phase matching
I2ω asymptotically coincides
with the input intensity. Out
of phase matching, I2ω is an
oscillatory function
Fig. 7.3 Scheme for

intracavity second-harmonic
generation
(7.15) shows that the input intensity can be fully converted into the second-harmonic
intensity, provided that the propagation distance is sufficiently long (the crystal length
L should be much larger than L SH ). The behavior of the SH intensity as a function
of the propagation distance is illustrated in Fig. 7.2.
The first nonlinear optical experiment was performed in 1961 by Franken and
coworkers by using quartz as the nonlinear crystal. The pump was a ruby laser pulse
at λ = 0.6943 µm. A very weak second-harmonic signal at λ = 0.347 µm was
detected. In that experiment the conversion efficiency was very low, about 10−8 ,
because the phase-matching condition was not satisfied.
SHG becomes particularly efficient if the nonlinear crystal is placed inside the
laser cavity, as schematically shown in Fig. 7.3. Intra-cavity SHG represents a very
convenient method to generate coherent green and blue light starting from near
infrared solid-state lasers. As an example, 532-nm green light can be generated by
frequency-doubling of a neodymium laser in a Fabry-Perot cavity in which both
mirrors are totally reflecting at 1064 nm and the output mirror is totally transmitting
at 532 nm. Once fixed the gain in the active medium, the type of nonlinear crystal,
and the cavity structure, the only free design parameter is the length L of the crystal.
This is a typical optimum-coupling problem: if L is too small there is little second-
harmonic generation, if L is too large there are strong cavity losses.
7.2.1 Second-Order Nonlinear Optical Materials
Similarly to the case of the Pockels effect described in Chap. 4, second-order non-
linear optical interactions can occur only in non-centrosymmetric materials, that
is, in materials that do not display inversion symmetry. Since liquids, gases, amor-
phous solids (such as glass), and even many crystals do, in fact, display inversion
symmetry, χ (2) vanishes identically for such media, and consequently they cannot
produce second-order nonlinear optical interactions. Second-order optical materials
are intrinsically anisotropic, thus electric susceptibilities of any order are tensorial
quantities. The i-th cartesian component of the vector P is expressed as:

3
3
Pi = εo χi j E j + χi jk E j E k (7.18)
j=1 j,k=1
Equation (7.18) represents an extension of the linear treatment developed in

Sect. 2.6 for optically anisotropic media. Assuming that the second-order tensor χi j
is diagonalized, the only non-zero components of the linear susceptibility are χ11 ,
χ22 , and χ33 . Furthermore, in order to simplify the notation, only crystals possessing
a symmetry axis, taken as the z-axis, will be considered. In such crystals, χ11 = χ22 .
Note that (7.18) has the same structure as (4.2) that was written to describe the
Pockels effect, and so the properties of the third-order tensor χ (2) are similar to those
of the electro-optic tensor, with an important difference concerning the symmetry
properties. Since the value of Pi cannot change when we exchange the position of E k
with that of E j inside (7.18), the symmetry relation χi jk = χik j holds in this case.
In the scientific and technical literature, the properties of second-order nonlinear
materials are usually expressed in terms of the di jk tensor that is defined by:
1
di jk = χi jk (7.19)
2
In the standard notation only two indices are used, by replacing the pair jk with one
index m running from 1 to 6. The following convention is used: m = 1 corresponds
to jk = 11, m = 2 to jk = 22, m = 3 to jk = 33, m = 4 to jk = 23, m = 5 to
jk = 13, and m = 6 to jk = 12. The nonlinear susceptibility tensor dim can then
be represented as a 3 × 6 matrix.
Crystalline structures are classified according to 32 distinct symmetry classes.
For each class it is possible to predict, using symmetry arguments, what tensor
components are non-zero. For instance, in the case of lithium niobate the non-zero
components are: d15 , d16 , d21 , d22 , d24 , d31 , d32 , and d33 .
In order to illustrate the basic aspects of nonlinear interactions, it is useful to
discuss some specific situations. Consider a pump beam propagating along the y
axis of a lithium niobate crystal. If the beam is an ordinary wave, linearly polarized
along x, the index m can only take the value 1, so that, among all the non-zero
components of dim , the only accessible ones are d21 and d31 . Since the propagation
is along y, the electric polarization cannot have a longitudinal component, so d21
does not play any role in the SHG process. The only surviving terms from the general
relation (7.18) are:
Table 7.1 Symmetry class, transparency window and nonlinear coefficients for some nonlinear
crystals at λ = 1064 nm
Crystal Class Window (µm) d (10−12 m/V)
Li N bO3 3m 0.35–5.20 d33 = 25.2, d31 = 4.6
d22 = 2.1
KT P mm2 0.35–4.50 d33 = 13.7, d31 = 6.5
d32 = 5.0
LBO mm2 0.16–2.30 d33 = 0.04, d31 = 0.85
d32 = 0.67
Lithium niobate (Li N bO3 ), potassium titanyl phosphate (K T i O P O4 ), known as KTP, lithium
triborate (Li B3 O5 ), known as LBO
P1 (ω) = εo χ11 E 1 (ω)

P3 (2ω) = 2εo d31 [E 1 (ω)]2 . (7.20)
The component P3 (2ω) generates a SH field polarized along z, i.e. an extraordi-

nary SH wave. Therefore the phase-matching condition becomes:
2ω e
Δk = (n − n oω ) = 0. (7.21)
c 2ω
If the pump beam is linearly polarized along z, the only surviving terms from the
general relation (7.18) are:
P3 (ω) = εo χ33 E 3 (ω)

P3 (2ω) = 2εo d33 [E 3 (ω)]2 (7.22)
In this case, both pump and SH fields are extraordinary waves, and the phase-
matching condition is:
2ω e
Δk = (n − n eω ) = 0. (7.23)
c 2ω
In order to prepare the description of the different phase-matching methods that
will be given in the next section, it is instructive to examine the properties of some
common nonlinear crystals—see Table 7.1.
Lithium niobate and KTP exhibit large nonlinear coefficients. LBO has smaller
coefficients, but its transparency window extends into the ultraviolet, and thus is the
best candidate to generate ultraviolet light by SHG.
7.2.2 Phase Matching
The practical importance of nonlinear optics comes from the fact that the new fields
generated by nonlinear interactions can take up, in principle, all the optical power of
the incoming fields. However, as shown in the preceding section, this is possible only
if phase-matching is satisfied. Phase matching should be seen not only as a constraint
that limits the applicability of nonlinear effects, but also as an effective method to
select a single nonlinear interaction, from among many possible interactions. The
structure of the equations is greatly simplified if one defines a priori which interaction
will be phase-matched. According to this criterion, only those terms relevant for SHG
have been selected in writing (7.3).
Birefringence phase matching. Since the index of refraction of any medium changes
with ω, it is generally not possible to satisfy the phase-matching condition (7.23).
In contrast, condition (7.21) suggests the possibility of obtaining phase matching
by exploiting the crystal birefringence to compensate for material dispersion (see
Fig. 7.4). As an example, lithium niobate is a uniaxial crystal presenting negative
birefringence (n o > n e ), as indicated in Table 2.3. Taking into account that both
indices grow with ω, it can be shown that there is a frequency ω pm (corresponding to
a wavelength λ pm ≈ 1.1 µm), for which: n o (ω pm ) = n e (2ω pm ). In such a situation,
an ordinary input beam can be efficiently converted into an extraordinarily polarized
second-harmonic beam.
For a given crystal the condition n o (ω pm ) = n e (2ω pm ) can be at most satis-
fied for only one wavelength. There are, however, some degrees of freedom that
can be exploited to obtain tunable phase matching. One possibility is offered by the
consideration that the refractive index of the extraordinary beam can be varied con-
tinuously between n o and n e by changing the angle between kω and the optical axis,
as described by (2.128). Therefore, one can tune ω pm by rotating the crystal around
the axis perpendicular to the plane containing kω and the z axis. In this case there is,
however, a non-zero spatial walk-off of the second-harmonic beam with respect to
the fundamental beam, because the two Poynting vectors are no longer parallel (see
Fig. 7.4 Birefringence

phase matching in lithium
niobate
Fig. 7.5 Nonlinear crystal

with periodic inversion of the
ferroelectric domains
(2.129) in Chap. 2). Since the spatial overlap between the two beams decreases with
the propagation distance, the walk-off effect introduces a limitation of the crystal
length. Noting that the refractive indices are weakly dependent on temperature, it is
also possible to finely tune ω pm by varying T .
The method so far discussed, in which two ordinary photons at frequency ω are
converted into one extraordinary photon at frequency 2ω is known as type I phase
matching. Other degrees of freedom are added to the SH experiment if two pump
beams having the same frequency, but different directions or different polarization,
are considered. A particularly interesting method, known as type I I phase matching,
uses one ordinary and one extraordinary photon to generate the SH photon. The
phase-matching condition for this method is:
n oω + n eω = 2n e2ω . (7.24)
Quasi-phase-matching. There are circumstances in which birefringence phase-

matching is not suitable for an efficient SHG. For instance, the nonlinear material
may possess insufficient birefringence to compensate for linear dispersion. Another
case is when the application requires the use of the d33 coefficient, which is much
larger than the off-diagonal coefficients for some crystals (see Table 7.1). The d33
coefficient can be accessed only if all the interacting waves have the same polariza-
tion, but the problem is that the phase-matching condition (7.24) cannot be satisfied
because of optical dispersion. There is a method, known as quasi-phase-matching
(QPM), that is alternative to birefringence phase-matching. This method makes use of
a periodically poled nonlinear crystal. As shown in Fig. 7.5, such a material consists
of a sequence of slabs parallel to the optical axis, each having a thickness Λ/2 and
presenting an alternate orientation of the optical axis. The inversion of the orientation
of the optical axis does not change the index of refraction, but has the consequence
of inverting the sign of the nonlinear susceptibility, so that now d33 is not constant
along y, but exhibits a square spatial oscillation with period Λ. Considering the
first harmonic of the function d33 (y), which is expressed by the sinusoidal function
(2/π )d33 exp[i(2π/Λ)y], and substituting inside (7.4) and (7.5), the phase-matching
condition becomes:
2ω e 2π
Δk = (n − n eω ) = . (7.25)
c 2ω Λ
Equation (7.25) implies that Λ = 2L c . Note that, in the case of QPM, the momen-
tum of the SH photon is not equal to twice the momentum of the pump photon. The
difference is taken up by the periodically poled crystal.
An intuitive explanation of the QPM can be given as follows. In the presence
of a wavevector mismatch, the amplitude of the generated SH wave oscillates with
propagation distance, as shown in Fig. 7.2. If the period Λ of the alternation of the
optical axis is set equal to twice L c , then each time the SH amplitude is about to
begin to decrease because of the wavevector mismatch, a reversal of the sign of d33
occurs which allows the SH amplitude to grow monotonically.
Equation (7.25) indicates that phase matching can be achieved at any desired
frequency by appropriately choosing Λ. An important property of the periodical-
poling approach to phase-matching is also that it is effective with any nonlinear
crystal, irrespective of the fact that the material might be birefringent or not. The
technical problem is how to fabricate a periodically poled crystal. If the nonlinear
optical material is a ferroelectric crystal, as is the case for lithium niobate and KTP,
the orientation of the optical axis can be defined by a sufficiently strong static electric
field. Therefore, periodically poled structures can be fabricated by applying electric
fields with a sequence of electrodes of alternating polarity. Typically, electric fields
of the order of 20 kV/mm are applied, with a period that can be as small as 5 µm.
As an example, consider a second-harmonic-generation experiment using a
neodymium laser (λ = 1064 nm) and a KTP crystal. By putting 2d31 instead of χ (2)
inside (7.17), and taking d31 = 6.5 pm/V, it is found that, in order to obtain a reason-
able value for L S H , say 1 cm, it is necessary to have an input field E ω (0) ≈ 4.7 MV/m,
corresponding to a laser intensity of about 5 MW/cm2 . This can be obtained by
focusing the laser output or by coupling the laser beam into a nonlinear single-mode
waveguide. The required intensity is however reduced by a factor 16, if QPM is used
to exploit the larger coefficient of d33 instead of d31 .
7.2.3 SHG with Ultrashort Pulses
Spectral acceptance. The conversion efficiency η2ω , defined by (7.12), reduces to

half of the peak value when Δk L ≈ π/2, as shown in Fig. 7.1. One can there-
fore define a phase-matching bandwidth Δk pm ≈ π/L, and, correspondingly, a
frequency bandwidth Δω pm , usually known as the spectral acceptance. A pump
beam at an angular frequency ω can efficiently generate a SH beam provided that
ω pm − Δω pm /2 ≤ ω ≤ ω pm + Δω pm /2. The expression for Δω pm can be derived
by inserting the first-order frequency-dependence of the refractive indices into the
relation
Δk = (2ω/c)n e2ω − (2ω/c)n o2ω = π/L ,

Fig. 7.6 Optical scheme of an autocorrelator for the measurement of duration of ultrashort pulses
and obtaining:
−1
λ pm dn oω dn e
Δω pm = − 2 2ω . (7.26)
L dω dω
The derivatives in (7.26) are calculated at ω = ω pm .

The concept of spectral acceptance is very useful when dealing with nonlinear
interactions involving optical pulses. As shown in in Sect. 3.11.3, a transform-limited
pulse having a duration τ p has a spectral broadening Δω p of the order of 1/τ p . All the
optical frequencies present in the pulse can contribute to SHG only if Δω p is smaller
than Δω pm . Considering that Δω pm is inversely proportional to L, the shorter is the
pulse, the smaller the crystal thickness should be.
Measurement of the duration of ultrashort pulses. The optical pulses generated
by mode-locked lasers usually have a duration much shorter than the response time
of the photodetectors (see Sect. 5.5), so it is not possible to measure their duration
by direct detection. The measurement can be performed through second-harmonic
generation by the method illustrated in Fig. 7.6. It involves dividing the pulse into two
paths using a beam-splitter, then introducing a variable delay between the pulses, and
then spatially overlapping the two pulses in an instantaneously responding nonlinear-
optical medium, such as a SHG crystal. The best performance is obtained using the
type-II phase matching described by (7.24). The linear polarization of one of the
two pulses is rotated, so that, when they recombine at the SHG crystal, one acts as
an ordinary wave and the other as an extraordinary wave. Type-II phase-matching
requires that second-harmonic photons are exclusively generated by using one photon
from each pulse, therefore the SH signal appears only if there is a temporal overlap
between the two pulses. The detector response time is much longer than the pulse
duration, so that the detected SH signal is an integral over time t. By sweeping
the delay τ between the two pulses, the measured SH power versus τ becomes
proportional to the autocorrelation function of the pulse intensity I (t):
∞
G(τ ) = I (t)I (t + τ )dt. (7.27)
−∞
√ pulse of duration τ p , G(τ ) is also Gaussian, with a width

In the case of a Gaussian
at half-height equal to 2τ p .
Since the delay is changed by changing the path length, and path length variations
can be controlled with an accuracy of the order of one micron, the time resolution of
the measurement is of the order of 1 fs.
7.3 Parametric Effects
Difference-frequency generation. If two beams at frequencies ω1 and ω3 , with ω3 >

ω1 , are sent into a second-order nonlinear crystal, then the nonlinear interactions can
give rise to new frequencies representing either the sum or the difference between
the input frequencies. By choosing a specific phase-matching condition, only one
process can be selected. In the case of difference-frequency generation (DFG), a
new beam at frequency ω2 = ω3 − ω1 is generated, provided that the condition
k2 = k3 − k1 is satisfied. All the phase-matching methods discussed in the previous
section in connection with SHG can also be applied to sum/difference-frequency
generation. If birefringence phase-matching is used, then each of the three waves can
be ordinary (o) or extraordinary (e). For instance, type I phase matching in a negative
birefringence crystal, like LBO or KTP, is described as o − o − e, indicating that
waves at ω1 and ω2 are ordinary, and the highest frequency ω3 beam is extraordinary.
In the DFG process, one photon at frequency ω3 is annihilated and two photons at
frequencies ω1 and ω2 are created, so that the process not only produces a new coher-
ent beam at ω2 , but also amplifies the beam at ω1 . Such an amplification process is
known as “parametric amplification”. Whereas standard optical amplifiers work only
at frequencies corresponding to transitions among specific energy levels, paramet-
ric amplifiers use transparent crystals and can amplify, in principle, any frequency
within the transparency window of the crystal, provided that phase matching can be
achieved.
Considering collinear propagation along the y axis, the total field is given by the
superposition of three fields:
1
E(y, t) = {E 1 (y) exp[−i(ω1 t − k1 y)] + c.c. + E 2 (y) exp[−i(ω2 t − k2 y)] + c.c.
2
+ E 3 (y) exp[−i(ω3 t − k3 y)] + c.c.} (7.28)
The nonlinear propagation equations, derived by using the same approach outlined
for SH generation, are:
d E 1 (y) ω1 χ (2)
=i E 3 E 2∗ exp(iΔky), (7.29)
dy 2cn 1
d E 2 (y) ω2 χ (2)
=i E 3 E 1∗ exp(iΔky), (7.30)
dy 2cn 2
d E 3 (y) ω3 χ (2)
=i E 1 E 2 exp(−iΔky), (7.31)
dy 2cn 3
where Δk = k3 − k1 − k2 .
The situation here considered is that of a high intensity ω3 beam, called pump,
and a low intensity ω1 beam, called signal. The generated ω2 beam is called idler.
Equations (7.29) and (7.30) can be analytically solved at phase matching by assuming
that the pump beam is not depleted by the DFG process, such that E 3 (y) is considered
as a constant, coinciding with E 3 (0). Fixing the initial conditions as E 1 (0) = 0 and
E 2 (0) = 0, the obtained solutions are:
E 1 (y) = E 1 (0) cosh(g p y)

n 1 ω2
E 2 (y) = i E 1 (0)sinh(g p y), (7.32)
n 2 ω1
where the parametric gain per unit length g p is given by:

1 ω1 ω2 (2)
gp = χ E 3 (0). (7.33)
2c n1n2
Equation (7.32) shows that both fields grow monotonically as functions of the
propagation distance. Since the annihilation of a photon at ω3 simultaneously pro-
duces a photon at ω1 and a photon at ω2 , the two intensities I1 and I2 grow in a
parallel way: I1 (y) = (ω1 /ω2 )I2 (y) + I1 (0). As an example, consider the case of a
lithium niobate crystal pumped by a 532-nm beam having an intensity of 5 MW/cm2 ,
producing a field E 3 (0) = 4.1 × 106 V/m. Taking χ (2) = 10 pm/V, n 1 = n 2 = 2.2,
ν1 = 3 × 1014 Hz, it is found: g p = 0.6 cm−1 .
Recalling that an amplifier becomes an oscillator once a positive feedback is
provided, the parametric amplifier can be transformed into a parametric oscillator
by using an optical cavity (see Fig. 7.7). The initial seed for parametric oscillation
comes from parametric fluorescence, that is the process by which a “pump” photon at
frequency ω3 spontaneously breaks down into two photons at frequencies ω1 (signal)
and ω2 (idler). Potentially, parametric fluorescence can yield any pair of frequencies
that satisfy energy conservation, but the choice must also satisfy the phase-matching
condition. Tuning of the parametric oscillator is usually achieved by rotating the
crystal or, in a finer scale, by changing the operating temperature.
Sum-frequency generation. Considering a situation in which the phase-matching
condition is satisfied for the sum-frequency generation (SFG), and assuming again
that the pump at ω3 is not depleted, the two equations describing the evolution of the
signal field at ω1 and the sum field at ω2 = ω1 + ω3 are the following:
7.3 Parametric Effects 207
Fig. 7.7 Oscillator cavity based on parametric amplification
d E 1 (y) ω1 χ (2) ∗
=i E (0)E 2
dy 2cn 1 3
d E 2 (y) ω2 χ (2)
=i E 3 (0)E 1 . (7.34)
dy 2cn 2
Without loss of generality, one can put E 3 (0) = E 3∗ (0), and the solutions are:
E 1 (y) = E 1 (0) cos(g p y)

n 1 ω2
E 2 (y) = −i E 1 (0)sin(g p y), (7.35)
n 2 ω1
where

1 ω1 ω2 (2)
g p = χ E 3 (0). (7.36)
2c n1n2
The intensity conversion ratio in a crystal of length L is:
I2 (L) ω2 2
= sin (g p L). (7.37)
I1 (0) ω1
In the case in which all the input ω1 photons are converted into ω2 photons, the
conversion ratio has the maximum value ω2 /ω1 . This is larger than one, reflecting
the fact that the intensity of the converted beam is the sum of I1 (0) plus the intensity
(ω3 /ω1 )I1 (0) provided by the pump beam.
SFG can be useful for converting an infrared beam into a more easily detectable
visible beam, or for generating ultraviolet light by mixing visible and infrared light.
As a final comment to this section, it should be noted that the linear electro-
optic effect (Pockels effect) of Sect. 4.1 can be considered as a particular case of
frequency mixing, in which the external electric field E at frequency ωm plays
the role of the field E1 . Since ωm is negligibly small with respect to the optical
frequency ω, phase-matching is always satisfied in the case of the electro-optic effect

for the sum- and the difference-frequency processes, so that both frequencies ω ±
ωm are generated. These frequencies are the two side-bands of the electro-optically
modulated signal.
7.4 Third-Order Effects
Considering a centrosymmetric nonlinear material, in which the induced polariza-

tion is the sum of a linear term plus a cubic term, and assuming that the material is
illuminated by a monochromatic plane wave having angular frequency ω and prop-
agating along the y-axis, the electric polarization inside the nonlinear medium can
be written as:
εo χ (1)
εo χ (3) 3 −3i(ωt−k y)
P(y, t) = E ω e−i(ωt−kω y) + c.c. + Eω e ω + c.c.

2 8
3εo χ (3) 2 ∗ −i(ωt−kω y)
+ Eω Eω e + c.c. . (7.38)
8
Many different nonlinear phenomena can be generated by the third-order nonlin-
earity. Here the description will be limited to phenomena occurring when a single
beam at a single frequency and direction is incident on the nonlinear medium. Wave
mixing processes involving more than one incident frequency will not be discussed.
Third-harmonic generation. The electric polarization of (7.38) contains terms
oscillating at ω and at 3ω. Once the expression for P(y, t) is inserted into the wave
equation, the 3ω term will give rise to third-harmonic generation (THG). Formally,
one could expect a behavior of the THG conversion efficiency similar to that found
for SHG. However, the practical use of THG is rather limited because it is difficult to
find materials simultaneously having a large χ (3) , a transparency window including
both ω and 3ω, and also offering the possibility of phase matching. In practice, THG
is usually obtained through a cascade of two second-order processes.
Optical Kerr effect. Neglecting the 3ω term, (7.38) can be re-written as P(y, t) =
Pω exp[−i(ωt − kω y)] + c.c., where:
3χ (3)
(1)
Pω = εo χ (1) + |E ω |2 E ω = εo χe f f E ω , (7.39)
4
Equation (7.39) is similar to (4.37), which describes the Kerr effect. The big
difference is that, in this case, the change in the susceptibility of the medium is due
to the electric field of the propagating wave, whereas, in the case treated in Sect. 4.2,
the effect was induced by a “low-frequency” external field created by an electronic
device.
7.4 Third-Order Effects 209
By using the relation between index of refraction and electric susceptibility, and
assuming that the nonlinear contribution is small in comparison with the linear one,
the effective refractive index is:

(1)
n eff = 1 + χeff ≈ n + n 2 |E ω |2 or n eff = n + n 2 I, (7.40)
where I is the intensity of the light beam and the nonlinear coefficient n 2 , measured
in m 2 /W , is given by:
3χ (3)
n2 = . (7.41)
4εo cn 2
Equation (7.40) expresses the optical Kerr effect. Values of n 2 are given in
Table 7.2 for three materials.
As already discussed for the Kerr effect in Sect. 4.2, a more detailed analysis
would show that an initially isotropic material becomes uniaxial in the presence
of a strong linearly polarized optical beam, with the direction of the optical axis
coinciding with the direction of the incident electric field. The index ellipsoid of the
isotropic medium, initially spherical, acquires an intensity-dependent deformation.
If the optical Kerr effect is only due to the displacement of electron clouds, the
induced birefringence almost instantaneously follows changes in the intensity of the
illuminating beam. Instead of considering a self-induced effect, one could imagine
a cross-induced situation, in which a high-intensity beam modulates a weak probe
beam through the optical Kerr effect, with a response time as short as the duration
of the pump pulse (less than 1 ps). By using the same schemes of Sect. 4.1, a very
fast amplitude- or phase-modulation could be achieved. In principle, this opens the
way to the realization of all-optical modulators and switches.
It is interesting to discuss a numerical example involving the common material
of silica glass, which has χ (3) = 1.8 × 10−22 m2 /V2 , corresponding to n 2 =
2.4 × 10−20 m2 /W, as reported in Table 7.2. A propagation along the distance
L inside the glass produces a nonlinear phase delay Δφ = (2π/λ)n 2 I L. Taking
L = 1 cm and λ = 1 µm, the light intensity required to induce a nonlinear phase
shift equal to π is: I = 2.1 × 1015 W/m2 . Such an intensity could be reached by
focusing a laser pulse with energy of 2 µJ and a duration of 1 ps into a focal spot of
diameter 30 µm. Such an optical switch is conceptually interesting from the point
Table 7.2 Two-photon absorption and Kerr coefficients for silica glass, silicon (Si) and gallium
arsenide (GaAs) at two distinct wavelengths
Material λ (nm) βa (cm/GW) n 2 (cm2 /W)
Silica glass 1064 2.4 × 10−16
Si 1540 0.8 0.45 × 10−13
GaAs 1540 10.2 1.59 × 10−13
of view of realizing all-optical logical circuits, but has little practical applicability,
because it is bulky and requires too large a pulse energy. However, the performance
can be greatly improved by using a material with a larger n 2 and confining the optical
beams inside a waveguide of small cross-section.
Two-photon absorption. In general terms, the third-order susceptibility can be a
(3)
complex quantity, written as: χ (3) = χ Re + iχ I(3)
m . In order to see the effect of the
imaginary part, consider a plane wave, with electric field (1/2)E(y) exp[−i(ωt −
ky)] + c.c., propagating through a medium with a Kerr nonlinearity. The equation
describing the evolution of the field as a function of y can be derived using the same
procedure as that adopted for the case of the second-order nonlinearity. The result is:
dE ω
= i 3χ (3) |E|2 E. (7.42)
dy c
Putting E(y) = A(y) exp[iφ(y)], where A(y) and φ(y) are real quantities, (7.42)
splits into the two equations:
dφ ω (3)
= 3χ Re A2 , (7.43)
dy c
and
dA ω (3)
= − 3χ I m A3 , (7.44)
dy c
describing the evolution of the nonlinear phase and of the field amplitude, respec-
tively. Recalling that the light intensity I is proportional to A2 , and multiplying both
sides of (7.44) by 2 A, it is found that I (y) satisfies the equation:
dI
= −βa I 2 , (7.45)
dy
where
3π χ I(3)
m
βa = , (7.46)
εo n 2 cλ
is known as the two-photon absorption coefficient. Equation (7.45) describes a

decrease of intensity during propagation in an otherwise transparent medium. The
fact that the derivative of the intensity is proportional to the intensity square suggests
that the loss is due to the simultaneous absorption of two photons. If this is the case,
βa can be different from 0 only when the photon energy is larger than half the energy
gap between ground state and excited state.
Integrating (7.45), it is found:
Io
I (y) = , (7.47)
1 + βa Io y
Many experiments performed with semiconductors have indeed verified the

behavior predicted by (7.47). As an example, the values of n 2 and βa for two well-
known semiconductors are reported in Table 7.2. Note that the energy of the incident
photon is larger than half the bandgap energy of the two semiconductors, whereas,
in the case of silica glass, the photon energy of 1.17 eV is too small to give rise to
two-photon absorption.
Self-focusing. If the propagating beam has a finite cross-section, like a Gaussian
spherical wave, then a new effect occurs, because the optical Kerr effect produces
a transverse refractive-index profile that follows the intensity profile of the beam. If
n 2 is positive, a greater index of refraction is induced on-axis than in the wings of
the beam, creating a positive lens that tends to focus the beam. This effect is known
as “self-focusing”. The focal length of the self-induced Kerr lens can be calculated
as follows. Assuming a radial intensity profile of the type Io exp[−2(r 2 /wo2 )], the
radially varying refractive index is given by:
r2
n(r ) = n o + n 2 Io exp[−2(r 2 /wo2 )] ≈ n o + n 2 Io 1 − 2 2 , (7.48)
wo
where this approximate expression is obtained by expanding the exponential in a

power series, and then truncating the expansion after the first order.
If the thickness of the medium is L, then the radially dependent phase delay is:
2π L r 2
φ(r ) = n o + n 2 Io 1 − 2 2 , (7.49)
λ wo
In order to understand the effect of a transverse distribution of the phase delay,

the simplest strategy is to compare (7.49) with the transverse distribution created by
a real lens having focal length f . The radial dependence of the phase delay due to a
lens, as derived from (2.79) in Sect. 2.6, is:
2π r 2
φ(r ) = φo − , (7.50)
λ 2f
By comparing (7.49) and (7.50), the focal length of the Kerr lens, f K , is found
to be:
wo2
fK = , (7.51)
4Io n 2 L
This lensing effect has a practical application: it is used to induce self-mode-

locking in solid-state lasers, such as titanium sapphire lasers. The mechanism is the
following. The multimode laser can operate either in a free-running regime, in which
the modes have random phases, or in a mode-locking regime. If the number of modes
is large, the peak intensity of the mode-locked pulses is much larger than the CW
intensity of the free-running laser. It is assumed that the third-order nonlinearity of
the active crystal gives rise to a Kerr lens in the mode-locking regime, but no Kerr lens
in free running. The effect of the Kerr lens is that of focusing the laser beam at some
point inside the cavity. If a pinhole is placed at that point, the free-running regime
will experience large losses, whereas the mode-locking regime will be unaffected.
As a consequence, the laser is forced to operate as a mode-locked oscillator. More
generally, Kerr-lens mode-locking can be explained by keeping in mind the definition
of cavity stability given in Sect. 3.8. If the cavity is designed to be stable in presence
of the Kerr lens and unstable in absence of the Kerr lens, the laser will choose to
operate in the mode-locking regime in presence of self-focusing. The technique is
known as “Kerr-lens mode-locking”.
When the material thickness becomes comparable to, or even larger than, f K , then
the description of the beam propagation becomes complicated because a nonlinear
wave equation needs to be solved. By increasing the input power, a situation can be
found in which self-focusing exactly balances diffraction, and so the beam propagates
through the nonlinear medium without changing its profile. This phenomenon is
known as “self-trapping”. If the input power is further increased, the beam will
catastrophically focus, causing optical damage inside the medium in many cases.
Self-phase-modulation. If an optical pulse of intensity I (t) is propagating into a
third-order nonlinear medium, the optical Kerr effect produces a time-dependent
change in the refractive index experienced by the same beam. This phenomenon is
known as “self-phase-modulation” (SPM). Neglecting for the moment any optical
dispersion, the phase delay over a propagation length L is given by:
2π L
φ(t) = ωo t − [n + n 2 I (t)] (7.52)
λo
SPM produces a broadening of the optical spectrum. The origin of broadening

can be explained by noting that the instantaneous frequency ωinst , defined as
dφ(t) 2π L d I
ωinst = = ωo − n2 , (7.53)
dt λo dt
sweeps from a value smaller than ωo at the pulse front, where d I /dt is positive, to
a value larger than ωo at the pulse tail.
In optical-communication systems, where short pulses propagate over long dis-
tances inside single-mode optical fibers, it is important to account for the effect of
SPM on the pulse shape. A nonlinear equation describing the pulse evolution along
the fiber can be derived from the wave equation by assuming that the pulse amplitude
A(y, t) is a slowly varying function of y and t (d A/dy
β1 A and d A/dt
ωo A),
and truncating to the second order the power series expansion of the propagation
constant β, as expressed in (6.19). The nonlinear equation obtained is then:
∂A ∂A 1 ∂2 A
+ β1 + iβ2 2 = iγ |A|2 A, (7.54)
∂y ∂t 2 ∂t
where γ is a constant proportional to n 2 :

n 2 ωo
γ = , (7.55)
c Ae f f
and Aeff is the mode area. A typical value in the case of a single-mode silica fiber is:
γ = 2 W−1 km−1 .
In general the combined effect of dispersion and nonlinearity is for a broadening of
the pulse both in time and frequency. However, if β2 and n 2 have opposite signs, as is
the case for a nonlinear medium exhibiting positive n 2 and anomalous group-velocity
dispersion, it is possible to find a situation in which the two effects compensate each
other. A pulse that can propagate unchanged is known as a temporal optical soliton,
and has the form:

t − β1 y
As (y, t) = Aso sech expiκ y . (7.56)
τs
Equation (7.56) describes a pulse traveling at the group velocity β1 −1 , with a peak
amplitude Aso and a duration τs that are mutually connected by the relation

|β2 |
Aso τs = , (7.57)
γ
and κ is given by:

|β2 |
κ= . (7.58)
2τs 2
Note that the pulse duration τs can take any value, but the shorter the pulse the
higher the peak intensity should be, according to (7.57). The pulse energy required,
which is proportional to A2so τs , is inversely proportional to the pulse duration.
In principle, the pulse described by (7.56) could propagate without distortion for
an arbitrary propagation length, and so could be of great interest for long-distance
optical communications. However, the presence of effects that have been neglected
in the derivation of (7.54), such as fiber losses and third-order dispersion, strongly
diminishes the practical applications of solitons.
7.5 Stimulated Raman Scattering
When a light beam at frequency ν p travels across a transparent medium, a usually

small fraction of its power is scattered in all directions because of the presence of
thermal fluctuations inside the medium. In addition to the main Rayleigh scattering
contribution at ν p , the spectrum of scattered light may also contain a much weaker
contribution at the down-shifted frequency νst = ν p − νv , where νv is the frequency
of a vibrational mode of the medium. This secondary process is known as “Raman
scattering”. The energy difference between the incident photon and the Raman scat-
tered photon is used to excite a molecular vibration. The phenomenon occurs in
molecular gases and liquids, and also in amorphous or crystalline solids. In the case
of gases, Raman scattering can occur also in connection with molecular rotational
levels.
Besides scattering events in which vibrational energy is transferred to the mole-
cule, usually called Stokes processes, there are also events in which vibrational
energy is taken from the molecule. This may happen if the molecule is already in an
excited vibrational state. For these anti-Stokes processes, the Raman scattering event
de-excites the molecule, and so the scattered photon has a frequency νast = ν p + νv
(see Fig. 7.8). The number of anti-Stokes scattering events is smaller than that of
Stokes scattering events because the number of excited molecules is smaller than
that of non-excited ones. At thermal equilibrium, the ratio between the anti-Stokes
and the Stokes intensity is given by:
Iast ν 4 hνv
ast
= exp − . (7.59)
Ist νst kB T
Since, typically, hνv is of the order of 0.1 eV, whereas k B T = 0.025 eV at room
temperature, spontaneous Raman scattering is dominated by Stokes emission.
Fig. 7.8 Raman emission

7.5 Stimulated Raman Scattering 215
Fig. 7.9 Raman shift in

silica glass
In the case of spontaneous Raman scattering, the Stokes photon is emitted with
a random direction and a random phase. If the medium is simultaneously illumi-
nated by a pump beam at frequency ν p and a second beam at the Stokes frequency
νst = ν p − νv , then there is a finite probability that a stimulated Raman scattering
(SRS) process occurs, where the scattered Stokes photon is emitted with the same
k-vector and the same phase of the incident Stokes photon. The stimulated process
absorbs one photon from the pump, adds one photon to the Stokes field and, at the
same time, excites a molecular vibration. Therefore, stimulated Raman scattering
amplifies the field at νst at the expense of the pump power. Note that such an ampli-
fication method can be applied to beams of any frequency. In fact, provided that
the medium is transparent to both pump and Stokes frequencies, it is not required
that ν p and νst be coincident with internal resonances of the atomic system, as is
required for a standard light amplifier. To obtain Raman gain the pump field and the
Stokes field need not necessarily be collinear, but the maximum effect is obtained
in the collinear configuration, considering that the experiments are performed with
Gaussian beams of finite transversal size. In the following, the discussion is limited
to SRS processes occurring in single-mode optical fibers, where propagation of the
two fields is necessarily collinear.
Fiber Raman amplifier. In the case of silica glass, Raman scattering offers a broad
spectrum (see Fig. 7.9) centered around a shift of 13 THz, instead of a sharp line. This
shift corresponds to the vibrational frequency of the Si–O bond. The Raman peak is
broadened because glass is a disordered medium, and so the vibrational frequency of
the Si–O bond is somewhat dependent on the random local environment. Assuming
that a strong pump wave at frequency ν p and a weak Stokes wave at a frequency
νst ≈ ν p − 13 THz are launched into the fiber, the Stokes wave gets amplified
through SRS. Since both pump wave and Stokes wave will be fiber modes, their
transversal distribution is fixed during propagation (no diffraction!). Consequently,
only the dependence of the field amplitudes on the propagation distance y has to be
considered. The equation describing the evolution of the Stokes field is:
d E st 3ωst (3) ∗ α
=i χ E p E p E st − E st , (7.60)
dy cn s 2
where α is the loss coefficient. It is evident from the structure of the gain term in (7.60)
that stimulated Raman scattering is a third-order nonlinear process. A very important
property of SRS is that phase-matching is always satisfied because the momentum
difference between incident and scattered photon is taken up by the medium.
The equation describing the propagation of the optical power, Pst , as derived from
(7.60), is the following:
d Pst gR
= Pp Pst − α Pst , (7.61)
dy Ae f f
where Pp is the pump power, Aeff the area of the fiber mode, and g R the Raman-gain
coefficient, defined by:
12ωst n p (3)
gR = χ . (7.62)
εo c2 n st I m
Note that the Raman gain depends on the imaginary part of χ (3) . This is not
surprising because processes in which there is an exchange of energy between wave
and medium generally depend on the imaginary part of the electric susceptibility, as
has already been seen in the case of two-photon absorption.
Assuming no pump-depletion, i.e. a constant Pp , the integration of (7.61) gives
the Raman gain in a fiber of length L:

gR
G = exp Pp − α L (7.63)
Aeff
For instance, at λ = 1.55 µm, taking g R /Aeff = 3.2 × 10−3 m−1 W−1 , Pp = 2 W,
αo = 0.2 dB/km, which means α = 0.46 × 10−4 m−1 , L = 500 m, it is found a
Raman gain of about 25. Signal amplification by SRS is commonly used in optical
communication links.
So far only stationary beams have been considered. What happens with short
pulses? It is found that the Raman gain has a build-up time that is related to the
spontaneous decay time of molecular vibrations. The precise value depends on the
specific vibrational transition, but a typical value is ≈1 ps. Therefore, for pulses
shorter than about 1 ps, the Raman gain becomes smaller and is essentially negligible
for femtosecond pulses.
If only the pump beam is present at the fiber input, then spontaneous Raman scat-
tering provides a weak signal that can act as a seed for growth of the Stokes wave.
If the first Stokes wave becomes sufficiently powerful, then it can in turn generate
the second Stokes, and so on. As a consequence, a large fraction of the input power
can be transferred to the Raman lines. Furthermore, it should be noted that stimu-
lated Raman scattering may be either forward- or backward-propagating in the fiber.
7.5 Stimulated Raman Scattering 217
Fig. 7.10 Scheme of the Raman fiber laser based on a nested cavity. In a laser designed to emit the
third Stokes wave, all the FBGs have high reflectivity, except for the λs3 output coupler
Backward-propagating stimulated scattering process can result in a severe attenua-

tion of the forward propagating wave. The spontaneous generation of a stimulated
Stokes wave is detrimental for Raman fiber amplifiers, and so it is important to know
what is the critical pump power, Pcr , above which significant stimulated effects are
generated even in absence of an input signal. It can be shown that
Aeff α
Pcr ≈ 16 . (7.64)
gR
A typical value for a single-mode fiber is: Pcr ≈ 1.5 W.

Fiber Raman laser. SRS is a very interesting method not only to amplify signals, but
also to generate new laser wavelengths by converting the Raman fiber amplifier into
an oscillator. The availability of high-power diode lasers as pump sources and the
development of high reflectivity fiber Bragg gratings makes it possible the realization
of nested cavities in which several Stokes lines can simultaneously resonate. In this
configuration the pump power can be very efficiently converted to the highest-order
Stokes line.
A Fabry-Perot nested cavity is shown in Fig. 7.10. The feedback elements are
fiber Bragg gratings (FBGs), written directly on the core of the fiber, and placed at
the input and output sides of the fiber. This nested scheme can generate a cascade of
n frequencies through stimulated Raman scattering: each intermediate Stokes line at
wavelength λsk oscillates in a cavity made by a pair of high-reflectivity FBGs, and
can be efficiently converted into the λsk+1 Stokes line. The cascade is terminated at
λsn by a low-reflectivity FBG, which couples a significant fraction of the λsn Stokes
power to the output. An additional grating centered at the pump wavelength ω p yields
a double-pass pumping scheme. The stimulated Raman effect couples waves which
are co- or counter-propagating in the same way. As an example, the pump source
could be a CW diode-pumped cladding-pumped Y b3+ fiber laser at 1117 nm. By
using such a pump, a cascade of five (1175, 1240, 1315, 1395 and 1480 nm) orders
of Stokes can produce the 1480 nm wavelength. The efficiency of such a laser can
be very large: typically 50 % of the input power at 1117 nm can be converted into
output power at 1480 nm. Such a source can be very useful as a pump for high-power
and remotely-pumped erbium amplifiers.
7.6 Stimulated Brillouin Scattering
In a dense medium another effect of thermal fluctuations is to generate spontaneous

acoustic waves oscillating at all possible frequencies and propagating in all directions.
As has been described in Chap. 4, the interaction between an optical wave and an
acoustic wave may give rise to a frequency-shifted diffracted wave. The phenomenon
by which an incident beam is diffracted by a randomly excited acoustic wave inside
condensed matter is known as Brillouin scattering.
Considering a plane wave with frequency νi and wave vector ki incident on the
medium, Brillouin scattering is observed in all possible direction. Selecting light
scattered at an angle α with respect to the direction of ki , the observed spectrum of
the scattered light is found to contain three peaks, one at νi and two at the frequencies
νi ± νs , where νs is the frequency of the spontaneous acoustic wave that produces a
diffracted beam in the chosen direction. Taking kd as the wave vector of the photon
diffracted at angle α, then the wave vector ks of the acoustic wave involved in the
process is given by (4.43) (see also Fig. 4.12). The peaks at the frequencies νi ± νs
are both found in the spectrum because both orientations of ks are present inside the
medium. The modulus of ks is derived from (4.40), taking into account that the angle
θ of (4.40) coincides with α/2. The corresponding acoustic frequency is:
ks ki
νs = = sin(α/2) (7.65)
2π u s π us
Similarly to the case of SRS, it is possible to have stimulated Brillouin scattering

(SBS) at the frequency νi − νs . The SBS wave is usually observed in the backward
direction (α = π ), because the maximum overlap between the incident and the
scattered wave occurs for that direction. In the forward direction ks = 0, so there
can be no Brillouin shift.
While the material nonlinearity of silica is actually not very high, the small effec-
tive mode area and long propagation length used in fibers tend to strongly favor
nonlinear effects. For silica fibers, the Brillouin frequency shift is of the order of
10 − 20 GHz. The dependence of the Brillouin frequency shift on the material com-
position and, to some extent, on the temperature and pressure of the medium can be
exploited for fiber-optic sensors.
If only the pump beam is present, then spontaneous Brillouin scattering provides
a weak signal at νi − νs that can act as a seed for growth of a stimulated Brillouin
scattering wave. Inside optical fibers the spontaneous generation of SBS is frequently
encountered when narrow-band optical signals (e.g. from a single-frequency laser)
are amplified in a fiber amplifier, or have just propagated through a passive fiber.
This effect is usually undesirable, and so the optical setup is designed to avoid the
generation and propagation of SBS.
Problems 219
Problems
7.1 The frequency of a Nd-YAG laser output (λ = 1064 nm) is to be doubled in

a lithium niobate crystal by type-I birefringence phase matching. Knowing that, for
Li N bO3 , n o (λ = 1064 nm) = 2.2347, n e (λ = 532 nm) = 2.2316, and n o (λ =
532 nm) = 2.3301, calculate the phase-matching angle between the optical axis and
the direction of the incoming beam.
7.2 Calculate the second-harmonic conversion efficiency for type-I harmonic gen-
eration in a perfectly phase-matched 2-cm long KTP crystal with an incident beam
at λ = 1064 nm having an intensity of 3 MW/cm2 . (For KTP, n ω = n 2ω ≈ 1.8, and
χ (2) = 13 pm/V).
7.3 The generation of the second-harmonic of 1064-nm light is performed by using

a periodically poled lithium niobate crystal (n e = 2.1550 at λ = 1064 nm and
n e = 2.2316 at λ = 532 nm). Calculate the period of periodical poling required for
phase matching of second-harmonic generation.
7.4 A parametric amplifier uses a 2-cm long KTP crystal (n 1 ≈ n 2 ≈ 1.77, χ (2) =
10 pm/V) to amplify light of wavelength 580 nm in a collinear wave configuration.
The pump wavelength is λ = 400 nm and its intensity is 4 MW/cm2 . Determine the
gain coefficient and the overall gain.
7.5 A Q-switched Nd-YAG laser produces an output pulse containing 10 mJ of

energy with a pulse duration of 10 ns. The pulse generates an optical Kerr effect
in a 2-cm thick carbon disulfide (n 2 = 3.2 × 10−14 cm2 /W) cell. Assuming that
wo = 100 µm, calculate the focal length due to the self-focusing effect.
7.6 A 1540-nm 2-ns laser pulse impinges on a 1-mm thick GaAs plate. Assuming
a spot size of 0.2 mm and a two-photon absorption coefficient βa = 10.2 cm/GW,
calculate the transmittance of the plate for the two cases in which the pulse energy
is 1 mJ and 100 µJ.
7.7 A laser pulse at λo = 1064 nm propagating into an optical fiber excites a cascade
of stimulated Raman scattering processes. Assuming that the Raman frequency shift
is 13 THz, calculate the wavelength of the second Raman Stokes line, λs2 .
Chapter 8
Applications
Abstract This chapter concisely describes the most important applications of pho-
tonics, which are based on the optical methods and devices discussed in Chaps. 1–7.
Section 8.1 deals with the applications to information and communication technol-
ogy. After introducing the design principles of digital information transmission and
recording, the structure and properties of an optical communication link are pre-
sented. This is followed by a description of optical memories and of the prospects
of integrated photonics. Section 8.2 is dedicated to optical metrology and sensing,
including the optical radar, the laser gyroscope, and fiber-optic sensing of temper-
ature, strain, and electric current. The achievements and the potentialities of laser
materials processing are reviewed in Sect. 8.3, with some examples of the many
possible biomedical applications treated in Sect. 8.4. The last three sections concern
topics of very wide interest, with an impact on everyday life: displays, lighting and
photovoltaic energy conversion. In all of these, photonic methods and devices play
an increasingly important role, even if lasers are not directly involved.
8.1 Information and Communication Technology
Photonics pervades all the aspects of the information and communication technol-
ogy. Optical-fiber transmission of data, voice, and video is the basic technology for
long-haul communications and internet traffic, and is now also reaching individual
users directly via fiber-to-the-home broadband networks. Optical discs are used for
distributing digitally encoded music and films, and for storing large amounts of data.
Printing and photocopying machines rely on lasers. High-speed data processing will
benefit from the development of integrated photonic circuits in the near future.
Before describing specific applications, it is useful to discuss some general aspects
concerning digital information recording and transmission.
A written text is a combination of alphabetical and mathematical symbols. Each
symbol is identified by a sequence of 7 binary units, known as bits, allowing 27 = 128
combinations (ASCII code). As an example, a page of 2000 characters corresponds
to a sequence of 14000 bits, that is, 14 kb.

DOI 10.1007/978-3-319-20627-1_8
222 8 Applications
An acoustic signal (voice and sound) induces a vibration of a receiving mem-

brane that is converted into an electric signal by a transducer. The electric signal is
digitalized by an analog-to-digital converter, which transforms an analog function of
continuous time into a numeric sequence of discrete times. The minimum sampling
interval is determined according to the sampling theorem of Nyquist and Shannon,
which states: if a function of time f (t) has a spectral width Δν, it is completely
determined by giving its values at a sampling rate of 2Δν. If the digitalization makes
use of 28 = 256 levels, then each sample is given by a sequence of 8 bits. The typical
bit rate required for a digital audio signal is about 60 kb/s.
A video signal consists in a sequence of 20 images per second. An image is a
matrix of about 700 × 500 = 350,000 pixel. This number needs to be multiplied by
a factor of 3 to account for the RGB (red, green, blue) color code. The required bit
rate can be considerably reduced by using very effective compression techniques.
8.1.1 Optical Communications
An optical communication system uses a transmitter, which encodes a message into

an optical signal, a waveguide, which carries the signal to its destination, and a
receiver, which converts the optical bit stream into the original electrical form. Opti-
cal fiber is the standard type of waveguide for optical communications. Free-space
optics is useful for communications between spacecrafts, but is not of much interest
for terrestrial links, because its range is limited and the quality of the link is too highly
dependent on atmospheric factors. The first fiber-optic communication systems have
been installed in the 1980s. In particular, the first transatlantic submarine fiber-optic
cable was laid down in 1988 in order to connect the United States and Europe. Two
years later a submarine connection linked the United States to Japan. The world-wide
deployment of fiber systems took place in the 1990s, following the development of
single-mode semiconductor lasers and optical fiber amplifiers. The carrier wave-
length, which must be close to the minimum-attenuation wavelength of silica-fibers,
varies from 1520 to 1620 nm (185 to 200 THz). The interval 1530–1570 nm is known
as the C band, while the L band corresponds to 1570–1620 nm.
A scheme of a point-to-point optical-communication system is shown in Fig. 8.1.
The most common light source for high-bit-rate optical communications is the single-
mode DFB semiconductor laser. The optical signal to be transmitted is a stream of
bits, created by an external modulator or by direct modulation of the laser pump
current. The signal encoding is typically simple amplitude modulation, in which the
presence of a pulse means 1 and the absence means 0. The direct-modulation method
is simple, but the generated pulses are usually not transform-limited, and so are
more strongly degraded by optical dispersion in long distance communications. The
external modulation by an electro-optic modulator of the type described in Chap. 4 not
only ensures a better pulse quality, but also offers a greater flexibility in the choice
of the modulation format. An interesting encoding method, which may substitute
amplitude modulation, is the phase modulation method known as “differential phase
8.1 Information and Communication Technology 223
Fig. 8.1 Scheme of a point-to-point fiber-optic link. A optical amplifier
shift keying” (DPSK). Phase modulation formats allow for more than two phase
levels per single bit, and thus provide an increase in bit-rate, obtainable without
changing the physical structure of the link.
The existing links are using single-mode fibers of type SMR or DS. The effects of
optical dispersion can be compensated for by inserting DIS-CO fibers into the link.
These have a large dispersion parameter, D, with a sign opposite to that of SMR or
DS fibers. Another possibility for compensation is through the use of chirped fiber
Bragg gratings, which are specifically designed to provide a large dispersion with
the appropriate sign.
Since the optical signal is attenuated during propagation along the fiber, there
is a need for periodic signal regeneration. The amplification is usually provided by
the erbium-doped fiber amplifiers (EDFAs), described in Sect. 6.6. Semiconductor
optical amplifiers (SOAs) can also be used for signal regeneration. These enjoy
different advantages and disadvantages with respect to EDFAs. The main advantages
are electrical pumping, small size, low cost, and a central wavelength that is selectable
through the choice of material. The main disadvantages are the low saturated output
power (which means limited linearity), the sensitivity to signal polarization and to
temperature, the high insertion loss, and high noise. In particular, since the gain
and refractive index of a SOA can evolve on the picosecond time scale, the amplified
pulse may be distorted in both amplitude and phase. In contrast, the gain dynamics of
EDFAs is slow, and so picosecond pulses are linearly amplified with a gain determined
only by the average pump power. As discussed in Sect. 7.6, signal regeneration can be
also performed using Raman fiber amplifiers, which are based on stimulated Raman
scattering. Raman fiber amplifiers can cover a very broad bandwidth. They are often
used in optical links, as amplifiers complementary to EDFAs.
At the receiver a photodetector, such as a p-i-n photodiode or an avalanche pho-
todiode, is used for optical-to-electrical conversion. The demodulation re-creates the
electrical bit stream used at the transmitter to modulate the optical carrier. In every
communication link there is some unavoidable noise added during transmission and
at the receiver, and this leads to errors. The performance of a digital communication
system is measured by the probability of error per bit, known as the “bit error rate”
224 8 Applications
Fig. 8.2 Wavelength division multiplexing technique
(BER). The BER is required to be <10−9 , which means one wrong bit every billion
bits. All receivers need a certain minimum power to operate reliably. This power
level is known as the receiver sensitivity.
One way of increasing the transmission bandwidth of an optical link is to employ
the technique of “wavelength division multiplexing” (WDM). In this scheme, signals
with different carrier wavelengths are mixed together (i.e. multiplexed) and then
transmitted simultaneously down a fiber, according to the scheme shown in Fig. 8.2.
This technique increases the capacity of an optical fiber system without changing
the fiber and without increasing the bit rate in the single carrier. In order to minimize
cross-talk, the frequency separation between two adjacent carriers (called “channels”
in the optical-communication jargon) must be greater than the modulation bandwidth
used by a single channel. A typical situation is that of a 40-channel WDM system,
using 40 lasers, all operating in the C band, with a wavelength spacing of 0.8 nm
(corresponding to a frequency spacing of 100 GHz). If the bit rate of each channel
is kept around 10 Gb/s, then interference among different channels is negligible.
This system has the same capacity of a single channel that transmits at 400 Gb/s. At
present, WDM systems with a capacity of several Tb/s are available. WDM can play
an important role in a network where a number of users are connected to a central
hub. Each user is assigned a particular transmission wavelength. The individual
wavelength required by each user is isolated by using a narrow-band optical filter.
8.1.2 Optical Memories
In an optical memory, data are stored in a digital format on an optical disc, and can be
read-out by a laser beam. An optical disc is a flat circular disc on which binary data
(bits) are encoded in the form of pits (binary value of 0, due to lack of reflection when
read) and lands (binary value of 1, due to a reflection when read) on a special material
on one of its flat surfaces. The diameter of a single pixel and the pixel separation
are about 1 µm. A laser source writes one bit of data at a time on the disc, typically
through a thermal mechanism. The pixels are written along a spiral developing from
an inner radius to an outer radius.
8.1 Information and Communication Technology 225
A variety of different materials and recording mechanisms can be used. Both

write-once read-many (WORM) and magneto-optic read/write discs are presently
available. In the case of WORM discs, a focused laser beam heats a spot on the surface
layer during the recording stage. The local reflectance of the surface is irreversibly
changed, either by creating a pit or by modifying the surface properties. In the case
of the read/write disc, light absorption reversibly transforms the crystalline structure
of the surface layer (typically made by the AgInSbTe or GeSbTe compounds) to
produce a large difference in reflectivity through a phase transition. If the material is
magneto-optic, such as some rare-earth transition-metal alloys of TbFeCo, the effect
of the phase transition is to rotate the polarization of reflected light. The rotation is
detected by crossed polarizers.
The first generation of the compact discs (CDs), developed in the mid-1980s, were
designed to hold up to 75 min of music which required 650 Mb of storage.
Optical discs are usually between 7.6 and 30 cm in diameter, with 12 cm being
the most common size. A typical disc is about 1.2 mm thick, while the track pitch
(distance from the center of one track to the center of the next) is typically 1.6 µm.
Optical discs are most commonly used for storing music (e.g. for use in a CD
player), video (e.g. for use in a Blu-ray player), or data and programs for personal
computers. For computer data backup and physical data transfer, optical discs are
gradually being replaced with faster, smaller solid-state devices, especially the USB
flash drive.
The objectives used to focus the light beams are characterized by their numerical
aperture (NA), defined in Sect. 2.2.9. When a light beam is focused by an objective
with a given NA, the diameter of the obtained focal spot is:
aλ
df = , (8.1)
NA
where a is a numerical constant of the order of 1. The area used to record one bit of
information is 2d f 2 . If λ = 0.8 µm, it is possible to memorize 150 Mb over 1 cm2 .
The basic digital memory unit is the byte (B), which is equivalent to 8 bit. To give
an example of the used notation: 106 Byte = 1 MegaByte = 1 MB.
An important technical aspect concerning the design of optical disc readers is that
the distance between lens and disc surface must remain constant to within an accuracy
of the order of the Rayleigh length (a few microns). This requires a sophisticated
real-time active stabilization system.
The most common optical discs have the following properties.
Compact Disc. A Compact Disc (CD) uses a GaAlAs semiconductor laser, emitting
infrared light of 780 nm, for writing and reading. The laser beam is focused with an
objective that has a NA = 0.45–0.50. The size of the focal spot on the disc surface is
1.6 µm. The CD capacity is of 680 MB.
Digital Versatile Disc. A Digital Versatile Disc (DVD) uses a GaAlAs semiconductor
laser, emitting red light of 650 nm, for writing and reading. The laser beam is focused
with an objective that has a NA = 0.60–0.65. The size of the focal spot on the disc
226 8 Applications
surface is 1.1 µm. The DVD capacity is of 4.7 GB, and can be doubled by using both
disc surfaces.
Blu-Ray. A Blu-Ray disc uses a Ga I n N semiconductor laser, emitting light of
405 nm (notwithstanding the name of the disc, this is violet light!), for writing and
reading. The laser beam is focused with an objective that has a NA = 0.85. The size
of the focal spot on the disc surface is 0.48 µm. The DVD capacity is of 25 GB, and
can be doubled by using both disc surfaces.
The capacity of optical memories could be significantly increased by using “vol-
ume” storage instead of “surface” storage. Such a possibility is actively being inves-
tigated, but, at present, no commercial products are available.
8.1.3 Integrated Photonics
Optical systems, as they have been described throughout this book, contain vari-
ous discrete elements, each made using different materials and technologies, with
dimensions several orders of magnitude larger than the wavelength of light. A major
challenge for the future will be the development of high-bandwidth, low energy con-
sumption optical circuits at low cost. This will be possible provided optical systems
evolve towards combining all devices and components for the generation, manipu-
lation, and detection of light onto a single substrate, i.e. a single “chip”. Such a chip
is the optical analogue of an electronic integrated circuit, and should possibly be
based on silicon, in order to exploit the technologies already available in microelec-
tronics. Silicon is abundant, non-toxic and environmentally friendly. Since silicon is
transparent to light of wavelength larger than 1.15 µm, silicon integrated photonics
operates in the infrared. Photonic integration will play a key role in reducing cost,
space and power consumption, and improving flexibility and reliability.
An important application of integrated photonics would be that of realizing on-
chip processors for short-distance optical communications, i.e. the so-called “local
area networks” (LAN). Another very important area concerns optical interconnec-
tions, through transceivers, which are devices that convert electric signals into optical
signals and transport the optical signal from one device to another device. Optical
data transportation is immune from induction effects, and may also reduce energy
dissipation through Joule heating. Indeed, it is believed that a major obstacle for the
realization of next-generation supercomputers will be energy dissipation due to data
transfer among microprocessors and memories. To save energy, power-consuming
parts of the electrical interconnects should be replaced by optical interconnects.
Research and development in the field of silicon photonics is being very actively
pursued. Promising results have been obtained in the design and fabrication of pas-
sive components, such as waveguides, mirrors, and filters. Various types of silicon
modulators have also been tested. The main difficulties concern light amplification,
because electron-hole recombination is not radiative in bulk silicon, as discussed in
Chap. 5.
8.2 Optical Metrology and Sensing 227
8.2 Optical Metrology and Sensing
In many industrial processes, tests of structural stability, and environmental studies,

there exists a need to monitor variables such as displacement, strain, temperature,
pressure, flow rate and chemical composition. Sensing is also important because the
data acquired in real-time may be used as input signals for automatic systems of
industrial production and environmental control.
Optical sensing is useful in a variety of areas, because of its high accuracy, immu-
nity to electromagnetic disturbances, little invasiveness, and the possibility to operate
in a hostile environment. In order to illustrate the potential of optical sensing, a few
applications are below described, without any pretension to be exhaustive.
8.2.1 Measurements of Distance
To measure distances, light is emitted towards and reflected (or scattered) from an
object. Using the speed of light, and the measured time-of-flight (or phase delay)
of the reflected light, it is possible to calculate the distance between laser emitter
and target object. Different optical methods are available, depending on the range of
distances and on the required accuracy.
Optical radar. Distance measurement with a pulsed-laser radar is based on the time-
of-flight principle: the distance is determined from the time needed by a short light
pulse to travel from the sender to the object, and back. In a range from a few meters to
several hundred meters, this is the method of choice for determining distances. In the
case of pulsed laser radars, every measurement unambiguously corresponds to the
distance to a target object, even if multiple targets are hit by the laser beam. A laser
radar can process fast variations in distance without dynamic errors. This is necessary
for fast scanning of scenes with pronounced distance variations, or measurements of
rapidly changing distances. Laser diodes are generally employed as the light sources.
Fast and sensitive avalanche photodiodes (APD) provide the detectors.
The largest distance ever measured by an optical radar is the distance between the
Earth and the Moon, L EM , which is about 3.8 × 105 km, corresponding to a delay τ
of 2.5 s. Of course, L E M is not constant because the Moon’s orbit around the Earth
is elliptical. It is instructive to discuss the problems encountered in making such a
measurement. The direct signal due to back-scattering from the lunar surface is too
weak to be detected by the receiver, so it was necessary to place a high-reflectance
mirror on the Moon. A normal mirror would be of little use, because it cannot reflect
back in exactly the opposite direction, independently from the value of the incidence
angle. The mirror brought by the astronauts of the first Apollo mission in July 1969
consisted of a two-dimensional matrix of corner cubes (see Sect. 2.2.6), with an area
of As ≈ 1 m2 . Assuming that the laser beam has a divergence of θo = 10−4 rad,
and using (1.40) and (1.41), the radius of the illuminated area on the lunar surface
228 8 Applications
is given by w L = L E M θo = 38 km. The power fraction intercepted by the installed

mirror is As /(π w L 2 ). If the back-reflected beam is collected over a receiver area
equal to As , the total fraction of power received by the detector is:
As
≈ 5 × 10−11 . (8.2)
4π w L 2
Such a small value would make the measurement difficult. However, the power
ratio of (8.2) can be increased by enlarging with a telescope the radius of the laser
beam.
Time-of-flight sensing in pulsed mode is the dominant technique for distances
longer than 50 m and is routinely used for many civilian applications of range sensing,
such as mapping and surveying.
An alternative approach to pulse illumination is to use amplitude modulated con-
tinuous light. In this case a phase shift in the modulation signal is measured between
the launched and the returned light, and the time-of-flight is determined by divid-
ing the phase shift by the modulation frequency ωm . The measurement may present
an ambiguity because the measured phase shift is a periodic function of the dis-
tance. This ambiguity can be eliminated by appropriately choosing the modulation
frequency or by measuring at more than one modulation frequency.
Interferometric method. The target is illuminated by a continuous laser beam,
exhibiting a very good amplitude and phase stability. The back-reflected beam is
superposed with a fraction of the incident beam using a Michelson interferome-
ter, as shown in Fig. 8.3. The light intensity at the detector is proportional to the
square modulus of the sum of the two fields. By using the same approach utilized
to describe the Mach-Zehnder interferometer in Sect. 4.1.2, it can be shown that the
output signal from the photodetector is proportional to cos2 (2π ΔL/λ), where ΔL
is the difference between the two arms of the interferometer. There is ambiguity in
such a result, because the addition of an arbitrary number of half-wavelengths to
Fig. 8.3 Michelson

interferometer
Fig. 8.4 Distance

measurement through
triangulation
ΔL would not change cos(Δφ). Generally, interferometry is more appropriate for

displacement monitoring than for absolute distance measurement, though multiwave-
length and scanning interferometry permit absolute distance measurement as well.
Interferometric techniques are particularly suitable to detect small displacements or
small-amplitude vibrations in structural and mechanical engineering investigations
or in geodetic surveys.
Triangulation. A collimated light source is used to illuminate the target whose
distance is to be measured. A camera lens optics, laterally displaced from the light
source, images the laser spot on the object onto a position-sensitive detector (typically
a linear array of CCD detectors), as shown in Fig. 8.4. By knowing the distance
between source and lens and measuring the position of the image of the light spot on
the array of detectors, the distance of the object can be deduced. It has been reported
that a similar method was used around 600 BC by the Greek mathematician Thales
of Miletus to measure the height of the pyramids of Giza and also to determine
the distance to a ship at sea. Triangulation is used to fix the focusing distance in
photographic cameras. Triangulation sensors offer the advantages of a low price
and fast measurement (data acquisition rates of tens or hundreds of kilohertz are
possible). Red laser light is normally used, but the eye-safe 1.5 µm infrared light
can also be employed. The useful range of distances for commercial instruments is
from one centimeter to a few meters. Adding the capability to scan the position of the
triangulation laser spot in the object plane opens up the possibility for object shape
sensing.
Self-mixing interferometry. Self-mixing interferometry uses a single-mode laser,
typically a simple semiconductor laser, and exploits the fact that any back-reflected
signal from the target that reenters the laser cavity will modulate the laser output. This
occurs because the target and the diode internal cavity can be regarded as forming
a compound cavity. Because of feedback from the object reflection, the laser power
output oscillates as the target object is displaced; one oscillation period corresponds
to a target displacement of half the wavelength. If the feedback is moderately strong,
then the oscillations are asymmetrically distorted, and it is possible to distinguish
230 8 Applications
between advancing and receding displacement. The self-mixing technique is inex-

pensive and offers high accuracy in displacement measurement, and can be extended
to velocity and vibration measurements. Absolute distance measurements have been
reported by observing self-mixing under current modulated operation.
8.2.2 Velocimetry
Successive time-of-flight distance measurements can determine the rate of change

in object position, that is, the object velocity. A typical application is in traffic speed
enforcement, using a laser speed gun, which repeatedly measures the distance to a
moving car at a high repetition rate.
Similarly, successive distance measurements can be used to monitor periodic
motion such as vibrations. This is subject to the following restrictions: (i) The ampli-
tude of motion is larger than the distance resolution of the measurement, and (ii) The
period of motion is significantly longer than the time required for each individual
distance measurement.
Laser Doppler velocimetry. An alternative optical technique for direct measure-
ment of velocity and vibration is optical Doppler sensing. This deduces the velocity
of moving objects by detecting the Doppler frequency shift of light reflected, or
scattered, from a moving object. In optical Doppler sensing the light source is a CW
laser, and detection is usually by a heterodyne technique. The detector simultane-
ously receives the launched frequency and the Doppler shifted frequency from the
moving object, and the addition of these two waves produces a beat at the difference
frequency (easily seen on an oscilloscope) at the detector output. Fourier transform
analysis of the detector signal, performed by an electronic spectrum analyzer, gives
the frequency shift.
The general expression of the Doppler shift is given by (4.41). In the particular
case of back-reflection (or back-diffusion), kd = −ki , the Doppler shift is:
u
Δν D = νd − νi = 2 νi , (8.3)
c
where νi is the laser frequency, νd the frequency of the back-reflected beam, and u
the projection of the target velocity on the beam direction. The projection is taken
as positive if it is directed toward the emitter. As an example, using a laser beam at
600 nm, corresponding to νi = 5 × 1014 Hz, and considering an object that travels
toward the emitter at a speed of 36 km/h = 10 m/s, a positive Doppler shift of
Δν D ≈ 33 MHz is calculated. It should be noted that the sign of the Doppler shift is
lost when measuring the beat frequency at the detector output. A simple technique
that allows for the recovery of the velocity direction consists of the introduction of
a controlled frequency shift (by using one of the modulation methods described in
Chap. 4) on that part of the laser beam which is sent to the detector. If the laser
frequency is shifted by νs , then the measured shift becomes Δν D = νd − νi − νs =

2(u/c)νi − νs . In this case, a positive u gives a shift different from that of a negative
u, and so the ambiguity is removed.
Laser Doppler velocimetry is a non-invasive method with useful applications in
fluidodynamics, where measurements of the velocity field in the laminar or turbu-
lent motion of a liquid inside a duct can be performed, and in aerodynamics, where
wind tunnel flow measurements can provide unique insights into complex aerody-
namic phenomena. In addition, laser velocimetry has health care applications, such
as arterial blood flow measurements. In all these cases, the return signal does not
come from the fluid itself, but is rather due to scattering from suspended particles
following the local fluid motion. If suitable particles are not already present in the
fluid, then seeding of controlled-size (typically of diameter around 1 µm) particles
can be carried out.
The most common technique utilized in fluidodynamics is that of dividing the
laser beam in two parts, which cross each other in the fluid region where the velocity
has to be measured, as shown in Fig. 8.5. Calling θ the angle between the propagation
directions of the two beams, a periodic set of parallel interference fringes is formed
inside the intersection region, with a fringe spacing given by:
λ
s= . (8.4)
2 sin(θ/2)
A particle suspended in the flowing fluid produces, when crossing the periodic
set of fringes, a scattered light intensity that is proportional, at each point to the
local light intensity. Therefore, the light intensity scattered by the flowing particle is
oscillating in time, with a period equal to the flight time between two consecutive
fringes (i.e. between an intensity maximum and the next intensity maximum). The
oscillation period is TD = s/u, where u is the velocity component in the direction
perpendicular to the fringe planes. Since TD is independent from the observation
Fig. 8.5 Scheme of a laser Doppler velocimeter

232 8 Applications
direction, the scattered light signal can be collected with a large aperture lens. The
period TD is derived by analyzing the oscillations in the electric current output of
the photodetector. If the full velocity vector is needed, then the measurement must
involve three sets of fringes with different orientations.
Gyroscope. Rotation sensors, known as gyroscopes, are essential instruments for
aircraft navigation. Optical gyroscopes rely on the Sagnac effect, which consists
in a phase shift induced between two light beams traveling in opposite directions
round a circular path when the platform on which the optical components are fixed
is rotating about an axis perpendicular to the platform. In a single-mode ring laser
cavity, the two oppositely directed traveling beams have the same frequency, which
is inversely proportional to optical path length in a cavity round trip, as discussed
in Sect. 3.8. Thus any physical mechanism that causes the optical paths to differ for
the two beams, results in the oscillation frequency being different. To calculate the
expected phase shift, assume that the cavity is a circular ring of radius R. Neglecting
the correction due to the fact that part of the path may not be in air, if there is no
rotation the optical path corresponding to one turn is equal to 2π R for both beams.
Therefore both beams will return to the starting point in a time T = 2π R/c. If,
however, the ring is rotating in a clockwise direction at a rate Ω rad/s, the clockwise
beam will arrive at the starting point in a time T + longer than T because the optical
path is now 2π R plus the extra path ΔL = Ω RT + due to the platform rotation.
From the equation
2π R + ΔL
T+ = , (8.5)
c
and the similar one that can be written for T − , the round-trip time taken by the
counter-clockwise beam, the two round-trip times are derived as:

+ 2π R RΩ − 2π R RΩ
T = ≈ T 1+ ; T = ≈ T 1− , (8.6)
c − RΩ c c + RΩ c
where it is assumed that RΩ c. By using (8.6), the frequency difference is found

to be proportional to the rotation rate of the cavity:
c T+ − T− 2RΩ
ν+ − ν− = = , (8.7)
λ T λ
where λ is the unperturbed laser wavelength.
Thus by measuring the frequency difference, the rotation of the laser cavity (and
any vehicle to which it is mounted) can be determined. The ring laser gyroscope is
currently utilized in the inertial navigation systems of military aircrafts, commercial
airliners, ships and spacecrafts.
8.2.3 Optical Fiber Sensors
Most optical sensors are based on the following principle: if the refractive index of
a material depends on an applied external field, then the amplitude of the external
field can be determined by measuring the change of phase delay suffered by a light
beam that is traveling across the material. Since the refractive index of a material
can depend on temperature or pressure or electric field (Pockels effect) or magnetic
field (Faraday effect), it becomes possible to design optical sensors measuring each
of these fields. Particularly interesting, because of their high sensitivity and design
simplicity are optical fiber sensors. In this section, a few examples are considered
with the aim of illustrating the possibilities of this approach.
Temperature and strain sensors. In principle, a measurement of phase delay
requires an interferometer, such as the Mach-Zehnder interferometer, described in
Chap. 4, or the Michelson interferometer, shown in Fig. 8.3. A much simpler scheme
can be utilized with optical fibers by exploiting the properties of the fiber Bragg
gratings (FBGs) described in Sect. 6.5. FBGs reflect only the wavelength of light
that corresponds exactly to the periodicity of the grating. The Bragg wavelength,
as expressed by (6.37), can be shifted by changing the grating period or the effec-
tive refractive index of the grating. Strain modifies the grating period. Temperature
variations affect both the refractive index and the grating period, but around room
temperature the effect of temperature on refractive index is about one order of mag-
nitude larger than that of thermal expansion (or contraction). The typical response
of the Bragg wavelength shift to strain is about 1 pm/µε in the near infrared. The
unit of strain is ε: 1 με corresponds to the elongation by 1 µm of a 1 m long fiber.
The typical temperature response is about 10 pm/◦ C. The Bragg sensor can be inter-
rogated by introducing a broadband light beam and monitoring the back-reflected
radiation with a spectrometer that includes an array of CMOS photodetectors, as
shown in Fig. 8.6. There are a number of situations where it is necessary to monitor
a parameter in many points along a length. Optical fibers can be installed over a
measurement path in a large structure, like a dam, a bridge, an aircraft to monitor the
spatial and temporal behavior of strain and temperature, with the fiber having several
Fig. 8.6 Fiber Bragg grating sensor

234 8 Applications
FBGs spaced at intervals along its length. If each grating has a different pitch, then
it becomes possible to monitor the environment along the whole fiber because each
reflection can be recognized by its unique wavelength.
Temperature sensing distributed over a fiber length of several kilometers can also
be performed by sending a short laser pulse along the sensing fiber and detecting the
backscattered Raman light with high temporal resolution. The power ratio between
anti-Stokes and Stokes light, expressed by (7.59), depends on the fiber temperature,
but it is independent of the laser power, the launch conditions and the composition
of the fiber (provided no fluorescence occurs). The only correction required is a
small allowance for differences in the fibre attenuation at the Stokes and antiStokes
wavelengths, when long lengths of fiber are used. By using a pulsed semiconductor
laser emitting at 1550 nm (the minimum loss wavelength for silica fibers), the Stokes
and anti-Stokes lines are found around 1650 and 1450 nm, respectively. The temper-
ature distribution over a few tens of kilometers can be derived from the measured
power ratio with a spatial resolution of a few meters and a temperature resolution of
a few degrees Kelvin. Stimulated Brillouin scattering may also be used for distrib-
uted temperature sensing. The sensing capability here arises from the dependence
of the local acoustic velocity in glass on both strain and temperature. As shown by
(7.65), the Brillouin frequency shift is inversely proportional to the acoustic veloc-
ity. The technique is more complicated than the case of the Raman sensor, because
the measurement of the Brillouin frequency shift requires a pulsed pump and a
counter-propagating continuous-wave probe. Positional information is obtained by
a time-domain analysis of the fiber output.
Electric current sensor. This type of sensor is based on the Faraday rotation effect,
described in Sect. 2.5.4, which states that the polarization of light waves is rotated
when propagating along (or opposite to) a magnetic field inside a magneto-optic
material. The sensor uses the propagation of a linearly polarized optical beam in a
bulk material or in an optical fiber exhibiting a non-zero Verdet constant. Compared
with the bulk devices, the Verdet constant of optical fibers is quite small, but their
optical path length can be increased to compensate for this, by winding the fiber with
a large number of turns around the current conducting element. The intensity of the
electric current i is obtained by measuring the rotation angle θ and using (2.153):
θ = C V μo N i, (8.8)
where N is the number of turns and C V is the Verdet constant of the fiber. As an
example, considering a cable that carries a current of 1 kA, choosing a fiber with
C V = 4 rad(T · m)−1 and putting N = 20, it is found that θ ≈ 5.6◦ . An impor-
tant feature of this approach is that the optical measurement does not interfere with
the electric configuration of the circuit, as would be the case for a standard electro-
magnetic sensor. This fiber sensor, which can be made temperature and vibration
insensitive, is especially well fit to measuring the high currents present in the high-
voltage substations of power distribution systems.
8.3 Laser Materials Processing 235
8.3 Laser Materials Processing
When an intense laser beam is even partially absorbed by a material, the absorbed
energy goes into heating. The local temperature may become high enough to liq-
uefy or vaporize the material. The time it takes to raise the temperature can be so
short that material is melted or removed before the heat can diffuse out from the
point of beam contact. It is then possible to perform materials processing opera-
tions, such as cutting, welding, drilling, and marking. Essentially, all materials are
processable, metals, plastics, glass, semiconductors, textiles, leather and cardboard.
The materials processing market is the second most important photonic market, after
the information and communication technology.
Lasers are very versatile sources for materials processing. The main advantages
of the laser as a working tool are the following:
• Since there is no physical contact with the processed material, the laser tool does
not wear out, and there is also a reduced probability of a contamination.
• There is a variety of available laser wavelengths, and thus it is always possible to
find the appropriate wavelength of operation for any type of material.
• The beam intensity can be easily modulated spatially and temporally. This allows
the integration of monitoring and control systems based on intelligent photonic
sensing techniques.
• Laser fabrication is environmentally attractive because it reduces energy consump-
tion and chemical waste.
The laser sources most utilized for materials processing are the following; the
diode-pumped neodymium laser and the ytterbium fiber laser, both emitting in the
near infrared around λ = 1 µm, the CO2 laser and the Er-YAG laser, both emitting in
the mid infrared (9–10 µm), and the excimer laser, which emits in the ultraviolet. In
some applications where intense green, blue, or ultraviolet light is needed, the emis-
sion of a near infrared laser is frequency-doubled or tripled, by using the nonlinear
processes described in Chap. 7.
Laser materials processing covers a large variety of applications, including the
possibility of machining materials that would otherwise be very difficult to process
with conventional tools. In the following some of the most important applications
are briefly described.
• Solid-state and fiber lasers operating on a single transverse mode produce a
Gaussian beam that can be focused to the smallest of spot sizes. The huge power
densities associated with the small spot produce excellent cuts in thin and thick
metals, and deep penetration for welding applications. Lasers are widely used for
the process of scribing, in which a pulsed laser is used to drill a sequence of blind
holes along designed fracture lines. Examples are the scribing of silicon solar cells
and of ceramic substrates for the electronics industry. Laser cutting processes are
extensively used in the industrial production of plastics, fabrics, and paper.
• Multi-mode lasers offer scalability of powers up to tens of kilowatts, while still
maintaining a very good beam quality. The wide range of applications with these
236 8 Applications
lasers include cutting of thick automotive parts, heat treating industrial compo-
nents, welding titanium panels for aerospace industry, deep penetration welding
for ship building industry, and clean cutting of thick stainless steels.
• Marking is the process through which an industrial product is identified with a
code, a series number, a date, a producer name, and so on. Q-switched solid-
state and fiber lasers offer excellent performance in laser marking applications.
The ability to focus the pulse power to small spot sizes enables high-speed, high-
quality marking on metals and almost any kind of surface.
• Additive manufacturing, also called three-dimensional printing, is a new process
of making a three-dimensional solid object of virtually any shape from a computer-
aided-design (CAD) model. In this process, successive layers of material are laid
down in different shapes, and a high-power laser (a CO2 or an ytterbium fiber
laser) is used to sinter or melt powder granules of a material to create fully dense
solid structures.
8.4 Biomedical Applications
The biomedical applications of photonics can be broadly divided into those based on
high-intensity processes, in which the laser beam melts or vaporizes the biological
material, and those based on low-intensity processes, in which the laser beam is a
diagnostic or manipulating tool. The interaction of light with biological materials
produces a variety of different effects. Thermal effects arise from the conversion of
the energy of absorbed light into heat through a combination of various non-radiative
processes. These do not show a strong wavelength dependence. For high-intensity
of illumination, the temperature at the laser spot on the biological tissue can become
large enough to cause melting or coagulation. Another effect, which can be induced
by ultraviolet light, is photoablation. Since the energy of an ultraviolet photon is larger
than the energy of some molecular bonds, cellular components can be decomposed,
and thus photofragmented species are released from the tissue. Thermal effects and
photoablation are the basis for laser surgery.
The area of biomedical applications of photonics, especially when the empha-
sis is on low-intensity processes, is now frequently called biophotonics. An effect
observed when molecular constituents of cells and tissues are excited to a vibrational-
electronic (vibronic) level by the absorption of a photon is fluorescent light emission.
Fluorescence is a process by which an excited molecule returns to the ground state in
two steps. Firstly, by quickly releasing in a non-radiative way its vibrational energy
and secondly by emitting a photon with an energy smaller than that of the absorbed
photon. Fluorescence has an important role in biophotonics, as a means to detect the
presence of specific molecular groups, or to monitor the interactions of a molecule
with its environment. Also, as will be discussed below, the main optical bioimaging
methods are based on fluorescence.
In some cases, instead of emitting a photon, the molecule excited to an upper
energy level uses its excitation energy to start a chemical reaction. Such a
8.4 Biomedical Applications 237
photochemical reaction is the basis, for example, of the procedure called photo-
dynamic therapy, in which light is used to activate a molecule (called photosensi-
tizer) that then triggers a chain of chemical reactions leading to the destruction of a
cancerous or diseased cell.
The use of photonics for diagnostics is important for biological research, and
may also lead to the development of easy-to-access, minimally invasive, low-cost
screening methods. One aim is to provide a means of early detection of diseases. As
an example, “in vivo” studies utilize optical-fiber endoscopy for the examination of
tissues inside the human body in order to spot the presence of tumors. Another topic
of biophotonic interest is trapping and manipulation of biological cells by using laser
Gaussian beams.
The following sections briefly treat the applications of photonics to ophtalmology
and bioimaging.
8.4.1 Ophtalmology
The first biomedical application of lasers, dating back to the early 1960s, was in
ophtalmology, concerning the cure of retinal detachment. This kind of treatment is
still widely utilized. Another very popular application is the correction of nearsight-
edness, or myopia.
Retinal detachment. The retina is a light-sensitive layer of tissue, positioned on the
inner surface of the eye. Light absorbed by the retina initiates a cascade of chemical
and electrical events that ultimately transmits electrical signals to the brain through
the optic nerve. A retinal detachment is a separation of the retina from the underlying
tissue within the eye. Retinal detachment is one of the leading causes of blindness.
Green laser light can travel through the eye lens and the vitreous humor without
being absorbed. The laser beam can be focused onto the retina, selectively treating
the desired area while leaving the surrounding tissues untouched. This approach
exploits the capability of the laser to perform the welding operation in a position
that is not accessible to a mechanical tool. Retinal detachment is treated by using an
argon laser (λ = 515 nm) or a frequency-doubled neodymium laser (λ = 532 nm).
These wavelengths are close to the sensitivity maximum of the eye, and thus are very
effectively absorbed by the retina.
Correction of myopia. Several visual defects, such as myopia or astigmatism, are
due to the fact that the eye lens does not correctly focus images at the retina position.
A cure for myopia involves using a laser to photo-ablate small sections of the cornea,
which is a thin layer of tissue covering the front of the eye, to correct its radius of
curvature so that light is less tightly focused through the eye lens and onto the retina.
The operation is performed by using an excimer laser, emitting ultraviolet light. In
this type of treatment, three main procedures are followed, known as photorefrac-
tive keratectomy (PRK), laser epithelial keratomileusis (LASEK), and laser in situ
keratectomy (LASIK).
238 8 Applications
Artificial retina. A revolutionary application that is presently at an early stage of

development is the realization of an artificial retina. This involves the effort to develop
an implantable microelectronic retinal prosthesis that restores useful vision to people
blinded by a retinal disease, such as diabetes, that makes inactive the light-sensing
cells in the retina. The device consists of a multi-electrode array in a one-inch pack-
age that allows the implanted electronics to wirelessly communicate with a camera
mounted on a pair of glasses. It is powered by a battery pack worn on a belt. The
device is intended to bypass the damaged eye structure: the array provides the elec-
trical signals which are transmitted to the brain through the optic nerve. This is a
technology that combines microelectronics, neurobiology and photonics.
8.4.2 Bioimaging
Some recently developed optical methods can be applied to the observation of a wide
range of biological objects at high resolution.
Fluorescence. In addition to traditional optical microscopy, bioimaging methods
based on fluorescence have become increasingly important. A number of cellular con-
stituents fluoresce when excited by visible light. This fluorescence is called endoge-
nous fluorescence. When no endogenous fluorophores are present, as it happens, for
instance, with DNA, labeling of biological structures with exogenous fluorophores
is needed for bioimaging. A very important fluorophore is the green fluorescent
protein (GFP) which is not naturally present in most cells, but can be generated in
situ by a specific biochemical method. The GFP has a major excitation peak at a
wavelength of 395 nm and a minor one at 475 nm. Its emission peak is at 509 nm,
which is in the lower green portion of the visible spectrum. By using different flu-
orophores specifically targeting different parts of the cell and by optically filtering
the fluorescent emission, it is possible to visualize separately the different parts of
a specimen. By using a scanning optical microscope full three-dimensional images
can be reconstructed.
High-resolution fluorescence. Further increased spatial resolution can be obtained
by two recently developed nonlinear optical methods, known as two-photon confocal
microscopy and stimulated emission depletion (STED) microscopy. In two-photon
microscopy a fluorophore is excited by simultaneously absorbing two photons, each
one having half of the energy required to bring the molecule to its excited state. Two
photon absorption, mentioned in Sect. 7.4, is a nonlinear process with a small cross-
section. Since the transition probability for simultaneous two-photon absorption is
proportional to the square of the instantaneous light intensity, as shown by (7.45),
a train of mode-locked ultrashort laser pulses with a large peak power (≈50 kW)
and a low average power is utilized. The average power must be kept very low to
minimize any thermal damage of the biological specimen. In two-photon microscopy
the excitation beam is tightly focused, and thus the region outside the focus has little
chance to be excited. This eliminates out-of-focus fluorescence, so that the excitation
8.4 Biomedical Applications 239
region is strongly localized. The whole sample can be sectioned by scanning the
focal point. Two-photon microscopy can use a near-infrared-wavelength laser as
the excitation source and produce fluorescence in the visible range. Using infrared
photons instead of ultraviolet or blue photons gives the advantages of less photo-
damage, and better penetration into the sample. The spatial resolution can be smaller
than the size of the focal region of the exciting beam. STED functions by deactivating
the fluorescence of specific regions of the sample while leaving a center focal spot
active to emit fluorescence. This is achieved by using a double-laser design, with a
doughnut shaped depletion beam. By the STED method a spatial resolution of the
order of 10 nm can be attained.
Endoscopy. An endoscope is a medical imaging device inserted into the patient’s
body to get a clear view of the inside. Endoscopy is a great medical breakthrough
because it helps to reduce diagnostic surgeries. In a fiber-optic endoscope light is
guided to the area under investigation by non-coherent fibre optic bundles (bundles
where the optical fibres are not lined up at both ends). The image is transmitted back
by a coherent fibre optic bundle (a bundle where the optical fibres are lined up at
both ends of the fibre so that an image can be transmitted).
Optical coherence tomography. Optical coherence tomography (OCT) is a reflec-
tion imaging technique similar to ultrasound imaging. A light beam (usually in the
near infrared) is back-scattered from the biological tissue under study, in order to pro-
duce an image. OCT is based on an interferometric scheme similar to that illustrated
in Fig. 8.3. In general terms, the phase relation between successive wavefronts of the
optical signal is maintained within a distance known as “coherence length”, lc , which
is given by the velocity of light divided by the bandwidth Δνs of the source. The
interferometer produces a well-defined interference pattern only if the path-length
difference between reference beam and back-scattered beam is within the coherence
length. In the case of OCT the source bandwidth is very broad, so that lc ≈ 1 µm.
By scanning the path-length of the reference beam, an interference signal appears
only in those positions corresponding to path differences within lc . The interference
pattern contains the information about the refractive index variation of the sample,
which thus non-invasively provides the optical image.
8.5 Liquid Crystal Displays
Liquid crystal displays (LCD) are found in a variety of electronic devices, including
cell phones, tablets, laptop computers, computer monitors, TV screens and digital
watches. As a consumer technology, LCD has been in widespread use since the early
1970s when it first appeared in digital watches.
Images in an electronic display are made up of pixels, which are the smallest
distinct element in an image. In color displays the pixel is made of three subpixels,
red, green and blue, which can form any color, when added together at different levels
of intensities. Each individual pixel (or subpixel) is made-up of a light source and an
240 8 Applications
electronic valve which can turn the light on, off, or to some intermediate level. The
light valve consists of a thin liquid crystal film, sandwiched between two crossed
polarizers. The orientation of the liquid crystal molecules is controlled within each
pixel by applying an electric field. To this end, transparent electrodes are patterned
on the glass plates of the liquid crystal cell. The electrodes are made of a thin layer
of indium tin oxide, a semiconductor that has a good electric conductivity and, at
the same time, is transparent to visible light. In the case of the twisted-nematic cell
described in Chap. 4, the pixel is bright in absence of applied field, and becomes
opaque when the liquid crystal molecules are forced to align parallel to the applied
field.
In the case of simple displays, like those of digital watches or pocket calculators,
each pixel is driven separately. In the case of television screens, a matrix connection
is built: on a side of the liquid crystal panel all horizontal lines are interconnected,
whereas at the opposite side the vertical lines are interconnected, in such a way to
select a specific pixel by crossing an horizontal line with a vertical line.
At present there are two common methods of back-lighting in LCD flat panels:
cold-cathode fluorescent lamp and LED. Fluorescent lamps used to be the most
widespread method of back-lighting for LCD TVs, and consists of a series of tubes
laid horizontally down the screen. LED backlighting is now very common and has
been in use in TVs since 2004.
It is a general property of displays based on polarized light that the viewing angle
is limited. Viewing angle relates to how well you can see a display image when
looked at from the side of the display. The lateral visibility is greatly improved by
using the so-called in-plane switching (IPS), instead of the twisted nematic approach.
In IPS the molecular alignment and the orienting electric field are in the plane of the
panel.
8.6 LED Applications
LEDs are currently used in a large variety of applications, such as displays, street
lights, automotive lighting, ambient lighting, light sources for robotic vision, remote
control units, and so on.
LEDs have an important role in displays. They are used for three-color back-
lighting liquid crystals displays (see Sect. 8.5), and also for directly producing images
in the giant displays at sports stadium, for instance. In those mega-screens, the images
must be very bright, in order to be clearly seen even in presence of full sunlight. A
pixel inside a mega-screen is about 1 cm in size and includes tens of LEDs of the three
fundamental colors. The brightness and color of each pixel are fixed by controlling
the bias voltage of all the LEDs making-up the pixel. High brightness LEDs or
solid-state lasers are also used for projection displays.
Displays made of organic light-emitting diodes, also known as OLEDs, are now
finding more and more uses. Whereas OLED screens were initially limited to the
displays of certain cell phones and car radios, they are now entering into the TV
8.6 LED Applications 241
market as well. Compared to conventional liquid crystal displays (LCDs), OLED

displays can offer enhanced contrast and lower energy consumption. The current
LCD architecture is based on rigid sheets of glass and is therefore neither flexible
nor lightweight enough for newly developing applications, such as roll-up screens
in movie-theater format or large-area light-emitting wallpaper. Printed OLEDs are
components where the functional layer is applied to a substrate by means of a printing
process. Printing processes are cheaper and can operate with larger formats than the
processes typically used in the semiconductor industry.
LEDs are playing an increasing role in general illumination. Lighting by solid-
state light sources, concisely called “solid-state lighting” (SSL) has several attractive
properties. First of all, LEDs are more efficient than incandescent and fluorescent
lamps, and so visible light can be produced with reduced heat generation or parasitic
energy losses. Long-life LEDs can provide 50,000 h or more of operation, which can
reduce maintenance costs. In comparison, an incandescent light bulb lasts approxi-
mately 1,000 h. The best commercial white LED lighting systems provide twice the
luminous efficacy (lumens per watt) of fluorescent lighting. Considering that general
illumination uses about 20 % of the available worldwide electrical energy, solid-state
lighting could offer a significant contribution to energy saving. Colored LEDs are
especially advantageous for colored lighting applications because “lossy” filtering
of white light is not needed. Another interesting aspect of SSL technology is that the
spectral, spatial and temporal distribution of ambient lighting can be readily adjusted,
offering new interesting opportunities. LED systems are intrinsically safe, because
they use low voltage and are generally cool to the touch.
A last point to mention concerns ultraviolet LEDs based on AlGaN. These sources
have the potential to replace mercury lamps inside the sterilization systems, which
are used in various industries, such as the food industry, and in biomedicine.
8.7 Photovoltaic Cells
A photovoltaic cell is an electrical device that converts the energy of light directly
into electricity by the photovoltaic effect. The working principle is exactly the same
as that described for photodiodes in Sect. 5.5.2. A photon absorbed in the depletion
layer of a p-n junction generates an electron-hole pair. The migration of the photo-
generated carriers produces an open-circuit voltage difference across the junction, or
a closed-circuit electric current without being attached to any external voltage source.
The energy of solar light can therefore be converted directly into electrical energy.
The electric signal obtained is continuous, but can be converted into an alternating
signal by an inverter. Solar cells were first tested on satellites that were launched in
the late 1950s. The tests were very successful, but the cost of those early devices
made their utilization for civilian electrical energy generation unpractical.
The motivations for a massive development of electricity generation based on
solar energy conversion are quite obvious: (i) the worldwide energy consumption is
continuously growing; (ii) at present about 70 % of electrical energy is produced by
242 8 Applications
burning fossil fuels, and this should be lowered because of environmental problems
and also, on a long-term basis, because hydrocarbons are a non-renewable source
of energy; (iii) solar energy is abundant and is available everywhere on the Earth’s
surface, so that, in principle, each country (even each community) could become
self-sufficient in the production of photovoltaic electrical energy.
The first issue to be discussed in efficient conversion is the selection of the semi-
conductor material, bearing in mind that the incident photon is absorbed only if its
energy hν is larger than the bandgap energy E g . Silicon, exhibiting E g = 1.11 eV
(see Table 5.1), can absorb all visible wavelengths. However, if a visible photon is
absorbed by silicon, the fraction of photon energy exceeding 1.11 eV is lost to elec-
tric conversion, it goes into vibrational energy and is finally dissipated as thermal
energy.
Defining the conversion efficiency η as the ratio between the generated electric
power and the incident optical power, and considering that visible (solar) photons
cover the energy range 1.8–3.1 eV, the upper limit of η for a silicon cell is 30 %. A
higher efficiency can be obtained using GaAs cells. By a stack of photovoltaic cells
made with different semiconductors of decreasing E g in such a way to cover the
whole solar spectrum, the upper limit to η would become 68 %, but at the price of a
large increase in cost.
Currently, solar cells are produced from monocrystalline silicon, polycrystalline
silicon, and gallium arsenide (i.e., III–V solar cells). Although existing monocrys-
talline or polycrystalline silicon solar cells are economical, their photoelectric con-
version efficiencies only range between 15 and 20 %. By comparison, III–V solar
cells possess a conversion efficiency ranging between 40 and 50 %, but are more
expensive. Typically the semiconductor thickness required to absorb 90 % of the
incident power is about 30 µm. Second generation solar cells use thin-film technol-
ogy, which gives a lower efficiency, but is considerably cheaper. Thin-film solar cells
can employ not only crystalline silicon, but also amorphous silicon.
Another technology in use is that of a high concentration photovoltaic module,
in which the size of the photosensitive area is reduced by concentrating the incident
solar power with a lens. Systems containing large-area concentration lenses and small
III–V solar cells have been successfully tested.
Similarly to the case of light-emitting diodes, where organic compounds can sub-
stitute inorganic semiconductors, photovoltaic cells made of organic photovoltaics
based on semiconducting polymers can be prepared. This may represent a poten-
tially inexpensive means of generating electricity directly from sunlight. However,
at present, the conversion efficiency is lower than that of semiconductor solar cells.
The main attraction of organic photovoltaics is that they can be printed at high speeds
onto large areas of thin, flexible plastic substrate using roll-to-roll processing tech-
niques. This creates the prospect of being able to coat every roof and other suitable
building surface with photovoltaic materials at extremely low cost. Research is ongo-
ing to find organic compounds with a broad absorption spectrum and a sufficiently
long chemical stability.
Appendix A
System of Units and Relevant Physical
Constants
The international system of units (SI) is used in this book. The three fundamental
units are meter (m), second (s) and kilogram (kg). A prefix can be added to each unit
to change its magnitude by multiples of 10. A list of prefixes is given in Table A.1.
In engineering it is common to make use of decibel units, abbreviated as dB. Any
ratio R can be converted into decibels through the general definition
Ro (in dB) = 10 log10 R. (A.1)
For example, an amplifier gain of G = 104 corresponds to G o = 40 dB. As

R = 1 corresponds to 0 dB, ratios smaller than 1 are negative on the decibel scale.
For instance, if R = 0.8, which corresponds to a loss of 20 %, the ratio in decibel is
Ro = −1 dB.
Table A.1 Prefixes of units Prefix Power of 10 Symbol

zepto 10−21 z
atto 10−18 a
femto 10−15 f
pico 10−12 p
nano 10−9 n
micro 10−6 µ
milli 10−3 m
kilo 103 k
mega 106 M
giga 109 G
tera 1012 T
peta 1015 P
exa 1018 E
zetta 1021 Z

DOI 10.1007/978-3-319-20627-1
244 Appendix A: System of Units and Relevant Physical Constants
Table A.2 Relevant physical constants

Speed of light in vacuum c 2.9979 × 108 m/s
Vacuum permittivity εo 8.8542 × 10−12 F/m
Vacuum permeability μo 1.2566 × 10−6 H/m
Electric charge of the electron e 1.6022 × 10−19 C
Mass of the electron me 9.1094 × 10−31 kg
Boltzmann constant kB 1.3807 × 10−23 J/K
Planck constant h 6.6261 × 10−34 J·s
Table A.2 lists the values of the physical constants that appear more frequently
throughout the book.
Appendix B
List of Acronyms
Several acronyms are used throughout the book. Each acronym is defined the first
time it appears in a chapter. In any case, to offer a further help to the reader, acronyms
are listed here in alphabetical order (Table B.1).

DOI 10.1007/978-3-319-20627-1
246 Appendix B: List of Acronyms
Table B.1 List of acronyms APD Avalanche photodiode

CCD Charge-coupled device
CD Compact disc
CW Continuous wave
DFB Distributed feedback
DVD Digital video disc
EDFA Erbium-doped fiber amplifier
FBG Fiber Bragg grating
KDP Potassium di-hydrogenated phosphate
KTP Potassium titanyl phosphate
LBO Lithium triborate
LED Light-emitting diode
MQW Multiple quantum well
NA Numerical aperture
OLED Organic light-emitting diode
PMD Polarization-mode dispersion
QCL Quantum cascade laser
QPM Quasi phase matching
SH Second-harmonic
SHG Second-harmonic generation
SMR Single-mode reduced
SOA Semiconductor optical amplifier
SOI Silicon on insulator
SPM Self phase modulation
SRS Stimulated Raman scattering
VCSEL Vertical-cavity surface-emitting laser
WDM Wavelength division multiplexing
YAG Yttriun aluminum garnet
Reading List
General
B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley Interscience,
2nd ed. 2007.
R. S. Quimby, Photonics and Lasers: An Introduction, Wiley Interscience, 2006.
J-M. Liu, Photonic Devices, Cambridge University Press, 2005.
J. Wilson and J. Hawkes, Optoelectronics, Prentice Hall, 3rd ed. 1998.
Concerning Chaps. 1 and 2
E. Hecht, Optics, Addison-Wesley, 4th ed. 2002.
R. Guenther, Modern Optics, Wiley, 1990.
M. V. Klein and T. E. Furtak, Optics, Wiley, 2nd ed. 1986.
J. W. Goodman, Introduction to Fourier Optics, Roberts, 3rd ed. 2005.
Concerning Chaps. 3–5
A. Yariv, Quantum Electronics, Wiley, 1989.
O. Svelto, Principles of Lasers, Springer Verlag, 5th ed. 2010.
W. Koechner and M. Bass, Solid-State Lasers: A Graduate Text, Springer Verlag,
2003.
L. A. Coldren, S. W. Corzine, and M. L. Mashanovitch, Diode Lasers and Photonic
Integrated Circuits, Wiley, 2012.
S. Donati, Photodetectors, Prentice Hall, 2000.
Concerning Chaps. 6–8
J. Hecht, Understanding Fiber Optics, Prentice Hall, 5th ed. 2005.
R. W. Boyd, Nonlinear Optics, Academic Press, 2nd ed. 2003.
G. I. Stegeman and R. A. Stegeman, Nonlinear Optics, Wiley, 2012.
G. P. Agrawal, Nonlinear Fiber Optics, Academic Press, 5th ed. 2013.
G. P. Agrawal, Fiber-Optic Communication Systems, Wiley Interscience, 2010.
S. Donati, Electro-optical Instrumentation, Prentice-Hall, Upper Saddle River,
2004.
W. M. Steen and J. Mazumder, Laser Material Processing, Springer, 4th ed. 2010.
M. H. Niemz, Laser-Tissue Interactions, Springer, 4th ed. 2011.
DOI 10.1007/978-3-319-20627-1
Index
A Circular polarization, 5, 69
Absorption cross-section, 88 Corner cube, 41, 227
Acousto-optic deflection, 142
Acousto-optic effect, 137
Acousto-optic filter, 142 D
ADP, 122 3D printing, 236
Airy function, 20, 47 DFB laser, 153, 222
Amplitude modulation, 129, 130, 168, 222 Dielectric constant, 3
acousto-optic, 141 Difference-frequency generation, 205
electro-optic, 128 Diffraction gratings, 55
Anti-reflection coating, 36, 159 blazed, 58
Argon laser, 237 Diffraction theory
Artificial retina, 238 Fraunhofer approximation, 17, 52
Attenuation coefficient, 28, 176 Fresnel approximation, 15
Autocorrelation function, 204 Dirac delta function, 21, 91, 113
Avalanche breakdown, 167 Dispersive prism
Avalanche photodiode, 167, 223, 227 resolving power, 55
Distributed feedback laser, 153
Divergence angle, 13, 103, 115, 227
B Doppler effect, 138, 230
Beam splitter, 40, 126 DVD, 225
Biophotonics, 236
Birefringence, 66, 183, 201
Birefringence phase matching, 201 E
Bit error rate, 223 EDFA, 187, 223
Blackbody, 82, 91 Electric permittivity, 27, 63
Blu-ray disc, 226 Electric susceptibility, 26, 121, 199
Bolometer, 162 Electro-absorption modulators, 167
Bragg wavelength, 186, 233 Electro-optic coefficients, 122
Brewster angle, 31, 32 Electro-optic effect, 121
Brillouin scattering, 218, 234 Electro-optic modulators
half-wave voltage, 128
KDP, 125, 132
C lithium niobate, 130
Calcite, 66, 67 response time, 129
Cat’s eye, 42 Electro-optic switch, 128
Charge-coupled device (CCD), 167, 229 liquid crystal, 136
DOI 10.1007/978-3-319-20627-1
250 Index
Electro-optic tensor, 122 Kerr effect, 121, 133

Evanescent wave, 43, 175 KTP, 200
Excimer laser, 186, 237
F L
Fabry-Perot interferometer, 59, 98 Laser
Faraday effect, 73, 233, 234 carbon dioxide, 108, 110, 235
Fiber Bragg grating, 186, 191, 217, 233 divergence, 103
chirped, 187, 223 efficiency, 101
Fiber lasers, 190 Er-YAG, 235
ytterbium, 235 excimer, 108, 235
Fiber Raman laser, 217 Fabry-Perot cavity, 96, 99, 101, 150,
Fluorescence, 236, 238 191
Focal length, 46, 211 four-level, 106
Franz-Keldish effect, 167 gas, 107
Fraunhofer diffraction, 17, 56 Helium-Neon, 107
Fresnel diffraction, 15, 52 linewidth, 116
Fresnel number, 104 longitudinal modes, 98
loop gain, 97
neodymium, 106, 107, 110, 198, 235
G
ring cavity, 96, 191
Gaussian spherical waves, 9, 48, 102
complex radius of curvature, 10 ruby, 101, 105
focalization, 49 solid-state, 105
radius of curvature of the wavefront, 11 three-level, 105
Germanium, 165 threshold condition, 97
Group velocity, 180 titanium sapphire, 106, 113
transverse modes, 103
unstable cavity, 103, 212
H Laser Doppler velocimetry, 231
Half-wave plate, 68 Laser gyroscope, 232
Helmholtz equation, 6, 76, 174 Laser pulses
Hermite-Gauss spherical waves, 13, 103 transform-limited, 118
Laser velocimetry, 230
Lead sulfide, 165
I Lens, 45, 51, 52
Index ellipsoid, 64, 123 converging, 46
Index of refraction, 27 diverging, 46
anisotropic materials, 64, 122 Light-emitting diode (LED), 159, 240, 241
dielectric materials, 36 AlGaN, 241
metals, 34
GaAsP, 159, 160
Indium antimonide, 165
InGaN, 160
In-plane switching, 240
white, 161
Iridescence, 40
Light scattering, 176, 214, 218
Limit angle, 32, 171, 173
J Liquid crystals, 135, 239
Jones matrices, 68, 132 Lithium niobate, 122, 124, 199, 200, 206
Lithium tantalate, 122
Lithium triborate (LBO), 200
K Littrow configuration, 59
KDP, 122, 123 Loop gain, 85, 98
Kerr constant, 133 Lorentzian function, 88
Index 251
M polarization-maintaining, 183
Mach-Zehnder interferometer, 126, 228, 233 single mode, 175, 182
Magnesium fluoride, 37, 39 SMR, 223
Magneto-optic effect, 73, 225 step-index, 173, 174
Maser, 83 ZBLAN, 190
Matrix optics, 50 zero-dispersion wavelength, 182
Maxwell equations, 3, 25 Optical fiber sensors, 233
Michelson interferometer, 233 Optical isolator, 74, 188
Mode-locking, 111, 133, 142, 158, 191 Optical Kerr effect, 134, 209
Kerr-lens, 114, 212 Optical radar, 158, 227
Multilayer mirror, 38, 154, 186 Optical soliton, 213
Multiple quantum well, 155, 168 Optical waveguides, 76, 130, 151, 171
propagation constant, 77
silicon, 226
N single-mode, 78
Neodymium laser, 107, 156, 198, 237 Oscillator, 84
Nonlinear optical materials, 200 parametric, 206
Numerical aperture
lens, 47, 225
optical fiber, 173 P
Parametric interactions, 205
Paraxial approximation, 8, 17, 46, 103
O Paraxial equation, 7, 10
OLED, 159, 240 Phase gratings, 57, 137
Optical activity, 71 Phase matching, 197, 205
Optical amplifier Phase modulation, 123, 222
erbium-doped fiber, 187, 223 Phonon, 138, 147
parametric, 205 Photoablation, 236
Raman fiber, 223 Photodetectors
semiconductor, 159, 223 germanium, 165
Optical coherence tomography, 239 photoconductive, 165
Optical communications, 177, 222 photoelectric, 163
Optical dispersion, 27, 178, 213 silicon, 165
Optical fiber Photodiodes, 165
attenuation coefficient, 176 p-i-n, 166, 223
cladding pumped, 189, 191 Photoelastic effect, 137
couplers, 185 Photoelastic tensor, 140
current sensor, 234 Photoelectric cell, 163
DIS-CO, 223 Photomultiplier tube, 164
dispersion, 178 Photon, 3, 138, 146, 197, 206
dispersion length, 180 Photonic crystals, 40
DS, 182, 223 Photovoltaic effect, 166, 241
erbium-doped amplifier, 187 Piezoelectric effect, 123
graded-index, 173 Pixel, 167, 222, 224, 239, 240
group velocity, 180 Plasma frequency, 34
intermodal dispersion, 178 Plastic optical fibers, 185
microstructured, 184 Pockels effect, 121, 124, 199, 207, 233
multimode, 182 Polarizers
normalized frequency, 174 Glan–Thompson, 67
numerical aperture, 173 Polaroid, 67
photonic-crystal, 184 Polaroid, 67
plastic, 185 Population inversion, 89
PMD, 183 Potassium niobate, 122
252 Index
Potassium titanyl phosphate, 200 GaInAs, 156

Power apectrum, 54 GaInAsP, 156
Poynting vector, 4, 65, 201 homojunction, 150, 151
Prism InGaN, 156
dispersive, 54 internal efficiency, 157
total-internal-reflection, 41 longitudinal modes, 158
modulation, 158
multiple quantum well, 155
Q quantum cascade, 156
Q-switching, 109 stripe geometry, 153
active, 133, 142 vertical-cavity surface-emitting, 154
passive, 111 Semiconductor optical amplifier (SOA),
Quantum cascade laser, 156 159, 223
Quarter-wave plate, 68, 132 Semiconductors
Quartz, 66, 72 direct-bandgap, 147
Quasi-phase-matching, 202 II-VI, 156
indirect-bandgap, 146
n-type, 148
R p-type, 148
Raman gain coefficient, 216 Silicon, 242
Raman scattering, 214, 234 Silicon photonics, 226
Rate equations, 93, 100 Snell’s law, 29, 31, 38, 66, 173
Rayleigh length, 11, 49, 225 Solid-state lighting, 241
Rayleigh scattering, 177, 214 Spatial filter, 53
Rectangle function, 113 Spatial frequency, 18, 53
Reflectance, 31 Spherical mirror, 46, 52, 102
metallic surface, 33 Spherical wave, 8
Reflection grating, 57 focalization, 47
Resolving power, 54, 61 Spontaneous emission, 96, 100
RGB color code, 222 Spontaneous emission coefficient, 92
Rotation matrix, 70 Stimulated Brillouin scattering, 218, 234
Rotatory power, 73 Stimulated emission, 83, 92
Ruby laser, 83 Stimulated emission cross-section, 89
Stimulated Raman scattering, 190, 215
Sum-frequency generation, 206
S
Sagnac effect, 232
Second-harmonic generation, 196 T
conversion efficiency, 196, 203 TGG, 74
spectral acceptance, 203 Third-harmonic generation, 208
Seebeck effect, 162 Three-level system, 93
Self-focusing, 211 Transmittance, 31
focal length, 211 Two-photon absorption, 210
Self-phase-modulation, 212
Semiconductor laser, 83, 188
AlGaAs, 107 U
DFB, 222 Uniaxial medium, 64
distributed feedback, 153
double-heterojunction, 151
edge emitting, 154 V
efficiency, 157 VCSEL laser, 154
GaAlAs, 158 Verdet constant, 73, 234
Index 253
W Z
Walk-off angle, 65, 201 ZBLAN glasses, 177
Wave equation, 3, 26, 195 Zinc sulfide, 39
Wavelength division multiplexing, 224
WOLED, 162

Photonics, A Short Course - Degiorgio

Uploaded by

Copyright:

Available Formats

Photonics, A Short Course - Degiorgio

Uploaded by

Document Information

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

Photonics, A Short Course - Degiorgio

Uploaded by

Copyright:

Available Formats

What is the purpose of the Undergraduate Lecture Notes in Physics (ULNP) series?

What is the purpose of the Undergraduate Lecture Notes in Physics (ULNP) series?

What are some of the topics covered in the document text?

What are some of the topics covered in the document text?

Undergraduate Lecture Notes in Physics

ULNP titles must provide at least one of the following:

More information about this series at http://www.springer.com/series/8917

ISSN 2192-4791 ISSN 2192-4805 (electronic)

Library of Congress Control Number: 2015947401

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

Pavia, Italy Vittorio Degiorgio

Preface to the First Edition

properties of materials is adequate. The description of optical ampliﬁcation pro-

Pavia, Italy Vittorio Degiorgio

2 Optical Components and Methods. . . . . . . . . . . . . . . . . . . . . . . . . 25

2.4.3 Reflection Grating . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5 Semiconductor Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

6 Optical Fibers . . . . . . . . . . . . . . . ............... . . . . . . . . . 171

7 Nonlinear Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

8.2 Optical Metrology and Sensing . . . . . . . . . . . . . . . . . . . . . . . . 227

Appendix A: System of Units and Relevant Physical Constants . . . . . . 243

Appendix B: List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Reading List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

Abstract Light is an electromagnetic phenomenon that is described, in common

1.1 Spectrum of Electromagnetic Waves

© Springer International Publishing Switzerland 2016 1

Table 1.1 Spectrum of electromagnetic waves

Table 1.2 Visible

1.2 Electromagnetic Waves in Vacuum

The vacuum propagation of electromagnetic waves is described by Maxwell’s

where c = (εo μo )−1/2 is the propagation velocity of the electromagnetic wave. It

E = Eo cos (ωt − k · r + φ) = Eo Re{exp[−i(ωt − k · r + φ])}, (1.9)

where k is the propagation vector, with modulus k = 2π/λ = ω/c, and φ is a

1.3 Polarization of Light

The direction of polarization of an electromagnetic wave is the direction of the vector

E y = E yo cos(ωt − kz + φ), (1.14)

where φ is the phase shift between the two components. Putting X = E x /E xo =

By squaring both sides of (1.15) one obtains:

X 2 + Y 2 − 2X Y cos φ = sin2 φ, (1.16)

A generic polarization state can also be described as a superposition of two circular

1.4 Paraxial Approximation

Representing the field of the monochromatic wave as: E(r, t) = E(r)exp(−iωt),

E(x, y, z) = U (x, y, z)exp(ikz), (1.19)

the Helmholtz equation becomes:

It is now assumed that U is a slowly varying function of z, where slowly varying

A consequence of (1.21) is that the second derivative of U with respect to z is

Fig. 1.2 Paraxial approximation

In order to better clarify the meaning of the paraxial approximation, it is useful to

U (x, y, z) = Uo {1 + exp[ik(cos α − 1)z][exp(ikx sin α) + exp(−ikx sin α)]}.

The first derivative with respect to z is:

1.4.1 Spherical Waves

The origin of the spherical wave is at point ro having coordinates xo , yo , z o .

The paraxial approximation is applied to (1.24) by assuming that the fraction

To use (1.25) instead of (1.24) is analogous to approximating the spherical wave-

1.4.2 Gaussian Spherical Waves

By substituting (1.27) inside (1.22), one obtains:

By using the initial conditions q(z o ) = qo , A(z o ) = Ao , the following solutions

It is therefore demonstrated that the Gaussian spherical wave is a solution of the

Note that R + T = 1, whereas the sum |ρ|2 + |τ |2 is, in general, = 1.