Principles of Photonics

Principles of Photonics
With this self-contained and comprehensive text, students will gain a detailed understanding of
the fundamental concepts and major principles of photonics. Assuming only a basic background in optics, readers are guided through key topics such as the nature of optical elds, the
properties of optical materials, and the principles of major photonic functions regarding the
generation, propagation, coupling, interference, amplication, modulation, and detection of
optical waves or signals. Numerous examples and problems are provided throughout to enhance
understanding, and a solutions manual containing detailed solutions and explanations is
available online for instructors.
This is the ideal resource for electrical engineering and physics undergraduates taking introductory, single-semester or single-quarter courses in photonics, providing them with the
knowledge and skills needed to progress to more advanced courses on photonic devices,
systems, and applications.
Jia-Ming Liu is Distinguished Professor of Electrical Engineering and Associate Dean for
Academic Personnel of the Henry Samueli School of Engineering and Applied Science at the
University of California, Los Angeles. Professor Liu has published over 250 scientic papers
and holds 12 US patents, and is the author of Photonic Devices (Cambridge, 2005). He is a
fellow of the Optical Society of America, the American Physical Society, the IEEE, and the
Guggenheim Foundation.
Downloaded from Cambridge Books Online by IP 131.111.164.128 on Sat Aug 20 20:12:45 BST 2016.
http://ebooks.cambridge.org/ebook.jsf?bid=CBO9781316687109
Cambridge Books Online Cambridge University Press, 2016
ENDORSEMENTS FOR LIU, PRINCIPLES OF PHOTONICS
With much thoughtfulness and a rigorous approach, Prof. Jia-Ming Liu has put
together an excellent textbook to introduce students to the principles of photonics. This
book covers a comprehensive list of subjects that allow students to learn the
fundamental properties of light as well as key phenomena and functions in photonics.
Compared to other textbooks in classical optics, this book places the necessary
emphasis on photonics for readers who want to learn about this eld. Compared to other
textbooks introducing photonics, this book is carefully and well written, with ample
examples, illustrations, and well-designed homework problems. Instructors will nd
this book very helpful in teaching the subjects, and students will nd themselves
gaining solid understanding of the materials by reading and working through the book.
Lih Lin, University of Washington
For a long while the photonics community has been waiting for a new textbook which
is informative, comprehensive, and also contains practical examples for students; in
other words, one which describes fundamental concepts and provides working
principles in optics. Professor Jia-Ming Lius book, Principles of Photonics, serves very
well for these purposes it covers optical phenomena and optical properties of
materials, as well as the basic principles behind light emitting, modulation,
amplication and detection devices that are commonly used nowadays in
communications, displays, and sensing. A distinguishing feature of this book is its
seamless use of additional space to ensure that each concept is sufciently explained
in words, coupled with mathematics, simple yet illustrative gures, and/or examples.
Each chapter ends with questions/problems followed by key references, making it very
self-contained and very easy to follow.
Paul Yu, University of California, San Diego
A pedagogical tour-de-force. Professor Liu covers the principles of photonics with

extreme attention to notation, completeness of derivations, and clear examples matched
to the concepts being taught. This is a book one can really learn from.
Jeffrey Tsao, Sandia National Lab
Principles of
Photonics
JIA-MING LIU
University of California
University Printing House, Cambridge CB2 8BS, United Kingdom

Cambridge University Press is part of the University of Cambridge.
It furthers the Universitys mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9781107164284
Jia-Ming Liu 2016
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2016
Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall
A catalog record for this publication is available from the British Library
Library of Congress Cataloging-in-Publication data
Names: Liu, Jia-Ming, 1953- author.
Title: Principles of photonics / Jia-Ming Liu.
Description: Cambridge, United Kingdom : Cambridge University Press, [2016] | Includes bibliographical
references and index.
Identiers: LCCN 2016011758 | ISBN 9781107164284 (Hard back : alk. paper)
Subjects: LCSH: Photonics.
Classication: LCC TA1520 .L58 2016 | DDC 621.36/5dc23 LC record available at
https://lccn.loc.gov/2016011758
ISBN 978-1-107-16428-4 Hardback
Additional resources for this publication at www.cambridge.org/9781107164284
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To Vida and Janelle
CONTENTS
Preface
Partial List of Symbols
Basic Concepts of Optical Fields

1.1 Nature of Light
1.2 Optical Fields and Maxwells Equations
1.3 Optical Power and Energy
1.4 Wave Equation
1.5 Harmonic Fields
1.6 Polarization of Optical Fields
1.7 Optical Field Parameters
Problems
Bibliography
Optical Properties of Materials

2.1 Optical Susceptibility and Permittivity
2.2 Optical Anisotropy
2.3 Resonant Optical Susceptibility
2.4 Optical Conductivity and Conduction Susceptibility
2.5 KramersKronig Relations
2.6 External Factors
2.7 Nonlinear Optical Susceptibilities
Problems
Bibliography
Optical Wave Propagation

3.1 Normal Modes of Propagation
3.2 Plane-Wave Modes
3.3 Gaussian Modes
3.4 Interface Modes
3.5 Waveguide Modes
3.6 Phase Velocity, Group Velocity, and Dispersion
3.7 Attenuation and Amplication
Problems
Bibliography
page xi
xiii
1
1
4
8
10
11
13
18
20
21
22
22
24
32
38
44
44
55
60
65
66
66
73
86
92
108
122
129
132
139
viii
Contents
4 Optical Coupling
4.1 Coupled-Mode Theory
4.2 Two-Mode Coupling
4.3 Codirectional Coupling
4.4 Contradirectional Coupling
4.5 Conservation of Power
4.6 Phase Matching
Problems
Bibliography
5 Optical Interference
5.1 Optical Interference
5.2 Optical Gratings
5.3 FabryProt Interferometer
Problems
Bibliography
6 Optical Resonance
6.1 Optical Resonator
6.2 Longitudinal Modes
6.3 Transverse Modes
6.4 Cavity Lifetime and Quality Factor
6.5 FabryProt Cavity
Problems
Bibliography
7 Optical Absorption and Emission

7.1 Optical Transitions
7.2 Transition Rates
7.3 Attenuation and Amplication of Optical Fields
Problems
Bibliography
8 Optical Amplication
8.1 Population Rate Equations
8.2 Population Inversion
8.3 Optical Gain
8.4 Optical Amplication
8.5 Spontaneous Emission
Problems
Bibliography
141
141
147
154
156
159
160
165
168
169
169
183
191
200
203
204
204
207
211
214
216
221
223
224
224
234
241
245
248
249
249
251
259
265
267
270
273
Contents
Laser Oscillation
9.1 Conditions for Laser Oscillation
9.2 Mode-Pulling Effect
9.3 Oscillating Laser Modes
9.4 Laser Power
Problems
Bibliography
10
Optical Modulation
10.1 Types of Optical Modulation
10.2 Modulation Schemes
10.3 Direct Modulation
10.4 Refractive External Modulation
10.5 Absorptive External Modulation
Problems
Bibliography
11
Photodetection
ix
274
274
277
279
285
293
296
297
297
298
308
319
344
353
361
11.1 Physical Principles of Photodetection

11.2 Photodetection Noise
11.3 Photodetection Measures
Problems
Bibliography
362
362
375
382
391
395
Appendix A
Appendix B
Appendix C
Appendix D
Index
396
403
405
406
409
Symbols and Notations

SI Metric System
Fundamental Physical Constants
Fourier-Transform Relations
Cambridge Books Online

http://ebooks.cambridge.org/
Jia-Ming Liu
Book DOI: http://dx.doi.org/10.1017/CBO9781316687109
Online ISBN: 9781316687109
Hardback ISBN: 9781107164284
Chapter
Preface pp. xi-xii
Chapter DOI: http://dx.doi.org/10.1017/CBO9781316687109.001
Cambridge University Press
PREFACE
The eld of photonics has matured into an important discipline of modern engineering and
technology. Its core principles have become essential knowledge for all undergraduate students
in many engineering and scientic elds. This fact is fully recognized in the new curriculum of
the Electrical Engineering Department at UCLA, which makes the principles of photonics a
required course for all electrical engineering undergraduate students. Graduate students studying in areas related to photonics also need this foundation.
The most fundamental concepts in photonics are the nature of optical elds and the properties
of optical materials because the entire eld of photonics is based on the interplay between
optical elds and optical materials. Any photonic device or system, no matter how simple or
sophisticated it might be, consists of some or all of these functions: the generation, propagation,
coupling, interference, amplication, modulation, and detection of optical waves or signals.
The properties of optical elds and optical materials are addressed in the rst two chapters of
this book. The remaining nine chapters cover the principles of the major photonic functions.
This book is written for a one-quarter or one-semester undergraduate course for electrical
engineering or physics students. Only some of these students might continue to study advanced
courses in photonics, but at UCLA we believe that all electrical engineering students need to
have a basic understanding of the core knowledge in photonics because it has become an
established key area of modern technology. Many universities already have departments that
are entirely devoted to the eld of photonics. For the students in such photonics-specic
departments or institutions, the subject matter in this book is simply the essential foundation
that they must master before advancing to other photonics courses. Based on this consideration,
this book emphasizes the principles, not the devices or the systems, nor the applications.
Nevertheless, it serves as a foundation for follow-up courses on photonic devices, optical
communication systems, biophotonics, and various subjects related to photonics technology.
Because this book is meant for a one-quarter or one-semester course, it is kept to a length that
can be completed in a quarter or a semester. Because it likely serves the only required
undergraduate photonics course in the typical electrical engineering curriculum, it has to cover
most of the essential principles. The chapters of this book are organized based on the major
principles of photonics rather than based on device or system considerations. These attributes
are the key differences between this book and other books in this eld.
Through my teaching experience on this subject over many years, I nd a need for a textbook
that has the following features.
1. It is self-contained, and its prerequisites are among the required core courses in the typical
electrical engineering curriculum.
2. It covers the major principles in a single book that can be completely taught in a one-quarter
or one-semester course. And it treats these subjects not supercially but to a sufcient depth
http://dx.doi.org/10.1017/CBO9781316687109.001
xii
Preface
for a student to gain a solid foundation to move up to advanced photonics courses, if the
student stays in the photonics eld, or for a student to gain a useful understanding of
photonics, if the student moves on to a different eld.
3. It has ample examples that illustrate the concepts discussed in the text, and it has plenty of
problems that are closely tied to these concepts and examples.
This book is written with the above features to serve the need for a book covering a core
photonics course in a modern electrical engineering curriculum.
There are two prerequisites for a course that uses this book: (1) basic electromagnetics up to
electromagnetic waves and (2) basic solid-state physics or solid-state electronics. No advanced
background in optics beyond what a student normally learns in general physics is required. At
UCLA, this course is taught as a required course in the Electrical Engineering Department to
undergraduate juniors and seniors. The materials of this book have been test taught for a few
years in this one-quarter course, which has 38 hours of lectures, excluding the time for the
midterm and nal exams. This course is followed by elective courses on photonic devices and
circuits, photonic sensors and solar cells, and biophotonics.
Carefully designed examples are given at proper locations to illustrate the concepts discussed
in the text and to help students apply what they learn to solving problems. Each example is tied
closely to one or more concepts discussed in the text and is placed right after that text; its
solution does not simply give the answer but presents a detailed explanation as part of the
teaching process. An ample number of problems are given at the end of each chapter. The
problems are labeled with the corresponding section numbers and are arranged in the sequence
of the material presented in the text. The entire book has 100 examples and 247 problems.
The materials in this book are selected and structured to suit the purpose of a course on the
principles of photonics. Besides the newly written materials, text and gures are adopted from
my book Photonic Devices wherever suitable. All examples and problems, except for the very
few that illustrate key concepts, are newly designed specically to meet the pedagogical
purpose of this book.
This book was developed through test teaching a course in the new curriculum at UCLA. In
this process, I received much feedback from my colleagues and my students. I would like to
thank my editor, Julie Lancashire, for her help at every stage during the development of this
book, and my content manager, Jonathan Ratcliffe, for taking care of the production matters of
this book. I would like to express my loving appreciation to my daughter, Janelle, who took a
special interest in this project and shared my excitement in it. Special thanks are due to my wife,
Vida, who gave me constant support and created an original oil painting for the cover art of
this book.
PARTIAL LIST OF SYMBOLS
Symbol
Unit
Meaning; derivatives
References1
none
round-trip intracavity eld amplication factor
(6.4)
aE , aM
none
asymmetry factors for TE and TM modes
(3.130)
~
A, A
W1=2
mode amplitude
(4.23), (4.26)
Av
W1=2
amplitude of mode
(4.3)
A21
s1
Einstein A coefcient
(7.21)
m2
area
(11.59)
confocal parameter of Gaussian beam
(3.69)f
none
normalized guide index
(3.129)
none
linewidth enhancement factor
(9.39)
~
B, B
W1=2
mode amplitude
(4.24), (4.27)
Hz
bandwidth
(11.1)
B12 , B21
m3 J1 s1
Einstein B coefcients
(7.19), (7.20)
real magnetic induction in the time domain
(1.3)
B, B
complex magnetic induction
(1.41)
m s1
speed of light in free space
(1.1)b, (1.39)
cv
none
overlap coefcient between modes v and
(4.19)
cijkl
m2 A2
quadratic magneto-optic coefcient
(2.77)
thickness or distance; d g , dQW
(3.127)
d, d0
beam spot size diameter, d 2w, d 0 2w0
(3.69)b
dE , dM
effective waveguide thicknesses for TE and TM modes
(3.138), (3.143)
none
group-velocity dispersion; D1 , D2 , D
(3.167)
W1
detectivity
(11.58)
xiv
(cont.)
Symbol
Unit
References1
s m2
group-velocity dispersion; D D=c
(3.168)
m Hz1=2 W1
specic detectivity
(11.59)
C m2
real electric displacement in the time domain
(1.2)
D, D
C m2
complex electric displacement; De , Do , D , D
(1.42), (1.51)
D, D
C m2
slowly varying amplitudes of D and D; De , D0
(3.57)
DR
dB
dynamic range
(11.62)
electronic charge
(2.30)f
ê
none
unit vector of electric eld polarization; ê e , ê o , ê , ê
(1.61)
E1 , E2
eV
energies of levels j1i and j2i
(7.1)
Ec , Ev
eV
conduction-band and valence-band edges
(10.106)
EF
eV
Fermi energy
(11.5)b
Eg
eV
bandgap
(10.105), (11.7)
Eth
eV
threshold photon energy
(11.5)
V m1
real electric eld in the time domain
(1.2)
E0 , E 0
V m1
static or low-frequency electric eld
(2.54)
Ee , Eh
V m1
electric elds seen by electrons and holes
(10.106)
E, E
V m1
complex electric eld
(1.40)
Ev , E v
V m1
complex electric eld of mode v
(3.1)
E, E
V m1
slowly varying amplitudes of E and E; E e , E o , E , E
(1.52)
Ev, E v
V m1
complex electric eld prole of mode v
(3.1)
^v
E
V m1 W1=2
normalized electric mode eld distribution, E v Av E^ v
(3.18)
ER
dB
extinction ratio
(10.18)
Hz
acoustic or modulation frequency, f =2
(2.79)b, (10.27)b
f 3dB
Hz
3-dB modulation bandwidth or cutoff frequency
(10.31), (11.64)
fK
Kerr focal length
(10.115)
f ijk
m A1
linear magneto-optic coefcient, Faraday coefcient
(2.76), (10.77)
xv
(cont.)
Symbol
Unit
References1
fr
Hz
relaxation resonance frequency, f r r =2
(10.41)
none
excess noise factor
(11.38)
none
nesse of interferometer or optical cavity
(5.49), (6.12)
Fz; z0
none
forward-coupling matrix for codirectional coupling
(4.48)
g, g v
m1
gain coefcient, amplication coefcient
(3.183)f, (7.46)
g0
m1
unsaturated gain coefcient
(8.22)
g th
m1
threshold gain coefcient
(9.9), (9.19)
g^ v
lineshape function
(7.2)
none
degeneracy factor; g1 , g2
(7.1)f, (7.28)
s1
gain parameter
(9.18)
g0
s1
unsaturated gain parameter
(9.22)
gn
m3 s1
differential gain parameter
(10.36)
gp
m3 s1
nonlinear gain parameter
(10.36)
gth
s1
threshold gain parameter
(9.20), (10.34)
none
cavity round-trip eld gain; Gc , Gmn , Gcmn
(6.4)
none
photodetector current gain
(11.4)f, (11.36)
G, G0
none
optical amplier power gain, G0 for unsaturated gain
(8.39)
h,
Js
Plancks constant, h=2
(1.1)
h1 , h2 , h3
m1
transverse oscillation parameters of mode eld
(3.104), (3.133)
height of acousto-optic transducer
(10.89)
H
none
Heaviside step function
(2.24)
H m
none
Hermite function
(3.73)f
A m1
real magnetic eld in the time domain
(1.3)
H0 , H 0
A m1
static or low-frequency magnetic eld
(2.68)
H, H
A m1
complex magnetic eld
(1.42)
Hv , H v
A m1
complex magnetic eld of mode v
(3.2)
xvi
(cont.)
Symbol
Unit
References1
H, H
A m1
slowly varying amplitudes of H and H
(3.5)
Hv , Hv
A m1
complex magnetic eld prole of mode v
(3.2)
^
H
A m1 W1=2
normalized electric mode eld distribution,

^
H AH
(3.18)
none
p
1
current; ib , id , in , iph , is
(11.4)f
injection current; I 0 , I m , I th
(10.22)
W m2
optical intensity; I 0 , I i , I in , I out , I r , I t
(1.56)
I0
reverse current
(11.15)
I v
W m2 Hz1
optical spectral intensity distribution
(7.17)
Ip, Is
W m2
optical pump and signal intensities
(8.36)
I sat
W m2
saturation intensity
(8.22)
J, J
A m2
real current density
(1.5)
A m2
complex current density
(2.35)
m1
propagation constant, wavenumber; k 0 , k i , k r , k t
(1.84)
kB
J K1
Boltzmann constant
(7.14), (7.25)
ke , ko
m1
propagation constants of extraordinary and ordinary

waves
(3.57)
k 0 , k 00
m1
real and imaginary parts of k, k k 0 ik 00
(3.180)
kx , ky , kz
m1
propagation constants of x, y, and z polarized elds
(2.15)
kX , kY , kZ
m1
propagation constants of X, Y, and Z polarized elds (2.67)
k , k
m1
propagation constants of circularly polarized elds
(2.21)
k^
none
unit vector in the k direction
(1.84)
m1
wavevector; ki , kr , kt , kq
(1.1)b, (1.52)
ke , ko
m1
wavevectors of extraordinary and ordinary waves
(3.56)f, (3.57)
kx , ky , kz
m1
wavevectors of x, y, and z polarized elds
(3.48)f
k , k
m1
wavevectors of circularly polarized elds
(10.74)
xvii
(cont.)
Symbol
Unit
References1
m1
wavenumber of acoustic wave or grating, K 2=
(2.79)f, (4.35)
K factor of a semiconductor laser
(10.43)
m1
wavevector of acoustic wave
(2.79)
length or distance
(3.185)
lc
coupling length; lPM

c
(4.56)
lRT
round-trip optical path length
(6.1)
l=4 , l=2
quarter-wave and half-wave lengths
(3.49), (3.50)
length of acousto-optic transducer
(10.89)
none
transverse mode index associated with x
(3.1)f
none
modulation index
(10.27)
m0
kg
free electron rest mass
Fig. 11.1
kg
effective mass of carriers
(2.31)
m
e , mh
kg
effective masses of electrons and holes
(10.107)
kg
atomic or molecular mass
(7.14)
M TE , M TM
none
numbers of guided TE and TM modes
(3.152), (3.153)
Ms
A m1
saturation magnetization
(10.78)
A m1
real magnetic polarization in the time domain
(1.3)
M0 , M 0
A m1
static or low-frequency magnetization
(2.70)
none
transverse mode index associated with y
(3.1)f
none
index of refraction; n , n
(1.84)
m3
electron concentration
(11.9)
n0
m3
equilibrium concentration of electrons
(11.9)f
n1 , n2 , n3
none
refractive indices of waveguide layers, n1 > n2 > n3
(3.125)
n2
m2 W1
coefcient of intensity-dependent index change
(10.101)
ne , no
none
extraordinary and ordinary indices of refraction
(2.15)f, (3.56)
nx , ny , nz
none
principal indices of refraction
(2.14)
xviii
(cont.)
Symbol
Unit
References1
nX , nY , nZ
none
new principal indices of refraction
(2.66)
n , n
none
principal indices of refraction for circular polarized

modes
(2.20)
n? , njj
none
indices of second-order magneto-optic effect
(2.16)
n0 , n00
none
real and imaginary parts of refractive index,

n n0 in00
(3.181)
n^
none
unit normal vector
(1.23)
none
some number
(5.21)
none
group index; N 1 , N 2 , N
(3.171)
m3
carrier density
(2.31)
m3
effective population inversion
(8.4)
N1, N2,
Nt
m3
population densities in levels j1i, j2i, and all levels
(7.26), (8.12)
N sp
none
spontaneous emission factor
(9.14)
none
number of charge carriers
(11.3)
NEP
noise equivalent power
(11.55)
none
probability
(11.18)
none
cross-section ratio for pumping
(8.13)
m3
hole concentration
(11.9)
p0
m3
equilibrium concentration of holes
(11.9)f
pijkl p0ijkl
none
elasto-optic and rotation-optic coefcients
(2.83)
pvk
Hz1
probability density function
(7.10)
power; Pa , Pin , Pout , Ppk , Pth , Pv
(3.17)
Pp , Ps
tr
in
out
pump and signal powers; Pth
p , Pp , Ps , Ps
(9.27), (8.37)
Psat
saturation power
(8.37)
Psp
spontaneous emission power
(8.44)
Ptrsp
critical uorescence power
(8.46)
xix
(cont.)
Symbol
Unit
References1
^ sp
P
W m3
spontaneous emission power density
(8.43)
p^trsp
W m3
critical uorescence power density
(8.45)
C m2
real electric polarization in the time domain
(1.2)
P, P
C m2
complex electric polarization
(1.50)
Pn
C m2
nth-order nonlinear real electric polarization
(2.91)
Pn , Pn
C m2
nth-order nonlinear complex electric polarization
(2.91)
Pres
C m2
complex electric polarization from resonant transition (7.47)b
none
longitudinal mode index
(5.47), (6.9)
none
order of coupling or diffraction
(4.36), (5.24)
charge
(2.30)
qz
complex radius of curvature of a Gaussian beam
(3.75)
none
quality factor of resonator; Qmnq
(6.26), (6.30)
none
acousto-optic diffraction parameter
(10.83)
radial coordinate, radial distance
none
reection coefcient; r1 , r 2 , rp , rs
(3.91), (4.67)
none
pumping ratio of a laser
(9.26)
rijk , rk
m V1
linear electro-optic coefcients, Pockels coefcients
(2.58), (2.60)
rf , r
none
complex modulation response function
(10.29), (10.40)
spatial vector
(1.2)
none
reectance, reectivity; R1 , R2 , Rp , Rs
(3.93)
resistance; Ri , RL
(11.16)
m3 s1
effective pumping rate for population inversion
(8.6)
R1 , R2
m3 s1
pumping rates for levels j1i and j2i
(8.1), (8.2)
Rf
none
electrical power spectrum of modulation response
(10.30), (10.44)
Rz; 0; l
none
R, Rij
none
reverse-coupling matrix for contradirectional coupling (4.59)

(2.82)
rotation tensor and elements, R Rij
xx
(cont.)
Symbol
Unit
References1
radius of curvature; R1 , R2
(3.71), (6.31)
none
amplitude of rotation; Rij
(2.87)
A W1
responsivity of photodetector with current output; R0
(11.50)
V W1
responsivity of photodetector with voltage output
(11.51)
separation, spacing; se
Fig. 4.2, (10.69)
none
signal; sn
(11.18)
sijkl , skl
m2 V2
quadratic electro-optic coefcients, Kerr coefcients
(2.58), (2.60)
m3
photon density
(9.21)
Ssat
m3
saturation photon density
(9.24)
W m2
real Poynting vector
(1.32)
W m2
(1.54)
S, Sij
none
complex Poynting vector; Se , So

strain tensor and elements, S Sij
(2.81)
none
amplitude of strain; S ij
(2.87)
none
number of photons
(11.2)
SNR
none, dB
signal-to-noise ratio
(11.26)
time
none
transmission coefcient; tp , ts
(3.92)
tr , tf
risetime and falltime
(11.63)b
temperature
(7.14)
time interval
(1.53)
round-trip time of optical cavity
(6.1)
none
transmittance, transmissivity; T p , T s
(3.94), (10.108)
u, u0
J m3
electromagnetic energy density
(7.16), (1.33)
uv
J m3 Hz1
spectral energy density
(7.16)
u, ui
elastic deformation wave and its components
(2.79), (2.81)
optical energy; U mode
(9.28)
xxi
(cont.)
Symbol
Unit
References1
amplitude of elastic wave
(2.79)
voltage; v n , v out , v s
(11.16)
m s1
velocity
Fig. 11.1
va
m s1
acoustic wave velocity
(2.80)b
vg
m s1
group velocity; v g
(3.165)
vp
m s1
phase velocity; v p
(3.162)
none
normalized frequency and waveguide thickness, V

number
(3.128)
rad A1
Verdet constant
(10.77)
voltage; V m , V , V =2
(10.51), (11.15)
Vc
none
cutoff V number; V cm
(3.147)
m3
volume; V gain , V mode
(1.31)b, (6.2)
w, w0
Gaussian beam radius, spot size
(3.69), (3.70)
width of acousto-optic cell
(10.91)
s1
transition probability rate; W 12 , W 21 , W p , W sp
(7.22)(7.24)
W p, W m
W m3
power densities expended by EM eld on P and M
(1.34), (1.35)
W v
none
transition rate per unit frequency;

W 12 v, W 21 v,W sp v
(7.19)(7.21)
spatial coordinate
^x
none
unit coordinate vector or principal dielectric axis
^
spatial coordinate along X
^
X
none
new principal dielectric axis
spatial coordinate
^y
none
spatial coordinate along Y^
Y^
none
spatial coordinate
(1.62), (2.13)b
(2.65)b
(1.62), (2.13)b
(2.65)b
xxii
(cont.)
Symbol
Unit
References1
^z
none
(3.16), (2.13)b
zR
Rayleigh range of a Gaussian beam
(3.69)
spatial coordinate along Z^
Z^
none
(2.65)b
rad
eld polarization angle
(1.64)
rad
walk-off angle of extraordinary wave
(3.60)
, v
m1
attenuation coefcient, absorption coefcient
(3.180), (7.45)
m1
unsaturated absorption coefcient
(10.110)
m1
propagation parameter for contradirectional coupling
(4.61)
none
bottleneck factor
(8.7)
m1
propagation constant of a mode; mn , TE , TM
(3.1)
0 , 00
m1
real and imaginary parts of , 0 i00
(3.184)
m1
propagation parameter for codirectional coupling
(4.50)
s1
relaxation rate, decay rate; 21 , i , out
(2.23)
1 , 2 , 3
m1
transverse decay parameters of mode eld
(3.118), (3.131)
s1
acoustic decay rate
(10.93)
s1
cavity decay rate, photon decay rate; cmnq
(6.25)
s1
differential carrier relaxation rate
(10.37)
s1
nonlinear carrier relaxation rate
(10.37)
s1
total carrier relaxation rate
(10.42)
s1
spontaneous carrier relaxation rate
(10.42)
none
overlap factor
(6.2)
m1
phase mismatch parameter for phase mismatch of 2
(4.31)
mnq
rad s1
frequency shift of mode pulling
(9.12)
n, p
m3
excess electron and hole concentrations
(11.10)
C m2
change in electric polarization
(4.8)
xxiii
(cont.)
Symbol
Unit
References1
pulsewidth or time duration; t ps
(10.117), (11.1)
F m1
change or variation in electric permittivity
(2.55), (4.12)
~
, ~
F m1
amplitudes of and
(2.88)
, ij
none
change or variation in relative impermeability
(2.58)
rad
divergence angle of a Gaussian beam
(3.72)
spectral width; g
Table 7.1
Hz
optical linewidth, bandwidth; vD , vg , vinh , vh
(7.4)
vc
Hz
longitudinal mode linewidth
(6.18)
vL
Hz
longitudinal mode frequency spacing
(6.17)
vmnq
Hz
oscillating laser mode linewidth
(9.13)
vST
Hz
SchawlowTownes linewidth of laser mode; vST

mnq
(9.14)
rad
phase shift or phase retardation
(10.13)
rad
phase width of a cavity resonance peak
(6.11)
rad
phase spacing between cavity resonance peaks
(6.10)
none
change or variation in electric susceptibility
(2.54)
rad s1
optical linewidth, bandwidth, 2v; inh , h (7.3)f, (7.13)
F m1
electric permittivity
(2.11), (3.4)
F m1
electric permittivity of free space
(1.2)
0 , 00
F m1
real and imaginary parts of , 0 i 00
(3.179)
x, y, z
F m1
principal dielectric permittivities
(2.13)
X , Y , Z
F m1
new principal dielectric permittivities
(2.65)
,
F m1
principal dielectric permittivities of circular

polarizations
(2.17)
r, t
F m4 s1
real permittivity tensor in the real space and time

domain
(1.21)
, ij
F m1
complex permittivity tensor in the frequency domain
(1.60)
res
F m1
permittivity of resonant transition
(6.36)
xxiv
(cont.)
Symbol
Unit
References1
rad
ellipticity of polarization ellipse
(1.68)
mn z
rad
phase variation of Gaussian mode eld; RT

mn
(3.76)
none
coupling efciency; PM
(4.55)
none
power conversion efciency
(9.37)
coll
none
collection efciency
(11.48)
none
external quantum efciency
(10.24), (11.48)
none
internal quantum efciency
(11.48)
inj
none
injection efciency
(10.22)
none
slope efciency, differential power conversion

efciency
(9.38)
none
transmission efciency
(11.48)
, ij ,
none
relative impermeability tensor and its elements,

ij
(2.57)
rad
angle, spherical angular coordinate
(3.51)
rad
orientation of the polarization ellipse
(1.69)
rad
Brewster angle or Bragg angle
(3.100), (10.88)
rad
critical angle
(3.102)
rad
angle of diffraction
(10.87)
def
rad
deection angle
Example 10.9
rad
Faraday rotation angle
(10.75)
i , r , t
rad
angles of incidence, reection, and refraction

(transmitted)
(3.88)
m1
coupling coefcient; v
(4.13)
m1
coupling coefcient; ~v
(4.20)
optical wavelength in free space
(1.1)
cutoff wavelength; cm
(3.151)
th
threshold wavelength
(11.5)
acoustic wavelength or grating period
(2.79)b, (4.35)
(cont.)
Symbol
Unit
References1
H m1
magnetic permeability of free space
(1.3)
e , h
m2 V1 s1
electron and hole mobilities
(11.9)
Hz
optical frequency
(1.1)
v0
Hz
center optical frequency
(2.27)f, (7.12)
v21
Hz
resonance frequency between levels j1i and j2i
(7.1)
none
duty factor
Fig. 4.3
, M 0z
none
permittivity tensor elements for circular birefringence
(2.16), (2.78)
C m3
charge density
(1.6)
rad m1
specic Faraday rotation
(10.79)
S m1
conductivity; 0
(2.33), (11.9)
12 , 21
m2
transition cross sections
(7.36), (7.37)
a , e
m2
absorption and emission cross sections
(7.38), (7.39)
2s
none
variance of s
(11.19)
lifetime, decay time, or time constant
(2.30), (7.6)
1, 2
uorescence lifetimes of levels j1i and j2i
(7.6), (7.8)
photon lifetime; cmnq
(6.23)
saturation lifetime or spontaneous carrier lifetime
(8.23), (10.23)
sp
spontaneous radiative lifetime
(7.32)
rad
azimuthal angle, azimuthal angular coordinate
(3.52)
work function potential; e work function
(11.6)
rad
phase or phase shift
(1.63), (1.83)
none
electric susceptibility
(2.11)
electron afnity potential; e electron affinity
(11.7)
res
none
resonant electric susceptibility
(2.25), (2.26)
x, y, z
none
principal dielectric susceptibilities
(2.15)f
0 , 00
none
real and imaginary parts of , 0 1 00
(2.7)b
xxv
xxvi
(cont.)
Symbol
Unit
References1
r; t
m3 s1
real susceptibility tensor in the real space and time

domain
(1.20)
, ij
none
complex susceptibility tensor in the frequency domain (1.59)
m V1
second-order nonlinear susceptibility in the frequency (2.98), (2.100)

domain
3 , ijkl
m2 V2
third-order nonlinear susceptibility in the frequency

domain
(2.99), (2.101)
rad
spatial phase of mode eld distribution
(3.107)
rad
angle between Se and optical axis of crystal
(3.60)
rad s1
optical angular frequency; 2v
(1.1)b
rad s1
center optical angular frequency; 0 2v0
(2.22), (7.13)
21
rad s1
resonance angular frequency between levels j1i and

j2i
(2.22)
rad s1
cutoff frequency; cm
(3.151)
rad s1
acoustic or modulation angular frequency; 2f
(2.79), (10.27)
rad s1
relaxation resonance frequency; r 2f r
(10.41)
2 , ijk
Sufxes, f forward and b backward, on the equation number indicate symbols explained for the rst
time in the text immediately after or before the equation cited.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
1 - Basic Concepts of Optical Fields pp. 1-21
1
1.1
NATURE OF LIGHT
..............................................................................................................
Photonics addresses the control and use of light for various applications. Light is electromagnetic radiation of frequencies in the range from 1 THz to 10 PHz, corresponding to wavelengths
between 300 m and 30 nm in free space, which is generally divided into the infrared,
visible, and ultraviolet regions. In this spectral region, the electromagnetic radiation exhibits the
dual nature of photon and wave. The photon nature has to be considered in the generation,
amplication, frequency conversion, or detection of light, whereas the wave nature is important
in all processes but especially in the propagation, transmission, interference, modulation, or
switching of light.
1.1.1 Photon Nature of Light

The energy of a photon is determined by its frequency or, equivalently, its angular frequency
2. Associated with its particle nature, a photon has a momentum determined by its
wavelength or, equivalently, its wavevector k. These characteristics are summarized below for
a photon in free space:
speed
energy
momentum
c ;
h pc;
p h=c h=,
p k.
The energy of a photon that has a wavelength of in free space can be calculated using the
formula:
h
1:2398
1239:8
m eV
nm eV:
(1.1)
The photon energy at the optical wavelength of 1 m is 1.2398 eV, and its frequency is
300 THz.
EXAMPLE 1.1
The visible spectrum ranges from 700 nm wavelength at the red end to 400 nm wavelength at
the violet end. What is the frequency range of the visible spectrum? What are the energies of
visible photons?
Solution:
The 700 nm optical wavelength at the red end has a frequency of
red
c
red
3 108 m s1
429 THz
700 nm
and a photon energy of

hred
1239:8
1239:8
nm eV
eV 1:77 eV:
red
700
The 400 nm optical wavelength at the violet end has a frequency of

violet
c
violet
3 108 m s1
750 THz
400 nm
and a photon energy of

hviolet
1239:8
1239:8
nm eV
eV 3:10 eV:
violet
400
Therefore, the frequency range of the visible spectrum is from 429 THz to 750 THz. Visible
photons have energies in the range from 1.77 eV to 3.10 eV.
The energy of a photon is determined only by its frequency or, equivalently, by its free-space
wavelength, but not by the light intensity. The intensity, I, of monochromatic light is related to
the photon ux density, or the number of photons per unit time per unit area, by
photon flux density
I
I
:
h
The photon ux, or the number of photons per unit time, of a monochromatic optical beam is
related to the beam power P by
photon flux
P
P
:
h
EXAMPLE 1.2
Find the photon ux of a monochromatic optical beam that has a power of P 1 W by taking
its wavelength at either end of the visible spectrum. What are the momentum carried by a red
photon and the momentum carried by a violet photon? What is the total momentum carried by
the beam in a time duration of t 1 s?
1.1 Nature of Light
Solution:
From Example 1.1, the photon energy of the 700 nm wavelength at the red end is
hred 1:77 eV, and that of the 400 nm wavelength at the violet end is hviolet 3:10 eV.
Therefore, the photon ux of a beam that has a power of P 1 W at the 700 nm red
wavelength is
red photon flux
P
1
s1 3:53 1018 s1 ,

hred 1:77 1:6 1019
and the photon ux of a beam that has a power of P 1 W at the 400 nm violet wavelength is
violet photon flux
P
hviolet
1
s1 2:02 1018 s1 :
3:10 1:6 1019
The momentum carried by a red photon is

pred
hred 1:77 1:6 1019

N s 9:44 1028 N s,
c
3 108
and that carried by a violet photon is

pviolet
hviolet 3:10 1:6 1019
N s 1:65 1027 N s:
c
3 108
The total momentum carried by an optical beam that has a power of P during a time duration of
t is independent of the optical wavelength:
total momentum photon fluxpt
P h
Pt
t
:
h c
c
Therefore, irrespective of whether the wavelength of the beam is at the red or the violet end, the
total momentum carried by the beam in a time duration of t 1 s is
total momentum
Pt
11
3:33 109 N:
c
3 108
1.1.2 Wave Nature of Light

An optical wave is characterized by the space and time dependence of the optical eld, which is
composed of coupled electric and magnetic elds governed by Maxwells equations. It varies
with time at an optical carrier frequency, and it propagates in a spatial direction determined by a
wavevector. The behavior of an optical wave is strongly dependent on the optical properties of
the medium. An optical eld is a vectorial eld characterized by ve parameters: polarization,
magnitude, phase, wavevector, and frequency. Polarization and wavevector are vectorial
quantities; magnitude, frequency, and phase are scalar quantities. The general properties of
optical elds are described in the following sections.
1.2
OPTICAL FIELDS AND MAXWELLS EQUATIONS
..............................................................................................................
An electromagnetic eld in a medium is characterized by four vectorial elds:
electric eld
electric displacement
magnetic eld
magnetic induction
Er; t
Dr; t
H r; t
Br; t
V m1 ,
C m2 ,
A m1 ,
T or Wb m2 :
The response of a medium to an electromagnetic eld generates the polarization and the
magnetization:
polarization (electric polarization)
magnetization (magnetic polarization)
Pr; t C m2 ,
M r; t A m1 :
The electric eld Er; t and the magnetic induction Br; t are the macroscopic forms of the
microscopic elds seen by the charge and current densities in the medium. The polarization
Pr; t and the magnetization M r; t are the macroscopically averaged densities of microscopic
electric dipoles and magnetic dipoles that are induced by the presence of the electromagnetic
eld in the medium. These macroscopic forms are obtained by averaging over a volume that is
small compared to the dimension of the optical wavelength but is large compared to the atomic
dimension. The electric displacement Dr; t and the magnetic eld H r; t are macroscopic
elds dened as
Dr; t 0 Er; t Pr; t,
(1.2)
and
H r; t
1
Br; t M r; t ,
0
(1.3)
where 0 1=36 109 F m1 8:854 1012 F m1 is the electric permittivity of free
space and 0 4 107 H m1 is the magnetic permeability of free space. In addition to the
induced charge density and current density that respectively generate electric dipoles and
magnetic dipoles for Pr; t and M r; t , an independent charge or current density, or both,
from external sources may exist:
charge density
current density
r; t C m3 ,
J r; t A m2 :
The behavior of a space- and time-varying electromagnetic eld in a medium is governed by

space- and time-dependent macroscopic Maxwells equations:
E
H
B
,
t
D
J,
t
Faradays law;
Ampères law;
(1.4)
(1.5)
D ,
B 0,
Gausss law; Coulombs law;
(1.6)
absence of magnetic monopoles:
(1.7)
Note that Gausss law in the form of (1.6) is equivalent to Coulombs law because one can be
derived from the other. The current and charge densities are constrained by the continuity
equation:
J
0,
t
conservation of charge:
(1.8)
The total current density in an optical medium has two contributions: the polarization current
from the bound charges of the medium and the current from free charge carriers, thus
Jtotal J bound J free . The free-carrier current has two possible origins, one from the response
of the conduction electrons and holes of the medium to the optical eld and the other from an
external current source: J free J cond J ext . Both J bound and J cond are induced by the optical
eld; thus
J total J bound J free J bound Jcond J ext Jind J ext ,
(1.9)
where J ind J bound J cond : Similarly, the total charge density can be decomposed as
total bound free bound cond ext ind ext :
(1.10)
In an optical medium, charge conservation requires that an increase of charge density induced
by an optical eld at a location is always accompanied by a reduction at another location,
resulting in no net macroscopic induced charge density. Therefore, ind 0 and total ext for
a macroscopic optical eld. By contrast, an induced macroscopic current density of J ind 6 0
can exist in an optical medium.
In an optical medium that is free of external sources, J ext 0 and total ext 0, but
Jtotal J bound J cond J ind 6 0: Both J bound and Jcond are induced currents in response to an
optical eld. The bound-electron polarization current J bound is a displacement current that is
always included in the D=t term but not in the J term in (1.5). The conduction current J cond is
also an induced current, but it is carried by free charge carriers in the medium. In the case when
both external current and external charge are absent, the form of Maxwells equations depends
on how the conduction current is treated. There are generally two alternatives.
1. Being an induced current, J cond can be considered as a displacement current to be included
in the D=t term so that J 0 in (1.5). Then, Maxwells equations are
E
H
B
,
t
D
,
t
(1.11)
(1.12)
D 0,
(1.13)
B 0,
(1.14)
where D is the electric displacement that includes optical-eld-induced responses from all
bound and conduction charges in the medium.
2. Being a current carried by free charge carriers, J cond can be separated from the D=t term so
that J J cond in (1.5). Then, Maxwells equations have the form:
E
H
B
,
t
Dbound
J cond ,
t
(1.15)
(1.16)
Dbound 0,
(1.17)
B 0,
(1.18)
with J cond 0, where Dbound is the electric displacement that includes only the contribution from bound charges and excludes that from the conduction current.
These two alternative forms of Maxwells equations are equivalent. The form using (1.16) is
taken only when a specic effect of the conduction current is considered, as in Section 2.4.
Otherwise, the form using (1.12) is generally taken. Therefore, we use the general form given in
(1.11)(1.14) unless the situation calls for specic attention to a conduction current.
1.2.1 Transformation Properties

Maxwells equations and the continuity equation are the basic physical laws that govern the
behavior of electromagnetic elds. They are invariant under the transformation of space
inversion, in which the spatial vector r is changed to r0 r, i.e., r ! r, or x; y; z !
x; y; z, and under the transformation of time reversal, in which the time variable t is
changed to t 0 t, i.e., t ! t: This means that the form of these equations is not changed
when we perform the space-inversion transformation or the time-reversal transformation, or
both together.
The eld quantities that appear in Maxwells equations, however, do not have to be invariant
under space inversion or time reversal. Their transformation properties are summarized as
follows.
1. Electrical elds: The electric eld vectors E, D, and P are polar vectors associated with the
charge-density distribution. They change sign under space inversion but not under time
reversal.
2. Magnetic elds: The magnetic eld vectors B, H, and M are axial vectors associated with
the current-density distribution. They change sign under time reversal but not under space
inversion.
3. Charge density: The charge density is a scalar. It does not change sign under either space
inversion or time reversal.
4. Current density: The current density J is a polar vector that is the product of charge density
and velocity: J v. It changes sign under either space inversion or time reversal following
the sign change of the velocity vector under either transformation.
1.2.2 Optical Response of a Medium

Polarization and magnetization are generated in a medium by the response of the medium to the
electric and magnetic elds, respectively: Pr; t depends on Er; t , and M r; t depends on
Br; t: At an optical frequency, the magnetization vanishes: M 0: Therefore, it is always true
for an optical eld that
Br; t 0 Hr; t:
(1.19)
Because 0 is a constant that is independent of the medium, the magnetic induction Br; t can
be replaced by 0 H r; t for any equations that describe optical elds, including Maxwells
equations, thus effectively eliminating one eld variable. Note that this is not true at DC or low
frequencies, however, because a nonzero DC or low-frequency magnetization, M 6 0, can exist
in any material. Indeed, it is possible to change the optical properties of a medium through a
magnetization induced by a DC or low-frequency magnetic eld, leading to the functioning of
magneto-optics. It should be noted that even for magneto-optics, the magnetization is induced
by a DC or low-frequency magnetic eld that is separate from the optical eld. No magnetization is induced by the magnetic component of the optical eld.
The optical properties of a material are completely determined by the relation between Pr; t
and Er; t: This relation is generally characterized by an electric susceptibility tensor, ,
through the following denition for electric polarization,
t
Pr; t 0
r r0 ; t t 0 Er0 , t 0 dr0 dt 0:
(1.20)
all r0
The relation between Dr; t and Er; t is characterized by the electric permittivity tensor, , of
the medium:
t
Dr; t 0 Er; t Pr; t
r r0 ; t t 0 Er0 , t 0 dr0 dt 0:
(1.21)
all r0
From (1.20) and (1.21), the relationship between and in the real space and time domain is
r; t 0 rt I r; t,
(1.22)
where I is the identity tensor that has the form of a 3 3 unit matrix and the delta functions are
Dirac delta functions: all r rdr and tdt 1. The relation in (1.22) indicates that and
contain exactly the same information about the medium: one is known when the other is known.
Because and, equivalently, represent the response of a medium to an optical eld and thus
completely characterize the macroscopic electromagnetic properties of the medium, (1.20) and
(1.21) can be regarded as the denitions of Pr; t and Dr; t , respectively.
1.2.3 Boundary Conditions

At the interface of two media of different optical properties, as shown in Fig. 1.1, the optical
eld components must satisfy certain boundary conditions. These boundary conditions can be
Figure 1.1 Boundary between two media of different optical properties.
derived from Maxwells equations given in (1.11)(1.14). From (1.11) and (1.12), the tangential components of the elds at the boundary satisfy
n^ E1 n^ E2 ,
(1.23)
n^ H 1 n^ H2 ,
(1.24)
where n^ is the unit vector normal to the interface as shown in Fig. 1.1. From (1.13) and (1.14),
the normal components of the elds at the boundary satisfy
n^ D1 n^ D2 ,
(1.25)
n^ B1 n^ B2 :
(1.26)
The tangential components of E and H are continuous across an interface, while the normal
components of D and B are continuous. Because B 0 H at an optical frequency, as discussed
above, (1.24) and (1.26) also imply that the tangential component of B and the normal
component of H are also continuous. Consequently, all of the magnetic eld components in
an optical eld are continuous across a boundary. Possible discontinuities in an optical eld
exist only in the normal component of E or in the tangential component of D.
1.3
OPTICAL POWER AND ENERGY
..............................................................................................................
Taking the dot product of H and (1.4) and that of E and (1.5) yields
H E H
E H E
B
,
t
D
E J:
t
(1.27)
(1.28)
Using the vector identity B A A B A B, (1.27) and (1.28) can be

combined to give
1.3 Optical Power and Energy
Figure 1.2 Boundary surface enclosing a volume element.
E H E J E
D
B
H
:
t
t
Using (1.2) and (1.3) and rearranging (1.29), we obtain

P
0 2 0
M
2
E J E H
0 H
:
jEj jH j E
2
t 2
t
t
(1.29)
(1.30)
Recall that power in an electric circuit is given by voltage times current and has the unit of
W V A (watts = volts amperes). Similarly, in an electromagnetic eld E J is the power
density and has the unit of V A m3 , or W m3 . From (1.30), the total power dissipated by an
electromagnetic eld in a volume of V is simply the integral of E J over the volume:

0 2 0
P
M
2
E JdV E H n^da
E 0 H
dV , (1.31)
jEj jH j dV
2
2
t
t
t
V
where the rst term on the right-hand side is a surface integral over the closed surface A of the
volume V and n^ is the outward-pointing unit normal vector of the surface, as shown in Fig. 1.2.
Each term in (1.31) has the unit of power, and each has an important physical meaning.
1. The vectorial quantity
SEH
(1.32)
is called the Poynting vector of the electromagnetic eld. It represents the instantaneous
magnitude and direction of the power ow of the eld.
2. The scalar quantity
u0
0 2 0
jEj jH j2
2
2
(1.33)
has the unit of energy per unit volume and is the energy density stored in the propagating
eld. It consists of two components, thus accounting for energies stored in both electric and
magnetic elds at any instant of time.
3. The last term in (1.31) also has two components associated with electric and magnetic elds,
respectively. The quantity
Wp E
P
t
(1.34)
10
is the power density expended by the electromagnetic eld on the polarization. It is the rate
of energy transfer from the electromagnetic eld to the medium on inducing the electric
polarization in the medium. Similarly, the quantity
W m 0 H
M
t
(1.35)
is the power density expended by the electromagnetic eld on the magnetization.

With these physical meanings attached to the terms in (1.31), it can be seen that (1.31) simply
states the law of conservation of energy in any arbitrary volume element V in the medium. The
total electromagnetic energy in the medium equals that contained in the propagating eld plus
that stored in the electric and magnetic polarizations.
For an optical eld, E J 0 and W m 0 because J 0 and M 0, as discussed above.
Then, (1.31) becomes
S n^da
u0 dV W p dV ,
(1.36)
t
V
which states that the total optical power owing into volume V through its boundary surface A
is equal to the rate of increase with time of the energy stored in the propagating elds in V plus
the power transferred to the polarization of the medium in this volume.
1.4
WAVE EQUATION
..............................................................................................................
By applying to (1.11) and using (1.19) and (1.12), we obtain the wave equation:
E 0
2 D
0:
t 2
(1.37)
By using (1.2), the wave equation can be expressed as

E
1 2 E
2 P

,
0
c2 t 2
t 2
(1.38)
where
1
c p 3 108 m s1
0 0
(1.39)
is the speed of light in free space.

The wave equation in (1.38) describes the space-and-time evolution of the electric eld of the
optical wave. Its right-hand side can be regarded as the driving source for the optical wave; that
is, the polarization in a medium drives the evolution of an optical eld. This wave equation can
take on various forms depending on the characteristics of the medium, as will be seen on
various occasions later. Here we leave it in this general form.
1.5 Harmonic Fields
1.5
11
HARMONIC FIELDS
..............................................................................................................
Optical elds are harmonic elds that vary sinusoidally with time. The eld vectors dened in
the preceding section are all real quantities. For harmonic elds, it is always convenient to use
complex elds. We dene the space- and time-dependent complex electric eld, Er; t, through
its relation to the real electric eld, Er; t :1
Er; t Er; t E r; t Er; t c:c:,
(1.40)
where c.c. means the complex conjugate. In our convention, Er; t contains the complex eld
components that vary with time as exp it with having a positive value, while E r; t
contains those components that vary with time as exp it with positive . The complex elds
of other eld quantities are similarly dened.
With this denition for the complex elds, all of the linear eld equations retain their forms.
In terms of complex optical elds, Maxwells equations in the form of (1.11)(1.14) are
E
H
B
,
t
D
,
t
(1.41)
(1.42)
D 0,
(1.43)
B 0;
(1.44)
and those in the form of (1.15)(1.18) are

E
H
B
,
t
Dbound
Jcond ,
t
(1.45)
(1.46)
Dbound 0,
(1.47)
B 0:
(1.48)
The wave equation in terms of the complex electric eld is

E
1 2 E
2 P

,
0
c2 t2
t 2
(1.49)
In some literature, the complex eld is dened through a relation with the real eld as Er; t Er; t E r; t=2,
which differs from our denition in (1.40) by the factor 1=2. The magnitude of the complex eld dened through this
alternative relation is twice that of the complex eld dened through (1.40). As a result, expressions for many quantities
may be different under the two different denitions. An example is the time-averaged Poynting vector given in (1.53),
which would be changed to S ReE H =2 in this alternative denition of the complex eld.
12
while
t
Pr; t 0
r r0 ; t t 0 Er0 , t 0 dr0 dt0
(1.50)
all r0
and
t
Dr; t 0 Er; t Pr; t
all
r r0 ; t t0 Er0 , t0 dr0 dt0:
(1.51)
r0
It is important to note that while E, D, and P are complex, r r0 ; t t 0 and r r0 ; t t0

in (1.50) and (1.51) are always real functions of space and time and are the same as those in
(1.20) and (1.21).
The complex electric eld of a harmonic optical eld that has a carrier wavevector of k and a
carrier angular frequency of can be further expressed as
Er; t E r; t exp ik r it ê E r; t exp ik r it ,
(1.52)
where E r; t is the space- and time-dependent amplitude of the eld, and ê is the unit
polarization vector of the eld. The vectorial eld amplitude E r; t is generally a complex
vectorial quantity that has a magnitude, a phase, and a polarization. Other complex eld
quantities, such as Dr; t , Br; t , and Hr; t , can be similarly expressed. The space- and
time-dependent phase factor in (1.52) indicates the direction of wave propagation:
ik r it
for a wave propagating in the k direction;
ik r it for a wave propagating in the k direction.
1.5.1 Light Intensity

The light intensity, or irradiance, is the power density of the harmonic optical eld. It can be
calculated by time averaging the Poynting vector over one wave cycle:
T

1
S
Sdt 2Re E H ,
T
(1.53)
where Re means taking the real part. We can dene a complex Poynting vector:
S E H
(1.54)
so that
S S S S S ,
(1.55)
which has the same form as the relation between the real and complex elds dened in (1.40)
except that the Poynting vector in this relation is time averaged. In the case of a coherent
monochromatic wave, E H E H ; then, (1.55) can be written as S S S . The
13
light intensity, I, on a surface is simply the magnitude of the real time-averaged Poynting vector
projected on the surface:

I S n^ S S n^,
(1.56)
where n^ is the unit normal vector of the projected surface and I is in watts per square meter.
1.5.2 Fields in Momentum Space and Frequency Domain

For harmonic optical elds, it is often useful to consider the complex elds in the momentum
space and frequency domain dened by the following Fourier-transform relations:

Er; t exp ik r itdrdt,
Ek;
for > 0,
(1.57)
all r
Er; t
1
2 4

Ek; exp ik r it dkd:
(1.58)
0 all k
Note that Ek; in (1.57) is only dened for > 0; therefore, the integral for the time
dependence of Er; t in (1.58) only extends over positive values of . This is in accordance
with the convention we used to dene the complex eld Er; t in (1.40). All other space- and
time-dependent quantities, including other eld vectors and the permittivity and susceptibility
tensors, are transformed in a similar manner.
Through the Fourier transform, the convolution integrals in real space and time become
simple products in the momentum space and frequency domain. Consequently, we have
Pk; 0 k; Ek;
(1.59)
Dk; 0 1 k; Ek; k; Ek; :
(1.60)
and
Note that in the real space and time domain Pr; t and Dr; t are connected to Er; t through
convolution integrals in space and time, whereas in the momentum space and frequency domain
Pk; and Dk; are connected to Ek; through direct products.
1.6
POLARIZATION OF OPTICAL FIELDS
..............................................................................................................
The polarization state of an optical eld is determined by the vectorial nature of the optical eld.
It is characterized by the unit polarization vector ê of the complex electric eld expressed in
(1.52). Consider a monochromatic plane optical wave that has a complex electric eld of
Er; t E exp ik r it ê E exp ik r it,
(1.61)
14
where E is a constant independent of r and t, and ê is its unit vector. The polarization state of
the optical eld is characterized by the unit vector ê . The optical eld is linearly polarized, also
called plane polarized, if ê can be expressed as a constant, real vector. Otherwise, the optical
eld is elliptically polarized in general, and is circularly polarized in some special cases.
For the convenience of discussion, we take the direction of wave propagation to be the z
direction so that k k^z and assume that both E and H lie in the xy plane. Then, we have

E ^x E x ^y E y ^x jE x jeix ^y E y eiy ,
(1.62)
where E x and E y are space- and time-independent complex amplitudes, with phases x and y ,
respectively. The polarization state of the wave is completely characterized by the phase
difference and the magnitude ratio between the two eld components E x and E y :
y x ,
< ,
(1.63)
0 :
2
(1.64)
and
1
tan

E y
,
jE x j
Because only the relative phase matters, we can set x 0 and take E jE j to be real in the
following discussion. Then E from (1.62) can be written as
E ê E,
with ê ^x cos ^y ei sin :
(1.65)
Using (1.40), the space- and time-dependent real eld is

Ez; t 2E ^x cos cos kz t ^y sin cos kz t :
(1.66)
At a xed z location, say z 0, we see that the electric eld varies with time as
Et 2E ^x cos cos t ^y sin cos t :
(1.67)
1.6.1 Elliptic Polarization

In general, E x and E y have different phases and different magnitudes. Therefore, the values of
and can be any combination. At a xed point in space, both the direction and the magnitude of
the eld vector E in (1.67) can vary with time. Except when the values of and fall into one
of the special cases discussed below, the tip of this vector generally describes an ellipse, and the
wave is said to be elliptically polarized. Note that we have assumed that the wave propagates in
the positive z direction. When we view the ellipse by facing against this direction of wave
propagation, we see that the tip of the eld vector rotates counterclockwise, or left handedly, if
> 0; and it rotates clockwise, or right handedly, if < 0: Figure 1.3 shows the ellipse traced
by the tip of the rotating eld vector at a xed point in space. Also shown in the gure are the
relevant parameters that characterize elliptic polarization.
In the description of the polarization characteristics of an optical eld, it is sometimes
convenient to use, in place of and , a set of two other parameters, and , which specify
the orientation and ellipticity of the ellipse, respectively. The orientational parameter is the
15
Figure 1.3 Ellipse described by the tip of

the eld of an elliptically polarized
optical wave at a xed point in space.
Also shown are relevant parameters
characterizing the state of polarization.
The propagation direction is assumed to
be the positive z direction, and the ellipse
is viewed by facing against this
direction.
directional angle measured from the x axis to the major axis of the ellipse. Its range is taken to
be 0 < for convenience. The ellipticity is dened as
b
tan1 ,
a
,
4
4
(1.68)
where a and b are the major and minor semiaxes, respectively, of the ellipse. The plus sign for
> 0 is taken to correspond to > 0 for left-handed polarization, whereas the minus sign
for < 0 is taken to correspond to < 0 for right-handed polarization. The two sets of
parameters ; and ; have the following relations:
tan 2 tan 2 cos ,
(1.69)
sin 2 sin 2 sin :
(1.70)
Either set is sufcient to completely characterize the polarization state of an optical eld.
Elliptic polarization can be considered as the general polarization state for any combination of
and values, whereas linear polarization and circular polarization are special cases of elliptic
polarization for specic combinations of and values.
1.6.2 Linear Polarization

An optical eld is linearly polarized when 0 or for any value of . It is also characterized
by 0 and , if 0; or by 0 and , if . Clearly, the ratio E x =E y is
real in this case; therefore, linear polarization is described by a constant, real unit vector as
ê ^x cos ^y sin :
(1.71)
It follows that Et described by (1.67) reduces to

Et 2Eê cos t:
(1.72)
16

Figure 1.4 Field of a linearly polarized
optical wave.
The tip of this vector traces a line in space at an angle of with respect to the x axis, as shown in
Fig. 1.4.
1.6.3 Circular Polarization

An optical eld is circularly polarized when =2 or =2, and =4. It is
also

characterized by =4 or =4, and 0. Because =4, we have jE x j E y
p
E= 2. There are two different circular polarization states.
1. Left-circular polarization: For =2, also =4, the wave is left circularly polarized
if it propagates in the positive z direction. The complex eld amplitude in (1.65) becomes
^x i^y
E E p Eê ,
2
and Et described by (1.67) reduces to
p
Et 2E ^x cos t ^y sin t :
(1.73)
(1.74)
As we view against the direction of propagation ^z , we see that the eld vector Et rotates
counterclockwise at an angular frequency of . The tip of this vector describes a circle. This
is shown in Fig. 1.5(a). This left-circular polarization is also called positive helicity. Its unit
vector is
^x i^y
ê
p :
2
(1.75)
2. Right-circular polarization: For =2, also =4, the wave is right circularly
polarized if it propagates in the positive z direction. We then have
^x i^y
E E p Eê
2
(1.76)
and
Et
p
2E ^x cos t ^y sin t :
(1.77)
17
Figure 1.5 (a) Field of a left circularly polarized wave. (b) Field of a right circularly polarized wave.
The tip of this eld vector rotates clockwise in a circle, as shown in Fig. 1.5(b). This rightcircular polarization is also called negative helicity. Its unit vector is
^x i^y
ê
p :
2
(1.78)
As can be seen, neither ê nor ê is a real vector. Note that the identication of ê , dened
in (1.75), with left-circular polarization and that of ê , dened in (1.78), with right-circular
polarization are based on the assumption that the wave propagates in the positive z direction.
For a wave that propagates in the negative z direction, the handedness of these unit vectors
changes: ê becomes right-circular polarization, while ê becomes left-circular polarization.
1.6.4 Orthogonal Polarizations

Two polarizations are orthogonal if they are normal to each other. The unit polarization vector ê
can be either a real vector, for a linearly polarized wave, or a complex vector, for a circularly or
elliptically polarized wave. Each unit polarization vector is normalized to be a unit vector
according to the relation:
ê ê 1:
(1.79)
Two polarizations, ê 1 and ê 2 , are orthogonal if

ê 1 ê
2 0:
(1.80)
Note that normalization is not performed by ê ê 1, and orthogonality is not dened by

ê 1 ê 2 0.
EXAMPLE 1.3
Consider the two circularly polarized unit vectors ê and ê that are given in (1.75) and (1.78),
respectively. Show that they are normalized unit vectors that are orthogonal to each other.
18
Solution:
Using (1.79) for normalization, we nd that

^x i^y
^x i^y
^x i^y
^x i^y
p p
p p
ê ê
1
2
2
2
2
and
ê ê

^x i^y
^x i^y
^x i^y
^x i^y
p p
p p
1:
2
2
2
2
Therefore, both ê and ê are normalized unit vectors. Using (1.80) for orthogonality, we nd
that

^x i^y
^x i^y
^x i^y
^x i^y
p p
p p
ê ê
0
2
2
2
2
and
ê ê

^x i^y
^x i^y
^x i^y
^x i^y
p p
p p
0:
2
2
2
2
Therefore, ê and ê are normalized unit vectors that are orthogonal to each other. The two
circular polarizations are orthogonal to each other. Note that ê ê ê ê 0 6 1 and
ê ê ê ê 1 6 0, which can be easily veried.
1.7
OPTICAL FIELD PARAMETERS
..............................................................................................................
As stated in Section 1.1, an optical eld is characterized by the ve parameters of polarization
ê , magnitude jE j, phase E , wavevector k, and frequency :
Er; t E r; t exp ik r it
ê E r; t exp ik r it
(1.81)
ê jE r; t jeiE r;t exp ik r it,

where E ê E is the vectorial complex eld amplitude that contains the eld polarization ê and
the scalar complex eld amplitude E. The scalar complex eld amplitude E jEjeiE has a
magnitude of jE j and a phase of E . Note that in general, jE j and E can vary with space and
time, as indicated above in (1.81). Among the ve parameters, ê and k are vectors, while jE j,
E , and are scalars.
The unit polarization vector ê fully characterizes the polarization state of an optical eld. It
can be real, for linearly polarized light, or complex, for elliptically or circularly polarized light.
The details are discussed in the preceding section.
1.7 Optical Field Parameters
19
The magnitude jE j of the complex eld amplitude denes the strength of the optical eld. For
simplicity of discussion, consider a linearly polarized wave so that the unit polarization vector ê
is a real vector. Then the complex eld given in (1.81) yields the following real eld,

(1.82)
Er; t Er; t E r; t 2jE r; t jê cos k r t E r; t :
Therefore, under our denition of the complex eld through (1.40), the amplitude of the real
eld is 2jE r; t j. Note that this eld amplitude can be a function of space and time to describe
the modulation on the eld strength in space and time. It describes an envelope of the eld on
the optical carrier.
The phase E of the complex eld amplitude is the phase shift with respect to the space- and
time-varying phase factor, k r t. As seen in (1.82), the total phase of the eld is
r; t k r t E r; t :
(1.83)
In the case when E is a constant that is independent of both space and time, it has physical
meaning only when it is compared to a reference, such as the phase of another eld. An
unreferenced constant phase can be eliminated by redening the origin of the space or time
coordinate. Nevertheless, as expressed in (1.81) and (1.82), this phase can be a function of
space or time, or both: E r; t: The spatial dependence of E r; t leads to a shift of the
wavevector from the carrier wavevector k; the temporal dependence of E r; t leads to a shift
of the frequency from the carrier frequency :
The wavevector k denes the spatial variation and the propagation direction of the optical
carrier eld. Its value, k, known as the propagation constant or the wavenumber, is determined
by the wavelength, or equivalently the frequency, of the optical wave and the refractive index of
the medium:
2n ^ n ^
k kk^
k
k,
(1.84)
where n is the refractive index of the medium. From (1.82), it can be seen that k denes the
spatial variation of the optical carrier eld. The propagation direction of a wave is dened as
the direction normal to the wavefront of the wave, and a wavefront is the surface of a constant
phase: r; t constant: With r; t k r t E r; t from (1.83), the space-dependent
wavevector is
kr k E :
(1.85)
Thus, the space-dependent wave propagation direction can be found as

kr
k^r
:
k r
(1.86)
In the case when E is independent of space so that E 0, such as the case of a plane wave,
the wave propagates with a space-independent propagation constant k in a space-independent
propagation direction dened by the constant unit vector k^ k=k. In the case when E varies
across space so that E 6 0, such as the case of a spatially diverging or converging wave,
20
either one or both of the propagation constant kr and the propagation direction dened by
k^r kr=k r vary from one spatial location to another.
The frequency denes the temporal variation of the optical carrier eld. It is the optical
angular frequency that is related to the eld oscillation frequency as 2; has the unit
of hertz Hz while has the unit of radians per second rad s1 . As an optical wave
propagates through different media of different refractive indices, its wavelength, thus the
value of k, changes with the changing refractive indices, but its frequency remains unchanged.
The angular frequency of a wave is dened by the temporal variation of its phase. With
r; t k r t E r; t from (1.83), the angular frequency can be found as
t
E:
t
t
(1.87)
The frequency of the wave is the constant in the case when E is independent of time so that
E =t 0, such as the case of a monochromatic wave. In the case when E varies with time,
such as the case of a phase-modulated wave, the frequency t is a function of time with a shift
of E =t from the constant frequency .
Problems
1.1.1 At room temperature, diamond transmits optical waves of wavelengths longer than
227 nm but absorbs shorter wavelengths. What is the bandgap energy of diamond at
room temperature?
1.1.2 At room temperature, the bandgap energy of Ge is 0.66 eV. It absorbs photons of energies
above its bandgap and transmits those of energies below its bandgap. What is the cutoff
wavelength for light to be transmitted through a thick piece of pure Ge?
1.1.3 Find the wavelength and photon energy of a terahertz wave at a frequency of 5 THz.
1.1.4 The optical window for long-distance optical communications is at the 1.55 m wavelength. What are the optical frequency and the photon energy?
1.1.5 A red laser pointer emits a red beam of P 1 mW power at the 635 nm wavelength.
What are the photon energy, the photon momentum, and the photon ux of this beam? If
it illuminates a totally absorbing surface, what is the force exerted by the beam on the
absorbing surface? If it illuminates a totally reecting surface, what is the force exerted
by the beam on the reecting surface?
1.2.1 Verify that Maxwells equations and the continuity equation, given in (1.4)(1.8), are
invariant under (a) the transformation of space inversion, (b) the transformation of time
reversal, and (c) the simultaneous transformation of space inversion and time reversal.
1.4.1 Derive the optical wave equation given in (1.37) in the case when J 0 so that
Maxwells equations take the form of (1.11)(1.14). Show that in this case the optical
wave equation can be expressed in the form of (1.38).
Bibliography
21
1.4.2 In the case when a conduction current Jcond is explicitly separated from the D=t term so
that Maxwells equations take the form of (1.15)(1.18), rewrite the optical wave
equation given in (1.37) and that given in (1.38) to explicitly account for Jcond .
1.5.1 By taking the Fourier transform on the relation given in (1.50) between Pr; t and Er; t
in the real space and time domain, verify the relation given in (1.59) between Pk; and
Ek; in the momentum space and frequency domain.
1.6.1 As discussed in the text, any polarization state in the xy plane can be generally considered
as elliptic polarization represented by the unit polarization vector ê ^x cos ^y ei sin
given in (1.65) with proper choices of and for a particular polarization state. Because
the xy plane is a two-dimensional space, a basis set of unit polarization vectors consists of
two orthonormal vectors. Find the other unit polarization vector ê that forms a basis
together with ê .
1.6.2 The circularly polarized unit vectors ^
e and ê given in (1.75) and (1.78) are each
expressed in terms of the linearly polarized unit vectors ^x and ^y . Each pair form a basis
for representing any polarization state in the xy plane. Show that each of the linearly
polarized unit vectors ^x and ^y can be represented in terms of a linear superposition of two
circularly polarized components on the basis of ê and ê .
1.6.3 Express the general linearly polarized unit vector ^
e ^x cos ^y sin given in (1.71) as
a linear superposition of two circularly polarized components on the basis of the
circularly polarized unit vectors ê and ê given in (1.75) and (1.78), respectively.
Bibliography
Born, M. and Wolf, E., Principles of Optics: Electromagnetic Theory of Propagation, Interference and
Diffraction of Light, 7th edn. Cambridge: Cambridge University Press, 1999.
Fowler, G. R., Introduction to Modern Optics, 2nd edn. New York: Dover, 1975.
Iizuka, K., Elements of Photonics in Free Space and Special Media, Vol. I. New York: Wiley, 2002.
Jackson, J. D., Classical Electrodynamics, 3rd edn. New York: Wiley, 1999.
Liu, J. M., Photonic Devices. Cambridge: Cambridge University Press, 2005.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
2 - Optical Properties of Materials pp. 22-65
2
2.1
OPTICAL SUSCEPTIBILITY AND PERMITTIVITY
..............................................................................................................
The electric susceptibility, , and the electric permittivity, , of an optical medium characterize
the intrinsic response of the medium to an optical eld. They are respectively dened in (1.20) for
the relation between Pr; t and Er; t and in (1.21) for the relation between Dr; t and Er; t:
t
Pr; t 0
r r0 ; t t 0 Er0 ; t 0 dr0 dt 0,
(2.1)
all r0
t
Dr; t 0 Er; t Pr; t
r r0 ; t t0 Er0 ; t 0 dr0 dt0:
(2.2)
all r0
These relations can be expressed in terms of the complex eld:

t
Pr; t 0
r r0 ; t t0 Er0 ; t 0 dr0 dt 0
(2.3)
all r0
t
Dr; t 0 Er; t Pr; t
r r0 ; t t0 Er0 ; t 0 dr0 dt0:
(2.4)
all r0
The relations in the momentum space and frequency domain, obtained by taking the Fourier
transform on (2.3) and (2.4), are direct products, given in (1.59) and (1.60):
Pk; 0 k; Ek;
(2.5)
Dk; 0 1 k; Ek; k; Ek; :
(2.6)
The real-space and time-domain relations given in (2.1)(2.4) are convolution integrals over
real space and time. The convolution in time accounts for the fact that the response of a medium
to the stimulation by an electric eld is generally not instantaneous, or local, in time and does
not vanish for some time after the stimulation is over. Because time is unidirectional, causality
exists in physical processes. An earlier stimulation can inuence the property of a medium at a
later time, whereas a later stimulation does not have any effect on the medium at an earlier time.
Therefore, the upper limit in the time integral is t, not innity. By contrast, the convolution in
space accounts for the spatial nonlocality of the material response. Stimulating a medium at a
2.1 Optical Susceptibility and Permittivity
23
Figure 2.1 Nonlocal responses in (a) time and (b) space.
location r0 can result in a change in the property of the medium at another location r: For
example, the property of a semiconductor at one location can be changed by electric or optical
excitation at another location through carrier diffusion. There is generally no spatial causality
because space is not unidirectional; therefore, spatial convolution is integrated over the entire
space. Figure 2.1 shows the temporal and spatial nonlocality of responses to electromagnetic
excitations. The temporal nonlocality of the optical response of a medium makes the optical
property of the medium dependent on the optical frequency, a phenomenon known as frequency
dispersion, whereas the spatial nonlocality makes the optical property of the medium dependent
on the optical wavevector, a phenomenon known as momentum dispersion. The frequency
dispersion and the momentum dispersion of a medium are respectively characterized by the
frequency dependence and the momentum dependence of k; and k; . Because k;
and k; are respectively the Fourier transforms of r; t and r; t , it is clear that the
frequency dispersion and the momentum dispersion of a medium respectively originate from
the temporal nonlocality and the spatial nonlocality of its response to an optical stimulation.
The susceptibility tensor r; t and the permittivity tensor r; t of real space and time are
always real quantities though the optical elds in the real space and time domain can be
expressed either as real elds, as in (2.1) and (2.2), or as complex elds, as in (2.3) and (2.4).
This statement is true even when the medium exhibits an optical loss or gain. However, the
susceptibility tensor k; and the permittivity tensor k; in the momentum space and
frequency domain are generally complex. If an eigenvalue i of k; is complex, the
corresponding eigenvalue i of k; is also complex, and their imaginary parts have the
same sign because k; 0 1 k; . The signs of the imaginary parts of such eigenvalues tell whether the medium provides an optical gain or loss. In our convention, we write, for
example, i 0i i 00i in the frequency domain. Then, 00i > 0 indicates an optical loss or
absorption, while 00i < 0 represents an optical gain or amplication.
The fact that r; t and r; t are real quantities leads to the following symmetry relations
for the tensor elements of k; and k; :
ij k; ij k; and ij k; ij k; ,
(2.7)
24
which are called the reality condition. The reality condition implies that 0ij k; 0ij k;
and 00ij k; 00ij k; : The real and imaginary parts of ij k; have similar properties.
Therefore, the real parts of ij k; and ij k; are even functions of k and , whereas the
imaginary parts are odd functions of k and . If a tensor element ij k; or ij k; has any
constant term that is independent of k and , the constant term can only appear in its real part
because a constant value is an even function of k and . As a result, the imaginary part is always a
function of k or , or both. The optical loss, or gain, in a medium is associated with the imaginary
part of an eigenvalue of k; or k; ; consequently, a medium that absorbs or amplies light
is inherently dispersive. Any other effect that can be described by the imaginary part of an eigenvalue
of k; or k; is also inherently dispersive in either momentum or frequency, or both.
In addition to the nonlocality of medium response, it is also important to consider the
inhomogeneity of a medium, in both space and time. Spatial inhomogeneity exists in every
optical structure, such as an optical waveguide, where the optical property is a function of
space. Temporal inhomogeneity exists when the optical property of a medium varies with time,
for example, because of modulation by a low-frequency electric eld or by an acoustic wave.
The space and time variables characterizing nonlocality are relative space and time of the
medium response with respect to an optical stimulation, whereas those characterizing inhomogeneity are absolute space and time measured with respect to a reference point in space and a
reference point in time. When both response nonlocality and medium inhomogeneity are
considered, the response nonlocality is commonly characterized in the momentum space and
frequency domain as a function of k and by taking the Fourier transform on the relative space
and time, whereas the medium inhomogeneity is characterized in the real space and time
domain as a function of the absolute space and time variables r and t; therefore, k; ; r; t
and, correspondingly, k; ; r; t .
In a linear medium, changes in the wavevector of an optical wave, or coupling between
waves of different wavevectors, can occur only if the optical property of the medium in which
the wave propagates is spatially inhomogeneous such that k; ; r; t is a function of space.
Likewise, changes in the frequency of an optical wave, or coupling between waves of different
frequencies, are possible in a linear medium only if the optical property of the medium is time
varying such that k;;r;t varies with time. A change in the wavevector of an optical wave
^ as in the case of reection or
can take the form of a change in the wave propagation direction k,
diffraction of an optical wave, or in the propagation constant k through a change in the optical
wavelength, as in the case when a wave propagates from one part of the medium to another part
of a different refractive index. A change in the frequency of an optical wave results in the
generation of other frequencies or the conversion to a completely different frequency.
2.2
OPTICAL ANISOTROPY
..............................................................................................................
In general, both and are tensors because the P and D vectors are not necessarily parallel to
the E vector due to material anisotropy. In the case of an isotropic medium, both and reduce
to the scalars and , respectively. In the case of a linear anisotropic medium, both and are
second-order tensors. They can be expressed in the matrix form:
11
@
21
31
12
22
32
1
0
13
11
A
@
and 21
23
33
31
1
13
23 A:
33
12
22
32
25
(2.8)
Each of the relationships P 0 E and D E is carried out as the product between a

tensor and a column vector:
0
1
0
P1
11
@ P2 A 0 @ 21
P3
31
12
22
32
10 1
0 1 0
E1
D1
11
13
23 A@ E 2 A and @ D2 A @ 21
33
E3
D3
31
12
22
32
10 1
E1
13
23 A@ E 2 A: (2.9)
33
E3
In general, the matrices in (2.8) representing the and tensors are not diagonal when they are
expressed using an arbitrarily chosen coordinate system. When optical eld vectors are
projected on the axes of this coordinate system, a component of P or D does not necessarily
contain only the corresponding component of E but can also contain one or both of the other
two E components. For example, P1 and D1 are functions of E 2 or E3 , or both, unless
12 13 0, in which case 12 13 0 as well, because P1 0 11 E 1 12 E 2 13 E 3
and D1 11 E 1 12 E 2 13 E 3 .
Because and are physical quantities, they are diagonalizable matrices that can always be
diagonalized by a proper set of eigenvectors, yielding
0
1
@0
0
0
2
0
1
0
1
0
0 A and @ 0
3
0
0
2
0
1
0
0 A:
3
(2.10)
Here i and i are, respectively, the eigenvalues of and with corresponding eigenvectors ê i
such that
ê i i ê i and ê i i ê i , for i 1, 2, 3:
(2.11)
The characteristics of the eigenvalues i and i , as well as their eigenvectors ê i , depend on the
symmetry properties of and . The two matrices representing and have the same symmetry
properties because 0 1 , where 1 has the form of a 3 3 identity matrix when it
is added to the tensor. Therefore, and are diagonalized by the same set of eigenvectors.
When an optical eld is projected on these eigenvectors, each component of P or D depends
only on the corresponding component of E but not on the other two E components; that is,
Pi 0 i E i and Di i E i .
The three eigenvectors ê i dene the principal polarization states for proper decomposition of
optical eld vectors so that each component has a well-dened susceptibility i and permittivity
i . They are the principal normal modes of polarization satisfying the orthonormality condition:

1, for i j;
ê i ê j ij
(2.12)
0, for i 6 j:
As discussed in Section 1.6, a real eigenvector represents linear polarization, while a complex
eigenvector represents elliptic or circular polarization. The characteristics of these eigenvectors
26
are determined by the symmetry properties of and , which are determined by the properties
of the medium. Because and have the same properties and the same eigenvectors, only is
mentioned in the following discussion while all conclusions apply equally to .
2.2.1 Reciprocal Media

Nonmagnetic materials that are not subject to an external magnetic eld are reciprocal media.
In a reciprocal medium, the Lorentz reciprocity theorem of electromagnetics holds; consequently, the source and the detector of an optical signal can be interchanged for the same
function of an optical system. If such a material is not optically active, its optical properties are
described by a symmetric tensor: ij ji . The eigenvectors ê i of a symmetric tensor are
always real vectors. They can be chosen to be ^x , ^y , and ^z of a rectilinear coordinate system in
real space. This is true even when is complex.
1. If a nonmagnetic medium does not have any optical loss or gain, its tensor is Hermitian,
i.e., ij
ji . A symmetric Hermitian tensor is real and symmetric: ij ij ji ji : The
eigenvectors ê i are real vectors representing linear polarization states, and all three eigenvalues i have real values.
2. If a nonmagnetic medium has an optical loss or gain, its tensor is still symmetric but is
complex and non-Hermitian: ij ji but ij 6
e i are real
ji : Then, the eigenvectors ^
vectors representing linear polarization states, but at least one of the eigenvalues i is
complex. The sign of the imaginary part, 00i , indicates whether the medium has a loss or
gain for an ê i -polarized optical wave: 00i > 0 for a loss and 00i < 0 for a gain, as discussed
in Section 2.1 in terms of 00i .
3. If a nonmagnetic medium is optically active, it is still reciprocal although its tensor is not
symmetric. The eigenvectors ê i are complex vectors representing elliptic or circular polarization states, but the eigenvalues can be real, if the medium has no loss or gain, or complex,
if the medium has an optical loss or gain.
2.2.2 Nonreciprocal Media

Magnetic materials, and nonmagnetic materials that are subject to an external magnetic
eld, are nonreciprocal media. In such a medium, no symmetry exists when the source and
the detector of an optical signal are interchanged. The tensor describing the optical
properties of such a material is not symmetric: ij 6 ji . The eigenvectors ê i of a nonsymmetric matrix are generally complex vectors. Therefore, they are not ordinary coordinate axes
in real space.
1. For a magnetic medium that has no optical loss or gain, is Hermitian: ij
ji : The
eigenvalues i are real even though the eigenvectors ê i are complex vectors representing
elliptic or circular polarization states.
2. For a magnetic medium that has an optical loss or gain, is nonsymmetric and nonHermitian. The eigenvectors ê i and the eigenvalues i are generally complex.
27
EXAMPLE 2.1
At a given optical wavelength, the permittivity tensors of several optical materials are obtained
with respect to an arbitrary set of rectilinear coordinates in real space. From each of the
permittivity tensors shown below, identify each material as being (i) reciprocal or nonreciprocal
and (ii) lossless or lossy. Here lossless means having no loss or gain, and lossy means
having a loss or gain.
0
2:3 i0:3
C
A;
3:2 i0:1
3:4 i0:2 0:7 i0:1
B
A : 0 @ 0:7 i0:1
0
0
2:25
B
C : 0 @ i0:35
0
i0:35
2:20
0
C
0 A;
4:79
0:17
B
B : 0 @ 0:17
4:49
0
0
B
D : 0 @ 0:02 4:88
2:30
0
4:91 0:02
B
E : 0 @ 0:20 i0:18
0:20 i0:18
2:72
i0:22
C
0:05 A;
0:05
5:01
0:01
C
A;
0:01 4:58 i0:02
2:74
C
i0:22 A:
2:38
Solution:
The permittivity tensor of a reciprocal material is symmetric with ij ji , and that of a lossless
medium is Hermitian with ij
ji . The properties of each material can be determined by
examining its permittivity tensor using these two characteristics. A, nonreciprocal and lossy; B,
reciprocal and lossless; C, nonreciprocal and lossless; D, reciprocal and lossy; E, nonreciprocal
and lossless.
2.2.3 Linear Birefringence and Linear Dichroism

For a reciprocal material that is not optically active, the eigenvectors ê i of for proper
decomposition of optical eld vectors are real unit vectors representing three linearly polarized
principal normal modes. These three orthogonal real unit vectors can be labeled as ^x , ^y , and ^z ,
which can be used to dene the axes of a rectilinear coordinate system in real space.
Noncrystalline materials are generally isotropic, for which the choice of the orthogonal coordinate axes ^x , ^y , and ^z is arbitrary. For a crystal, these unique ^x , ^y , and ^z coordinate axes are called
the principal dielectric axes, or simply the principal axes, of the crystal. In the coordinate
system dened by these principal axes, is diagonalized with eigenvalues x , y , and z , known
as the principal permittivities. The properly decomposed components of D and E along these
axes have the following simple relations,
Dx x E x , Dy y E y , Dz z E z :
(2.13)
28
The values x = 0 , y = 0 , and z = 0 are the eigenvalues of the dielectric constant tensor, = 0 ,
and are called the principal dielectric constants. They dene three principal indices of
refraction:
r
r
r
x
y
z
nx
, ny
, nz
:
(2.14)
0
0
0
The propagation constants for the ^x , ^y , and ^z principal normal modes of polarization are,
respectively,
kx
nx
ny
nz
, ky
, kz
:
c
c
c
(2.15)
When is diagonalized, is also diagonalized along the same principal axes with corresponding principal dielectric susceptibilities, x , y , and z . The principal dielectric susceptibilities of
any dielectric material of no loss or gain always have real, positive values; therefore, the
principal dielectric constants of a lossless dielectric material are always greater than unity.
In an anisotropic crystal, the properly decomposed optical eld components in two different
principal normal modes of polarization dened by two different eigenvectors ê i and ê j have
different indices of refraction, i.e., ni 6 nj , and thus different propagation constants, i.e.,
k i 6 kj , when the eigenvalues i and j are different for the two polarization states. This
phenomenon is known as birefringence. A crystal that shows birefringence is a birefringent
crystal. Two principal normal modes of polarization experience different degrees of optical loss
or gain when their principal dielectric constants have different imaginary parts. This phenomenon is known as dichroism.
The birefringence of an anisotropic nonmagnetic crystal causes two different linearly polarized principal normal modes to propagate with different propagation constants; this is known as
linear birefringence. The dichroism of an anisotropic nonmagnetic crystal appears between two
linearly polarized principal normal modes; this is known as linear dichroism.
The state of polarization of an optical wave generally varies along its path of propagation
through an anisotropic crystal unless it is linearly polarized in the direction of a principal axis.
However, in an anisotropic crystal with nx ny 6 nz , a wave propagating in the z direction
does not see the anisotropy of the crystal because in this situation the x and y components of the
eld have the same propagation constant. This wave maintains its original polarization as it
propagates through the crystal. Evidently, this is true only for propagation along the z axis in
such a crystal. Such a unique axis in a crystal along which an optical wave can propagate with
an index of refraction that is independent of its polarization state is called the optical axis of
the crystal.
An anisotropic crystal that has only one distinctive principal index among its three principal
indices is called a uniaxial crystal because it has only one optical axis, which coincides with the
axis of the distinctive principal index of refraction. It is customary to assign ^z to this unique
principal axis such that nz is the distinctive index with nx ny 6 nz . The two identical principal
indices of refraction are called the ordinary index, no , and the distinctive principal index of
refraction is called the extraordinary index, ne . Thus, nx ny no and nz ne . The crystal is
called positive uniaxial if ne > no ; it is negative uniaxial if ne < no . A birefringent crystal of
29
three distinct principal indices of refraction is called a biaxial crystal because it has two optical
axes, neither of which coincides with any of the principal axes.
EXAMPLE 2.2
At the 1 m optical wavelength, the permittivity tensor of the KDP crystal represented in an
arbitrarily chosen Cartesian coordinate system dened by ^x 1 , ^x 2 , and ^x 3 unit vectors, with
^x 1 ^x 2 ^x 3 to satisfy the right-hand rule, is found to be
0
1
2:19
0
0:05196
A:
0
2:28
0
0@
0:05196
0
2:25
Find the principal indices of refraction and the corresponding principal axes ^x , ^y , and ^z in terms
of the coordinate axes ^x 1 , ^x 2 , and ^x 3 . Is KDP uniaxial or biaxial? If it is uniaxial, is it positive or
negative uniaxial?
Solution:
The given tensor is symmetric and Hermitian because KDP is a nonmagnetic dielectric crystal
that has a negligible optical loss at the 1 m optical wavelength. Diagonalization of the matrix
yields the eigenvalues 2.28, 2.28, and 2.16 for the principal dielectric constants. Thus, the
crystal is uniaxial. By convention we assign the distinctive dielectric constant of 2.16 to be
associated with the z principal axis. The principal indices of refraction and the corresponding
principal axes are
p
nx 2:28 1:51, ^x 0:500^x 1 0:866^x 3 ;
p
ny 2:28 1:51, ^y ^x 2 ;
p
nz 2:16 1:47, ^z 0:866^x 1 0:500^x 3 :
Note that ^x ^y ^z to satisfy the right-hand rule. The KDP crystal is negative uniaxial because
nx ny > nz so that no > ne .
The optical anisotropy of a crystal depends on its structural symmetry. Crystals are classied
into seven systems according to their symmetry. The linear optical properties of these seven
systems are summarized in Table 2.1. Some important remarks regarding the relation between
the optical properties and the structural symmetry of a crystal are as follows.
1. A cubic crystal does not have an isotropic structure although its linear optical properties are
isotropic. For example, most IIIV semiconductors, such as GaAs, InP, InAs, AlAs, etc., are
cubic crystals with isotropic linear optical properties. Nevertheless, they have well-dened
^ and ^c . They are also polar semiconductors, which have anisotropic
crystal axes, a^, b,
nonlinear optical properties.
2. Although the principal axes may coincide with the crystal axes in certain crystals, they are
^ and
not the same concept and are not necessarily the same. The crystal axes, denoted by a^, b,
30
Table 2.1 Linear optical properties of crystals

Crystal symmetry
Optical property
Cubic
Isotropic: nx ny nz
Trigonal, tetragonal, hexagonal
Uniaxial: nx ny 6 nz
Orthorhombic, monoclinic, triclinic
Biaxial: nx 6 ny 6 nz
^c , are dened by the structural symmetry of a crystal, whereas the principal axes, denoted by
^x , ^y , and ^z , are determined by the symmetry of . The principal axes of a crystal are
orthogonal to one another, but the crystal axes are not necessarily so.
2.2.4 Circular Birefringence and Circular Dichroism

For a nonreciprocal material or an optically active reciprocal material, the eigenvectors ê i of
for proper decomposition of optical eld vectors are generally complex unit vectors representing orthogonal elliptic polarization states, which may reduce to linear or circular polarization
states in particular cases. Optical activity is the phenomenon that a linearly polarized optical
wave remains linearly polarized but with its plane of polarization rotating about the direction of
propagation as it travels through a material. Natural optical activity that appears in a nonmagnetic reciprocal material not subject to a magnetic eld was rst discovered in quartz. It
occurs in many organic materials such as solutions of sugar or amino acids. Nonreciprocal
materials of interest in photonics can be magnetic with an intrinsic magnetization, M 0 , or
nonmagnetic but subject to a static or low-frequency external magnetic eld, H 0 ; these
materials exhibit magnetically induced optical activity for magneto-optics applications, such
as optical isolation and optical circulation.
Consider, for simplicity, a nonsymmetric that has only two off-diagonal elements:
0
n2
0 @ i
0
i
n2
0
1
0
0 A,
n2k
where n , nk , and can be real or complex. The eigenvalues are

0 n2 , 0 n2 , z 0 n2k ;
(2.16)
(2.17)
and the corresponding eigenvectors are

1
1
ê p ^x i^y , ê p ^x i^y , ^z :
2
2
(2.18)
The complex eigenvectors, ê and ê are respectively the left and right circularly polarized unit
vectors dened in (1.75) and (1.78). These two eigenvectors are complex unit vectors because
the tensor in (2.16) is not symmetric. If n , nk , and are all real, the eigenvalues are all real
31
because then is Hermitian. If n , nk , or is complex, the eigenvalues are also complex

because then is non-Hermitian.
It is clearly not possible to attach the meaning of the principal axes in real space to the
complex eigenvectors given in (2.18). Nonetheless, these eigenvectors still dene the principal
normal modes of polarization for proper decomposition of optical eld components:
D E , D E , Dz z E z :
(2.19)
Therefore, = 0 , = 0 , and z = 0 are the principal dielectric constants for the three normal
modes. They dene the following three principal indices of refraction:
n
q
q
n2 n
, n n2 n
, nz nk ,
2n
2n
(2.20)
where the approximate expansion of the square root is valid for =2n n : The propagation
constants for the principal normal modes of polarization are
k
n
n
nz
, k
, kz
:
c
c
c
(2.21)
When an optical wave propagates along the z axis, in either the positive z or the negative z
direction, the principal normal modes of polarization are the circularly polarized modes ê and
ê , which have different propagation constants k and k , respectively. This phenomenon that
the two circularly polarized modes have different propagation constants is called circular
birefringence. In the presence of an optical loss or gain, both n and n become complex no
matter whether the optical loss or gain is characterized by the nonzero imaginary part of a
complex n or , or both. When the imaginary parts of n and n have different values, the two
circularly polarized normal modes experience different degrees of optical loss or gain. This
phenomenon is called circular dichroism, as distinct from the linear dichroism between two
linearly polarized modes.
Circular birefringence caused by the magneto-optic effect in a magnetic material or in a
nonmagnetic material subject to a magnetic eld is known as magnetic circular birefringence.
Circular birefringence in a nonmagnetic reciprocal material that has natural optical activity is
known as natural circular birefringence. Circular dichroism caused by a loss or gain associated
with the magneto-optic effect in a magnetic material or in a nonmagnetictic material subject to a
magnetic eld is known as magnetic circular dichroism. Circular dichroism due to a loss or
gain in a nonmagnetic reciprocal material that has natural optical activity is known as natural
circular dichroism.
The similarities between the two phenomena of natural optical activity and magnetically
induced optical activity are that both have circularly polarized normal modes and both can
cause circular birefringence and circular dichroism. In both cases, the plane of polarization of a
linearly polarized wave can be rotated as the wave travels through the material. The fundamental difference between the two phenomena is that natural optical activity is reciprocal, so that a
round trip through the medium cancels the polarization rotation, whereas magnetically induced
optical activity is nonreciprocal, so that a round trip through the medium does not cancel but
doubles the polarization rotation. In the simplest case of the nonsymmetric tensor of the form
32
given in (2.16), natural optical activity can be described by k^ ^z , which depends on the
propagation direction k^ and on a characteristic constant of the medium, whereas magnetically
induced optical activity is described by M 0z or H 0z , which is a linear function of M 0z or
^ Whereas all materials exhibit magneticH 0z but is independent of the propagation direction k.
ally induced optical activity in the presence of a magnetization or a magnetic eld, natural
optical activity cannot exist in centrosymmetric materials. In an otherwise centrosymmetric
medium, such as a liquid, the addition of molecules, such as sugar molecules, that cause optical
activity breaks the centrosymmetry of the system.
2.3
RESONANT OPTICAL SUSCEPTIBILITY
..............................................................................................................
Frequency dispersion of a medium is caused by the fact that the response of the medium to an
optical eld does not end instantaneously but relaxes over time after the optical stimulation. The
root of the optical response is the interaction between the electrons in the material and the
optical eld. The electrons in a material can be either valence electrons, which are localized
bound electrons, or conduction electrons, which are nonlocalized free electrons. The electrons
in atoms and molecules are bound electrons that have discrete energy levels. In a condensed
matter, such as a solid material, the electronic states form energy bands. Separate impurity
atoms or molecules that are embedded in an insulator or a semiconductor as dopants can have
discrete energy levels inside an energy band or in the gap between two energy bands of the
host solid. The electrons in a valence band of a solid material, which can be an insulator,
a semiconductor, or a metal, are bound electrons. An electron in a conduction band of a
semiconductor or a metal behaves like a free electron, but it has an effective mass that is
determined by the structure of the conduction band and is different from the electron mass in
free space. A hole in a valence band of a semiconductor behaves like a free positive charge
carrier with an effective mass that is determined by the structure of the valence band.
Resonant interaction involves the transition of an electron, stimulated by an optical eld,
between two discrete energy levels or between two energy bands. Nonresonant interaction can
take place between an electron in a conduction band, or a hole in a valence band, and an optical
eld while the electron or hole stays in the same band without making a transition to another
band. Both resonant and nonresonant interactions contribute to material dispersion, but their
characteristics are different. In this section, the salient characteristics of resonant interactions
involving valence electrons are considered. The dispersion characteristics of nonresonant
interactions involving free charge carriers are considered in the next section.
A given material generally has many transition resonances across the electromagnetic
spectrum; each resonance is characterized by a resonance frequency, 0 , and a relaxation rate, .
A resonant interaction involves two separate energy states: a lower energy state j1i of energy
E1 and population density N 1 , and an upper energy state j2i of energy E2 and population
density N 2 . The energy states j1i and j2i are discrete energy levels in an atom or molecule, or
specic states in different energy bands of a condensed matter. The population densities N 1 and
N 2 are the number of electrons per unit volume in states j1i and j2i, respectively. When a
material is in thermal equilibrium with its background environment, i.e., in its normal state, the
33
Figure 2.2 Discrete energy levels for

resonant interaction.
laws of population distribution require that its lower energy level be more populated than its
upper energy level such that N 1 > N 2 : Population inversion with N 2 > N 1 is possible only
when a material is sufciently pumped to bring it far away from thermal equilibrium.
Because the focus of this section is on the salient features of resonant susceptibility, we
consider the simple case of the resonant interaction involving two discrete energy levels as
shown in Fig. 2.2. The transition resonance frequency is determined by the energy separation of
the two levels,
0 21
E2 E1
,
(2.22)
and the relaxation rate is the total susceptibility relaxation rate contributed by various relaxation
mechanisms involving the two energy levels,
21 :
(2.23)
Note that the susceptibility relaxation rate 21 discussed here is the rate of relaxation of the
optical polarization induced by the optical eld, which is generally different from the population decay rates of the two energy levels. The details of such differences are discussed in
Section 7.1.
The resonant susceptibility associated with two discrete energy levels can be obtained by
quantum mechanical calculation through the density matrix formalism. Quantum mechanical
calculation allows the accurate treatment of the susceptibility as a tensor; it can be extended to a
complex system that has multiple energy levels or energy bands. A classical Lorentz model that
describes the single-resonance system as a one-dimensional damped oscillator is often used to
obtain the key features of the resonant susceptibility. (See Problem 2.3.1.)
The quantum mechanical result of the resonant susceptibility tensor as a function of the
response time t with respect to an optical excitation at time zero is
2N 1 N 2 p12 p12 21 t
e
sin 21 t H t
0
8
< 2N 1 N 2 p12 p12 21 t
sin 21 t, t 0;
e
0
: 0,
t < 0;
res t; 21
(2.24)
where the Heaviside step function H t has the values of H t 1 for t 0 and H t 0 for
t < 0; and p12 h1jp^j2i is the matrix element of the electric-dipole operator p^ e^
x for the
transition between states j1i and j2i, where e is the electronic charge and x^ is the displacement
34
operator. We consider the eigenvalue of the susceptibility tensor for a normal mode of
polarization ê . For simplicity, we express it in terms of 0 and by applying (2.22) and (2.23):
8
< 2Np2 t
2
2Np t
(2.25)
res t; 0
e sin 0 t H t 0 e sin 0 t, t 0;
:
0
0,
t < 0;
where N N 2 N 1 is the population difference between the upper and the lower energy
levels, and p p12 ê is the electric-dipole strength of the resonant transition. Note that
res t 0 for t < 0 because a medium can respond only after, but not before, an excitation.
This is the causality condition, which applies to all physical systems.
The Fourier transform of (2.25) to the frequency domain yields
res ; 0 res t; 0 eit dt

Np2
1
1
0 0 i 0 i
Np2
1
:

0 0 i
(2.26)
In (2.26), we have taken the so-called rotating-wave approximation by keeping only the
resonant term that contains 0 in the denominator and dropping the nonresonant term that
contains 0 in the denominator because for a frequency in the optical spectral region it is
always valid that 0 j 0 j near resonance. The real and imaginary parts of this
resonant susceptibility are
0res
Np2
0
Np2
00
,

,
res
2
2
0 0
0 0 2 2
(2.27)
which are plotted in Fig. 2.3.
Figure 2.3 Real and imaginary parts, 0res and 00res , respectively, of susceptibility for a medium that
shows (a) a loss and (b) a gain near a resonance frequency at 0 .
35
The imaginary part 00res of the resonant susceptibility has a Lorentzian lineshape, which has
a full width at half-maximum (FWHM) of 2. In terms of the frequency =2, the
lineshape has a center frequency at 0 0 =2 and a FWHM of =2 =. The sign
of 00res depends on that of N. When the material is in its normal state in thermal equilibrium
with the surrounding, the lower energy level is more populated than the upper level so that
N < 0; thus, the material shows an optical loss near resonance with 00res > 0. This characteristic results in the absorption of light at the resonance frequency 0 when the material is
in thermal equilibrium with its background environment. When population inversion is accomplished so that N > 0, the material shows an optical gain with 00res < 0, resulting in the
amplication of light at 0 due to stimulated emission, such as in the case of an optical
amplier or a laser. Note that both 0res and 00res are proportional to N. Therefore, when
00res changes sign with N, 0res also changes sign. When 00res > 0, for N < 0, 0res
is positive for < 0 and negative for > 0 , as is shown in Fig. 2.3(a); when 00res < 0, for
N > 0, 0res is negative for < 0 and positive for > 0 , as is shown in Fig. 2.3(b).
A medium generally has many resonance frequencies, each corresponding to an absorption
frequency for the medium in its normal state. The permittivity of the medium due to all bound
electrons is the sum of all resonance susceptibilities:
"
#

X
X N i p2
1
1
i
: (2.28)
res ; 0i 0

bound 0 1
0i ii 0i ii
i
i
Note that the rotating-wave approximation is not taken in the above expression because a frequency
near one resonance frequency can be very far away from another resonance frequency. For this
reason, the rotating-wave approximation is not generally valid across a broad spectrum. The
characteristics of the real and imaginary parts of bound for a medium in its normal state as a
function of over a spectral range covering a few resonances are illustrated in Fig. 2.4. Some
important dispersion characteristics of res and bound are summarized below.
1. It can be seen from Fig. 2.3(a) that for a material in its normal state, 0res < 0 is always
larger than 0res > 0 . Therefore, around any single resonance frequency, 0bound at
any frequency on the low-frequency side has a value greater than that at any frequency on
the high-frequency side.
2. From (2.28), it is found that
bound 0 0
X N i p2 20i
i
> 0 and bound 0 :
20i 2i
i
(2.29)
We see that because N i < 0 for a material in thermal equilibrium, the DC susceptibility
contributed by all bound electrons in a material is real and positive so that the DC
permittivity bound 0 due to all bound electrons is always real and larger than 0 . At a very
high frequency that is well above all resonance frequencies, such as one in the hard X-ray
region, all bound electrons stop responding to the high-frequency eld so that the medium
behaves much like free space to the high-frequency eld; thus bound 0 . At a nite
frequency of that is far away from any resonance frequency, 00bound 0 so that
bound 0bound and bound 0 > bound . Therefore, the permittivity of an insulator,
36
Figure 2.4 Real and imaginary parts of bound as a function of for a medium in its normal state over a spectral
range covering a few resonance frequencies.
which does not have free charge carriers, at a frequency that is far away from all resonances
is always smaller than its DC permittivity.
3. A medium is said to have normal dispersion in a spectral region where 0 increases with
frequency so that d 0 =d > 0. It is said to have anomalous dispersion in a spectral region where
0 decreases with increasing frequency so that d 0 =d < 0. Because dn=d and d 0 =d
have the same sign, the index of refraction also increases with frequency in a spectral region of
normal dispersion and decreases with frequency in a spectral region of anomalous dispersion.
4. It can be seen from Fig. 2.4 that when a material is in its normal state in thermal equilibrium,
normal dispersion appears everywhere except in the immediate neighborhood within the
FWHM of a resonance frequency, where anomalous dispersion occurs. This characteristic
can be reversed near a resonance frequency where resonant amplication, rather than
absorption, takes place due to population inversion.
37
5. In most materials that are transparent in the visible spectral region, such as glass and water,
normal dispersion appears in the visible region and may extend to the near-infrared and nearultraviolet regions.
Only transitions between discrete energy levels are considered above. In a solid material
where electronic states form energy bands, transitions between separate energy bands, called
band-to-band transitions or interband transitions, contribute to the resonant susceptibility of the
material. The susceptibility is found by integrating over the electronic states in the two bands
involved in the transitions; the integration takes into account the population distribution probability of electrons in each band. The general concepts described above are still valid, except that
the susceptibility contributed by band-to-band transitions does not show the characteristic sharp
resonance peaks of transitions between discrete energy levels seen in Figs. 2.3 and 2.4.
EXAMPLE 2.3
An atomic absorption spectral line associated with an optical transition from the ground state to
an excited state is found to appear at a center wavelength of 800 nm with a FWHM spectral
width of 0:48 nm. Find the energy of the excited state above the ground state. Find the
resonance frequency and the polarization relaxation rate associated with this transition. Where
can we nd anomalous dispersion caused by this atomic transition when the atoms are in their
normal state in thermal equilibrium with the surrounding?
Solution:
The energy of the excited state above the ground state is the photon energy of the absorption
wavelength at 800 nm:
E2 E1 h
1239:8
1239:8
nm eV
eV 1:55 eV:
800
The resonance frequency is

c 3 108 m s1
0
375 THz ;
800 109 m
0 20 2:36 1015 rad s1 :
Because , we can use the approximation =0 = to nd that

0:48
0
375 THz 225 GHz:
800
Thus, the relaxation rate is

7:07 1011 s1 :
When the atoms are in their normal state in thermal equilibrium with the surrounding, the
ground state is more populated than the excited state. In this situation, anomalous dispersion
caused by this transition is found within the FWHM of the spectral line, in the wavelength
range of
=2 800
0:24 nm, corresponding to the frequency range of 0
=2
375 THz
112:5 GHz.
38
2.4
OPTICAL CONDUCTIVITY AND CONDUCTION SUSCEPTIBILITY
..............................................................................................................
An electron in a conduction band of a semiconductor or a metal behaves like a free electron
with an effective mass, while a hole in a valence band of a semiconductor behaves like a free
positive charge carrier with an effective mass. The response of these free charge carriers to an
optical eld can be treated using quantum mechanics by considering induced transitions within
a band, known as intraband transitions, or using a classical Drude model by considering an
induced conduction current J cond as discussed in Section 1.1. Because the quantum mechanical
approach involves the consideration of the band structure, we use the classical Drude model for
simplicity. In this classical approach, the effective mass m of the charge carrier accounts for
the effect of the energy band; clearly, the value of m depends on the structure of the energy
band on which the charge carrier lies.
In the Drude model, conduction electrons, and holes in a semiconductor, are treated as
independent free charge carriers. The momentum, p, of a charge carrier is driven by the force
of an electric eld, F qE, and is damped by random collisions with the ions of the medium,
characterized by an average momentum relaxation time . Therefore,
dp
p
qE ,
dt
(2.30)
where q e for an electron and q e for a hole. The conduction current density is
J cond Nqv
Nqp
,
m
(2.31)
where N is the density of the free charge carriers. By combining (2.30) and (2.31), we have the
equation for the conduction current that is induced by an electric eld:
dJ cond J cond Ne2
E,
dt
(2.32)
where q2 e2 is used for the charge carriers to be either electrons or holes.

The general solution of (2.32) can be expressed as a convolution integral:
t
J cond t
t t 0 Et 0 dt 0,
(2.33)
where
8
2
< Ne t=
Ne t=
e ,
t e H t
m
:
m
0,
2
for t 0;
(2.34)
for t < 0:
Note that Jcond t and Et are real elds in the real space and time domain. The relation in
(2.33) denes the optical conductivity t in the real space and time domain, as seen in (2.34).
For simplicity, their spatial dependence is ignored. In terms of the complex eld,
t
Jcond t
t t 0 Et0 dt 0,
39
(2.35)
where t is the same as that in (2.34). The frequency domain relation is obtained by taking the
Fourier transform on (2.35):
Jcond E,
(2.36)
where
teit dt
Ne2 1
:
m 1 i
(2.37)
This frequency-dependent optical conductivity, also known as the AC conductivity, can be

expressed in terms of the DC conductivity:

0
,
1 i
(2.38)
Ne2
:
m
(2.39)
where 0 is the DC conductivity,

0
As discussed in Section 1.1, there are two alternative, but equivalent, ways to described the
optical response of free charge carriers: (1) by treating it as part of the total susceptibility and
total permittivity in the total displacement D, as in (1.12); or (2) by treating it as an optical
conductivity through an explicit conduction current Jcond , as in (1.16). The discussion above
follows the second alternative, which allows us to nd the optical conductivity in (2.38). By
equating the two alternative approaches, the conduction susceptibility, cond , due to the free
charge carriers can be found.
Equating (1.12) and (1.16) but expressing them in complex elds, we have
D Dbound
Jcond :
t
t
(2.40)
Converting this relation to the frequency domain, we nd

iD iDbound Jcond :
(2.41)
By using the relations D E, Dbound bound E, and Jcond

E from (2.36), we nd the total permittivity that includes all contributions from bound
and free charges in a material:
bound
i
0
bound
,
(2.42)
where bound is the permittivity from bound charges discussed in Section 2.3. Therefore, we
can identify the conduction susceptibility due to the free charge carriers:
40

Figure 2.5 Real and
imaginary parts, 0cond and
00cond , respectively, of the
conduction susceptibility,
normalized to 0= 0 , as a
function of .
cond
i
0
1

:
0
0 i
(2.43)
The real and imaginary parts of this conduction susceptibility are

0cond
0
1
0
1
, 00cond
,
2
2
2
0 1
0 2 1
(2.44)
which are plotted in Fig. 2.5.

At an optical frequency that is far away from any resonance transition frequency,
00
bound 0 so that bound 0bound . In this case, the real and imaginary parts of the
total permittivity given in (2.42) are
0 bound
0
,
1
2 2
00
0
:
2 2 1
(2.45)
We nd that due to the effect of the conduction electrons, the real part of the total susceptibility
vanishes, i.e., 0 0, at the frequency p , known as the plasma frequency:
2p
0
bound
1
Ne2
1
0
Ne2

:
2 bound m 2 bound bound m
(2.46)
Because it is almost always true that p 1 for most conducting materials, the plasma
frequency is generally dened by neglecting the 1= 2 term in (2.46). The permittivity bound
in (2.46) is taken to be a constant that has the value in the frequency range of interest. In terms
of 2p , the total permittivity can be expressed as
"
bound 1
2p 2
i
"
bound 1
2p 2
2 2 1
2p 2
2 2 1
The real and imaginary parts of this total permittivity are plotted in Fig. 2.6.
#
:
(2.47)
41
Figure 2.6 Real and imaginary parts, 0 and 00 , respectively, of the total permittivity, normalized to
bound , as a function of frequency showing (a) low-frequency characteristics and (b) high-frequency
characteristics. The value of p 10 is used for this plot.
Some important characteristics are summarized below.

1. For all frequencies, the real part 0cond of the conduction susceptibility is negative, and the
imaginary part 00cond is positive. Thus the conduction susceptibility only contributes to
optical loss and never contributes to optical gain, and it makes possible a negative real part
for the permittivity, as discussed below.
2. At low frequencies for which 1, 0 = bound 1 2p 2 approaches a constant
but 00 = bound 2p = becomes inversely proportional to frequency so that j 00 j
j 0 j. Then,
!
2
p
:
(2.48)
bound 1 2p 2 i
These low-frequency characteristics are seen in Fig. 2.6(a).

3. At high frequencies for which 1, 0 = bound 1 2p =2 and 00 0 so that
j 0 j j 00 j. Then,
!
2p
bound 1 2 :
(2.49)
These high-frequency characteristics are seen in Fig. 2.6(b).

4. At all frequencies, the imaginary part of the permittivity is positive because 00cond is
positive: 00 > 0 for all .
5. At frequencies below the plasma frequency, the real part of the permittivity is negative:
0 < 0 for < p . This leads to high reectivity on the surface and low penetration of
the optical eld in the medium, which are the common properties of metallic surfaces.
42
6. At frequencies above the plasma frequency, the real part of the permittivity is positive while
the positive imaginary part decreases quickly with increasing frequency. Consequently, the
contribution of the conduction susceptibility quickly diminishes. Then the medium behaves
optically like an insulator, allowing a high-frequency optical eld to penetrate through with
little attenuation except when the optical frequency comes close to a transition resonance.
7. For a perfect conductor, only free conduction electrons contribute to the optical response
so that the permittivity has no contribution from bound electrons; thus, bound 0 . For this
reason, it is a good approximation to take bound 0 for a metal that has a high conductivity, such as Ag, Au, Cu, and Al. For such a metal, it is also a good approximation to take the
effective electron mass as the free electron mass, m m0 , when applying (2.46).
8. For a semiconductor where electrons and holes both contribute to the conduction susceptibility, the total permittivity is
bound
e 0
h 0

,
e i h i
(2.50)
where
e 0
N e e2 e
N h e2 h
and
:
h
m
m
e
h
(2.51)
The plasma frequency is found at 0 0 to be

2p
e 0
1
h 0
1
N e e2
N h e2
2
2
:
bound m
bound e e bound h h bound m
e
h
(2.52)
EXAMPLE 2.4
Silver is one of the best conductors such that the free-electron Drude model describes its optical
properties reasonably well. In this model, the free electron density of Ag is found to be N
5:86 1028 m3 . The DC conductivity of Ag at T 273 K is 0 6:62 107 S m1 . Find
the plasma frequency p and the relaxation time for Ag at T 273 K. Also nd the cutoff
optical frequency p and the cutoff wavelength p . For what optical wavelengths is Ag expected
to be highly reective? For what wavelengths is it expected to become transmissive?
Solution:
For Ag, it is a good approximation to take bound 0 and m m0 . Then, using (2.46), we
nd that
2p

2
5:86 1028 1:6 1019
Ne2
rad2 s2 1:86 1032 rad2 s2

12
31
0 m
8:854 10
9:1 10
) p 1:36 1016 rad s1 ,
0
6:62 107
s1 4:02 1014 s 40:2 fs:

0 2p 8:854 1012 1:86 1032
43
The cutoff frequency and cutoff wavelength are those at the plasma frequency:
p
p
2:17 PHz,
2
c
138 nm:
p
Ag is highly reective for > p , corresponding to < p ; it becomes transmissive for < p ,
corresponding to > p .
EXAMPLE 2.5
GaAs is a direct-gap semiconductor that has an electron effective mass of m
e 0:067m0 and a
hole effective mass of m
0:52m
,
where
m
is
the
mass
of
a
free
electron.
Its low-frequency
0
0
h
dielectric constant is 10.9. Find the plasma frequency, the cutoff frequency, and the cutoff
wavelength for (a) an n-type GaAs sample that has an electron density of N e 1 1024 m3 ,
(b) a p-type GaAs sample that has a hole density of N h 1 1024 m3 , and (c) a GaAs sample
that is injected with an equal electron and hole density of N e N h 1 1024 m3 .
Solution:
As we will see below, the plasma frequency is much lower than the bandgap frequency of GaAs,
which corresponds to a wavelength of g 871 nm. Therefore, the low-frequency dielectric
constant is used for bound 10:9 0 . Then, the plasma frequency is found using (2.52).
(a) For the n-type GaAs with N e 1 1024 m3 , the hole density is negligibly small so that
2p
N e e2
bound m
e

2
1 1024 1:6 1019
rad2 s1 4:35 1027 rad2 s2 :

10:9 8:854 1012 0:067 9:1 1031
Therefore, p 6:60 1013 rad s1 , p 10:5 THz, and p 28:6 m.
(b) For the p-type GaAs with N h 1 1024 m3 , the electron density is negligibly small so that
2p
N h e2
bound m
h

2
1 1024 1:6 1019
rad2 s1 5:60 1026 rad2 s2 :

10:9 8:854 1012 0:52 9:1 1031

(c) For the injected GaAs with N e N h 1 1024 m3 ,
2p
N e e2
N h e2
bound m
bound m
e
h
4:35 1027 rad2 s2 5:60 1026 rad2 s2 4:91 1027 rad2 s2 :
44
2.5
KRAMERSKRONIG RELATIONS
..............................................................................................................
It can be seen from the above discussion that the real and imaginary parts of , or those of
, are not independent of each other. The susceptibility of any physical system has to satisfy
the causality requirement in the time domain. This requirement leads to a general relationship
between 0 and 00 in the frequency domain:
2 0 00 0 0
2
0 0
00
d
,

d0 ,
P
P 0 2
2
0 2 2
0
(2.53)
where the principal values are taken for the integrals. These relations are known as the Kramers
Kronig relations. They are valid for any that represents a physical process, such as the
resonant susceptibility res in Section 2.3 and the conduction susceptibility cond in
Section 2.4. Therefore, once the real part of for any physical process is known over the
entire spectrum, its imaginary part can be found, and vice versa. Note that the relations in (2.53)
are consistent with the fact that 0 is an even function of and 00 is an odd function of ,
as discussed in Section 2.1. The contradiction to this statement seen in (2.27) for 0res and
00res is only apparent but not real; it is caused by the rotating-wave approximation taken in
(2.26). There is no contradiction when the rotating-wave approximation is removed and exact
expressions are used for 0res and 00res . For 0cond and 00cond given in (2.44), it is clear
that 0cond is an even function of and 00cond is an odd function of .
2.6
EXTERNAL FACTORS
..............................................................................................................
The optical property of a material is inuenced by external factors, such as an electric eld, a
magnetic eld, an acoustic wave, an injection current, a pressure, or a temperature change. The
dependence of the optical property on an externally controllable factor allows the active control
and modulation of an optical wave; this is the basis for active photonic devices. On the other
hand, this characteristic is passively used in a photonic sensor which optically senses the
parameter that causes a change in the optical property of the sensor material. Some of the
major effects that cause changes in the permittivity of an optical material are discussed below.
2.6.1 Electro-optic Effect

The optical property of a dielectric material can be changed by a static or low-frequency electric
eld E0 through an electro-optic effect. The result is a eld-dependent susceptibility and thus a
eld-dependent permittivity:
P; E0 0 ; E0 E 0 E 0 ; E0 E
(2.54)
D; E0 ; E0 E E ; E0 E,
(2.55)
and
45
where the eld-independent susceptibility ; E0 0 and permittivity

; E0 0 represent the intrinsic linear response of the material at the optical frequency ,
whereas ; E0 and ; E0 represent the changes induced by the electric eld E0 . We
can dene the electric-eld-induced polarization change as P; E0 0 ; E0 E to
express the total eld-dependent displacement as D; E0 D P; E0 . The total
permittivity of the material in the presence of the electric eld is then
; E0 ; E0 0 ; E0 :
(2.56)
In the discussion of electro-optic effects, it is necessary to introduce the relative impermeability tensor, which is the inverse of the dielectric constant tensor:
1
:
(2.57)
0
The reason for considering the relative impermeability tensor is historical because electro-optic
effects are traditionally not expressed as ; E0 or ; E0 but are dened in terms of the
changes in the elements of the relative impermeability tensor as E0 E0 , which is
expanded as
X
X
ij E0 ij ij E0 ij
r ijk E 0k
sijkl E 0k E 0l ,
(2.58)
k
k, l
where the rst term ij is the eld-independent component, the elements of the third-order rijk
tensor are the linear electro-optic coefcients known as the Pockels coefcients, and those of
the fourth-order sijkl tensor are the quadratic electro-optic coefcients known as the electrooptic Kerr coefcients. The rst-order electro-optic effect characterized by the linear dependence of ij E0 on E0 through the coefcients r ijk is called the linear electro-optic effect, also
known as the Pockels effect. The second-order electro-optic effect characterized by the quadratic eld dependence through the coefcients sijkl is called the quadratic electro-optic effect,
also known as the electro-optic Kerr effect. In (2.58), indices i and j are associated with optical
elds, whereas indices k and l are associated with the low-frequency applied eld. Because the
tensor of a nonmagnetic electro-optic material is symmetric, the tensor as dened in (2.57)
is also symmetric; thus ij ji and ij ji . The symmetric indices i and j can be
contracted to reduce the double index ij to a single index using the index contraction rule:
ij :
or ij :
:
11 22 33 23, 32 31, 13 12, 21

xx yy zz yz, zy zx, xz xy, yx
1 2 3
4
5
6
Using index contraction, (2.58) is expressed as

X
X
E0 E0
r k E 0k
skl E 0k E 0l ,
k
k, l
(2.59)
(2.60)
where 1, 2, . . . , 6 and k, l 1, 2, 3 or x, y, z:
The Pockels effect does not exist in a centrosymmetric material, which is a material that
possesses inversion symmetry. The structure and properties of such a material remain
46
unchanged under the transformation of space inversion, which changes the signs of all
rectilinear spatial coordinates from x; y; z to x; y; z, and the signs of all polar vectors.
As discussed in Section 1.1, an electric eld vector is a polar vector that changes sign under the
transformation of space inversion. By simply considering the effect of space inversion, it is
clear that the electro-optically induced changes in the optical property of a centrosymmetric
material are not affected by the sign change in the applied eld from E0 to E0 , meaning that
ij E0 ij E0 . As can be seen from (2.58), this condition requires that the Pockels
coefcients r ijk vanish, but it does not require the electro-optic Kerr coefcients sijkl to vanish.
Consequently, the Pockels effect exists only in noncentrosymmetric materials, whereas the
electro-optic Kerr effect exists in all materials, including centrosymmetric ones. Structurally
isotropic materials, including all gases, liquids, and amorphous solids such as glass, show no
Pockels effect because they are centrosymmetric.
The majority of electro-optic devices are based on the Pockels effect because the electro-optic
Kerr coefcients are generally very small. For this reason, practical electro-optic applications
usually require noncentrosymmetric crystals in order to make use of the Pockels effect. Among
the 32 point groups in the 7 crystal systems, 11 are centrosymmetric, and the remaining 21 are
noncentrosymmetric. It is important to note that the linear optical property of a crystal is
determined only by its crystal system, as mentioned in Section 2.2 and summarized in Table 2.1,
but the electro-optic property further depends on its point group.
Because the electro-optic coefcients are traditionally dened through the changes in the
relative impermeability tensor, as expressed in (2.58), the eld-induced changes in the permittivity tensor have to be found through the relation between ij E0 and ij E0 . Using the
relation = 0 1, the relation between and can be found:

1
1
and :
0
0
(2.61)
As discussed in Section 2.2, the intrinsic permittivity tensor of a crystal in the absence of
the electric eld is diagonal with eigenvalues x , y , and z in the coordinate system dened by
the intrinsic principal dielectric axes ^x , ^y , and ^z , which are determined by the structural
symmetry of the crystal lattice. In this coordinate system, the relations in (2.61) can be written
explicitly as
ij
ij
ij
0 n2i n2j ij and ij 0

,
(2.62)
i j
ij
0 n2i n2j
p
where i 0 = i are the eigenvalues of the tensor and ni i = 0 are the principal indices
of refraction.
ij 0
EXAMPLE 2.6
LiNbO3 is a negative uniaxial crystal with nx ny no > nz ne . Being a crystal of the 3m
symmetry group, it has eight nonvanishing Pockels coefcients of four distinct values:
r 12 r22 , r 13 , r 22 , r 23 r 13 , r 33 , r42 , r 51 r42 , r 61 r22 . Find the eld-induced permittivity change E0 for an applied DC electric eld of E0 E 0x ^x E 0y ^y E 0z^z .
47
Solution:
According to (2.58), the eld-induced impermeability change due to the Pockels effect is
X
E0
rk E 0k ,
k
which can be expressed in the matrix form as

0
r 12
r 11
B
C B
B 2 C B r 21
B
C B
B C B r 31
B 3C B
B
CB
B 4 C B r 41
B
C B
B
C B
@ 5 A @ r 51
6
r 52
C
r 33 C0
1
C E0x
C
r 33 CB
C
C@ E 0y A:
r 43 C
C E
0z
C
r 53 A
r 62
r 63
r 22
r 32
r 42
r 61
r 13
Using the given nonvanishing Pockels coefcients for LiNbO3 , we have

0
B
C B
B 2 C B
B
C B
B C B
B 3C B
B
CB
B 4 C B
B
C B
B
C B
@ 5 A @
6
r22
r22
r42
r 42
r 22
r 13
r 22 E 0y r13 E 0z
B
C
r 13 C0
1 B r 22 E 0y r13 E 0z
B
C E 0x
B
r 33 C
r 33 E 0z
C B
CB
C@ E 0y A B
B
0 C
r 42 E 0y
B
C E
0z
B
C
0 A
r 42 E 0x
@
1
C
C
C
C
C
C:
C
C
C
A
r 22 E 0x
By the index contraction rule, 1 xx , 2 yy , 3 zz , 4 yz zy ,

5 zx xz , 6 xy yx . Using (2.62), we nd
xx 0 n4x xx 0 n4o r 22 E 0y 0 n4o r 13 E 0z ,
yy 0 n4y yy 0 n4o r 22 E 0y 0 n4o r 13 E 0z ,
zz 0 n4z zz 0 n4e r 33 E 0z ,
yz zy 0 n2y n2z yz 0 n2o n2e r 42 E 0y ,
zx xz 0 n2x n2z yz 0 n2o n2e r 42 E 0x ,
xy yx 0 n2x n2y xy 0 n4o r22 E 0x :
Expressed in the matrix form, the eld-induced permittivity change is
0 4
1
no r 22 E 0y n4o r 13 E 0z
n4o r22 E 0x
n2o n2e r 42 E 0x
B
C
E0 0 @
n4o r 22 E 0x
n4o r22 E 0y n4o r13 E 0z n2o n2e r 42 E 0y A:
n2o n2e r42 E 0x
n2o n2e r 42 E 0y
n4e r 33 E 0z
48
The electric-eld-induced changes represented by ; E0 usually generate off-diagonal

elements besides changing the diagonal elements:
0
1
0
1
xy
xz
x xx
x 0 0
y yy
yz A (2.63)
@ 0 y 0 A while ; E0 @ yx
zx
zy
z zz
0 0 z
in the coordinate system of the principal ^x , ^y , and ^z axes. As discussed in Section 2.2, of a
nonmagnetic material is a symmetric tensor. This remains true for a nonmagnetic material
subject to an applied electric eld; thus, for ; E0 in (2.63),
ij ; E0 ji ; E0 and ij ; E0 ji ; E0 :
(2.64)
Because the eld-dependent nondiagonal permittivity tensor is symmetric, it can be

diagonalized to nd a new set of eigenvalues X , Y , and Z with corresponding real
^ , Y^ , and Z^ , which represent a new set of linearly polarized principal normal
eigenvectors X
modes for dening the new principal dielectric axes of the material in the presence of the
electric eld E0 . In general, the new principal axes depend on the direction and, in certain
cases, the magnitude of E0 . Thus,
0
1
X 0 0
(2.65)
; E0 @ 0 Y 0 A:
0 0 Z
The propagation characteristics of an optical wave in the presence of an electro-optic effect are
then determined by X , Y , and Z , which dene the principal indices of refraction,
r
r
r
X
Y
Z
, nY
, nZ
,
(2.66)
nX
0
0
0
and the propagation constants,
kX
nX
nY
nZ
, kY
, kZ
,
c
c
c
(2.67)
^ Y^ , and Z^ principal normal modes of polarization. Note that these three new principal
for the X,
normal modes of polarization are linearly polarized. Therefore, electrically induced
birefringence and dichroism due to an electro-optical effect are linear birefringence and linear
dichroism.
EXAMPLE 2.7
At the 1 m optical wavelength, LiNbO3 has the refractive indices of no 2:238 and
ne 2:159. The four distinct values of its Pockels coefcients are r 13 8:6 pm V1 ,
r 22 3:4 pm V1 , r 33 30:8 pm V1 , and r 42 28 pm V1 . Use the results from Example
2.6 to answer the following questions. Is it possible to apply a DC electric eld to change the
principal indices of refraction through the Pockels effect without rotating the principal axes? If
this is possible, nd the changes in the principal indices of refraction caused by an applied
electric eld of E 0 5 MV m1 .
49
Solution:
For the Pockels effect to cause only changes in the principal indices of refraction without rotating
the principal axes, an applied electric eld has to generate changes only in the diagonal elements,
but not in the off-diagonal elements, of E0 . By examining E0 obtained in Example 2.6
for LiNbO3 , we nd that this is possible if the DC electric eld is applied only along the direction
of the z principal axis such that E0 E 0^z for E 0z E 0 and E0x E 0y 0. Then,
0 2
1
no n4o r13 E 0
0
0
A:
E0 E0 0 @
0
n2o n4o r 13 E 0
0
2
4
0
0
ne ne r33 E 0
Because E0 is diagonal in the coordinate system of the original principal axes, all principal
axes remain unchanged:
^ ^x , Y^ ^y , Z^ ^z :
X
Using (2.65) and (2.66), we nd the new principal indices of refraction:
nX nY n2o n4o r 13 E 0 1=2 no
n3o r 13
n3 r 33
E 0 , nZ n2e n4e r 33 E 0 1=2 ne e E 0 :
2
2
Clearly, the crystal remains negative uniaxial. The changes in the principal indices of refraction
caused by an applied electric eld of E 0 5 MV m1 are
nX nY no
n3o r13
2:2283 8:6 1012
E0
5 106 2:41 104
2
2
for the ordinary index and

nZ ne
n3e r 33
2:1593 30:8 1012
E0
5 106 7:75 104
2
2
for the extraordinary index.
2.6.2 Magneto-optic Effect

A material can be either diamagnetic or paramagnetic. A diamagnetic material does not contain
intrinsic magnetic dipole moments; a paramagnetic material consists of atoms or ions that have
intrinsic magnetic dipole moments. A paramagnetic material can be either magnetically disordered, when its intrinsic magnetic dipole moments are randomly oriented, or magnetically
ordered. A magnetically ordered material is ferromagnetic if all of its intrinsic dipole moments
line up in the same direction; it is ferrimagnetic if it contains different types of intrinsic dipole
moments that line up in alternating antiparallel directions but do not cancel each other; it is
antiferromagnetic, also called antiferrimagnetic, if different types of intrinsic dipole moments line
up in alternating antiparallel directions and cancel each other. Below a critical temperature, known
as the Curie temperature for a ferromagnetic material and the Nel temperature for a ferrimagnetic
50
material, the magnetic ordering in a ferromagnetic or ferrimagnetic material generates a spontaneous magnetization M 0 . No spontaneous magnetization exists in a diamagnetic material, in a
magnetically disordered paramagnetic material, or in an antiferromagnetic material.
As mentioned in Section 1.1, at an optical frequency 0 and thus B 0 H; the
response of a material, irrespective of whether it is magnetic or nonmagnetic, to an optical eld
at an optical frequency of is fully described by its electric susceptibility and, equivalently, by its electric permittivity . Nevertheless, a nonmagnetic material that does not have
a spontaneous magnetization still responds to a static or low-frequency magnetic eld, H 0 . Its
optical property can be changed by H 0 , resulting in a magnetic-eld-dependent susceptibility
and a magnetic-eld-dependent permittivity:
P; H 0 0 ; H 0 E 0 E 0 ; H 0 E
(2.68)
D; H 0 ; H 0 E E ; H 0 E,
(2.69)
and
where ; H 0 0 and ; H 0 0 represent the intrinsic properties of the

material in the absence of the static or low-frequency magnetic eld. In the case of a ferromagnetic or ferrimagnetic material, in which a static magnetization M 0 exists, the properties of the
material at an optical frequency are dependent on M 0 . Then, instead of (2.68) and (2.69), we
have magnetization-dependent susceptibility and magnetization-dependent permittivity:
P; M 0 0 ; M 0 E 0 E 0 ; M 0 E
(2.70)
D; M 0 ; M 0 E E ; M 0 E:
(2.71)
and
While and are changed by H 0 or M 0 , the magnetic permeability of the material at an
optical frequency remains the constant 0 , and the relation between B and H remains
independent of H 0 or M 0 : B 0 H. Therefore, magneto-optic effects are completely
characterized by ; H 0 , if no internal magnetization is present, or by ; M 0 , if an internal
magnetization is present. In general, these effects are weak perturbations to the optical properties of the material. The rst-order magneto-optic effect, or linear magneto-optic effect, is
characterized by a linear dependence of on H 0 or M 0 , and the second-order magneto-optic
effect, or quadratic magneto-optic effect, causes a quadratic dependence of on H 0 or M 0 .
We rst consider the magneto-optic effects in a material that has no spontaneous magnetization, i.e., a diamagnetic material, a magnetically disordered paramagnetic material, or an antiferromagnetic material. In such a material, operation of the time-reversal transformation yields
ij ; H 0 ji ; H0
(2.72)
when the material is subject to an external magnetic eld H 0 . This relation is generally true
regardless of the symmetry property of the material. If the material is lossless, then its dielectric
permittivity tensor is Hermitian:
ij ; H 0
ji ; H 0 :
(2.73)
If we express the real and imaginary parts of explicitly by writing ij 0ij i 00ij , we nd by
combining these two relations that
51
0ij ; H 0 0ij ; H 0 0ji ; H 0 0ji ; H 0 ,
(2.74)
00ij ; H 0 00ij ; H0 00ji ; H 0 00ji ; H 0 :
(2.75)
As a result, the magneto-optic effects in a lossless material that has no spontaneous magnetization can be generally described as
X
X
f ijk H 0k 0
cijkl H 0k H 0l ,
(2.76)
ij H 0 ij ij H 0 ij i 0
k
k, l
where f ijk and cijkl are real quantities that satisfy the following relations:
f ijk f jik ,
cijkl cjikl cijlk cjilk :
(2.77)
Because magnetic elds have transformation symmetry properties that are very different
from those of electric elds, magneto-optic effects also have properties very different from
those of electro-optic effects.
1. Because a magnetic eld does not change sign under space inversion, the linear magnetooptic effect does not vanish, thus f ijk 6 0, in a centrosymmetric material. By comparison, the
linear electro-optic effect vanishes, thus r ijk 0, in a centrosymmetric material because an
electric eld changes sign under space inversion.
2. Because a magnetic eld changes sign under time reversal, the linear magneto-optic effect is
nonreciprocal, thus f ijk f jik . By comparison, the linear electro-optic effect is reciprocal,
thus rijk r jik , because an electric eld does not change sign under time reversal.
3. Because the product of two electric eld components, E 0k E 0l , and the product of two
magnetic eld components, H 0k H 0l , both do not change sign under space inversion or time
reversal, the quadratic electro-optic effect and the quadratic magneto-optic effect both exist
in centrosymmetric materials and are both reciprocal, thus sijkl sjikl sijlk sjilk and
cijkl cjikl cijlk cjilk .
4. Both linear and quadratic magneto-optic effects exist in all materials, i.e., f ijk 6 0 and
cijkl 6 0 in all materials, including all solids, liquids, and gases.
5. When a magnetically induced optical loss exists in the linear magneto-optic effect, f ijk
becomes complex with an imaginary part that characterizes the loss. When it exists in the
quadratic magneto-optic effect, cijkl becomes complex with an imaginary part that characterizes the loss.
The magneto-optic effects in magnetically ordered crystals have the same general properties as
discussed above, but their details can be rather complicated due to the magnetic symmetry
properties of such crystals.
In reality, the magneto-optic effects are relatively weak in comparison to, and tend to be
obscured by, any natural or structural birefringence that might exist in a material. Fortunately,
both rst- and second-order magneto-optic effects exist in nonbirefringent materials, which
have isotropic linear optical properties, including noncrystals and cubic crystals. For these
reasons, materials of particular interest and practical importance for magneto-optic effects and
their applications are those in which birefringence originating from other effects, such as
material anisotropy or inhomogeneity, does not exist or, if it exists, does not dominate the
52
particular magneto-optic effect of interest. Such materials include isotropic materials and, in
some cases, uniaxial crystals subject to a magnetic eld or a magnetization that is parallel to the
optical axis. For magneto-optic effects in these materials, we can take the direction of H 0 or M 0
to be the z direction without loss of generality, i.e., H0 H 0z^z or M 0 M 0z^z . Then, H 0 or
M 0 can be generally expressed in the form of (2.16):
0
n2
H 0 or M 0 0 @ i
0
i
n2
0
1
0
0 A,
n2k
(2.78)
where represents the rst-order effect, and n2 and n2k account for the second-order effect. In
the case of H 0 , f 123 H 0z , n2 n2o c1133 H 20z n2o c2233 H 20z , and n2k n2o c3333 H 20z .
In the case of M 0 , is linearly proportional to M 0z with the symmetry of M 0z
M 0z , and n2 and n2k are functions of M 20z .
The linear dependence of ij H 0 on the magnetic eld, or that of ij M 0 on the magnetization, appears only as antisymmetric components in the off-diagonal elements of the permittivity
tensor. In the absence of a magnetically induced optical loss, these off-diagonal elements are
purely imaginary; then this rst-order magneto-optic effect results in a magnetically induced
circular birefringence, discussed in Section 2.2. When this rst-order magneto-optic effect
induces an optical loss, these off-diagonal elements become complex, resulting in a magnetically induced circular dichroism, also discussed in Section 2.2. The linear magneto-optic effect
has two notable phenomena: the Faraday effect and the magneto-optic Kerr effect. The Faraday
effect is manifested in the propagation and transmission of an optical wave through a material
subject to a magnetic eld or a magnetization; the magneto-optic Kerr effect is manifested in
the reection of an optical wave from the surface of such a material. The rst-order magnetooptic effect and these phenomena resulting from it are nonreciprocal.
By contrast, the quadratic dependence on the magnetic eld or the magnetization appears as
symmetric components in the permittivity tensor elements. This second-order magneto-optic
effect is reciprocal and is called the CottonMouton effect. In the absence of a magnetically
induced optical loss, it causes a magnetically induced linear birefringence in the material and is
analogous to, but much weaker than, the electro-optic Kerr effect. When this second-order
magneto-optic effect causes an optical loss, the symmetric permittivity tensor elements are
complex, resulting in a magnetically induced linear dichroism.
2.6.3
Acousto-optic Effect
An acoustic wave in a medium is an elastic wave of space- and time-dependent periodic
deformation in the medium. A traveling plane acoustic wave of a wavelength 2=K and
a frequency f =2 can be expressed as
ur; t U cos K r t ,
(2.79)
where U is the amplitude vector of the deformation that denes the polarization of the acoustic
^ is the acoustic wavevector
wave, is the angular frequency of the acoustic wave, and K K K
53
^ describes the propagation direction and K 2= =v a is the propagation constant

where K
with v a being the acoustic velocity. A standing plane acoustic wave is a combination of two
contrapropagating traveling waves of equal amplitude, wavelength, and frequency:
ur; t U cos K r cos t:
(2.80)
An acoustic wave polarized in the direction of K is known as a longitudinal wave, while one
with a polarization perpendicular to K is called a transverse wave. For any given direction of
propagation in a medium, there are three orthogonal acoustic normal modes of polarization: one
longitudinal or quasi-longitudinal mode, and two transverse or quasi-transverse modes.
The mechanical strains associated with deformation are described by a symmetric strain

tensor, S Sij , dened by

1 ui uj
,
(2.81)
Sij
2 xj xi
where the indices i, j x, y, z. The three tensor elements Sxx , Syy , and Szz are tensile strains,
while the other elements Syz Szy , Szx Sxz , and Sxy Syx are shear strains. In addition, there

is an antisymmetric rotation tensor, R Rij , dened by

1 ui uj
Rij
:
(2.82)

2 xj xi
Clearly, Rxx Ryy Rzz 0, while Ryz Rzy , Rzx Rxz , and Rxy Ryx . For elastic
deformation caused by an acoustic wave, all of the strain and rotation tensor elements are
space- and time-dependent quantities.
Mechanical strain in a medium causes changes in the optical property of the medium due to
the photoelastic effect. The basis of acousto-optic interaction is the dynamic photoelastic effect
in which the periodic time-dependent mechanical strain and rotation caused by an acoustic
wave induce periodic time-dependent variations in the optical properties of the medium. The
photoelastic effect is traditionally dened in terms of changes in the elements of the relative
impermeability tensor:

X
ij S; R ij ij S; R ij
(2.83)
pijkl Skl p0ijkl Rkl ,
k, l
where pijkl are dimensionless elasto-optic coefcients, also called strain-optic coefcients or
photoelastic coefcients, and p0ijkl are dimensionless rotation-optic coefcients. Both are fourth
order tensors. Because ij ji and Skl Slk , the pijkl tensor is symmetric in i and j and in k
and l. Because ij ji and Rkl Rlk , the p0ijkl tensor is symmetric in i and j but is
antisymmetric in k and l.
The photoelastic effect exists in all matter, including centrosymmetric crystals and isotropic

materials, because the pijkl tensor never vanishes in any material though the p0ijkl tensor
vanishes in isotropic materials and cubic crystals. Acousto-optic interactions are not precluded

by any symmetry property of a material. The tensor form of pijkl for a crystal is determined by the
point group of the crystal. The p0ijkl tensor elements of a crystal are determined by the birefringence of the crystal. If the indices i, j, k, are l referenced to the principal axes of a crystal, we have
54
p0ijkl
!

1 1
1

ik
jl
il
jk
2 n2i n2j
(2.84)
where ni and nj represent the principal indices of refraction of the crystal. It is clear that p0ijkl
vanishes in an isotropic material or a cubic crystal.
It is desirable to formally express the photoelastic effect caused by strain and rotation in a
medium in terms of a change in the permittivity of the medium as
; S; R ; S; R 0 ; S; R,
(2.85)
where is the dielectric permittivity tensor of the medium in the absence of strain and
rotation elds. Using (2.62), the elements of can be found from those of in (2.83):

X
ij 0 n2i n2j ij 0 n2i n2j

pijkl Skl p0ijkl Rkl ,
(2.86)
k, l
where for an acoustic wave, Skl and Rkl are functions of space and time. For a traveling wave
characterized by a wavevector of K and a frequency of as described by (2.79), Skl and Rkl can
be found by using (2.81) and (2.82), respectively. They have the form:
Skl S kl sin K r t, Rkl Rkl sin K r t,
(2.87)
where S kl is the amplitude of the strain and Rkl is the amplitude of the rotation. Therefore, the
photoelastic permittivity tensor is a function of space and time:
e
sin K r t ,
where e
is the amplitude of , and its elements are

X
e
ij 0 n2i n2j
pijkl S kl p0ijkl Rkl :
k, l
(2.88)
(2.89)
EXAMPLE 2.8
Silica glass is an isotropic material. An acoustic wave propagating in any direction in silica
glass can have two transverse modes and one longitudinal mode. The two transverse modes
have the same acoustic wave velocity of v Ta 5:97 km s1 , whereas the longitudinal mode has
an acoustic wave velocity of v La 3:76 km s1 . Take the acoustic wave propagation direction
to be the z direction. How does each mode of an acoustic wave at a frequency of 500 MHz
modulate the optical permittivity in space and time?
Solution:
All three modes modulate the optical permittivity at the same frequency of f 500 MHz,
thus 1 109 rad s1 , in time, but they modulate the optical permittivity differently
in space. Because the wave propagates in the z direction, the longitudinal mode is polarized
in the z direction while the two transverse modes are polarized in the x and y directions,
respectively.
55
For the longitudinal mode, v La 3:76 km s1 . Thus,

L
v La 3:76 103
2
m 7:52 m and K L
8:36 105 m1 :
6
f
500 10
L
The wavevector of the longitudinal mode is K K L^z . The optical permittivity that is modulated by the longitudinal acoustic wave varies in space and time with K L 8:36 105 m1 and
1 109 rad s1 as
z; t e
sin K L z t :
For both transverse modes, v Ta 5:97 km s1 . Thus,
T
v Ta 5:97 103
2
m 11:94 m and K T
5:26 105 m1 :
6
f
T
500 10
The wavevectors of both transverse modes are K K T^z . The optical permittivity that is
modulated by either of the transverse acoustic waves varies in space and time with K T
5:26 105 m1 and 1 109 rad s1 as
z; t e
sin K T z t :
The permittivity tensor e
is a constant that does not vary with space or time, but it has different
forms for different acoustic modes.
2.7
NONLINEAR OPTICAL SUSCEPTIBILITIES
..............................................................................................................
The origin of optical nonlinearity is the nonlinear response of electrons in a material to an
optical eld as the strength of the eld is increased. Macroscopically, the nonlinear optical
response of a material is described by a polarization that is a nonlinear function of the optical
eld. In general, such nonlinear dependence on the optical eld can take a variety of forms. In
particular, it can be very complicated when the optical eld becomes extremely strong.
In most situations of interest, with the exception of saturable absorption, the perturbation
method can be used to expand the total optical polarization in terms of a series of linear and
nonlinear polarizations:
Pr; t P1 r; t P2 r; t P3 r; t ,
(2.90)
where P1 is the linear polarization, and P2 and P3 are the second- and third-order nonlinear
polarizations, respectively. Except in some special cases, nonlinear polarizations of the fourth
and higher orders are usually not important and thus can be ignored. Note that the space- and
time-dependent polarizations in (2.90) are complex polarizations dened with respect to the
corresponding real polarizations according to the denition of the complex eld in (1.40):
Pn r; t Pn r; t Pn r; t Pn r; t c:c:,
(2.91)
56
where Pn r; t is the nth-order real nonlinear polarization and Pn r; t is the nth-order

complex polarization.
The optical eld involved in a nonlinear interaction usually contains multiple, distinct
frequency components. Such a eld can be expanded in terms of its frequency components:
X

X

Er; t
Eq r exp iq t
E q r exp ikq r iq t ,
(2.92)
q
where E q r is the slowly varying amplitude and kq is the wavevector of the q frequency
component. The nonlinear polarizations also contain multiple frequency components and can
be expanded as
X

Pn r; t
Pqn r exp iq t :
(2.93)
q
Note that we do not attempt to further express Pqn r in terms of a slowly varying polarization
amplitude multiplied by a fast varying spatial phase term, as is done for Eq r. The reason is
that the wavevector that characterizes the fast varying spatial phase of a nonlinear polarization
Pqn r is not simply determined by the frequency q but is dictated by the elds that generate
the nonlinear polarization. In the discussion of nonlinear polarizations, we also use the

notations E q and Pn q dened respectively as

E q Eq r and Pn q Pqn r:
(2.94)
Field and polarization components of negative frequencies are interpreted as

E q E q and Pn q Pn q :
(2.95)
The frequency-domain nth-order nonlinear susceptibility characterizing the nonlinear

response of a material to optical elds at frequencies 1 , 2 , . . . , n is a function of these
optical frequencies: n 1 ; 2 ; ; n . In the momentum space and frequency domain, the
nonlinear susceptibility is also a function of wavevectors: n k1 ; 1 ; k2 ; 2 ; ; kn ; n . The
reality condition discussed in Section 2.1 and expressed explicitly in (2.7) for the linear
susceptibility can be generalized for each nonlinear susceptibility. This reality condition leads
to the following relation for a nonlinear susceptibility:
n k1 ; 1 ; k2 ; 2 ; ; kn ; n n k1 ; 1 ; k2 ; 2 ; ; kn ; n :
(2.96)
When expressing the nonlinear polarization that is generated at a frequency of q 1 2

n by the nonlinear optical interaction of the optical elds at frequencies 1 , 2 , . . . , n ,
the following notation for the nonlinear susceptibility is used:

n q 1 2 n n 1 ; 2 ; ; n :
(2.97)
Using the expansions of the complex elds and polarizations in (2.92) and (2.93), we have
the expressions for the second- and third-order nonlinear polarizations:
X

P2 q 0
2 q m n : Em En
m, n
(2.98)
57
and
X

3 q m n p : Em En E p :
P3 q 0
m, n, p
(2.99)
The summation is performed for a given q over all positive and negative values of frequencies
that satisfy the constraint of m n q in the case of (2.98) and the constraint of m
n p q in the case of (2.99). More explicitly, by performing the summation over
positive frequencies only and by expanding the tensor products, we have
X X h 2

2
ijk q m n E j m E k n
Pi q 0
j, k m , n >0

2
ijk q m n E j m E
k n
i

2
ijk q m n E
(2.100)
m
k
n
j
and
3
Pi
X X h 3

q 0
ijkl q m n p Ej m E k n E l p
j, k , l m , n , p >0

3
ijkl q m n p Ej m E k n E
l p

3
ijkl q m n p Ej m E
k n E l p

3
ijkl q m n p E
j m E k n E l p

3
ijkl q m n p Ej m E
k n E l p

3
ijkl q m n p E
j m E k n E l p

i
3
E
p :
ijkl q m n p E
m
n
l
j
k
(2.101)
Usually only a limited number of frequencies participate in a given nonlinear optical interaction. Consequently, only one or a few terms among those listed in (2.100) or (2.101)
contribute to a particular nonlinear polarization.
EXAMPLE 2.9
Three optical elds at the wavelengths of 1 300 nm, 2 750 nm, and 3 1500 nm,
corresponding to the frequencies of 1 2c=1 , 2 2c=2 , and 3 2c=3 , respectively, are involved in second-order nonlinear
optical interactions. The optical elds at the three
p
frequencies are E 1 E 1 ^x ^y = 2, E 2 E 2^z , and E3 E 3^z , where ^x , ^y , and ^z are
the x, y, and z principal axes of the nonlinear crystal. Find the nonlinear polarization P2 4 at
the frequency of 4 2c=4 where 4 375 nm. Express the components of P2 4
explicitly in terms of the elements of 2 and the given magnitudes, E 1 , E2 , and E3 , of the
three optical elds.
58
Solution:
1
1
1
1
Because 1
1 3 2 2 4 , we nd that 4 1 3 2 2 . Therefore, the
second-order nonlinear polarization at the frequency 4 is

P2 4 0 2 4 1 3 : E1 E 3 2 4 3 1 : E 3 E1
i
2 4 2 2 : E2 E2 :
Note that there are two terms from the mixing of 1 and 3 because of permutation, but there is
only one term from 2 mixing with itself. Using the given elds at the three frequencies, we can
express the components of P2 4 as

E1 E
E1 E
2
2
2
4 1 3 p3 xyz
4 1 3 p3
Px 4 0 xxz
2
2
E
E
2
2
3 E1
3 E1
xzx 4 3 1 p xzy 4 3 1 p
2
i 2
2
2
xzz 4 2 2 E 2 ,

Py2 4
E1 E
E1 E
2
2
0 yxz
4 1 3 p3 yyz
4 1 3 p3
2
2
E
E
E
2
2
3 1
3 E1
yzy
4 3 1 p
4 3 1 p
yzx
2
2
i
2
2
yzz 4 2 2 E 2 ,

Pz2 4
E1 E
E1 E
2
2
0 zxz
4 1 3 p3 zyz
4 1 3 p3
2
2
E3 E1
E
2
2
3 E1
zzy
zzx
4 3 1 p
4 3 1 p
2
2
i
2
zzz
4 2 2 E 22 :
As discussed in Section 2.2, the form of the linear susceptibility tensor is determined by the
symmetry property of the material. The forms of the nonlinear susceptibility tensors of a
material also reect the spatial symmetry property of the material structure. As a result, some
elements in a nonlinear susceptibility tensor may be zero and others may be related in one
way or another, greatly reducing the total number of independent tensor elements. The linear
susceptibility tensor has its form determined only by the crystal system of a material, whereas
the form of a nonlinear susceptibility tensor further depends on the point group of the
material.
Within the 7 crystal systems, there are 32 point groups. Among the 32 point groups, 21 are
noncentrosymmetric and 11 are centrosymmetric. All gases, liquids, and amorphous solids are
centrosymmetric. Centrosymmetric materials possess space-inversion symmetry. In the electricdipole approximation, nonlinear optical effects of all even orders, but not those of the odd
59
orders, vanish identically in a centrosymmetric material. Therefore, 2 contributed by electricdipole interaction is identically zero in a centrosymmetric material, whereas a nonzero 3
exists in any material. This fact can be easily veried by considering the effect of space
inversion on the nonlinear polarizations P2 and P3 given in (2.98) and (2.99), respectively.
The space-inversion transformation can be performed on a centrosymmetric material without
changing the properties of the material. Being polar vectors, P2 , P3 , and E all change sign
under such a transformation. From (2.98), we then nd that P2 P2 . Therefore, P2
cannot exist and 2 has to vanish identically in a centrosymmetric material. No such conclusion is drawn for P3 or 3 as we examine (2.99) following the same procedure. Comparing
the above discussion with that in Section 2.6 for the Pockels coefcients r ijk , which vanish in
centrosymmetric materials, and the electro-optic Kerr coefcients sijkl , which exist in any
material, we nd the similarity between the 2 and r ijk , and that between 3 and sijkl . Indeed,
they are directly related:
r ijk
2 2

2
ni n2j ijk
0
2 2
0
2
ni n2j kij
(2.102)
and
sijkl
3 3

2
ni n2j ijkl
0 0:
(2.103)
EXAMPLE 2.10
The BBO crystal structure belongs to the 3m point group, for which the only nonvanishing
2
2
2
2
2
2
2
2
2
2
2
2 elements are xzx
yzy
, xxz
yyz
, yyy
yxx
xxy
xyx
, zxx
zyy
, and zzz
.
If the nonlinear interaction considered in Example 2.9 takes place in a BBO crystal, what
are the expressions of the components of P2 4 in terms of the nonvanishing elements
of 2 ?
Solution:
By keeping the terms that contain only the nonvanishing 2 elements in each of the components of P2 4 obtained in Example 2.9, we nd that

E1 E
E
2
2
3
3 E1
0 xxz 4 1 3 p xzx 4 3 1 p ,
2
2

E1 E
E
2
2
2
3
3 E1
Py 4 0 yyz 4 1 3 p yzy 4 3 1 p ,
2
2
Px2 4
2
4 2 2 E22 :
Pz2 4 0 zzz
60
Problems
2.1.1 Verify the relations given in (2.7) that are required by the reality condition.
2.2.1 At a given optical frequency, the optical susceptibility tensors of several materials are
measured with respect to an arbitrary rectilinear coordinate system in space, as listed
below. Identify each material as (1) a dielectric or magnetic material and (2) an optically
lossless or lossy material.
1
0
1
2:3
0:1 i0:2
0
2:0 i0:1 i0:3
0
1 i0:2 0 A;
A : @ 0:1 i0:2
2:7
i0:2 A; B : @ i0:3
0
i0:2
2:4
0
0
1:5
0
1
0
1
1:59 0:13 0:16
1:9 0:2
0:3
@
A
@
A;
C:
0:13 1:59 0:11 ; D : 0:2 2:8
0:1
0:16 0:11 1:41
0:3 0:1 2:5 i0:2
0
1
1:30 i0:35
0
E : @ i0:35
1:25
0:15 A:
0
0:15
1:40
2.2.2 Represented in an arbitrarily chosen right-handed Cartesian coordinate system dened by
the unit vectors ^x 1 , ^x 2 , and ^x 3 , with ^x 1 ^x 2 ^x 3 , the permittivity tensor of a crystal at
0:50 m is
0
1
5:481
0
0
0@ 0
5:267 0:214 A:
0
0:214 5:267
(a) Find the principal indices and the corresponding principal axes of the crystal at this
wavelength.
(b) Is this crystal birefringent or nonbirefringent? If it is birefringent, is it uniaxial or
biaxial? If it is uniaxial, is it positive or negative uniaxial?
2.2.3 At the 1:300 m optical wavelength, the permittivity tensor of a LiNbO3 crystal
represented in an arbitrarily chosen x1 ; x2 ; x3 rectilinear coordinate system with
^x 1 ^x 2 ^x 3 is found to be
0
1
4:938
0
0
4:770 0:168 A:
0@ 0
0
0:168 4:770
(a) Find the principal indices and the corresponding principal axes of the LiNbO3 crystal
at this wavelength.
(b) Is it possible to send an optical wave at this wavelength through a LiNbO3 crystal of
arbitrary thickness in such a manner that the polarization of the wave is maintained
throughout its path no matter how the wave is initially polarized? If the answer is yes,
how can this be arranged? If the answer is no, why is it not possible?
Problems
61
2.2.4 Represented in an arbitrarily chosen right-handed x1 ; x2 ; x3 Cartesian coordinate system

with ^x 1 ^x 2 ^x 3 , the permittivity tensor of a KTP crystal at 1:0 m is
0
1
3:035
0
0
3:210 0:147 A:
0@ 0
0
0:147 3:210
(a) Find the principal indices and the corresponding principal axes of the crystal at this
wavelength.
(b) Is the crystal birefringent or nonbirefringent? If it is birefringent, is it uniaxial or
biaxial? If it is uniaxial, is it positive or negative uniaxial?
2.2.5 What is the difference between linear birefringence and circular birefringence?
2.2.6 What is the difference between linear birefringence and linear dichroism? What is the
difference between circular birefringence and circular dichroism?
2.2.7 In a properly chosen xyz Cartesian coordinate system, a particular medium has a
symmetric but non-Hermitian permittivity tensor of the form:
0 2
1
n i i 0
(2.104)
0 @ i n2 i 0 A,
2
0
0
nz
where n, , , and are all real and positive numbers with n , , . Find the principal
refractive indices and the corresponding principal normal modes of polarization. Show
that this medium is linearly birefringent and linearly dichroic.
2.2.8 In a properly chosen xyz Cartesian coordinate system, a particular medium has an
asymmetric and non-Hermitian permittivity tensor of the form:
0 2
1
n i i 0
(2.105)
0 @ i n2 i 0 A,
0
0
n2z
where n, , , and are all real and positive numbers with n , , . Find the principal
refractive indices and the corresponding principal normal modes of polarization. Show
that this medium is circularly birefringent and circularly dichroic.
2.3.1 Lorentz model: The resonant susceptibility given in (2.26) for an atomic system that has a
single resonance frequency at 0 and a relaxation rate of can be obtained using a classical
Lorentz model by considering a one-dimensional damped oscillator for the bound electrons
of this system. The system consists of N oscillating electrons, each of which has a charge of
q e and an effective mass of m . The displacement of the oscillating electron in
response to the force of an optical eld is described by the Lorentz oscillator equation:
d2 x
dx
F
2 20 x ,
dt2
dt
m
(2.106)
where xt xt ^x and Ft qEt eEeit c:c: eE^x eit c:c: The electricdipole polarization due to the electron displacement induced by the optical eld is dened as
62
Pt Next:
(2.107)
The induced electron displacement and the corresponding optical polarization in response
to the optical eld at the frequency can be expressed as xt xt ^x
x^x eit c:c: and Pt P^x eit c:c:
(a) Solve the Lorentz oscillator equation by using the above expression for xt to nd x.
(b) Use the denition of the electric-dipole polarization and the frequency-domain
relation P 0 E , as given in (1.59), between the optical polarization
and the optical eld to nd , which is the resonant susceptibility res ; 0 .
(c) Compare the result obtained in (b) with the resonant susceptibility given in (2.26) by
taking N N 2 N 1 N because the atomic system considered here is in the
thermal-equilibrium state so that it is almost all populated in the ground level.
Identify the electric-dipole moment p in (2.26) and express it in terms of the
parameters of the Lorentz oscillator model.
2.3.2 All susceptibilities and permittivities of physical materials have to satisfy the reality
condition given in (2.7).
(a) Show that the real and imaginary parts of the resonant susceptibility given in (2.27)
do not satisfy the reality condition. Explain this apparent issue.
(b) Show that the resonant susceptibility given in (2.26) satises the reality condition
before the rotating-wave approximation is applied but not after that.
2.3.3 The absorption spectral line of Yb3 : Al2 O3 due to the optical transition from the 2 F7=2
ground level to the 2 F5=2 upper level appears at a center wavelength of 974:5 nm with
a FWHM spectral width of 7:4 nm. Find the energy separation between the two
levels. Find the resonance frequency and the polarization relaxation rate associated with
this transition. Where is anomalous dispersion caused by this transition found when the
Yb3 ions are in their normal state in thermal equilibrium with the surrounding?
2.4.1 Drude model: The Drude model considers free-moving electrons or holes that, unlike
bound electrons, do not have resonant oscillation frequencies.
(a) Show that the Drude model given in (2.30) can be obtained by setting 0 0 and
2 1= for the Lorentz model in Problem 2.3.1.
(b) Show that cond given in (2.43) can be obtained from the expression of res
found in Problem 2.3.1 by setting 0 0 and 2 1=.
2.4.2 Show that the conduction susceptibility given in (2.43) and its real and imaginary parts
given in (2.44) all satisfy the reality condition.
2.4.3 Aluminum is a good conductor. The free-electron Drude model describes its optical
properties reasonably well with a free electron density of N 1:81 1029 m3 . The DC
conductivity of Al at T 273 K is 0 4:08 107 S m1 . Find the plasma frequency
p and the relaxation time for Al at T 273 K. Also nd the cutoff optical frequency
p and the cutoff wavelength p . For what wavelengths is Al expected to be highly
reective? For what wavelengths is it expected to become transmissive?
2.4.4 Si has an electron effective mass of m

e 1:08m0 and a hole effective mass of mh
0:56m0 , where m0 is the mass of a free electron. Its low-frequency dielectric constant is
Problems
63
11.8. Find the plasma frequency, the cutoff frequency, and the cutoff wavelength for
(a) an n-type Si sample that has an electron density of N e 1 1024 m3 , (b) a p-type Si
sample that has a hole density of N h 1 1024 m3 , and (c) a Si sample that is injected
with an equal electron and hole density of N e N h 1 1024 m3 .
2.5.1 Show that the KramersKronig relations given in (2.53) satisfy the reality condition.
2.5.2 Do the real part 0res and the imaginary part 00res of the exact res given in (2.26)
before making the rotating-wave approximation satisfy the KramersKronig relations?
Do the real and imaginary parts, given in (2.27), of the res obtained under the
rotating-wave approximation satisfy the KramersKronig relations?
2.5.3 Do the real part 0cond and the imaginary part 00cond of the conduction susceptibility
given in (2.44) satisfy the KramersKronig relations?
2.6.1 LiNbO3 is a negative uniaxial crystal with nx ny no > nz ne . Being a crystal of the
3m symmetry group, it has eight nonvanishing Pockels coefcients of four distinct
values: r12 r 22 , r 13 , r 22 , r 23 r13 , r 33 , r42 , r 51 r 42 , and r61 r 22 . At the 1 m
optical wavelength, no 2:238 and ne 2:159, and the four distinct values of its
Pockels coefcients are r13 8:6 pm V1 , r 22 3:4 pm V1 , r 33 30:8 pm V1 , and
r 42 28 pm V1 . Use the results from Example 2.6 to nd the new principal axes and
the changes in the principal indices of refraction caused by an electric eld of E 0
5 MV m1 that is applied along the y principal axis.
2.6.2 InP is a cubic crystal of the 43m symmetry group with nx ny nz no and three
nonvanishing Pockels coefcients of the same value: r 41 r52 r 63 . At the 1:55 m
optical wavelength, no 3:166 and r 41 1:6 pm V1 . Because of the symmetry among
the three principal axes, an electric eld applied along any principal axis results in a
similar effect. Consider a DC electric eld of E0 10 MV m1 applied along the z
principal axis. Find the new principal axes and the changes in the principal indices of
refraction caused by the applied eld due to the Pockels effect.
2.6.3 KTP is a biaxial crystal of the mm2 symmetry group with nx 6 ny 6 nz and ve
nonvanishing Pockels coefcients of distinct values: r 13 , r 23 , r 33 , r42 , and r 51 . Find the
eld-induced permittivity change E0 for an applied DC electric eld of
E0 E 0x ^x E 0y ^y E 0z^z .
2.6.4 At the 1 m optical wavelength, the principal indices of KTP are nx 1:742, ny 1:750,
and nz 1:832; the nonvanishing Pockels coefcients are r 13 8:8 pm V1 ,
r 23 13:8 pm V1 , r 33 35 pm V1 , r 42 8:8 pm V1 , and r51 6:9 pm V1 . Is it

possible to apply a DC electric eld to change the principal indices of refraction through
the Pockels effect without rotating the principal axes? If this is possible, nd the changes in
the principal indices of refraction caused by an applied electric eld of E 0 12 MV m1 .
2.6.5 Magneto-optic effect can lead to circular birefringence and circular dichroism. For
simplicity, consider a material for which the only optical loss is magnetically induced
so that ij
ji in the absence of a magnetic eld or a magnetization but
ij H0 6
ji H 0 in the presence of a magnetic eld and ij M 0 6 ji M 0 in the
presence of a magnetization.
64
(a) Show for the case of a magnetic-eld-induced loss that the relations in (2.76) and
(2.77) are still valid but f ijk or cijkl , or both, are complex. Thus, the magneto-optic
permittivity tensor given in (2.78) can be generalized to the form:
0
n2 i
0 @ i 0 00
0
i 0 00
n2 i
0
1
0
0 A,
n2k
(2.108)
where 0 f 0123 H 0z , 00 f 00123 H 0z , n2 n2o c01234 H 20z , and c001234 H 20z . The same
concept is applicable to a magnetization-induced optical loss for which 0 and 00 are
linearly proportional to M 0z , and n2 and are functions of M 20z .
(b) Show that the rst-order magneto-optic effect results in circular birefringence and, in
the situation when 00 6 0 with a magnetically induced loss, circular dichroism.
(c) Show, by setting 0 00 0 to mathematically turn off the rst-order magneto-optic
effect, that the second-order magneto-optic effect does not cause circular birefringence, or circular dichroism, but only linear birefringence or linear dichroism.
2.7.1 Three optical elds at the wavelengths of 1 1200 nm, 2 600 nm, and 3 800 nm,
corresponding to the frequencies of 1 2c=1 , 2 2c=2 , and 3 2c=3 ,
respectively, are involved in second-order nonlinear optical interactions. The
poptical
elds at the three frequencies are E 1 E 1 ^x , E 2 E 2 ^y ^z = 2, and
E 3 E 3^z , where ^x , ^y , and ^z are the x, y, and z principal axes of the nonlinear crystal.
(a) Find the nonlinear polarization P2 4 at the frequency of 4 2c=4 where
4 400 nm. Express each of the components of P2 4 explicitly in terms of

the elements of 2 and the given magnitudes, E1 , E2 , and E 3 , of the three optical
elds.
(b) If the nonlinear interaction takes place in a KTP crystal, what are the expressions of
the components of P2 4 in terms of the nonvanishing elements of 2 ? Note that
KTP belongs to the mm2 point group, for which the only nonvanishing 2 elements
2
2
2
2
2
2
2
are xzx
, xxz
, yyz
, yzy
, zxx
, zyy
, and zzz
.
2.7.2 Three optical elds at the wavelengths of 1 1200 nm, 2 600 nm, and 3 800 nm,
corresponding to the frequencies of 1 2c=1 , 2 2c=2 , and 3 2c=3 ,
respectively, are involved in second-order nonlinear optical interactions. The
poptical
^
^
elds at the three frequencies are E 1 E 1 x , E 2 E 2 y ^z = 2, and
E 3 E 3^z , where ^x , ^y , and ^z are the x, y, and z principal axes of the nonlinear crystal.

the elements of 2 and the given magnitudes, E1 , E2 , and E 3 , of the three optical
elds.
(b) If the nonlinear interaction takes place in a KTP crystal, what are the expressions of
the components of P2 4 in terms of the nonvanishing elements of 2 ? Note that
KTP belongs to the mm2 point group, for which the only nonvanishing 2 elements
2
2
2
2
2
2
2
are xzx
, xxz
, yyz
, yzy
, zxx
, zyy
, and zzz
.
Bibliography
65
2.7.3 Two optical elds at the wavelengths of 1 500 nm and 2 1500 nm, corresponding
to the frequencies of 1 2c=1 and 2 2c=2 , respectively, are involved in
second-order nonlinear optical interactions. The optical elds at the two frequencies are
E 1 E 1 ^x and E 2 E 2 ^y , where ^x , ^y , and ^z are the x, y, and z principal axes of the
nonlinear crystal.

the elements of 2 and the given magnitudes, E 1 and E 2 , of the two optical elds.
(b) If the nonlinear interaction takes place in a LiNbO3 crystal, what are the expressions
of the components of P2 3 in terms of the nonvanishing elements of 2 ? Note
that LiNbO3 belongs to the 3m point group, for which the only nonvanishing
2
2
2
2
2
2
2
2
2
2
yzy
, xxz
yyz
, yyy
yxx
xxy
xyx
, zxx
zyy
,
2 elements are xzx
2
and zzz
.
Bibliography
Altman, C. and Suchy, K., Reciprocity, Spatial Mapping and Time Reversal in Electromagnetics, 2nd edn.
Dordrecht: Springer, 2001.
Bloembergen, N., Nonlinear Optics, 4th edn. Singapore: World Scientic, 1996.
Boyd, R. W., Nonlinear Optics, 3rd edn. Boston, MA: Academic Press, 2008.
Butcher, P. N. and Cotter, D., The Elements of Nonlinear Optics. Cambridge: Cambridge University Press,
1990.
Davis, C. C., Lasers and Electro-Optics: Fundamentals and Engineering, 2nd edn. Cambridge: Cambridge
University Press, 2014.
Fox, M., Optical Properties of Solids, 2nd edn. Oxford: Oxford University Press, 2010.
Korpel, A., Acousto-Optics, 2nd edn. New York: Marcel Dekker, 1997.
Landau, L. D. and Lifshitz, E. M., Electrodynamics of Continuous Media. Oxford: Pergamon, 1960.
Nye, J. F., Physical Properties of Crystals. London: Oxford University Press, 1957.
Post, E. J., Formal Structure of Electromagnetics. Amsterdam: North-Holland, 1962.
Saleh, B. E. A. and Teich, M. C., Fundamentals of Photonics. New York: Wiley, 1991.
Sapriel, J., Acousto-Optics. New York: Wiley, 1979.
Shen, Y. R., The Principles of Nonlinear Optics. New York: Wiley, 1984.
Sugano, S. and Kojima, N., eds., Magneto-Optics. Berlin: Springer, 2000.
Wooten, F., Optical Properties of Solids. New York: Academic Press, 1972.
Zernike, F. and Midwinter, J. E., Applied Nonlinear Optics. New York: Wiley, 1973.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
3 - Optical Wave Propagation pp. 66-140
3
3.1
NORMAL MODES OF PROPAGATION
..............................................................................................................
The propagation of an optical wave is governed by Maxwells equations. The propagation
characteristics depend on the optical property and the physical structure of the medium. They
also depend on the makeup of the optical wave, such as its frequency content and its temporal
characteristics. In this chapter, we discuss the basic propagation characteristics of a monochromatic optical wave in three basic categories of medium: an innite homogeneous
medium, two semi-innite homogeneous media separated by an interface, and an optical
waveguide dened by a transverse structure. Some basic effects of dispersion and attenuation
on the propagation of an optical wave are discussed in Sections 3.6 and 3.7.
The optical property of a medium at a frequency of is fully described by its permittivity
, which is a tensor for an anisotropic medium but reduces to a scalar for an isotropic
medium. For a homogeneous medium, is a constant of space; for an optical structure, it is
a function of space variables. Without loss of generality, we designate the z coordinate axis to
be the direction of optical wave propagation in an isotropic medium; thus the longitudinal axis
of an optical waveguide that is fabricated in an isotropic medium is the z axis. For this reason,
has only transverse spatial variations that are functions of the transverse coordinates,
which are x and y in the rectilinear coordinate system, or and r in the cylindrical coordinate
system. We use the rectilinear coordinates for our general discussion. The exception is optical
wave propagation in an anisotropic crystal, for which the natural coordinate system is that
dened by its principal axes but an optical wave does not have to propagate along its principal
z axis.
For the following discussion in this section, we consider propagation in an isotropic medium,
which is not necessarily homogeneous in space. The wave propagates in the z direction, and the
possible inhomogeneity characterizing the optical structure is described by a scalar permittivity
x; y, as illustrated in Fig. 3.1. If the medium is homogeneous, then x; y is a constant
of space, as shown in Fig. 3.1(a). If the medium is inhomogeneous in only one transverse
dimension, then it has a planar optical structure, such as a planar interface shown in Fig. 3.1(b)
or a planar waveguide shown in Fig. 3.1(c); in these cases, we take the structural variation to be
in the x direction for x; y x to be independent of the y variable. If structural variations
exist in two dimensions, then the medium has a nonplanar optical structure with x; y being a
function of both x and y, such as the single-core nonplanar waveguide shown in Fig. 3.1(d).
In any event, there is no structural variation in the direction of propagation; therefore, x; y
is never a function of the z variable.
67
Figure 3.1 (a) Homogeneous medium. (b) Planar interface. (c) Planar waveguide. (d) Nonplanar waveguide.
The normal modes of propagation for an optical wave in a medium are the characteristic
solutions of Maxwells equations subject to the boundary conditions that are dened by the
physical structure of the medium and are fully described by x; y. Each characteristic solution
has an eigenvalue, which gives the propagation constant, and an eigenfunction, which gives the
eld pattern of the normal mode. Therefore, each normal mode is dened by a specic
propagation constant and a pair of specic electric and magnetic mode eld proles E x; y
and Hx; y. It is possible for two or more degenerate normal modes to have the same
propagation constant but different eld proles. By contrast, two normal modes of different
propagation constants cannot share the same eld prole. Because electric and magnetic elds
are vectorial elds, a mode eld is dened by a specic amplitude and polarization pattern of
E x; y and Hx; y. A mode index is used to label a mode when the optical structure supports
multiple normal modes. Therefore, the space- and time-dependent electric and magnetic elds
of a normal mode at a frequency of are expressed as
E r; t E x; y exp i z it ,
(3.1)
H r; t H x; y exp i z it,
(3.2)
where is the propagation constant of the mode. If the cylindrical coordinate system is used,
then the mode elds in (3.1) and (3.2) are expressed as functions of and r: E ; r and
H ; r .
The characteristic of the mode index depends on the transverse boundary conditions
imposed on the mode eld. For an optical medium that imposes two-dimensional boundary
conditions in the transverse xy plane, the mode eld proles are functions of two transverse
spatial variables: E x; y and H x; y. Therefore, the mode index consists of two parameters
for characterizing the variations of the mode elds in these two transverse dimensions. Then
represents two mode numbers or symbols: mn. This is the case for an optical structure that
68
provides two-dimensional transverse optical connement, such as the nonplanar waveguide in

Fig. 3.1(d). Another example is a collimated Gaussian mode in a homogeneous medium, which
has a two-dimensional transverse prole. For an optical medium that imposes boundary conditions in only one transverse direction, such as that in Fig. 3.1(b) or (c), the mode eld proles are
functions of only one transverse spatial variable: E x and H x. In this case, the mode index
consists of only one parameter for characterizing the variations of the mode elds in the transverse
dimension x. Then represents only one mode number or symbol: m. For discrete modes, i.e.,
modes of discrete propagation constants, the mode index numbers are discrete numbers, which
are normally integers. For continuous modes, i.e., modes of continuously distributed propagation
constants, the mode index numbers are continuously distributed numbers.
3.1.1 Mode Types

For an optical structure in an isotropic medium, which is characterized by a spatial permittivity
distribution of scalar x; y, Maxwells equations for wave propagation take the form:
E 0
H
H
,
t
E
:
t
(3.3)
(3.4)
For the mode elds of the form of (3.1) and (3.2), these two equations can be expressed in terms
of the components of the mode eld proles as
E z
iE y i0 Hx ,
y
E z
i0 Hy ,
x
E y E x

i0 Hz ,
x
y
iE x
(3.5)
(3.6)
(3.7)
and
Hz
iHy iE x ,
y
(3.8)
Hz
iE y ,
x
(3.9)
Hy Hx

iE z :
x
y
(3.10)
iHx
From these equations, the transverse components of the electric and magnetic mode elds can
be expressed in terms of the longitudinal components:

E z
Hz
k2 2 E x i
i0
,
x
y
(3.11)
69
2

E z
Hz
i0
,
k 2 E y i
y
x
(3.12)
2

Hz
E z
k 2 Hx i
i
,
x
y
(3.13)
2

Hz
E z
k 2 Hy i
i
,
y
x
(3.14)
k 2 2 0 x; y
(3.15)
where
is a function of x and y to account for the transverse spatial inhomogeneity of the structure.
The relations in (3.11)(3.14) are generally valid for a longitudinally homogeneous structure
of any transverse geometry and any transverse index prole, for which x; y is not a function
of z. In a structure of cylindrical symmetry, such as an optical ber, the x and y coordinates of
the rectilinear system can be transformed to the and r coordinates of the cylindrical system for
similar relations. It is clear from (3.11)(3.14) that once the longitudinal mode eld components, E z and Hz , are known, all mode eld components can be obtained. Therefore, a normal
mode can be classied based on the characteristics of its longitudinal eld components, as
follows.
1.
2.
3.
4.
A transverse electromagnetic mode, or TEM mode, has E z 0 and Hz 0.

A transverse electric mode, or TE mode, has E z 0 and Hz 6 0.
A transverse magnetic mode, or TM mode, has Hz 0 and E z 6 0.
A hybrid mode has both E z 6 0 and Hz 6 0.
Several comments can be made.

1. Any dielectric optical structure that has an inhomogeneous transverse prole does not
support TEM modes. For such an optical structure, k2 2 0 x; y is not a constant of
space but 2 is always a constant; therefore, all eld components vanish when E z 0 and
Hz 0, as can be seen from (3.11)(3.14).
2. TEM modes exist in (a) a homogeneous dielectric medium without any conductors, (b) the
outside of a single-conductor transmission line in a homogeneous dielectric medium, and (c)
a waveguide consisting of multiple separate conductors in a homogeneous dielectric
medium. For a TEM mode to exist, (3.11)(3.14) require that x; y be a constant
of space so that k2 2 . Therefore, the propagation constant of a TEM mode is simply that
p
of the dielectric medium: k 0 .
3. Only TE and TM modes are allowed in (a) a planar dielectric structure of x; y x and
(b) the inside of a hollow metallic waveguide.
4. TE and TM modes are allowed but are not the only modes in (a) a planar metallic waveguide
consisting of two parallel plates, which also supports TEM modes, and (b) a nonplanar
dielectric waveguide, which also supports hybrid modes.
5. Hybrid modes are allowed in nonplanar dielectric waveguides, but not in planar dielectric
structures. The HE and EH modes of optical bers are hybrid modes.
70
6. From the above discussion, planar dielectric optical structures only have TE and TM modes,
whereas nonplanar dielectric optical structures only have TE, TM, and hybrid modes. None
of them have TEM modes.
EXAMPLE 3.1
Find the general relations between the transverse components of the electric eld and those of
the magnetic eld for (a) a TEM mode, (b) a TE mode, (c) a TM mode, and (d) a hybrid mode.
Solution:
The general relations between the transverse electric-eld components, E x and E y , and the transverse
magnetic-eld components, Hx and Hy , for each type of mode can be found from (3.5)(3.10).
(a) TEM modes: For a TEM mode, E z 0 and Hz 0. Therefore,
Hx
E y E y,
0
Ex
E x:
0
p
From these relations, it is always true that 0 k for a TEM mode.
(b) TE modes: For a TE mode, E z 0 but Hz 6 0. Therefore,
Hy
Hx
E y 6 E y ,
0
E x 6
E x:
0
p
From these relations, it is always true that 6 0 for a TE mode.
(c) TM modes: For a TM mode, Hz 0 but E z 6 0. Therefore,
Hy
Hx
E y,
E y 6
E x:
E x 6
0
p
From these relations, it is always true that 6 0 for a TM mode.
(d) Hybrid modes: For a hybrid mode, E z 6 0 and Hz 6 0. Therefore,
Hy
Hx 6
E y 6 E y ,
0
E x 6
E x:
0
p
From these relations, it is always true that 6 0 for a hybrid mode.
Hy 6
71
3.1.2 Power and Orthonormalization of Modes

The intensity distribution of a normal mode projected on a transverse plane, which has a
surface normal of n^ ^z , is given by

z E H
z,
I S ^z S S
^
E H ^
(3.16)
which is a function of x and y. The power, P , of the mode is obtained by integrating I x; y

over the entire transverse cross-sectional plane. It can be seen from (3.16) that the longitudinal
components, E z and Hz , of the mode elds do not contribute to the mode intensity or the mode
power. Because different normal modes are orthogonal to each other, the mode elds of a
lossless isotropic structure satisfy the orthogonality relation:

^z dxdy P :
E H
E

H
(3.17)

where is the Kronecker delta function for discrete modes, with and representing discrete
numbers; but is the Dirac delta function for continuous modes, with and
representing continuous numbers. For a nonplanar structure, mn and m0 n0 ; hence
mm0 nn0 . For a planar structure, m and m0 ; then, mm0 .
The normal mode elds are normalized according to the following orthonormality relation:

^ H
^ H
Ê
^ ^z dxdy :
E
(3.18)

This orthonormality relation dened in terms of cross products based on the form of the
Poynting vector is valid for all types of modes. Simplied relations in terms of dot products
exist for TE, TM, and TEM modes.
For TE modes, (3.17) can be reduced to
2
0
TE
E E
dxdy P :
(3.19)

Therefore, as an alternative to (3.18), the orthonormality relation among TE modes can also be
written as
2
0
^ dxdy :
^ E
E
(3.20)
1
TM
H H
dxdy P :
x; y
(3.21)

For TM modes, (3.17) can be reduced to

2

72
As an alternative to (3.18), the orthonormality relation among TM modes can also be written as
2

1 ^ ^
H H dxdy :
x; y
(3.22)
The simplied relations for TE modes and those for TM modes are both valid for TEM modes
because a TEM mode is both TE and TM. As discussed above, a TEM mode exists only when
x; y is a constant of space. Therefore, for TEM modes,
2
0
E E
dxdy

TEM
H H
:
dxdy P
(3.23)

There are two equivalent dot-product orthonormality relations among TEM modes:
2
0
^ dxdy
^ E
E

2
and
^ dxdy :
^ H
H
(3.24)

The orthogonality relation in (3.17) and the orthonormality relation in (3.18) indicate that
power cannot be transferred between different normal modes in a linear, lossless structure
of isotropic dielectric medium. For anisotropic or lossy structures, (3.17) and (3.18) do
not apply, neither do the other simplied relations for TE, TM, and TEM modes. The
orthogonality conditions and orthonormality relations for modes of such structures have
other forms.
3.1.3 Mode Expansion

The normal modes are orthogonal and can be normalized with the general orthonormality
relation given in (3.18). They form a basis for linear expansion of any optical eld at a
frequency of propagating in the optical medium:
X
^ x; y exp i z it ,
Er; t
A E
(3.25)
Hr; t
^ x; y exp i z it ,
A H
(3.26)
where the summation symbol sums over all discrete indices of the discrete modes and
integrates over all continuous indices of the continuous modes. In a linear structure where
the normal modes are dened, these modes propagate independently without exchanging
power. Therefore, the expansion coefcients A are constants that are independent of x, y, and z.
According to (3.17) and (3.18), the normal modes are normalized such that the mode power
is simply
P jA j2 :
(3.27)
3.2
73
PLANE-WAVE MODES
..............................................................................................................
A plane wave has wavefronts of innite parallel planes. As dened in Section 1.7, a
wavefront is the surface of a constant phase, and the wavevector is the gradient of the phase,
which is normal to the wavefront. Therefore, a monochromatic plane wave that propagates
in a homogeneous medium is dened by one constant frequency and one constant
wavevector k:
Er; t E exp ik r it ,
(3.28)
Hr; t H exp ik r it ,
(3.29)
where both E and H are constants of space and time. The electric displacement and the magnetic
induction of the plane wave have similar forms: Dr; t Er; t D exp ik r it and
Br; t 0 Hr; t B exp ik r it , where D and B are constants of space and time.
When operating on the elds of a plane wave, the space operator always yields ik and the
time operator =t always yields i. Therefore, for a plane wave propagating in a homogeneous
medium, the following replacements can be made:
! ik,
! i:
t
(3.30)
A monochromatic plane wave is a normal mode of propagation in a homogeneous medium

because it has a well-dened wavevector, thus a well-dened propagation constant. In an
isotropic medium, the propagation constant of a plane wave does not depend on the polarization
of the wave; therefore, a plane wave of any polarization has the same well-dened propagation
constant and is a normal mode. In an anisotropic medium, only elds of certain polarizations
have well-dened propagation constants, as discussed in Section 2.2. Plane-wave normal
modes in a homogeneous anisotropic medium have specic polarization characteristics and
polarization-dependent propagation constants that are determined by both the property of the
medium and the direction of wave propagation.
In any event, for a monochromatic plane-wave normal mode, Maxwells equations as given
in (1.41)(1.44) can be expressed in the algebraic form:
k E 0 H,
(3.31)
k H D,
(3.32)
k D 0,
(3.33)
k H 0:
(3.34)
Note that the relation B 0 H, as is always true for optical elds, is used for the above
equations. The wave propagation direction is dened by the wavevector k, whereas the power
ow direction is dened by the Poynting vector from (1.54):
S E H :
(3.35)
74
By combining (3.31) and (3.32) to eliminate the magnetic eld H, the algebraic form of the
wave equation for a plane wave is obtained:
k k E 2 0 D 0:
(3.36)
A plane-wave normal mode is characterized by six vectors: E, D, H, B, k, and S. Their relations

found from (3.31)(3.35) are summarized as follows.
1. From (3.31) and (3.35), the three vectors E, H, and S are always mutually orthogonal for a
plane wave in any homogeneous medium.
2. From (3.32)(3.34), the three vectors D, H, and k are always mutually orthogonal for a plane
wave in any homogeneous medium.
3. In any optical medium BkH is always true because B 0 H. Both are orthogonal to all of
the other four vectors E, D, k, and S.
4. In a homogeneous isotropic medium, DkE because D E. Both are orthogonal to all of the
other four vectors H, B, k, and S.
5. In a homogeneous anisotropic medium, D is not necessarily parallel to E because D E.
Both D and E are orthogonal to H and B, but E is not necessarily orthogonal to k while D is
not necessarily orthogonal to S.
As expressed in (3.28) and (3.29), a true plane wave transversely extends to innity in
space, which is unrealistic. It is a good approximation if a medium is homogeneous in all
directions over dimensions that are very large compared to the wavelength. Because the
eld amplitude of every plane wave is a constant of space, the difference between two plane
waves of the same frequency that propagate in the same direction is only in their polarization characteristics. Orthogonality between two such plane-wave modes is determined
only by the orthogonality of their polarization states but not by the spatial integral
of their eld overlap. Therefore, for a given wave propagation direction, there are only
two orthogonally polarized plane-wave modes. Furthermore, because a plane wave has a
constant amplitude extending throughout the transverse plane, the integrals that dene
mode normalization in Section 3.1 cannot be performed. For these reasons, the actual
amplitude of each wave is used in the eld expansion though a unit polarization vector is
often used to represent the polarization state of a plane wave. The plane wave basis
for linear expansion of any optical eld that has a frequency of and propagates in the
k^ direction through a homogeneous optical medium consists of only two orthogonally
polarized elements:

Er; t E1 r; t E2 r; t E 1 exp i1 k^ r it E 2 exp i2 k^ r it ,
(3.37)

Hr; t H1 r; t H2 r; t H1 exp i1 k^ r it H2 exp i2 k^ r it ,
(3.38)
where E 1 , H1 , E 2 , and H2 are constants of space; 1 and 2 are the propagation constants of
the two plane-wave modes; and the two modes satisfy the polarization orthogonality relations:
E1 E
2 E 1 E 2 0 and H1 H2 H1 H2 0:
(3.39)
75
Figure 3.2 Relationships among the

directions of E, D, H, B, k, and S in free
space or in an isotropic medium.
In a homogeneous medium, the propagation constants are determined by the material properties
and the polarization states of the waves but not by any optical structure. Therefore, 1 k1 and
2 k2 . The two propagation constants are the same if the medium is isotropic, but they are
generally different if the medium is anisotropic, as discussed below.
3.2.1 Isotropic Medium

The permittivity tensor of a homogeneous isotropic medium reduces to a scalar that is
independent of spatial location and direction. Free space is a special case of homogeneous
isotropic medium with 0 . Figure 3.2 shows the relations among the six vectors E, D, H, B,
k, and S of a plane wave that propagates in a homogeneous isotropic medium. For this plane
wave, EkDk because D E. A plane-wave normal mode of a homogeneous isotropic
medium is a TEM wave because its E and H elds are both orthogonal to its wavevector k.
With Ek, we nd that k k E k2 E. By using this relation and D E, the wave
equation in (3.36) is reduced to
k2 E 2 0 E 0,
(3.40)
which yields the eigenvalue equation:

k 2 2 0 :
(3.41)
Therefore, the propagation constant of the wave in the medium is

p n 2n 2n
k 0
,
c
c
(3.42)
where is the frequency of the optical wave, is its wavelength,

1
c p
0 0
(3.43)
dielectric constant1=2
n
0
(3.44)
is the speed of light in free space, and
76
is the index of refraction, or refractive index, of the isotropic medium. Because k is proportional
to 1=, it is also called the wavenumber. In a medium that has an index of refraction of n, the
optical frequency is still , but the optical wavelength is =n, and the speed of light is v c=n.
Regardless of the propagation direction or the polarization state, all plane waves of the same
frequency in a homogeneous isotropic medium are degenerate and have the same propagation
^ any two orthogonally polarized
constant k found in (3.42). For any given propagation direction k,
plane waves that propagate in the k^ direction can be used as the basis for linear expansion. Both
^ as is seen in Fig. 3.2. Because
are TEM waves and are orthogonal to the propagation direction k,
the medium is isotropic, the coordinates can be chosen such that the z axis is in the direction of
^ Then the eld expansion of (3.37) and (3.38) takes the form:
wave propagation, i.e., ^z k.
Er; t E 1 exp ikz it E 2 exp ikz it E 1 E 2 exp ikz it,
(3.45)
Hr; t H1 exp ikz it H2 exp ikz it H1 H2 exp ikz it:
(3.46)
For propagation in the z direction with k^ ^z as considered here, any two orthogonal polarization
states in the xy plane can be used as the basis set for the eld expansion. For example, the basis
set can be formed by the two linearly polarized waves E x ^x and E y ^y , by the two circularly
polarized waves E ê and E ê , or by any two orthogonal elliptically polarized waves. It can
be seen from (3.45) and (3.46) that the linear superposition of two plane-wave normal modes of
a homogeneous isotropic medium is also a normal mode of the same propagation constant.
Hence any plane wave of a given frequency traveling in a homogeneous isotropic medium is
a normal mode with the same propagation constant k. This is not true for plane waves traveling
in a homogeneous anisotropic medium, which is discussed below.
EXAMPLE 3.2
GaAs is a cubic crystal. At the 900 nm wavelength, its principal indices of refraction
are nx ny nz 3:593. A circularly polarized wave and a linearly polarized wave at this
wavelength propagate along the z and x principal axes, respectively. What are the propagation
constants and the wavelengths of these two waves in the GaAs crystal?
Solution:
Though GaAs has well-dened principal axes, it is optically isotropic because nx ny
nz n. Therefore, a plane wave of any polarization state propagating in any direction
is a normal mode that has a refractive index of n. At 900 nm, n 3:593. For both waves,
we nd the propagation constant to be
k
2n 2 3:593
2:51 107 m1
900 nm
and the wavelength in GaAs to be

GaAs
900 nm
250:5 nm:
n
3:593
77
3.2.2 Anisotropic Medium

As discussed in Sections 2.2, 2.6, and 2.7, the anisotropy of a medium can be intrinsic, such as
that of an anisotropic crystal, or it can be induced by an external factor, such as that caused by
an electro-optic, magneto-optic, acousto-optic, or nonlinear optical effect. The principal normal
modes associated with linear or circular birefringence have already been discussed in Section
2.2. Here we consider only linear birefringence of an anisotropic crystal characterized by a
symmetric dielectric tensor whose eigenvectors dene the principal axes ^x , ^y , and ^z with
eigenvalues x , y , and z , respectively.
Plane-wave normal modes still exist for wave propagation in a homogeneous anisotropic
medium. However, their characteristics depend on the direction of propagation with respect to
the principal axes of the medium. In contrast to plane-wave normal modes in an isotropic
medium, all of which are degenerate with the same propagation constant, plane-wave normal
modes in an anisotropic medium are generally nondegenerate. Their polarization states and
propagation constants are specic to each propagation direction. Three general cases are
discussed in the following.
Propagation along an Optical Axis
In the special case of propagation along an optical axis, the crystal appears to be isotropic to the
wave. For a uniaxial crystal, the optical axis is one of the principal axes, taken to be the z
principal axis by convention. For a biaxial crystal, neither of the two optical axes is a principal
axis. In any event, by the denition of optical axis, a wave does not experience any birefringence when it propagates along an optical axis. Then the plane-wave normal modes have the
same characteristics as those discussed above for an isotropic medium. All plane waves
polarized in the plane normal to an optical axis are normal modes of propagation along this
optical axis, and any two of them that are orthogonally polarized can be used as the basis for
linear expansion.
EXAMPLE 3.3
LiNbO3 is a negative uniaxial crystal that has principal refractive indices of nx ny no
2:238 and nz ne 2:159 at the 1 m wavelength. Find the possible arrangements for (a)
a linearly polarized wave and (b) a circularly polarized wave to propagate through LiNbO3 with
a propagation constant dened by either no or ne . In each case, nd the propagation constant
and the wavelength for the wave in LiNbO3 .
Solution:
The refractive index seen by a wave is determined by the polarization of the wave. Then, the
possible direction of propagation is constrained by a given polarization. Because the z principal
axis of the uniaxial LiNbO3 crystal is an optical axis, a wave that propagates along the z
direction with its polarization in the xy plane sees the crystal as optically isotropic with no
without seeing ne .
78
(a) A linearly polarized wave at 1 m sees no 2:238 if it is polarized in any direction in

the xy plane. This is always true when the wave propagates along the z principal axis. Then,
it has
ko
2no 2 2:238
1:41 107 m1
1 m
and
1 m
446:8 nm:
no 2:238
A linearly polarized wave sees ne 2:159 if it is polarized along the z principal axis. This is
possible only when the wave propagates in a direction that lies in the xy plane. Then, it has
ke
2ne 2 2:159
1:36 107 m1
1 m
and
1 m
463:2 nm:
ne 2:159
(b) A circularly polarized wave at 1 m sees no 2:238 if its circular polarization lies in
the xy plane. For this to happen, the wave has to propagate along the z principal axis. It has
ko
2no 2 2:238
1:41 107 m1
1 m
and
1 m
446:8 nm:
no 2:238
There is no possible arrangement for a circularly polarized wave to propagate in a uniaxial

crystal with a propagation constant dened by ne .
Propagation along a Principal Axis

When an optical wave propagates in a direction other than that along an optical axis, the index
of refraction depends on the direction of its polarization. In this situation, there exist two normal
modes of linearly polarized waves, each of which has a unique index of refraction. If the
propagation direction is along a principal axis that is not an optical axis, the two normal modes
are simply the principal modes of polarization that are linearly polarized along the other two
principal axes. Each principal mode of polarization has its characteristic principal index of
refraction.
Without loss of generality, take the principal axis along which the wave propagates to be the z
^ z . In the case when the z principal axis is not an optical axis, the other
principal axis so that kk^
two principal axes ^x and ^y , which are orthogonal to the propagation direction, are birefringent
with different principal permittivities, x 6 y , thus different propagation constants: k x 6 k y ,
where kx nx =c and ky ny =c as dened in (2.15). Note that kx and ky are the propagation
constants of the x- and y-polarized principal normal modes, respectively, not to be confused
with the x and y components of a wavevector k, which are normally expressed as kx and ky :
These two plane wave principal normal modes are
E 1 ^x E 1 ^x E x ,
E 2 ^y E 2 ^y E y ,
H1 ^y H1 ^y Hy ,
H2 ^x H2 ^x Hx ,
k1 1 k^ k x ^z ,
k2 2 k^ k y ^z :
(3.47)
In the form of (3.37) and (3.38), these two normal modes form the basis for linear decomposition of any plane wave that propagates along the z principal axis.
79
Figure 3.3 Evolution of the polarization state of an optical wave propagating along the principal axis ^z of an
anisotropic crystal that has nx 6 ny . Only the evolution over one half-period is shown here. (a) The optical
wave is initially linearly polarized at an arbitrary angle with respect to the principal axis ^x . (b) The optical
wave is initially polarized at 45 with respect to ^
x.
For a plane wave propagating along ^z , the electric eld can be expressed as
Er; t E1 r; t E2 r; t ^x E x exp ikx z it ^y E y exp iky z it :
(3.48)
Because the wave propagates in the z direction, the wavevectors are kx kx ^z for the x-polarized
eld and ky k y ^z for the y-polarized eld. The eld expressed in (3.48) has the following
propagation characteristics.
1. If Er; t is originally linearly polarized along one of the principal axes, i.e., E y 0 for
Er; t E1 r; t k^x or E x 0 for Er; t E2 r; t k^y , it remains linearly polarized in the
same direction as it propagates.

2. If Er; t is originally linearly polarized at an angle of tan1 E y =E x with respect to the
x axis with E1 r; t 6 0 and E2 r; t 6 0, its polarization state varies periodically along z
with a period of 2=jk y kx j because the two normal modes propagate with different
propagation constants. In general, its polarization follows a sequence of variations from
linear to elliptic to linear in the rst half-period and then reverses the sequence back to linear
in the second half-period. At the half-period position, it is linearly polarized at an angle of
on the other side of the x axis. Thus the polarization is rotated by 2 from the original
direction, as shown in Fig. 3.3(a). In the special case when 45 , the wave is circularly
polarized at the quarter-period point and is linearly polarized at the half-period point with its
polarization rotated by 90 from the original direction, as shown in Fig. 3.3(b).
These characteristics have very useful applications. A plate of an anisotropic material that has
a quarter-period thickness of
l=4
1
2

y
x
4 jk k j 4 ny nx
(3.49)
80
is called a quarter-wave plate. It can be used to convert a linearly polarized wave to circular or
elliptic polarization, and vice versa. A plate that has a thickness of 3l=4 or 5l=4 , or any odd
integral multiple of l=4 , also has the same function. By contrast, a plate that has a half-period
thickness of
l=2
1
2

y
x
2 jk k j 2 ny nx
(3.50)
is called a half-wave plate. It can be used to rotate the polarization direction of a linearly
polarized wave by any angular amount by properly choosing the angle between the direction
of the incident linear polarization and the principal axis ^x , or ^y , of the crystal. A plate of a
thickness that is any odd integral multiple of l=2 has the same function. Note that though the
output from a quarter-wave or half-wave plate can be linearly polarized, the wave plates are not
polarizers. Wave plates and polarizers are based on different principles and have completely
different functions. For the quarter-wave and half-wave plates discussed here, nx 6 ny . Between
the two principal axes ^x and ^y , the one with the smaller index is called the fast axis, while the
other, with the larger index, is the slow axis.
EXAMPLE 3.4
At 1 m, the principal indices of refraction of the KTP crystal are nx 1:742,
ny 1:750, and nz 1:832. Is the crystal uniaxial or biaxial? If you want to propagate a
linearly polarized wave through it, how do you arrange it so that its linear polarization is
maintained throughout the propagation path in the crystal? If the crystal is used to make a
half-wave plate for 1 m, what is the minimum thickness of the plate? In which direction
must the wave propagate to use this half-wave plate? Note that there is only one possible
minimum thickness.
Solution:
Because nx 6 ny 6 nz , the crystal is biaxial. To maintain linear polarization throughout, the
wave has to be linearly polarized along one of the principal axes while propagating along
a direction that is perpendicular to its polarization direction. Its propagation constant is
determined by its polarization direction but not by its propagation direction. For example, it
can be polarized in the x direction while propagating in any direction in the yz plane. In this
case, the wave sees nx and has a propagation constant of kx 2nx =.
Because the largest difference between two principal refractive indices is nz nx
1:832 1:742 0:09, the wave must propagate along the y axis of the crystal and have
its polarization in the zx plane, but not along the x or z axis, to utilize this birefringence for
the minimum thickness of the half-wave plate:
l=2
1:00
m 5:56 m:
2jnz nx j 2j1:832 1:742j
81
Figure 3.4 Relationships among the direction of wave

propagation and the polarization directions of the
ordinary and extraordinary waves.
Propagation in a General Direction

In the general case when the propagation direction is neither along an optical axis nor
along a principal axis, there still exist two linearly polarized normal modes. For simplicity,
the propagation in a uniaxial crystal is considered. The z principal axis of the uniaxial
crystal is the optical axis, and the wave propagation direction k^ is at an angle of with
respect to the z principal axis and at an angle of with respect to the x principal axis, as
shown in Fig. 3.4.
One of the normal modes is the polarization that is perpendicular to the optical axis. This
normal mode is called the ordinary wave. We use ê o to indicate its direction of polarization.
The other normal mode is clearly perpendicular to ê o because the two normal-mode polarizations are orthogonal to each other. This normal mode is called the extraordinary wave,
and we use ê e to indicate its direction of polarization. Note that these are the directions
of D rather than those of E. For the ordinary wave, ê o kDo kEo . For the extraordinary wave,
ê e kDe =
kEe except when ê e is parallel to a principal axis. Both ê o and ê e , being the unit vectors
of Do and De , are perpendicular to the propagation direction k^ because D is always perpen^ From this understanding, both ê o and ê e can be found if both k^ and the optical
dicular to k.
axis ^z are known:
ê o
1 ^
k ^z ,
sin
^
ê e ê o k:
(3.51)
These vectors are illustrated in Fig. 3.4. They can be expressed as

k^ ^x sin cos ^y sin sin ^z cos ,
(3.52)
^ ,
ê o ^x sin y cos
(3.53)
ê e ^x cos cos ^y cos sin ^z sin :
(3.54)
82

Figure 3.5 Determination of the indices of refraction for the
ordinary and extraordinary waves in a uniaxial crystal using
index ellipsoid.
The indices of refraction associated with the ordinary and extraordinary waves can be found
by using the index ellipsoid dened as
x2 y2 z2
1:
n2x n2y n2z
(3.55)
The index ellipsoid for the uniaxial crystal under consideration is illustrated in Fig. 3.5 with
nx ny no and nz ne . The intersection of the index ellipsoid and the plane normal to k^ at
the origin of the ellipsoid denes an index ellipse. The principal axes of this index ellipse are in
the directions of ê o and ê e , and their half-lengths are the corresponding indices of refraction.
For a uniaxial crystal, the index of refraction for the ordinary wave is simply no . The index of
refraction for the extraordinary wave depends on the angle and is given by
1
cos2 sin2
2 ,
n2e
n2o
ne
(3.56)
which can be seen from Fig. 3.5. We see that ne 0 no and ne 90 ne . For 0 , the
propagation direction k^ is along the optical axis. For 90 , the propagation direction k^ lies
in the plane perpendicular to the optical axis; in a uniaxial crystal, this situation is the same as
when k^ is along a principal axis that is not the optical axis.
Each of the two normal modes has a well-dened propagation constant; the ordinary
wave has k o no =c and the extraordinary wave has ke ne =c. Maxwells equations
in the form of (3.31)(3.34) have to be separately written with different values of k for
the ordinary and the extraordinary normal modes; no such form applies to a wave that is a
^ for the extraordinary way,
mixture of the two modes. For the ordinary way, k ko ko k;
^
k ke k e k.
83
EXAMPLE 3.5
LiNbO3 is a negative uniaxial crystal that has principal refractive indices of nx ny no
2:238 and nz ne 2:159 at the 1 m wavelength. Find the polarization directions ê o and
ê e , and the corresponding propagation constants k o and ke , of the ordinary and extraordinary
normal modes for a propagation direction k^ that makes an angle of 30 with respect to the
x principal axis and an angle of 45 with respect to the z principal axis.
Solution:
With 30 and 45 , we nd by using (3.52)(3.54) that
p
p
p
p
p
p
p
3
1
6
2
2
6
2
2
^x
^y
ê o ^x
^x
^y
^y , ê e
^z ,
^z :
k^
4
4
2
4
4
2
2
2
At 45 , we nd by using (3.56) that
2
1=2
cos 45
sin2 45
2:197:
ne 45
2:2382
2:1592

Therefore, the propagation constants of the two normal modes are, respectively,
ko
ke
2no 2 2:238
1:41 107 m1 ,
1 m
2ne 45 2 2:197
1:38 107 m1 :
1 m
Because D is always perpendicular to the propagation direction, Dk for both ordinary and
extraordinary waves. For an ordinary wave, Eo ko because Eo kDo . Therefore, the relationships shown in Fig. 3.6(a) among the eld vectors for an ordinary wave in an anisotropic
medium are the same as those shown in Fig. 3.2 for a wave in an isotropic medium. For an
extraordinary wave, in general Ee k
= e because Ee =kDe ; thus Se is not necessarily parallel to ke .
This means that Ee is not transverse to ke but has a longitudinal component in the ke direction.
The only exception is when ê e is parallel to a principal axis. As a result, the direction of power
ow, which is that of Se , is not the same as the direction of wave propagation, which is that
of ke and is normal to the wavefronts, i.e., the planes of constant phase. Their relationship is
shown in Fig. 3.6(b) together with the relationships among the directions of the eld vectors.
Note that Ee , De , ke , and Se lie in the plane normal to He because Be kHe . Though it is still true
that Ee He because ke Ee kHe according to (3.31), ke He =kEe because ke He kDe
according to (3.32).
These two plane-wave normal modes have the following characteristics:
E o ê o E o ,
Do ê o Do ,
Ho ê e Ho ,
^
ko ko k;
^ k
E e ê e E
e kE e ,
De ê e De ,
He ê o He ,
^
ke ke k;
(3.57)
84
Figure 3.6 Relationships among the directions of E, D, H, B, k, and S in an anisotropic medium for (a) an ordinary
wave and (b) an extraordinary wave. In both cases, the vectors E, D, k, and S lie in a plane normal to H.
where E
e e and E ke E e k^ are, respectively, the transverse and longitudinal compone Ee ^
ents of the electric eld of the extraordinary wave. Note that only E e has a longitudinal
component, and this component vanishes when ê e is parallel to a principal axis. Note also that
Ho kk^ ê o ê e and He kk^ ê e ê o because 0 H k E for each mode, according to
(3.31). In the form of (3.37) and (3.38), these two normal modes form the basis for the linear
expansion of any plane wave propagating along the k^ direction:

Er; t Eo r; t Ee r; t E o exp ik o k^ r it E e exp ik e k^ r it ,
(3.58)

Hr; t Ho r; t He r; t Ho exp iko k^ r it He exp ik e k^ r it :
(3.59)
If the electric eld of an extraordinary wave is not parallel to a principal axis, its Poynting
vector is not parallel to its propagation direction because Ee is not parallel to De . As a result,
its energy ows away from its direction of propagation. This phenomenon is known as spatial
beam walk-off. If this characteristic appears in one of the two normal modes of an optical wave
propagating in an anisotropic crystal, the optical wave splits into two beams that have parallel
wavevectors but separate, nonparallel traces of energy ow.
Consider a plane wave that propagates in a uniaxial crystal along a general direction k^ at an angle
of with respect to the optical axis ^z ; this wave consists of both ordinary and extraordinary waves,
as described by (3.58) and (3.59). Clearly, there is no walk-off for the ordinary wave because
^ For the extraordinary wave, Se is not parallel to k^ but points in a direction at an
Eo kDo so that So kk.
angle of e with respect to the optical axis. Figure 3.7(a) shows the relationships among these
^ which is dened as e , is called the walk-off angle
vectors. The angle between Se and k,
of the extraordinary wave. Note that is also the angle between Ee and De , as is seen in Fig. 3.7(a).
Because neither Ee nor De is parallel to any principal axis, their relationship is found through their
projections on the principal axes: Dez 0 n2e E ez and Dex, y 0 n2o E ex, y . Using these two relations and
the denition of in Figs. 3.6(b) and 3.7(a), it is found that the walk-off angle is given by
2

no
e tan
tan :
n2e
1
(3.60)
If the crystal is negative uniaxial, as dened in Fig. 3.6(b) is positive. This means that k^
is between Se and ^z for a negative uniaxial crystal. If the crystal is positive uniaxial, is
85
Figure 3.7 (a) Wave propagation and walk-off in a uniaxial crystal. (b) Birefringent plate acting as a polarizing
beam splitter for a normally incident wave. The ^x , ^y , and ^z unit vectors indicate the principal axes of the
birefringent plate.
negative and Se is between k^ and ^z . No walk-off appears if an optical wave propagates along
any of the principal axes of a crystal.
A birefringent crystal can be used to construct a simple polarizing beam splitter by taking
advantage of the walk-off phenomenon. For such a purpose, a uniaxial crystal is cut into a plate
whose surfaces are at an oblique angle with respect to the optical axis, as shown in Fig. 3.7(b).
When an optical wave is normally incident on the plate, it splits into ordinary and extraordinary
waves in the crystal if its original polarization contains components of both polarizations.
The extraordinary wave is separated from the ordinary wave because of spatial walk-off, creating
two orthogonally polarized beams. Because of normal incidence, both ke and ko are parallel to k^
although they have different magnitudes. When both beams reach the other side of the plate, they
are separated by a distance of d l tan jj, where l is the thickness of the plate. After leaving the
plate, the two spatially separated beams propagate parallel to each other in the same k^ direction
because the directions of their wavevectors have not changed, as also shown in Fig. 3.7(b).
EXAMPLE 3.6
LiNbO3 is a negative uniaxial crystal that has principal refractive indices of nx ny
no 2:238 and nz ne 2:159 at the 1 m wavelength. Find the walk-off angle of
of the extraordinary wave in LiNbO3 for a propagation direction k^ that makes an angle
of 30 with respect to the x principal axis and an angle of 45 with respect to
the z principal axis. If a collimated optical beam that consists of both ordinary and
extraordinary components at this wavelength propagates in this direction through a
LiNbO3 plate, how thick must the plate be for the ordinary and extraordinary beams
to be separated by at least 100 m?
86
Solution:
The walk-off angle for 45 is found by using (3.60) to be

2
1 2:238

tan 45 45 2:06 :
tan
2:1592
For the ordinary and extraordinary beams to be separated by at least 100 m,
d l tan > 100 m ) l >
100 m
2:78 mm:
tan 2:06
Thus, the thickness of the plate has to be at least 2:78 mm.
3.3
GAUSSIAN MODES
..............................................................................................................
A monochromatic optical wave propagating in a homogeneous isotropic medium is
governed by Maxwells equations for wave propagation given in (3.3) and (3.4). In this
situation, is a scalar constant so that D E and E D= 0: Then,
E E r2 E r2 E. By using this relation while combining (3.3) and
(3.4), we obtain the simple wave equation that is specic for the propagation of a monochromatic wave in a homogeneous isotropic medium:
r2 E 2 0 E 0,
(3.61)
where the substitution of =t ! i is taken for the monochromatic wave at the frequency .
Because every term in (3.61) has the same constant unit vector, the vectorial wave equation can
be reduced to the scalar Helmholtz equation:
r2 E k2 E 0,
(3.62)
where k2 2 0 , as dened in (3.41). A similar equation can be written for the magnetic eld.
Clearly, a monochromatic plane wave of the form in (3.28) and (3.29) is a solution of the
equations for wave propagation given in (3.3) and (3.4), which in this case reduce to the simple
form of (3.31) and (3.32) with D E; thus, it is a solution of the wave equation in (3.61).
Therefore, plane waves are normal modes of propagation in a homogeneous isotropic medium.
They are not the only normal modes, however, as the equations that govern wave propagation
in such a medium have other normal-mode solutions.
One important set of modes is the Gaussian modes. Like plane waves, Gaussian modes are
normal modes of wave propagation in a homogeneous isotropic medium. Different from a plane
wave, a Gaussian mode has a nite cross-sectional eld distribution dened by its spot size. Being
an unguided eld that has a nite spot size, a Gaussian mode differs from a waveguide mode,
discussed in Section 3.5, in that its spot size varies along its longitudinal axis, taken to be the
z axis, of propagation though its pattern remains unchanged. Its transverse eld distribution also
changes with z though the eld pattern does not change. The beam has a nite divergence angle, .
3.3 Gaussian Modes
87
For a collimated Gaussian beam that has a small divergence angle such that the paraxial
approximation
sin 1
(3.63)
is valid, the propagation constant of the Gaussian normal mode is k. Therefore, rather
than those in (3.1) and (3.2), the electric and magnetic elds of a monochromatic Gaussian
mode at a frequency of can be expressed as
Emn r; t E mn x; y; z exp ikz it ê E mn x; y; z exp ikz it,
(3.64)
Hmn r; t Hmn x; y; z exp ikz it k^ ê Hmn x; y; z exp ikz it ,
(3.65)
where m and n are mode indices associated with the two transverse dimensions x and y,
respectively. The paraxial approximation requires that
2

E
E

k and E , E , E jkE j
(3.66)
z2
z
x y z
for the electric eld amplitude, and there are similar relations for the magnetic eld amplitude.
In this approximation, the Helmholtz equation in (3.62) reduces to
2 E 2 E
E
2 i2k
0
2
x
y
z
(3.67)
for the electric eld amplitude in (3.64). The magnetic eld amplitude in (3.65) satises an
equation in H of the same form.
In the paraxial approximation, a Gaussian mode eld is a TEM mode that has only transverse
electric and magnetic eld components; it has neither longitudinal electric nor longitudinal
magnetic eld components. Then, the unit polarization vector ê for the electric mode eld in
(3.64) is polarized in the transverse xy plane; the unit vector k^ ê for the magnetic mode eld
in (3.65) is also polarized in the transverse xy plane because k^ ^z . The paraxial approximation
is not valid when a Gaussian beam is very tightly focused to the extent that its spot size is on the
order of its optical wavelength. In this situation, the longitudinal electric and magnetic eld
components cannot be ignored; such a Gaussian mode eld is not truly TEM.
The electric mode elds of Gaussian modes in the paraxial approximation are eigenfunctions
of (3.67); the corresponding magnetic mode elds have the same form because they are
eigenfunctions of an equation of H that has the same form as (3.67). As TEM modes, they
can be normalized by the dot-product orthonormality relations given in (3.24):
2k
0
^ 0 0 x; y; zdxdy
^ mn x; y; z E
E
mn
2k
^ mnx; y; z H
^ 0 0 x; y; zdxdy mm0 nn0 :
H
mn
(3.68)
The Gaussian beam eigenfunctions of (3.67) in the paraxial approximation have several salient
characteristics. A Gaussian beam has a nite spot size that varies with location along the
88
Figure 3.8 Gaussian beam characteristics.
propagation axis. The location where the smallest spot size of the beam occurs is known as the
waist of the Gaussian beam. This beam waist location is taken to be z 0 for a beam that
propagates in the direction along the z axis. The minimum Gaussian beam spot size, w0 , is
dened as the e1 radius of the Gaussian beam electric eld magnitude prole, i.e., the e2
radius of the Gaussian beam intensity prole, at the beam waist. The diameter of the beam waist
is d 0 2w0 : As illustrated in Fig. 3.8, a Gaussian beam has a plane wavefront at its beam waist.
The beam remains well collimated within a distance of
zR
kw20 nw20
,
2
(3.69)
p
known as the Rayleigh range, on either side of the beam waist. In (3.69), k 0 2n=
is the propagation constant of the optical beam in a medium of a refractive index n. The
parameter b 2zR is called the confocal parameter of the Gaussian beam.
Because of diffraction, a Gaussian beam diverges away from its waist and acquires a
spherical wavefront at a far-eld distance, where jzj zR . As a result, both its spot size,
wz, and the radius of curvature, Rz, of its wavefront are functions of the distance z from its
beam waist:
"

1=2

#1=2
z2
2z 2
wz w0 1 2
w0 1
(3.70)
zR
kw20
and
"

2 2 #
z2R
kw0
Rz z 1 2 z 1
:
(3.71)
z
2z
p
We see from (3.70) that w 2w0 at z zR . At jzj zR , far away from the beam waist, we
nd that Rz z and wz 2jzj=kw0 . Therefore, the far-eld beam divergence angle is
2
wz
4
2
kw0 nw0
jzj
(3.72)
3.3 Gaussian Modes
89
For the far eld at jzj zR , we nd that the beam spot size wz is inversely proportional to
the beam waist spot size w0 but is linearly proportional to the distance jzj from the beam waist.
This characteristic does not exist for the near eld at jzj
zR :
From (3.72), it can be seen that the paraxial approximation sin 1 expressed in
(3.63) is valid when the beam is well collimated so that the spot size is much larger than the
optical wavelength in the medium: w0 =n. Then the Gaussian mode elds are TEM modes.
This is normally the case for Gaussian wave propagation. The Gaussian mode elds are not
TEM when the beam is tightly focused such that the spot size is on the order of the optical
wavelength. In this situation, w0 =n, and the paraxial approximation is invalid.
EXAMPLE 3.7
A Gaussian beam from a Nd:YAG laser at the 1:064 m wavelength propagates in free
space with a beam divergence of 1 mrad. Find the beam waist spot size, the Rayleigh range,
and the confocal parameter of the beam. What are the spot sizes and the radii of curvature of
the beam at the distances of 10 cm, 1 m, 10 m, and 1 km, respectively?
Solution:
Given 1:064 m and 1 mrad, we nd from (3.72) that the beam waist spot size is
w0
2
2 1:064 m
677 m:
1 103
From (3.69), the Rayleigh range and the confocal parameter are found:

2
w20 677 106
zR
m 1:35 m and b 2zR 2:7 m:
1:064 106
By using (3.70) and (3.71), the spot sizes and the radii of curvature at different locations are
found:
w 695 m
w 843 m
w 5:06 mm
w 50:1 cm
R 18:33 m at z 10 cm,
R 2:82 m
at z 1 m,
R 10:18 m at z 10 m,
R 1 km
at z 1 km:
Within the Rayleigh range, both the spot size and the radius of curvature vary nonlinearly with
distance; the spot size increases slowly, whereas the radius of curvature decreases with distance.
At a large distance, both the spot size and the radius of curvature increase approximately linearly
with distance as the Gaussian beam approaches a spherical wave.
A complete set of Gaussian modes in the paraxial approximation includes the fundamental
TEM00 mode and high-order TEMmn modes. The specic forms of the mode elds depend
on the transverse coordinates of symmetry: the mode elds are described by a set of
HermiteGaussian functions in the rectilinear coordinates, whereas they are described by the
90
LaguerreGaussian functions in the cylindrical coordinates. Both sets are equally valid in free
space or in a homogeneous isotropic medium because there is no structurally dened symmetry.
Usually the HermiteGaussian functions in the rectilinear coordinates are used. In a transversely isotropic and homogeneous medium, a normalized TEMmn HermiteGaussian mode
eld propagating along the z axis can be expressed as

p
p
2x
2y
Cmn
k x2 y2
^
Hm
exp i mn z
Hn
exp i
E mn x; y; z
wz
wz
wz
2 qz
(3.73)

p
p
2
2
2
2
2x
2y
C mn
x y
kx y
Hm
exp i
exp i mn z,
Hn
exp 2
wz
w z
wz
wz
2 Rz
^ mn x; y; z k E^ mn x; y; z,
H
0
(3.74)
1=2
is the normalization constant, H m is the Hermite

where Cmn 0 =k 1=2 2mn m!n!
polynomial of order m, qz is the complex radius of curvature of the Gaussian wave given by
qz z izR or
1
1
2
i 2 ,
qz Rz
kw z
and mn z is a mode-dependent on-axis phase variation along the z axis given by

2z
1 z
1
mn z m n 1tan
m n 1 tan
:
zR
kw20
(3.75)
(3.76)
The Hermite polynomials can be obtained using the following relation:

2
dm e
:
H m 1 e
d m
m 2
(3.77)
Some low-order Hermite polynomials are

H 3 8 3 12:
(3.78)

We see from (3.73) and (3.78) that the transverse eld distribution E^ 00 x; y of the
fundamental TEM00 Gaussian mode at a xed longitudinal location z is simply a Gaussian
1=2
function of the transverse radial distance r x2 y2 and that the spot size wz is the e1
radius of this Gaussian eld distribution at z. The transverse eld distribution of a high-order
TEMmn mode is the Gaussian function spatially modulated by the Hermite polynomials H m x
and H n y in the x and y directions, respectively. As a result, its eld distribution spreads out
radially farther than that of the fundamental TEM00 mode. In general, the higher the order of a
mode is, the farther its transverse eld distribution spreads out. The intensity patterns of some
low-order HermiteGaussian modes are shown in Fig. 3.9. The HermiteGaussian modes are
dened in the rectilinear x; y; z coordinates. Because a homogeneous isotropic medium is also
cylindrically symmetric with respect to the wave propagation direction, it is also possible to
dene a complete set of the TEM Gaussian modes, known as the LaguerreGaussian modes, in
the cylindrical r; ; z coordinates with z being the longitudinal wave propagation direction.
The HermiteGuassian modes have rectilinear symmetry in the transverse plane, whereas the
H 0 1,
H 1 2,
H 2 4 2 2,
3.3 Gaussian Modes
91
Figure 3.9 Intensity patterns of some low-order HermiteGaussian modes.
LaguerreGaussian modes have circular and radial symmetry in the transverse plane. Each set
is a complete set of modes for eld expansion, and one set can be mathematically transformed
to the other set by linear expansion.
EXAMPLE 3.8
Find the transverse intensity distribution of the fundamental Gaussian mode as a function of the
distance z from the beam waist. Given a fundamental Gaussian beam of a power P, nd the
intensity I 0 z at the beam center as a function of the distance z. Express P and I 0 z in terms of
the beam spot sizes w0 at the beam waist and wz at the location z.
Solution:
For the fundamental Guassian mode, m n 0. Because the zeroth-order Hermite function is
a constant, H 0 x H 0 y 1, we nd from (3.73) that the fundamental Guassian mode eld
1=2
varies with x and y as x2 y2 so that E^ 00 x; y; z E^ 00 r; z, where r x2 y2 is the
transverse radial coordinate variable. Because a Guassian mode is a TEM mode, its eld

2

intensity is I r; z / E^ 00 r; z . Then, using (3.73), we can express I r; z as

2r2
,
I r; z I 0 zexp 2
w z
where I 0 z is the intensity at the beam center r 0. The power of the beam is found by
integrating the intensity distribution over the transverse plane:

2r 2
w2 z
P I r; z2rdr I 0 z exp 2
2rdr
I 0 z:
w z
2
0
Note that the power of a beam is a constant that does not vary with the propagation distance z.
By contrast, the intensity at the beam center varies with z as
I 0 z
2P
:
w2 z
In terms of the parameters at the beam waist,

P
w20
w2
I 0 0 and I 0 z 2 0 I 0 0:
2
w z
92
For Gaussian beam propagation in a homogeneous isotropic medium along the longitudinal
coordinate axis ^z , any two mutually orthogonal unit polarization vectors ê 1 and ê 2 in the
transverse xy plane can be chosen as the polarization basis for linear decomposition of the wave
polarization. Thus, the linear expansion of a Gaussian beam eld can be expressed as
X
X
Er; t ê 1
Amn, 1 E^ mn x;y;z exp ikz it ê 2
Amn, 2 E^ mn x;y;z exp ikz it, (3.79)
m, n
m, n
Hr; t
k ^
k
^z Er; t,
k Er; t
0
0
(3.80)
where ê 1 ^z ê 2 ^z 0 and ê i ê
j ij .
The concept discussed above can be extended to Gaussian beam propagation in a homogeneous anisotropic crystal. For simplicity, consider the case when the propagation direction k^ is
along a principal axis ^z that is not an optical axis so that nx 6 ny . As discussed in Section 3.2,
the two principal modes of polarization, ^x and ^y , form the unique basis for polarization
decomposition of TEM waves propagating along the z axis, when the x and y principal axes
are birefringent. In this situation, the Gaussian eld is decomposed into two linearly polarized
components that propagate with different propagation constants: k x nx =c and ky ny =c
for the x and y polarizations, respectively. The linear expansion of such a Gaussian beam eld
can be expressed as
Er; t Ex r; t Ey r; t
X
X
^x
Amn, x E^ mn, x x; y; z exp ik x z it ^y
Amn, y E^ mn, y x; y; z exp iky z it ,
m, n
m, n
(3.81)
Hr; t
kx
ky
^z Ex r; t
^z Ey r; t:
0
0
(3.82)
Because all of the characteristic parameters dened in (3.69)(3.72) for a Gaussian mode
eld are functions of the refractive index n, the two polarization modes in (3.81) have different
Gaussian beam parameters besides having different propagation constants. Therefore, in addition to changing its polarization state along the propagation axis as was the case for the plane
wave discussed in Section 3.2, a Gaussian beam that propagates in an anisotropic medium can
have two different spot sizes, two different divergence angles, and two different radii of
curvature between the two principal polarization modes. The beam typically has an elliptic
cross-sectional prole. When focused by a spherical lens, the two polarization modes are
focused at different focal points with different beam waist spot sizes.
3.4
INTERFACE MODES
..............................................................................................................
The simplest optical structure is a planar interface separating two semi-innite homogeneous
media, as shown in Fig. 3.1(b). The coordinates are chosen as shown in Fig. 3.1(b), with
the interface located at x 0 such that x 1 for x > 0 and x 2 for x < 0. The
3.4 Interface Modes
93
permittivities 1 and 2 of the two media are scalar constants, whereas the permeabilities
are simply 0 at optical frequencies. As discussed in Section 3.1, only TE and TM modes
are possible for this structure. Take the z axis to be the wave propagation direction.
Then, because the index prole is independent of the y coordinate and the wavevector has
no y component, all eld components have no variations in the y direction: E=y 0 and
H=y 0.
1. TE mode: For any TE mode of a planar structure, E z 0. It can be seen from (3.11)(3.14)
that E x 0, and Hy 0 as well because Hz =y 0. The only nonvanishing eld
components are Hx , E y , and Hz . Once the only nonvanishing electric eld component E y
is found for a TE mode, the two nonvanishing magnetic eld components can be obtained
by using (3.5) and (3.7):
E y,
0
(3.83)
1 E y
:
i0 x
(3.84)
Hx
Hz
2. TM mode: For any TM mode of a planar structure, Hz 0. It can be seen from (3.11)
(3.14) that Hx 0, and E y 0 as well because E z =y 0. The only nonvanishing
eld components are E x , Hy , and E z . Once the only nonvanishing magnetic eld component
Hy is found for a TM mode, the two nonvanishing electric eld components can be obtained
by using (3.8) and (3.10):
Ex
Ez
Hy ,
(3.85)
1 Hy
:
i x
(3.86)
In the case of a planar structure, it is convenient to solve for the unique transverse eld
component rst: E y for a TE mode and Hy for a TM mode. The other eld components,
including the longitudinal component, then follow directly.
3.4.1 Reection and Refraction

We rst consider the simple case of reection and refraction of plane waves at the planar
interface of two media as shown in Fig. 3.1(b). With the coordinates described above, the
interface is located at x 0 and the plane of incidence is the xz plane so that all wavevectors
have no y component. We assume that the optical wave is incident from the medium of 1 with
a wavevector of ki , while the reected wave has a wavevector of kr and the transmitted wave
has a wavevector of kt .
Because an optical wave varies with exp ik r it , the condition
ki r kr r kt r
(3.87)
94

Figure 3.10 Reection and
refraction of a TE-polarized wave at
the interface of two isotropic
dielectric media. The three vectors
ki , kr , and kt lie in the plane of
incidence. The relationship between
i and t shown here is for the case
of n1 < n2 :
Figure 3.11 Reection and

refraction of a TM-polarized wave at
the interface of two isotropic
dielectric media. The three vectors
ki , kr , and kt lie in the plane of
incidence. The relationship between
i and t shown here is for the case
of n1 < n2 :
is required at the interface x 0 for the boundary conditions described by (1.23)(1.26) to be

satised at all points along the interface at all times. This condition implies that the three vectors
ki , kr , and kt lie in the same plane known as the plane of incidence, as shown in Figs. 3.10 and
3.11. The projections of these three wavevectors on the interface are all equal so that
ki sin i kr sin r kt sin t
(3.88)
where i is the angle of incidence, and r and t are the angle of reection and the angle of
refraction, respectively, for the reected and transmitted waves. All three angles are measured
with respect to the normal n^ of the interface, as is shown in Figs. 3.10 and 3.11. Because ki kr
and ki =kt n1 =n2 , (3.88) yields the relation
i r
(3.89)
3.4 Interface Modes
95
for reection, and the familiar Snells law for refraction:

n1 sin i n2 sin t :
(3.90)
By expressing H in terms of k E in the form of (3.31) with appropriate values of k for the
incident, reected, and refracted elds, respectively, the amplitudes of the reected and transmitted elds can be obtained from the boundary conditions n^ E1 n^ E2 and n^ H1 n^ H2
given in (1.23) and (1.24). There are two different modes of eld polarization.
TE Polarization (s Wave, Wave)
For the transverse electric (TE) polarization, or the perpendicular polarization, the electric
eld is linearly polarized in a direction perpendicular to the plane of incidence while the
magnetic eld is polarized parallel to the plane of incidence, as shown in Fig. 3.10. This wave is
also called s polarized, or polarized. For the TE-polarized wave, the reection coefcient, r,
and the transmission coefcient, t, of the electric eld are respectively given by the following
Fresnel equations:
p
E r n1 cos i n2 cos t n1 cos i n22 n21 sin2 i
p ,
rs
(3.91)
E i n1 cos i n2 cos t n1 cos i n22 n21 sin2 i
ts
Et
2n1 cos i
2n1 cos i
p 1 r s :
E i n1 cos i n2 cos t n1 cos i n22 n21 sin2 i
(3.92)
The intensity reectance and transmittance, R and T, which are also known as reectivity and
transmissivity, respectively, are given by

I r Sr n^ n1 cos i n2 cos t 2

Rs
jr s j2 ,
(3.93)

Ii
n1 cos i n2 cos t
Si n^

I t St n^
1 Rs 6 jt s j2 :
(3.94)
Ts
I i S n^
i
TM Polarization (p Wave, Wave)

For the transverse magnetic (TM) polarization, or the parallel polarization, the electric eld is
linearly polarized in a direction parallel to the plane of incidence while the magnetic eld is
polarized perpendicular to the plane of incidence, as shown in Fig. 3.11. This wave is also
called p polarized, or polarized. For the TM-polarized wave, the reection and transmission
coefcients of the electric eld are respectively given by the following Fresnel equations:
p
E r n2 cos i n1 cos t n22 cos i n1 n22 n21 sin2 i
p ,
rp
(3.95)
E i n2 cos i n1 cos t n22 cos i n1 n22 n21 sin2 i
tp

Et
2n1 cos i
2n1 n2 cos i
n1
p
2
1
r
:
p
E i n2 cos i n1 cos t n2 cos i n1 n22 n21 sin2 i n2
(3.96)
96
The intensity reectance and transmittance for the TM polarization are given, respectively, by

I r Sr n^ n2 cos i n1 cos t 2 2

Rp
rp ,
I i Si n^ n2 cos i n1 cos t
(3.97)

I t St n^
1 Rp 6 t p 2 :
Tp
I i Si n^
(3.98)
Several important characteristics of the reection and refraction of an optical wave at an

interface between two media are summarized below.
1. For both TE and TM polarizations, R jrj2 and R T 1, but T 6 jt j2 :
2. In the case when n1 < n2 , light is incident from a rare medium upon a dense medium; then,
the reection is called external reection. In the case when n1 > n2 , light is incident from a
dense medium on a rare medium; then, the reection is called internal reection.
3. Normal incidence: In the case of normal incidence, i t 0: Then, there is no difference
between TE and TM polarizations, and

n1 n2 2
, T 1 R 4n1 n2 :
R
n1 n2
n1 n2 2
(3.99)
In the case when both media are lossless so that the values of n1 and n2 are both real, there is
a phase change for the reected electric eld with respect to the incident eld for external
reection at normal incidence, but the phase of the reected eld is not changed for internal
reection at normal incidence. A phase change of a value between 0 and is possible when
either or both media have an optical loss or gain so that n1 or n2 or both have complex values.
In any event, the values of R and T do not depend on the side of the interface from which the
incident wave comes from.
4. Brewster angle: For a TE wave, Rs increases monotonically with the angle of incidence. For
a TM wave, Rp rst decreases then increases as the angle of incidence increases. For the
interface between two lossless media, Rp 0 at an angle of incidence of i B , where
B tan1
n2
n1
(3.100)
is known as the Brewster angle. When i B , the angle of refraction for the transmitted
wave is
t
B :
2
(3.101)
It can be shown that this angle is the Brewster angle for the same wave incident from the
other side of the interface. Thus, the Brewster angles from the two sides of an interface are
complementary angles. Figure 3.12 shows, for both the external reection and the internal
reection, the reectances of TE and TM waves as functions of the angle of incidence at
the interface between two media of refractive indices of 1 and 3.5. These characteristics are
very useful in practical applications. At i B , a TM-polarized incident wave is totally
3.4 Interface Modes
97
Figure 3.12 Reectances of TE and TM waves at an interface of lossless media as functions of the angle of
incidence for (a) external reection and (b) internal reection. The reective indices of the two media used for
these plots are 1 and 3.5.
transmitted, resulting in a perfect transmitting window for the TM polarization. Such

windows are called Brewster windows and are useful as laser windows. For a wave of any
polarization that is incident at i B , the reected wave is completely TE polarized. Linearly
polarized light can be produced by a reection-type polarizer based on this principle.
5. Critical angle: In the case of internal reection with n1 > n2 , total internal reection occurs
if the angle of incidence i is larger than the angle
c sin1
n2
,
n1
(3.102)
which is called the critical angle. The reectances of TE and TM waves as functions of the
angle of incidence for internal reection at the interface between two media of refractive
indices of 1 and 3.5 are shown in Fig. 3.12(b). Note that the Brewster angle for internal
reection is always smaller than the critical angle.
6. At the interface of two lossless dielectric media, both of which have real refractive indices,
the transmitted eld has the same phase as the incident eld for both TE and TM polarizations because both ts and tp have positive, real values. For external reection of a TE wave,
the reected eld has a phase change at any incident angle. For internal reection of a TE
wave, the reected eld has no phase change at any incident angle smaller than the critical
angle. For external reection of a TM wave, the reected eld has no phase change at any
incident angle smaller than the Brewster angle, i < B , but has a phase change at any
incident angle larger than the Brewster angle, i > B . For internal reection of a TM wave,
the reected eld has a phase change at any incident angle smaller than the Brewster angle,
i < B , but it has no phase change at any incident angle larger than the Brewster angle but
smaller than the critical angle, B < i < c . (See Problem 3.4.1.)
7. The relations for the reection and transmission coefcients and those for the reectance and
transmittance, given in (3.91)(3.98), remain valid if one or both media have an optical loss
98
or gain so that the refractive indices have complex values. In this situation, each of the
reection and transmission coefcients of TE and TM waves has a phase that is different
from 0 or .
8. If one or both media have a loss or gain, the indices of refraction become complex. In this
situation, the reectance of the TM wave has a minimum value that does not reach zero. This
minimum value is determined by the imaginary parts of the refractive indices of both media.
9. For wave propagation in a general direction in an anisotropic medium, there are two normal
modes that have different indices of refraction. The refracted elds of these two normal
modes can propagate in different directions, resulting in the phenomenon of double
refraction. Meanwhile, the Poynting vector of a normal mode in the anisotropic medium
does not have to be in the plane of incidence.
10. Optical media are generally dispersive. Therefore, reectance and transmittance, as well as
the direction of the refracted wave, are generally frequency dependent.
EXAMPLE 3.9
The index of refraction of water is n 1:33. The index of refraction of ordinary glass depends
on its composition and the optical wavelength but is approximately n 1:5. The refractive
indices of semiconductors, such as Si, GaAs, and InP, vary signicantly with the optical
wavelength and the material composition, as well as with temperature, but they usually fall
in the range between 3 and 4. Take a nominal value of n 3:5 for the typical semiconductor.
For each material at its interface with air, nd the reectivity at normal incidence, the Brewster
angle for external reection, and the critical angle.
Solution:
Using (3.99), the reectivities at normal incidence are found to be R 0:02 for water, R 0:04
for glass, and R 0:31 for the semiconductor. Using (3.100), the Brewster angles for external
reection are found to be B 53:1 for water, B 56:3 for glass, and B 74 for the
semiconductor. Using (3.102), the critical angles are found to be c 48:8 for water, c
41:8 for glass, and c 16:6 for the semiconductor.
3.4.2 Radiation Modes

In the above, we considered the reection and refraction at a planar interface. Here we consider
the mode elds of this structure in the form of (3.1) and (3.2) with the characteristic propagation
constants in the z direction along the interface but with the mode eld proles E x and
H x being functions of only the x coordinate. The normal modes of a single interface are
radiation modes that have a continuous spectrum of eigenvalues, i.e., continuously distributed
values of propagation constants. From (3.87), we nd that the propagation constant in the z
direction is that of the common longitudinal z component of ki , kr , and kt :
k i, z k r , z k t, z :
(3.103)
3.4 Interface Modes
99
We assume that the two media are dielectric with 1 > 2 so that k1 n1 =c > k 2 n2 =c:
There are two different cases: (1) k 1 > > k 2 and (2) k 1 > k2 > , discussed below.
One-Sided Radiation Modes: k1 > > k2
This is the case when total internal reection occurs with i > c sin1 n2 =n1 , as
discussed above. Because ki, z k1 sin i and kr, z k1 sin r , the condition
k2i, x k2i, z k2r, x k2r, z k 21 requires that the transverse x components of ki and kr have the
same real value: h1 ki, x kr, x k1 cos i . However, no real solution of t exists for kt, z
k2 sin t and k t, x k 2 cos t to be valid because > k2 in this case; therefore, no real value
for the transverse x component of kt can be found. Instead, the condition k 2t, x k2t, z k22
requires that k t, x i2 be purely imaginary. Therefore, positive real parameters h1 and 2 can
be dened for the transverse eld proles in media 1 and 2, respectively, as
h21 k21 2 ,
22 2 k22 :
(3.104)
Using the two parameters h1 and 2 , the reection coefcients found in (3.91) and (3.95) for
the TE and TM polarizations can be expressed respectively as
n22 h1 in21 2
:
(3.105)
n22 h1 in21 2
2
As expected for total internal reection, Rs jr s j2 1 and Rp r p 1. However, from
(3.105), it is found that total internal reection has the following phase shifts for the TE and
TM polarizations, respectively,
r TE r s
h1 i2
,
h1 i2
TE s 2 tan1
2
,
h1
r TM r p
TM p 2 tan1
n21 2
:
n22 h1
(3.106)
As commented in the preceding subsection, for external reection at any incident angle or
internal reection at an incident angle smaller than the critical angle, the reection coefcient
of a TE or TM wave at an interface between two lossless dielectric media can only have a phase
of either 0 or . By contrast, (3.106) indicates that total internal reection of a TE or TM wave
can have a phase shift between 0 and .
The fact that ki, x and kr, x both have the real value of k i, x kr, x h1 means that the transverse
eld prole in medium 1 has sinusoidal variations extending to innity in the positive x
direction. By contrast, k t, x i2 means that the transverse eld prole in medium 2 decays
exponentially in the negative x direction away from the interface. This is a one-sided radiation
mode which is a radiation wave in medium 1 but is evanescent in medium 2, as illustrated in
Fig. 3.13. The penetration depth of the evanescent tail into medium 2 is 1
2 .
For the TE mode, it is only necessary to nd E y ; then the other two nonvanishing components
Hx and Hz can be found by using (3.83) and (3.84), respectively. The boundary conditions
require that E y , Hx , and Hz be continuous at the interface, which dictates that E y and E y =x
be both continuous at x 0. The eld prole satisfying these boundary conditions is

cos h1 x , x > 0,
E y x
(3.107)
cos exp 2 x, x < 0,
100

Figure 3.13 Total internal reection and
transverse eld prole of one-sided
radiation mode. The fact that r i
is shown.
where
tan1
2
1
TE :
h1
2
(3.108)
Note that the mode eld prole E y given in (3.107) is not normalized because it extends
to innity in the positive x direction. For x > 0, E y in (3.107) is the superposition of
an incident eld of an amplitude E i ^y ei =2 and a wavevector ki h1 ^x ^z and a
totally reected eld of an amplitude E r E i eiTE and a wavevector kr h1 ^x ^z so
that the total space- and time-varying electric eld is Er; t E i exp iki r it
E r exp ikr r it ^y E y x exp iz it.
For the TM mode, it is only necessary to nd Hy ; then the other two nonvanishing
components E x and E z can be found by using (3.85) and (3.86), respectively. The boundary
conditions require that Hy , E x , and E z be continuous at the interface, which dictates that Hy
and 1 Hy =x, i.e., n2 Hy =x, be both continuous at x 0. The eld prole satisfying these
boundary conditions is

Hy x
x > 0,
cos h1 x ,
cos exp 2 x, x < 0,
(3.109)
n21 2
1
TM :
2
2
n2 h1
(3.110)
where
tan1
Again, the mode eld prole Hy given in (3.109) is not normalized because it extends to innity in the
positive x direction. For x > 0, Hy in (3.109) is the superposition of an incident eld of an amplitude
Hi ^y ei =2 and a wavevector ki h1 ^x ^z and a totally reected eld of an amplitude Hr
Hi eiTM and a wavevector kr h1 ^x ^z so that the total space- and time-varying magnetic eld is
Hr; t Hi exp iki r it Hr exp ikr r it ^y Hy x exp iz it .
3.4 Interface Modes
101
EXAMPLE 3.10
A glass plate has a refractive index of 1.5 at the 1 m wavelength. Find the parameters
of the radiation modes at the airglass interface corresponding to internal reection at the two
different incident angles of 45 and 75 , respectively. What is the penetration depth of the
evanescent tail into the air if a radiation mode is found to be a one-sided radiation mode at a
particular incident angle? What are the phase shifts on reection at the interface for TE and TM
waves, respectively?
Solution:
In this problem, n1 1:5 and n2 1 so that the critical angle of the interface is c
sin1 1=1:5 41:8 . Because i > c for both incident angles, the radiation modes for both
cases are one-sided radiation modes. At 1 m,
k1
2n1
9:42 106 m1
and k2
2n2
6:28 106 m1 :
For i 45 > c , the radiation mode is a one-sided radiation mode; the parameters of this
radiation mode are
k1 sin i 6:66 106 m1 ,
q
6
1
h1 k 1 cos i 6:66 10 m , 2 2 k22 2:22 106 m1 :
The penetration depth of the evanescent tail into the air is 1
2 451 nm. The phase shifts on
reection at the interface for TE and TM waves are
TE 2 tan1
2
n2
0:64 rad 0:20, TM 2 tan1 21 2 1:29 rad 0:41:
h1
n2 h1
radiation mode are
k1 sin i 9:10 106 m1 ,
q
6
1
h1 k1 cos i 2:44 10 m , 2 2 k22 6:59 106 m1 :
The penetration depth of the evanescent tail into the air is 1
TE 2 tan1
2
n2
2:43 rad 0:77, TM 2 tan1 21 2 2:82 rad 0:90:
h1
n2 h1
Two-Sided Radiation Modes: k1 > k2 >

This is the case when partial reection accompanied by refracted transmission occurs for an
incident angle of i < c . In this case, k i, z k1 sin i and kr, z k1 sin r so that the
102
condition k 2i, x k2i, z k2r, x k2r, z k 21 requires that the transverse x components of ki and kr
have the same real value: h1 k i, x kr, x k 1 cos i . Meanwhile, because k 2 > , a real
solution of t exists for k t, z k 2 sin t so that the transverse x component of kt also has
a real value: h2 kt, x k2 cos t . Therefore, positive real parameters h1 and h2 can be dened
for the transverse eld proles in media 1 and 2, respectively, as
h21 k21 2 ,
h22 k 22 2 :
(3.111)
Note that h1 > h2 because k1 > k 2 .

Using the two parameters h1 and h2 , the reection coefcients found in (3.91) and (3.95)
for the TE and TM polarizations can be respectively expressed as
n22 h1 n21 h2
:
(3.112)
n22 h1 n21 h2
2
As expected for partial reection, Rs jr s j2 6 1 and Rp r p 6 1. Because h1 > h2 , there is
no phase shift in reection for the TE polarization: TE s 0. The phase shift in reection
for the TM polarization ips at the Brewster angle: TM p for i < B , but TM
p 0 for i > B . (See Problem 3.4.1.)
The real parameters h1 ki, x kr, x and h2 kt, x characterize a two-sided radiation mode
eld prole that has sinusoidal variations extending to innity in both positive and negative x
directions, as illustrated in Fig. 3.14. This eld pattern is the superposition of the incident,
reected, and transmitted elds on each side from two incident waves, one from medium 1 and
the other from medium 2, as also illustrated in Fig. 3.14 and discussed below.
For the TE mode, the E y eld prole satisfying the boundary conditions that E y and E y =x
are continuous at x 0 is

x > 0,
cos 2 cos h1 x 1 ,
E y x
(3.113)
x < 0,
cos 1 cos h2 x 2 ,
r TE r s
h1 h2
,
h1 h2
r TM r p
where the two phase factors 1 and 2 are related by

h1 tan 1 h2 tan 2 :
(3.114)
Figure 3.14 Partial reection

and transmission, and transverse eld
prole of two-sided radiation mode. The
fact that r i and t > i for incidence
from medium 1 is shown.
3.4 Interface Modes
103
The nonvanishing magnetic eld components Hx and Hz of the TE mode are found from E y by
using (3.83) and (3.84), respectively. The mode eld E y in (3.113) is not normalized because it
extends to innity in both positive and negative x directions. For all x, E y in (3.113) is the
superposition of the incident, reected, and transmitted elds resulting from two incident
waves: one from medium 1 that has a eld amplitude of E i1 ^y cos 2 ei 1 =2 and a wavevector
of ki1 h1 ^x ^z , and the other from medium 2 that has E i2 ^y cos 1 ei 2 =2 and
ki2 h2 ^x ^z . Note that (3.114) eliminates one free phase parameter so that the phase relation
between the two incident waves in the composition of the TE mode eld is determined.
For the TM mode, the Hy eld prole satisfying the boundary conditions that Hy and
2
n Hy =x are continuous at x 0 is

x > 0,
cos 2 cos h1 x 1 ,
(3.115)
Hy x
x < 0,
cos 1 cos h2 x 2 ,
where the two phase factors 1 and 2 are related by
h1
h2
tan 1 2 tan 2 :
2
n1
n2
(3.116)
The nonvanishing electric eld components E x and E z of the TM mode are found from Hy by
using (3.85) and (3.86), respectively. The mode eld Hy in (3.115) is not normalized because it
extends to innity in both positive and negative x directions. For all x, Hy in (3.115) is
the superposition of the incident, reected, and transmitted elds resulting from two incident
waves: one from medium 1 that has a eld amplitude of Hi1 ^y cos 2 ei1 =2 and a wavevector
of ki1 h1 ^x ^z , and the other from medium 2 that has Hi2 ^y cos 1 ei 2 =2 and
ki2 h2 ^x ^z . The relation in (3.116) eliminates one free phase parameter so that the phase
relation between the two incident waves in the composition of the TM mode eld is determined.
EXAMPLE 3.11
The glass plate with a refractive index of 1.5 at the 1 m wavelength given in Example 3.10 is
now immersed in water, which has a refractive index of 1.33. Find the parameters of the radiation
modes at the waterglass interface corresponding to internal reection at the two different incident
angles of 45 and 75 , respectively. What is the penetration depth of the evanescent tail into the
water if a radiation mode is found to be a one-sided radiation mode at a particular incident angle?
What are the phase shifts on reection at the interface for TE and TM waves, respectively?
Solution:
In this problem, n1 1:5 and n2 1:33 so that the critical angle of the interface is c
sin1 1:33=1:5 62:5 and the Brewster angle for internal reection is B tan1
1:33=1:5 41:6 < c . At 1 m,
k1
2n1
2n2
9:42 106 m1 and k2
6:28 106 m1 :
104
For i 45 < c , the radiation mode is a two-sided radiation mode; the parameters of this
radiation mode are
k 1 sin i 6:66 106 m1 ,
q
6
1
h1 k1 cos i 6:66 10 m , h2 k 22 2 5:05 106 m1 :
Because this mode is a two-sided radiation mode, it extends to innity on both the glass and
water sides. Because i 45 > B , the phase shifts of the internal reection at the interface for
TE and TM waves are
TE 0, TM 0:
radiation mode are
k 1 sin i 9:10 106 m1 ,
q
6
1
h1 k 1 cos i 2:44 10 m , 2 2 k22 3:59 106 m1 :
The penetration depth of the evanescent tail into the water is 1
TE 2 tan1
2
1:95 rad 0:62,
h1
TM 2 tan1
n21 2
2:16 rad 0:69:
n22 h1
3.4.3 Surface Plasmon Mode

In the above, we have seen that an interface between two isotropic dielectric media supports
only radiation modes. At most, it supports a one-sided radiation mode that has a localized
transverse eld distribution on only one side of the interface. No localized, guided surface mode
is supported by this type of interface. Guided surface modes do exist in certain types of
interface, such as that between an isotropic dielectric medium and an anisotropic dielectric
medium or that between an isotropic dielectric medium and a plasma medium.
We consider the interface between an isotropic dielectric medium of a permittivity 1 and an
isotropic plasma medium of a permittivity 2 , as shown in Fig. 3.15. For simplicity, we take the
limit that 1 so that the permittivity of the plasma medium is that given in (2.49):
!
2p
2 b 1 2 ,
(3.117)
where b bound is the background permittivity due to bound electrons and p is the plasma
frequency dened in (2.46). The plasma medium can be any medium that has free charge
carriers, such as a doped semiconductor or a metal. For simplicity, we neglect the absorption
3.4 Interface Modes
105
Figure 3.15 Surface plasmon mode at the

interface between a dielectric medium of
1 and a plasma medium of 2 .
loss in the dielectric medium and that due to bound electrons in the plasma medium so that
both 1 and b are real and positive: 1 > 0 and b > 0. However, as discussed in Section 2.4
and seen from (3.117), at any frequency below the plasma frequency, the permittivity of
the plasma medium is negative: 2 < 0 for < p . The opposite signs of 1 and 2 in this
situation create the possibility of a guided surface plasmon mode that is supported by the
interface.
The surface plasmon mode between a dielectric medium and a plasma medium is a TM mode.
To be guided by the interface, it has to be transversely localized near the interface. Thus, it has
to decay exponentially away from the interface in both positive and negative x directions with
characteristic parameters 1 and 2 , respectively:
21 2 k21 ,
22 2 k 22 :
(3.118)
Because the surface plasmon mode is a TM mode, we nd Hy with the boundary conditions
that Hy and 1 Hy =x are continuous at the interface located at x 0.
The guided TM mode can be normalized using (3.22). The normalized eld prole of the
surface plasmon mode that satises the boundary condition for the continuity of Hy is

exp 1 x, x > 0,
^
H y x C
(3.119)
x < 0,
exp 2 x,
where

1=2
1 2 1 2 1=2
:
C
1 1 2 2
(3.120)
The boundary condition for the continuity of 1 Hy =x at x 0 yields the eigenvalue

equation:
1 2
0:
1 2
(3.121)
^y
The nonvanishing mode electric eld components are E^ x and E^ z , which can be found from H
by using (3.85) and (3.86), respectively.
106
Figure 3.16 Dispersion curve for surface plasmon mode showing (a) propagation constant as a function of
frequency and (b) frequency as a function of propagation constant. At a low frequency, the surface plasmon
propagation constant approaches the propagation constant k 1 in the dielectric medium. As the frequency
increases towards sp , becomes much larger than k 1 and approaches innity. The example in this gure ispplotted
with 1 0 and b 0 for the surface of a perfect metal in free space. In this special case, sp p = 2:
Because 1 > 0, 2 > 0, and 1 > 0, it is necessary that 2 < 0 for the eigenvalue equation
to have a solution. Using the relations in (3.118), with k21 2 0 1 and k22 2 0 2 , the
eigenvalue equation (3.121) can be solved to nd

12
0
1 2
1=2
0 21
1
1 2
1=2
0 22
2
1 2
1=2
:
(3.122)
The condition for 1 , 2 , and in (3.122) to have real and positive solutions is that
2 < 0 and 1 2 < 0
2 < 1 < 0:
This condition limits the surface plasmon mode to the frequency range:
r
b
< sp
p ,
1 b
(3.123)
(3.124)
where sp is known as the surface plasma frequency.

Figure 3.16 shows the relation between and for the surface plasmon mode. At a low
p
frequency such that sp , 0 1 k1 so that the surface plasmon propagation
constant approaches the propagation constant k1 in the dielectric medium. As the frequency
increases, increases and gradually becomes much larger than k 1 , k1 , approaching innity
as the frequency approaches sp . Note that sp < p , as is also shown in Fig. 3.16. The cutoff
frequency and cutoff wavelength of a surface plasmon mode are sp sp =2 and
sp c=sp 2c=sp , respectively. The surface plasmon mode can be excited only by a
TM-polarized wave of < sp and > sp .
3.4 Interface Modes
107
EXAMPLE 3.12
A surface plasmon mode can exist at the interface between a silver plate and free space. The plasma
frequency of Ag found in Example 2.4 is p 1:36 1016 rad s1 . What is the surface plasma
frequency of this interface? What are the cutoff frequency and cutoff wavelength of the surface
plasmon mode? Does the surface plasmon mode exist at the 500 nm wavelength? If it exists,
nd its propagation constant and characteristic parameters. Find the penetration depths of the mode
into the free space and into the silver to nd its connement at the interface.
Solution:
At the interface between free space and Ag, 1 0 for free space and 2 is that of Ag. For Ag,
b 0 so that
!
!
!
2p
2p
2
2 b 1 2 0 1 2 0 1 2 :
p
Given p 1:36 1016 rad s1 for Ag, the surface plasma frequency is
r
r
p
b
0
p
p p 9:62 1015 rad s1 :
sp
0 b
0 0
2
Therefore, the cutoff frequency and cutoff wavelength are, respectively,
sp
c
196 nm:
1:53 1015 Hz 1:53 PHz, sp
sp
2
sp
The surface plasmon mode exists at the 500 nm wavelength because > sp .
For p 1:36 1016 rad s1 , we nd p 138 nm. Therefore, for 500 nm,
!

2
5002
12:13 0 :
2 0 1 2 0 1
1382
p
Then, by using (3.122), we nd

12
0
1 2
1=2

2 1 = 0 2 = 0 1=2
2
12:13 1=2 1
m
1 = 0 2 = 0
500 109 1 12:13
1:31 107 m1 ,

"
#1=2

1=2

1=2
0 21
2
1 = 0 2
2
1
m1
1
1 2
1 = 0 2 = 0
500 109 1 12:13
3:77 106 m1 ,
"
#1=2

1=2
2
2 = 0 2
2
12:132
m1
9 1 12:13
1 = 0 2 = 0
500 10
7 1
4:57 10 m :

0 22
2
1 2
1=2
1
The penetration depths are 1
1 265 nm into the free space and 2 22 nm into the silver.
1
Therefore, the connement of the surface plasmon mode at the interface is 1
1 2 287 nm.
108
3.5
WAVEGUIDE MODES
..............................................................................................................
The basic structure of a dielectric optical waveguide consists of a longitudinally extended
high-permittivity, thus high-index, optical medium, called the core, which is transversely
surrounded by low-permittivity, thus low-index, media, called the cladding. We consider a
straight waveguide whose longitudinal direction is taken to be the z direction, as shown in
Figs. 3.1(c) and (d).
In a planar waveguide, which has optical connement in only one transverse dimension, the
core is sandwiched between cladding layers in only one dimension, designated the x dimension,
with a permittivity prole of x, thus an index prole of nx, as shown in Fig. 3.1(c). The core
of a planar waveguide is also called the lm, while the upper and lower cladding layers are called
the cover and the substrate, respectively. Optical connement is provided only in the x dimension
by the planar waveguide. A waveguide in which the index prole has abrupt changes between the
core and the cladding is called a step-index waveguide, while one in which the index prole varies
gradually is called a graded-index waveguide. Figure 3.17 shows examples of step-index and
graded-index planar waveguides. In a nonplanar waveguide of two-dimensional transverse
optical connement, the core is surrounded by the cladding in all transverse directions, with
x; y and nx; y being functions of both x and y coordinates. A nonplanar waveguide can
also have a step-index or graded-index prole. As discussed in Section 3.1, a planar dielectric
waveguide supports only TE and TM modes, whereas a nonplanar dielectric waveguide supports
TE, TM, and hybrid modes. No TEM modes exist in dielectric waveguides.
To get a general idea of the modes of a dielectric waveguide, it is instructive to consider
the qualitative behavior of an optical wave in the asymmetric planar step-index waveguide
shown in Fig. 3.17(a), where n1 > n2 > n3 . For an optical wave of an angular frequency and
a free-space wavelength , the media in the three different regions of the waveguide dene three
propagation constants:
k1
n1 2n1
,
k2
n2 2n2
,
k3
n3 2n3
,
(3.125)
where k1 > k 2 > k3 .
Figure 3.17 Index proles of (a) a step-index planar waveguide and (b) a graded-index planar waveguide.
3.5 Waveguide Modes
109
An intuitive picture of waveguide modes can be obtained from studying ray optics by
considering the path of an optical ray, or a plane optical wave, in the waveguide, as shown
in the central column of Fig. 3.18. There are two critical angles associated with the internal
reections at the lower and upper interfaces:
c2 sin1
n2
n1
and
c3 sin1
n3
,
n1
(3.126)
respectively, where c2 > c3 because n2 > n3 . The characteristics of the reection and
refraction of the ray at the interfaces depend on the incident angle and the polarization of
the wave.
Guided Modes
For a ray that has an incident angle of > c2 > c3 at the interfaces of the waveguide,
the wave inside the core is totally reected at both interfaces and is trapped by the core,
resulting in a guided mode when the resonance condition described below is satised. As the
wave is reected back and forth between the two interfaces, it interferes with itself. A guided
mode can exist only when a transverse resonance condition is satised so that the repeatedly
reected wave constructively interferes with itself. In the core region, the x component of
the wavevector is h1 k1 cos , and the z component is k 1 sin . The phase shift caused
by a round-trip transverse passage of the eld in the core that has a thickness of d is
2h1 d 2k1 dcos . In addition, the internal reection at the lower interface causes a localized
phase shift of 2 as given in (3.106), and that at the upper interface causes a phase shift of 3 ,
which can be found by replacing 2 with 3 in (3.106). The phase shifts 2 and 3 are
functions of the incident angle ; for a given i > c2 > c3 , each of them has different
values for TE and TM waves.
The transverse resonance condition for constructive interference is that the total phase shift in
a round-trip transverse passage is
2h1 d 2 3 2k1 d cos 2 3 2m,
(3.127)
where m is an integer. Because m takes only integral values, only certain discrete values of
satisfy (3.127). This condition results in discrete values of the propagation constant m for
guided modes identied by the mode number m. From (3.106), we nd that < 2 < 0
and < 3 < 0 so that 2 < 2 3 < 0. Therefore, the smallest value of m for (3.127) to
have a solution is m 0; no negative values of m are allowed. The guided mode with m 0 is
the fundamental mode, and those with m 6 0 are high-order modes.
Though the critical angles, c2 and c3 , do not depend on the polarization of the wave, the
phase shifts, 2 and 3 , caused by internal reection at a given angle depend on
the polarization, as seen in (3.106). Therefore, (3.127) have different solutions for TE and TM
waves, resulting in different values of m and different mode characteristics for TE and
TM modes of a given mode number m. Because TM < TE < 0 as seen from (3.106), the
TM
solution of (3.127) yields TE > TM for a given value of m; thus, TE
m > m .
For a given polarization, the solution of (3.127) yields a smaller value of and a correspondingly smaller value of m for a larger value of m. Therefore, among guided modes of different
110
Figure 3.18 Modes of an asymmetric planar step-index waveguide where n1 > n2 > n3 . The range of the
propagation constants, the zig-zag ray pictures, and the eld patterns are shown correspondingly for
(a) the guided fundamental mode, (b) the guided rst high-order mode, (c) a substrate radiation mode for
1:3k3 , and (d) a substratecover radiation mode for 0:3k3 . The waveguide structure is chosen so that
it supports only two guided modes. The mode eld proles are calculated mode eld distributions that are
normalized to their respective peak values.
3.5 Waveguide Modes
111
orders but of the same polarization that are supported by a waveguide, the fundamental mode
has the largest propagation constant 0 ; that is, 0 > 1 > . . . for a given polarization, as shown
in Figs. 3.18(a) and (b).
Substrate Radiation Modes
When c2 > > c3 , total reection occurs only at the upper interface and not at the lower
interface. As a result, an optical wave incident from either the core or the substrate is refracted
and transmitted at the lower interface. This wave is not conned to the core, but is transversely
extended to innity in the substrate. It is called a substrate radiation mode. In this case, the
angle is not dictated by a resonance condition like (3.127) but can take any value in the range
of c2 > > c3 . As a result, the allowed values of form a continuum between k2 and k3
such that the modes are not discrete. The characteristics of a substrate radiation mode are
illustrated in Fig. 3.18(c).
SubstrateCover Radiation Modes
When c2 > c3 > , no total reection occurs at either interface. An optical wave incident
from either side is refracted and transmitted at both interfaces; thus, it transversely extends
to innity on both sides of the waveguide, resulting in a substratecover radiation mode. These
modes are not discrete; their values of form a continuum between k 3 and 0. The characteristics of a substratecover radiation mode are illustrated in Fig. 3.18(d).
In addition to the three types of modes discussed above, there are also evanescent radiation
modes, which have purely imaginary values of that are not discrete. Their elds decay
exponentially along the z direction. Because the dielectric waveguide considered here is lossless
and does not absorb energy, the energy of an evanescent mode transversely radiates away from
the waveguide. A lossless waveguide cannot generate energy, either. Therefore, evanescent
modes do not exist in a perfect, longitudinally innite waveguide. They exist at a longitudinal
junction or imperfection of a waveguide, as well as at the terminals of a realistic waveguide
that has a nite length. By comparison, a substrate radiation mode or a substratecover
radiation mode has a real ; therefore, its energy does not diminish as it propagates. Like a
plane wave, its power ows in the z direction, though its eld transversely extends to innity
because the power owing away from the center of the waveguide in the transverse direction is
equal to that owing toward the center.
The approach of ray optics used above gives an intuitive picture of the waveguide modes and
their key characteristics. Nevertheless, this approach has many limitations. In more sophisticated waveguide geometries such as that of a circular ber, the idea of using the resonance
condition based on total internal reection to nd the allowed values of for the guided modes
does not necessarily yield correct results. For a complete description of the waveguide elds,
rigorous electromagnetic analyses as illustrated below are required.
3.5.1 Step-Index Planar Waveguides

A step-index planar waveguide is also called a slab waveguide. The general structure and
parameters of a three-layer slab waveguide are shown in Fig. 3.17(a), which has a core
112
thickness of d and a step-index prole of n1 > n2 > n3 . In the above, the approach of ray
optics was used to illustrate an intuitive picture and some basic mode characteristics of a
slab waveguide. Further understanding requires quantitative analyses of the mode elds
discussed below.
Normalized Waveguide Parameters
The mode properties of a waveguide are commonly characterized in terms of a few dimensionless normalized waveguide parameters. The normalized frequency and waveguide thickness,
also known as the V number, of a step-index planar waveguide is dened as
2
d
q
q
n21 n22 d n21 n22 ,

c
(3.128)
where d is the thickness of the waveguide core. The propagation constant can be represented
by the following normalized guide index,
2 k22 n2 n22
b 2
,
k1 k22 n21 n22
(3.129)
where n c= =2 is the effective refractive index of the waveguide mode that has
a propagation constant of . The measure of the asymmetry of the waveguide is represented by
an asymmetry factor a, which depends on the polarization of the mode under consideration:
aE
n22 n23
for TE modes,
n21 n22
aM
n41 n22 n23

for TM modes:
n43 n21 n22
(3.130)
Note that aM > aE for a given asymmetric structure. For a symmetric waveguide, aM aE 0
because n3 n2 .
Mode Parameters
For a guided mode, positive real parameters h1 , 2 , and 3 exist such that
h21 k21 2 ,
22 2 k22 ,
23 2 k23
(3.131)
because k1 > > k2 > k3 . From the ray-optics approach discussed above and from (3.131),
the transverse component of the wavevector in the core region of a refractive index n1 is
h1 k 1 cos . For a guided mode, the transverse components of the wavevectors in the

1=2

1=2
substrate and cover regions are h2 k22 2
i2 and h3 k 23 2
i3 , respectively, which are purely imaginary because > k 2 > k3 . Thus, the eld of the guided mode has
to exponentially decay in the transverse direction with decay constants 2 and 3 in the substrate
and cover regions, respectively.
For a substrate radiation mode, h2 can be chosen to be real and positive because
k 1 > k 2 > > k 3 ; thus, (3.131) is replaced by
h21 k 21 2 ,
h22 k22 2 ,
23 2 k23 :
(3.132)
3.5 Waveguide Modes
113
For a substratecover radiation mode, both h2 and h3 are real and positive because
k1 > k2 > k 3 > ; thus, (3.131) is replaced by
h21 k21 2 ,
h22 k22 2 ,
h23 k23 2 :
(3.133)
The transverse eld pattern of a mode is characterized by the transverse parameters h1 , 2 (or
h2 ), and 3 (or h3 ). Because k1 , k 2 , and k3 are specied parameters of a given slab waveguide,
the only parameter that has to be determined for a particular waveguide mode is the longitudinal
propagation constant . Once the value of is found, all parameters that characterize the
transverse eld pattern are completely determined. Therefore, a waveguide mode is completely
specied by its . Alternatively, because of the denite relations between and the transverse
parameters, a mode is completely specied, and the value of its determined, if any one of the
transverse parameters is known. In most cases, rather than directly solving for , it is more
convenient to solve an eigenvalue equation for h1 , as seen below.
EXAMPLE 3.13
A step-index planar waveguide of the structure shown in Fig. 3.17(a) is made of glass of
slightly different compositions for the core and the substrate so that n1 1:54 for the core and
n2 1:47 for the substrate. The cover is simply air so that n3 1:00. The exact values of the
parameters for the guided modes depend on the core thickness, but the propagation constant of
any guided mode at a given wavelength is bounded within a range irrespective of the core
thickness. In what range can the propagation constant of a guided mode, if it exists, be found
at the 1 m wavelength? For what wavelengths can a guided mode be found to have a
propagation constant of 1:5 107 m1 ? What will happen to the answers if the structure
is immersed in water so that n3 1:33? What will happen if it is immersed in benzene so that
n3 1:50? What will happen if it is immersed in CS2 so that n3 1:63?
Solution:
With n1 1:54, n2 1:47, and n3 1:00, we have k 1 > k2 > k3 so that the propagation
constant of any guided mode, if it exists, has to be in the range of k1 > > k2 . At
1 m, we nd that
2n1
2n2
>>
9:68 106 m1 > > 9:24 106 m1 :
The wavelength of a guided mode that has a propagation constant of 1:5 107 m1 falls in
the range:
2n1
2n2
>>
645:1 nm > > 615:8 nm:
If the structure is immersed in water so that n3 1:33, we still nd that k1 > k 2 > k3
because n1 > n2 > n3 . Therefore, there are no changes in the answers obtained above.
If the structure is immersed in benzene so that n3 1:50, then k 1 > k3 > k2 because
n1 > n3 > n2 . Then, at 1 m,
114
2n1
2n3
>>
9:68 106 m1 > > 9:43 106 m1 :
And the wavelength of a guided mode that has a propagation constant of 1:5 107 m1
falls in the range:
2n1
2n3
>>
645:1 nm > > 632:8 nm:
If the structure is immersed in CS2 so that n3 1:63, then k 3 > k 1 > k2 because n3 > n1 > n2 .
In this situation, the structure does not have any guided mode because the core has a lower
refractive index than the cover. Only cover radiation modes and substratecover radiation
modes can be found for this structure.
Guided TE Modes
For a TE mode, it is only necessary to nd E y ; then the other two nonvanishing eld
components Hx and Hz can be found by using (3.83) and (3.84), respectively. The boundary
conditions require that E y , Hx , and Hz be continuous at the interfaces at x d=2 between
layers of different refractive indices. From (3.83) and (3.84), it can be seen that these boundary
conditions are equivalent to requiring E y and E y =x be continuous at these interfaces.
For a guided mode, we know that the transverse eld patterns in the core, substrate, and cover
regions are respectively characterized by the transverse eld parameters h1 , 2 , and 3 , given in
(3.131). A guided TE mode eld distribution that satises the boundary conditions for the
continuity of E y at x d=2 has the form:
8
< cos h1 d=2 exp 3 d=2 x, x > d=2,
(3.134)
d=2 < x < d=2,
E^ y CTE cos h1 x ,
:
cos h1 d=2 exp 3 d=2 x, x < d=2:
Application of the other two boundary conditions for the continuity of E y =x at x d=2
yields two eigenvalue equations:
h1 2 3
h21 2 3
(3.135)
h1 2 3
:
h21 2 3
(3.136)
tan h1 d
and
tan 2
A guided TE mode can be normalized using the orthonormality relation in (3.20) for
r
0
,
(3.137)
C TE
d E
where
dE d
1 1
2 3
is the effective waveguide thickness for a guided TE mode.
(3.138)
3.5 Waveguide Modes
115
Guided TM Modes
For a TM mode, it is only necessary to nd Hy ; then the other two nonvanishing eld
components E x and E z can be found by using (3.85) and (3.86), respectively. The boundary
conditions require that Hy , E x , and E z be continuous at the interfaces at x d=2 between
layers of different refractive indices. From (3.85) and (3.86), it can be seen that these boundary
conditions are equivalent to requiring Hy and 1 Hy =x, or n2 Hy =x, be continuous at
these interfaces.
For a guided mode, we know that the transverse eld patterns in the core, substrate, and cover
regions are respectively characterized by the transverse eld parameters h1 , 2 , and 3 , given in
(3.131). A guided TM mode eld distribution that satises the boundary conditions for the
continuity of Hy at x d=2 has the form:
8
< cos h1 d=2 exp 3 d=2 x, x > d=2,
^ y C TM cos h1 x ,
d=2 < x < d=2,
H
:
cos h1 d=2 exp 3 d=2 x, x < d=2:
(3.139)
Application of the other two boundary conditions for the continuity of n2 Hy =x at x d=2
yields two eigenvalue equations:

h1 =n21 2 =n22 3 =n23
(3.140)
tan h1 d
2
h1 =n21 2 3 =n22 n23
and

h1 =n21 2 =n22 3 =n23
tan 2
:
2
h1 =n21 2 3 =n22 n23
(3.141)
A guided TM mode can be normalized using the orthonormality relation in (3.22) for
CTM
s
0 n21
,
d M
(3.142)
where the effective waveguide thickness for a guided TM mode is

dM d
1
1
2 2
, where q2 2 2 1 and
2 q2 3 q3
k1 k2
q3
2 2
1:
k 21 k23
(3.143)
Modal Dispersion
Guided modes have discrete allowed values of . They are determined by the allowed values of
h1 because and h1 are directly related to each other through (3.131). Because 2 and 3 are
uniquely determined by through (3.131), they are also uniquely determined by h1 :
22 d 2 2 d 2 k22 d 2 V 2 h21 d 2 ,
(3.144)
23 d 2 2 d 2 k23 d 2 1 aE V 2 h21 d 2 :
(3.145)
116
Figure 3.19 Allowed values of normalized guide index b as a function of the V number and the asymmetry
factor aE for the rst three guided TE modes. The cutoff value V c for a mode is the value of V at the intersection
of its dispersion curve with the horizontal axis.
Figure 3.20 Propagation constants of guided modes as functions of optical frequency for a given step-index
dielectric waveguide.
Therefore, there is only one independent variable h1 in the eigenvalue equations. The solutions
of (3.135) yield the allowed parameters for guided TE modes, while those of (3.140) yield the
parameters for guided TM modes. A transcendental equation such as (3.135) or (3.140) is usually
solved numerically, or graphically by plotting its left- and right-hand sides as a function of
h1 d while using (3.144) and (3.145) to replace 2 and 3 by expressions in terms of h1 d. The
solutions yield the allowed values of , or the normalized guide index b, as a function of the
parameters a and V. The results for the rst three guided TE modes are shown in Fig. 3.19.
For a given waveguide, a guided TE mode has a larger propagation constant than the TM
mode of the same order:
TM
TE
m > m :
(3.146)
TM
However, the difference between TE
m and m is very small for modes of an ordinary dielectric
waveguide, where n1 n2 n1 . Then Fig. 3.19 can be used approximately for TM modes
with a aM .
3.5 Waveguide Modes
117
For a given waveguide, the values of aE and aM , as well as those of d and n21 n22 , are
completely specied. Then, of any guided mode is a function of the optical frequency
because V is a function of . Figure 3.20 illustrates the typical relation between and for
guided modes of different orders.
Comparing , k 1 , and k2 in Fig. 3.20, it is seen that the propagation constant of a waveguide
mode has a frequency dependence that is contributed by the structure of the waveguide besides
that due to material dispersion. This extra contribution also causes different modes to have
different dispersion properties, resulting in the phenomenon of modal dispersion. Polarization
dispersion also exists because TE and TM modes generally have different propagation constants.
Polarization dispersion is very small in a weakly guiding waveguide for which n1 n2 n1 .
Cutoff Conditions
As discussed above, 2 and 3 of a guided mode are real and positive so that the mode eld
exponentially decays in the transverse direction outside the core region and remains bound to
the core. This characteristic of a guided mode is equivalent to the condition that > c2 > c3
in the ray optics picture illustrated in Fig. 3.18 so that the ray in the core is totally reected by
both interfaces. Because c2 > c3 , the transition from a guided mode to an unguided radiation
mode occurs when c2 . This transition point corresponds to the condition that k2 and
2 0. As can be seen from the mode eld solutions given in (3.134) and (3.139), the
eld extends to innity on the substrate side when 2 0. This denes the cutoff condition
for a guided mode. The cutoff condition is determined by 2 0, rather than by 3 0, because
3 > 2 so that 2 reaches zero rst as their values are reduced.
At cutoff, V V c . The cutoff value V c of a particular guided mode is the value of V at
the point where the curve of its b versus V dispersion relation, shown in Fig. 3.19, intersects
with the horizontal axis b 0. From (3.144) and (3.145), we nd by setting 2 0 that, at
cutoff,
p
h1 d V c and 3 d aE V c :
(3.147)
Substituting (3.147) and 2 0 into (3.135) for a guided TE mode yields

p
tanV c aE :
(3.148)
Therefore, the cutoff condition for the mth guided TE mode is

p
V cm m tan1 aE , m 0, 1, 2, . . . :
(3.149)
Substituting (3.147) and 2 0 into (3.140) yields the cutoff condition for the mth guided
TM mode:
p
V cm m tan1 aM , m 0, 1, 2, . . . :
(3.150)
Using the denition of the V number given in (3.128), we can write
V cm
2
d
cm
q
q
c
n21 n22 m d n21 n22
c
(3.151)
where cm is the cutoff wavelength and cm is the cutoff frequency of the mth mode. The mth
mode is not guided at a wavelength longer than cm , or a frequency lower than cm .
118
For given waveguide parameters, (3.149) and (3.150) can be used, respectively, to determine the
cutoff wavelengths, and the corresponding cutoff frequencies, of TE and TM modes from (3.151).
For a given optical wavelength, they can be used to determine the waveguide parameters that allow
the existence of a particular guided mode. For given waveguide parameters and optical wavelength,
they can be used to determine the number of guided modes for the waveguide. Therefore, the total
number of guided TE modes supported by a given waveguide at a given optical wavelength is

V 1
1 p
M TE
aE ,
(3.152)
tan

int
and that of guided TM modes is
M TM

V 1
1 p
tan
aM ,

int
(3.153)
where int takes the nearest integer larger than the value in the bracket.
Because aM > aE 6 0 for an asymmetric waveguide, the value of V cm for the mth-order TM
mode is larger than that for the mth-order TE mode. Furthermore, both TE0 and TM0 modes
p
p
have cutoff: V cTE0 tan1 aE for the TE0 mode and V cTM0 tan1 aM for the TM0 mode,
with V cTM0 > V cTE0 . An asymmetric waveguide of a V number such that V cTM0 > V cTE0 > V
supports no guided modes, neither TE nor TM. An asymmetric waveguide of a V number such
that V cTM0 > V > V cTE0 supports the TE0 mode but not the TM0 mode. For V > V cTM0 > V cTE0 ,
both TE0 and TM0 modes are supported. As the V number increases, additional high-order
modes are supported in the sequence: TE1 , TM1 , TE2 , TM2 , . . .. As the V number decreases,
the highest order TM mode is cut off before the TE mode of the same order.
A waveguide that supports only one mode is called a single-mode waveguide. A waveguide
that supports more than one mode is a multimode waveguide. From the above discussion, a truly
single-mode asymmetric waveguide is one that supports only the TE0 mode but not the TM0
mode. However, a waveguide that supports only the fundamental TE0 and TM0 modes is often
called a single-mode waveguide, particularly in the situation of a symmetric waveguide, for
which the two fundamental modes both have no cutoff, as discussed below.
EXAMPLE 3.14
The step-index planar glass waveguide considered in Example 3.13 has n1 1:54 for the core, n2
1:47 for the substrate, and n3 1:00 for the cover. Consider the 1 m wavelength. What is the
range of core thickness for the waveguide to support the TE0 mode but not the TE1 mode? What is
the range of core thickness for the waveguide to support the TM0 mode but not the TM1 mode? What
is the range of core thickness for the waveguide to support the TE0 mode but not the TM0 mode?
Solution:
With n1 1:54, n2 1:47, and n3 1:00, we nd that
q
2
V d n21 n22 2:884d, where d is in m;
aE
n22 n23
5:51,
n21 n22
aM
n41 n22 n23

31:
n43 n21 n22
3.5 Waveguide Modes
119
For the waveguide to support the TE0 mode but not the TE1 mode,
p
p
p
V 1
tan1 aE
1 ) tan1 aE < V
tan1 aE

1:168
4:310
m < d
m
1:168 < V
4:310 )
2:884
2:884
405 nm < d
1:494 m:
M TE 1
)
)
0<
For the waveguide to support the TM0 mode but not the TM1 mode,
p
p
p
V 1
tan1 aM
1 ) tan1 aM < V
tan1 aM

1:393
4:535
1:393 < V
4:535 )
m < d
m
2:884
2:884
483 nm < d
1:572 m:
M TM 1
)
)
0<
For the waveguide to support the TE0 mode but not the TM0 mode,
p
p
M TE 1 and M TM 0 ) tan1 aE < V < tan1 aM ) 405 nm < d
483 nm:
3.5.2 Symmetric Slab Waveguides

For a symmetric slab waveguide, n3 n2 , aE aM 0, and 3 2 . Then, it can be seen from
(3.136) and (3.141) that for both TE and TM modes,tan 2 0 so that
m
,
2
m 0, 1, 2, . . . :
(3.154)
Therefore, the mode eld patterns of a symmetric waveguide given by (3.134) and (3.139) are
either even functions of x, varying in space as cos h1 x in the core region d=2 < x < d=2, for
even values of m, or odd functions of x, varying in space as sin h1 x in the core region
d=2 < x < d=2, for odd values of m. This characteristic is expected because the mode eld
pattern in a symmetric structure is either symmetric or antisymmetric. Figure 3.21 shows the
eld patterns and the corresponding intensity distributions of the rst few guided modes of a
symmetric slab waveguide.
By using the identity tan 2 2 tan =1 tan2 2 cot = cot2 1 while equating 3
to 2 , the eigenvalue equation in (3.135) for guided TE modes can be transformed to two
equations:
tan
h1 d 2
,
h1
2
for even modes;
cot
h1 d 2
,
h1
2
for odd modes:
(3.155)
These two equations can be combined in one eigenvalue equation for all guided TE modes:

h1 d m
tan

2
h1
2
2
q
V 2 h21 d 2
h1 d
m 0, 1, 2, . . . ,
(3.156)
120
Figure 3.21 (a) Field patterns and (b) intensity distributions of the rst few guided modes of a symmetric slab
waveguide.
Figure 3.22 Graphic solutions for the eigenvalues of guided TE and TM modes of a symmetric waveguide of
V 5. The intersections of dashed and solid curves yield the values of h1 d for eigenmodes.
where m is the same mode number as the one in (3.154). Using (3.140), a similar procedure
yields the eigenvalue equation for all guided TM modes:
q

2 2
2
2
2
h1 d m
n1 2 n1 V h1 d
tan
, m 0, 1, 2, . . . :
(3.157)

2 2
2
2
h1 d
n2 h1 n2
For a given value of the waveguide parameter V, the solutions of (3.156) yield the allowed values
of h1 d for both even and odd TE modes, and those of (3.157) yield the allowed values of h1 d for
both even and odd TM modes. Figure 3.22 shows an example for V 5. Because n1 > n2 , it can
be seen from comparing (3.156) with (3.157) and from the graphic solution shown in Fig. 3.22
TE
TM
TM
that for modes of the same order, hTE
1 < h1 ; thus m > m . This observation is consistent
with the conclusion obtained from the above general discussion on asymmetric waveguides.
3.5 Waveguide Modes
121
Because aE aM 0, TE and TM modes of a symmetric waveguide have the same cutoff

condition:
V cm m
(3.158)
for the mth TE and TM modes. This can also be seen in Fig. 3.22. Because m 0 for the
fundamental modes, neither the fundamental TE mode nor the fundamental TM mode of a
symmetric waveguide has cutoff. Any symmetric planar dielectric waveguide supports at least
one TE and one TM mode. The number of TE modes supported by a given symmetric
waveguide is the same as that of the TM modes, which is simply

V
M TE M TM
:
(3.159)
int
For this reason, a symmetric waveguide is never truly single mode because it supports at least
both TE0 and TM0 modes no matter how small its V number is, as long as V > 0. Often, a
symmetric slab waveguide that has V < is loosely called a single-mode waveguide because it
supports only the fundamental TE0 and TM0 modes. These conclusions are unique to symmetric waveguides. They are not true for an asymmetric waveguide. For example, an asymmetric
slab waveguide might not support any guided mode at a given optical wavelength because both
its fundamental TE and TM modes have a nonzero cutoff.
EXAMPLE 3.15
The step-index planar glass waveguide considered in Example 3.14 is made symmetric by using
the substrate material for the cover so that n2 n3 1:47 for the substrate and the cover while
keeping n1 1:54 for the core. Consider the 1 m wavelength. What is the range of core
thickness for the waveguide to support the TE0 mode but not the TE1 mode? What is the range
of core thickness for the waveguide to support the TM0 mode but not the TM1 mode? What is
the range of core thickness for the waveguide to support the TE0 mode but not the TM0 mode?
Solution:
With n1 1:54 and n2 n3 1:47, we nd that
q
2
V d n21 n22 2:884d, where d is in m; aE 0,
For the waveguide to support the TE0 mode but not the TE1 mode,
aM 0:
V
0<
1 ) 0<V
) 0<d
m ) 0 < d
1:089 m:
2:884
For the waveguide to support the TM0 mode but not the TM1 mode,
M TE 1
V
0<
1 ) 0<V
) 0<d
m ) 0 < d
1:089 m:
2:884
It is not possible for a symmetric waveguide to support the TE0 mode but not the TM0 mode
because they both have no cutoff.
M TM 1
122
3.6
PHASE VELOCITY, GROUP VELOCITY, AND DISPERSION
..............................................................................................................
Phase velocity, group velocity, and dispersion are important parameters that characterize the
propagation of an optical wave. Phase velocity determines the rate of phase variation in wave
propagation. Group velocity determines the speed of transmission of an optical signal. Dispersion is the primary cause of limitation on the bandwidth of the transmission of optical signals.
As discussed in Chapter 2, the susceptibility , thus the permittivity , of a medium is
a function of the optical frequency. This is the origin of material dispersion. In a homogeneous
anisotropic medium, normal modes of different polarizations have different characteristic
refractive indices, and thus different propagation constants, resulting in polarization dispersion.
In an optical structure, there are waveguide dispersion and modal dispersion besides material
dispersion. Both material dispersion and waveguide dispersion are examples of chromatic
dispersion because both are frequency dependent. Waveguide dispersion is caused by the
frequency dependence of the propagation constant of a specic mode due to the waveguiding
effect. The combined effect of material dispersion and waveguide dispersion for a particular
mode alone is called intramode dispersion. Modal dispersion is also called intermode
dispersion because it is caused by the variation in propagation constant between different
modes. Modal dispersion appears only when more than one mode is excited in a multimode
waveguide; it exists even when chromatic dispersion disappears.
To illustrate the concepts of phase velocity, group velocity, and dispersion, we rst consider a
plane-wave normal mode of a homogeneous medium that has a characteristic propagation
constant of k n=c, where n is the frequency-dependent characteristic refractive
index of the normal mode. Without loss of generality, the z coordinate direction is taken to be
along the propagation direction. The electric eld of such a monochromatic plane optical wave
can be written as
E E exp ikz it ,
(3.160)
where E is a constant vector independent of space and time.

The eld expressed in (3.160) represents a sinusoidal wave that has a phase varying with z and t as
kz t:
(3.161)
A point of constant phase on the space- and time-varying eld is dened by constant, thus
d kdz dt 0. If we track this point of constant phase as the wave propagates, we nd
that it moves with a velocity of
vp
dz
:
dt
k
(3.162)
This is called the phase velocity of the wave. Note that the phase velocity is a function of
the optical frequency because the refractive index n is a function of frequency. There is
phase-velocity dispersion due to the fact that dn=d 6 0. In the case of normal dispersion,
dn=d > 0 and dn=d < 0; in the case of anomalous dispersion, dn=d < 0 and dn=d > 0.
As discussed in Section 2.3, normal dispersion and anomalous dispersion are associated with
resonant transitions in a material.
123
Figure 3.23 Wave packet composed of two frequency components showing the carrier and the envelope. The
carrier travels at the phase velocity, whereas the envelope travels at the group velocity.
In practice, a propagating optical wave rarely contains only one frequency. It usually consists of
many frequency components that are grouped around some center frequency, 0 . For the simplicity
of illustration, we consider a wave packet traveling in the z direction that is composed of two plane
waves of equal real amplitude E. The frequencies and propagation constants of the two components are
1 0 d, k1 k0 dk,
2 0 d, k2 k0 dk:
(3.163)
The space- and time-dependent total real eld of the wave packet is then given by
E E exp ik 1 z i1 t c:c: E exp ik 2 z i2 t c:c:
n

o
2E cos k0 dkz 0 dt cos k 0 dk z 0 dt
(3.164)
4E cos dkz dt cos k0 z 0 t:

As illustrated in Fig. 3.23, the resultant wave packet has a carrier, which has a frequency of 0 and
a propagation constant of k0 , and an envelope, which varies in space and time as cosdkz dt.
Therefore, a xed point on the envelope is dened by dkz dt constant, which travels with a
velocity of
d
vg
:
(3.165)
dk
This is the velocity of the wave packet and is called the group velocity.
Because the energy of a harmonic wave is proportional to the square of its eld amplitude,
the energy carried by a wave packet that is composed of many frequency components is
concentrated in the regions where the amplitude of the envelope is large. Therefore, the energy
in a wave packet is transported at the group velocity v g . Because a wave package carries an
optical signal, thus information, optical signals and optical information are transmitted at the
group velocity. The constant-phase wavefront travels at the phase velocity, but optical energy
and information are transmitted at the group velocity.
In reality, the group velocity is usually a function of the optical frequency. Then,
d2 k
d 1
v 6 0,
2
d
d g
(3.166)
124
which represents group-velocity dispersion. A dimensionless coefcient for group-velocity

dispersion is dened as
D c
d2 k 2c2 d2 k
:
d2
d2
(3.167)
Group-velocity dispersion is an important consideration in the propagation of optical pulses, which

can represent information bits of an optical signal. It can cause broadening of an individual pulse, as
well as changes in the time delay between pulses of different frequencies. The sign of the groupvelocity dispersion can be either positive or negative. In the case of positive group-velocity
dispersion, d2 k=d2 > 0 and D > 0, a long-wavelength, or low-frequency, pulse travels faster
than a short-wavelength, or high-frequency, pulse. By contrast, a short-wavelength pulse travels
faster than a long-wavelength pulse in the case of negative group-velocity dispersion, d2 k=d2 < 0
and D < 0. In a given material, the sign of D generally depends on the spectral region of concern.
Group-velocity dispersion and phase-velocity dispersion discussed above have different meanings.
When measuring the transmission delay or the broadening of optical signals or pulses due
to the dispersion in a medium that has a large transmission length, such as an optical ber,
another group-velocity dispersion coefcient dened as
D
2c d2 k
D

2 d2
c
(3.168)
is usually used. This coefcient is generally expressed as a function of wavelength in the unit

of picoseconds per kilometer per nanometer ps km1 nm1 . It is a direct measure of the
chromatic pulse transmission delay over a unit transmission length.
To summarize, the propagation constant of a plane-wave normal mode is
k
n:
c
(3.169)
vp
c
,
k n
(3.170)
d c
,
dk N
(3.171)
dn
dn
n
d
d
(3.172)
Therefore, the phase velocity is
and the group velocity is

vg
where
N n
is called the group index. Using (3.167) and (3.168), the group-velocity dispersion coefcient
can be expressed as
D 2
d2 n
d2 n
or
D

:
c d2
d2
(3.173)
Figure 3.24 shows, as an example, the dispersion properties of pure silica glass and germania
silica glass.
125
Figure 3.24 (a) Index of refraction n and group index N and (b) group-velocity dispersion D as functions of
wavelength for pure silica (solid curves) and germaniasilica containing 13.5 mol% GeO2 (dashed curves).
Zero group-velocity dispersion appears at 1:284 m for pure silica.
EXAMPLE 3.16
The index of refraction of pure silica in the wavelength range between 1:0 and 1:6 m varies
with wavelength approximately as
n 1:4507 0:003012 0:003322 :
(a) Within this wavelength range, where does silica have normal dispersion? Where does it
have anomalous dispersion?
(b) Within this wavelength range, where does silica have positive group-velocity dispersion?
Where does it have negative group-velocity dispersion?
(c) Find the refractive index, the group index, and the group-velocity dispersion of silica at the
three wavelengths of 1:0 m, 1:3 m, and 1:6 m.
(d) Express the group-velocity dispersion as D in the unit of ps km1 nm1 .
Solution:
With the given wavelength dependence of the refractive index, we nd
dn
0:006023 0:00664,
d
N n
dn
1:4507 0:009032 0:003322 ,
d
D 2
d2 n
0:018062 0:006642 :
2
d
126
(a) From the above, we nd that dn=d < 0 for all wavelengths in the wavelength range
between 1:0 and 1:6 m. Therefore, silica has normal dispersion throughout this
wavelength range.
(b) The wavelength dependence of D obtained above indicates that it can be zero at the
wavelength:
D 0 ) 1:284 m:
It is found that silica has positive group-velocity dispersion with D > 0 for < 1:284 m,
and it has negative group-velocity dispersion with D < 0 for > 1:284 m.
(c) Using the wavelength dependence of each parameter obtained above, we nd
1:0 m
1:3 m
1:6 m
n
N
D
1:450 1:463
0:01142
1:447 1:462 0:00054
1:443 1:463 0:00994:
(d) Using (3.168) and the values of D obtained in (c), we nd
1:0 m
1:3 m
1:6 m
D
0:01142
0:00054
0:00994
D
38 ps km1 nm1
1:4 ps km1 nm1
21 ps km1 nm1 :
3.6.1 Waveguide Dispersion

The propagation constant of a mode of an optical structure is determined both by the
parameters of the optical structure and by the material properties. As seen in Figs. 3.16 and
3.20, due to the waveguiding effect, the frequency dependence of can be very different from
that of the k constants of the materials that form the optical structure. Therefore, of a mode has
mixed contributions from both material dispersion and waveguide dispersion. It is in fact more
convenient to directly consider the combined effect. To do so, we only have to replace k of a
plane-wave normal mode in all of the formulas obtained in the above by of the waveguide
mode under consideration, thus dening the effective refractive index n , the effective group
index N , and the effective group-velocity dispersion D for the mode:
n
N c
c
,
(3.174)
dn
d
,
n
d
d
(3.175)
2
d2
2 d n
:
d2
d2
(3.176)
D c
127
Figure 3.25 (a) Effective index of refraction and group index and (b) group-velocity dispersion of the
fundamental mode as a function of wavelength. The solid curves show the effective parameters of the mode
with both material and waveguide contributions. The dashed curves show only the material contribution
to the core and cladding regions, labeled 1 and 2, respectively.
The phase velocity and group velocity of the mode are, respectively,
c
,
n
(3.177)
d
c
:
d N
(3.178)
v p
and
v g
As an example of the contributions of the waveguiding effect to the dispersion parameters,

Fig. 3.25 shows n , N , and D of the fundamental mode of a circular optical ber in
comparison to the parameters of its core and cladding materials.
3.6.2 Modal Dispersion

The frequency dependence of the propagation constant of a mode discussed above is the total
intramode dispersion that includes material and waveguide contributions for the mode. Different normal modes of an anisotropic medium or an optical structure have different propagation
constants at a given optical frequency. Such differences lead to modal dispersion among
different modes, which is intermode dispersion.
For plane waves or Gaussian modes propagating in a homogeneous anisotropic medium,
modal dispersion exists due to different propagation constants for normal modes of different
polarizations, such as k x , ky , and k z of the linearly birefringent principal normal modes of
polarization given in (2.15), k and k of the circularly birefringent principal normal modes
of polarization given in (2.21), or k o and ke of the ordinary and extraordinary waves in (3.57).
Such modal dispersion causes polarization dispersion.
128
For normal modes in an optical structure, such as an interface or a waveguide, modal

dispersion exists among modes of the same polarization but of different order, such as the
different propagation constants of TE modes of different orders shown in Fig. 3.19. In general,
at a given frequency, a lower order mode has a larger propagation constant, as seen in Figs. 3.19
and 3.20. This dispersion is not caused by polarization or frequency but is purely imposed by
the optical structure. This type of modal dispersion is mode-order dispersion. Modal dispersion
in an optical structure also exists among modes of the same order but of different polarizations,
such as that between TEm and TMm modes of a planar waveguide. As discussed in Section 3.5
TM
and expressed in (3.146), TE
m > m for any given order m. This type of modal dispersion is
polarization-mode dispersion.
EXAMPLE 3.17
An optical pulse has a pulse duration of t ps 20 ps and a spectral width of ps 0:1 nm. It
is transmitted through a silica ber over a distance of 10 km. Use the data of silica obtained in
Example 3.16 for the silica ber to nd the transmission time and the temporal broadening of
the pulse due to group-velocity dispersion at the transmission end in the case when the center
wavelength of the pulse is at 1:0 m, 1:3 m, or 1:6 m. How does the group-velocity
dispersion temporally spread the pulse spectrum in each case?
Solution:
For a transmission distance of l, the transmission time ttr is
t tr
l
N
l
vg
c
and the temporal pulse broadening tGVD due to group-velocity dispersion is

t GVD jD jps l:
At 1:0 m, N 1:463 and D 38 ps km1 nm1 . Thus, for l 10 km,
ttr
N
1:463
10 103 s 48:8 s,
l
c
3 108
tGVD jD jps l 38 0:1 10 ps 38 ps:

At 1:3 m, N 1:462 and D 1:4 ps km1 nm1 . Thus, for l 10 km,
ttr
N
1:462
10 103 s 48:7 s,
l
8
c
3 10
tGVD jD jps l 1:4 0:1 10 ps 1:4 ps:

At 1:6 m, N 1:463 and D 21 ps km1 nm1 . Thus, for l 10 km,
ttr
N
1:463
10 103 s 48:8 s,
l
c
3 108
tGVD jD jps l 21 0:1 10 ps 21 ps:
129
We nd that the transmission time is about the same for all three wavelengths because the
group index is about the same for all three wavelengths. However, the temporal pulse
broadening varies much among the three wavelengths because of the different values of
group-velocity dispersion. At the low group-velocity dispersion point of 1:3 m, the pulse is
only slightly broadened. At the other two wavelengths, the broadening is larger than the original
pulse duration. Group-velocity dispersion causes frequency chirping in an optical pulse. At
1:0 m, the broadening causes the long-wavelength component of the pulse to move to the
temporal leading edge of the pulse because of positive group-velocity dispersion with D > 0
and D < 0, making the pulse positively chirped with its frequency increasing with time within
the pulse. At 1:3 m and 1:6 m, the broadening causes the short-wavelength component
of the pulse to move to the temporal leading edge of the pulse because of negative groupvelocity dispersion with D < 0 and D > 0, making the pulse negatively chirped with
its frequency decreasing with time within the pulse.
3.7
ATTENUATION AND AMPLIFICATION
..............................................................................................................
As discussed in Section 2.1, a complex eigenvalue of , thus that of , signies an
optical loss or gain for the corresponding principal mode of polarization of the medium, with
00 > 0 and 00 > 0 for optical loss, and 00 < 0 and 00 < 0 for optical gain. For a plane-wave
normal mode characterized by a complex eigenvalue ,
k 2 2 0 2 0 0 i 00 :
(3.179)
Therefore, the propagation constant k becomes complex:
k k0 ik00 k0 i :
2
(3.180)
The index of refraction also becomes complex:
r
0 i 00
:
(3.181)
n n in
0
The relation k n=c between k and n is still valid.
If we choose k0 to be positive, the sign of is the same as that of 00 . Then, k0 and n0 are both
positive, and k 00 and n00 also have the same sign as 00 . Taking the z coordinate direction to be
along the propagation direction, the electric eld of a monochromatic plane optical wave as
expressed in (3.160) is
0
00
E E exp ikz it E ez=2 exp ik0 z it :
(3.182)
It can be seen that the wave has a phase that varies sinusoidally with a period of 2=k0 along z.
However, because of the nonvanishing imaginary part k00 =2 of the propagation constant,
the magnitude jEj of the electric eld is not constant but varies exponentially with z.
130
The intensity of an optical eld projected on a surface is dened in (1.56):

I S n^ S S n^, where n^ is the unit normal vector of the projected surface. Note
that the intensity of a given optical eld depends on the projected surface on which the intensity
is measured. Note further that for an extraordinary wave in an anisotropic medium, the
^ These factors have
Poynting vector S is not generally parallel to the propagation direction k.
to be considered when calculating the intensity. For a monochromatic plane-wave normal mode
that has an optical eld given in (3.182), we can use the relation k E 0 H given in (3.31)
to nd that its intensity projected on the surface that is normal to the propagation direction k^
can be expressed as
I
2k 0 jE j2 2k 0 jE j2 z
e ,
0
0
(3.183)

where E E E k^ k^ is the component of the optical eld that is transverse to the

^ For a plane wave in an isotropic
propagation direction dened by k^ and E E E k^ k.
medium or an ordinary wave in an anisotropic medium, E E because E k^ 0. For an
extraordinary wave in an anisotropic medium, E 6 E because E k^ 6 0. In any event, the
optical intensity varies exponentially with z when 6 0.
Clearly, k 0 is the wavenumber in this situation, and the sign of determines the attenuation
or amplication of the optical wave.
1. If 00 > 0, then 00 > 0 and > 0. As the optical wave propagates, its eld amplitude and
intensity decay exponentially along the direction of propagation. Therefore, is called the
absorption coefcient or attenuation coefcient.
2. If 00 < 0, then 00 < 0 and < 0. The eld amplitude and intensity of the optical wave grow
exponentially. Then, we dene g as the gain coefcient or amplication coefcient.
Both and g have the unit of per meter, often also quoted per centimeter.
EXAMPLE 3.18
A Si crystal has a complex refractive index of n 4:30 i0:073 at the 500 nm wavelength. Find the absorption coefcient and the absorption depth of Si at this wavelength. What
is the complex susceptibility?
Solution:
From (3.180), the absorption coefcient is
2k00
4n00 4 0:073 1
m 1:835 106 m1 :
500 109
The absorption depth is 1 545 nm. Because 1 = 0 n2 , the complex susceptibility is

n2 1 4:30 i0:0732 1 17:48 i0:628:
131
3.7.1 Attenuation and Amplication of Waveguide Modes

Several factors contribute to the attenuation of the power of an optical wave propagating in an
optical structure. Besides the loss or gain contributed by the material as discussed above, the
imperfections of an optical structure, such as the roughness of its interfaces and the irregularity
of its geometric shape, cause additional losses. Furthermore, the distribution of optical loss or
gain might not be uniform across an optical structure because different regions of an optical
structure generally have different optical properties. In any event, the normal mode of an optical
structure is characterized by a unique, well-dened propagation constant . The attenuation or
amplication of the normal mode while it propagates through the structure is characterized by a
complex in the same manner as the complex k for a plane wave. Thus,
0 i00 0 i :
2
(3.184)
As described above, a positive is the absorption coefcient or attenuation coefcient of the

mode, whereas g is the gain coefcient or amplication coefcient of the mode.
For a guided mode, attenuation or amplication affects the mode across its entire prole even
though it does not have a uniform eld prole across the transverse plane. Therefore, the
attenuation or amplication of a guided mode is measured with respect to the change of its mode
power rather than its intensity: Pz / ez . The attenuation of optical power over a propagation
distance of l in an optical structure for a mode that has an attenuation coefcient of is given by
Pout Pin el :
(3.185)
The input and output powers of the mode, Pin and Pout , respectively, are measured in watts,
while is given per meter. The power is often measured in milliwatts or microwatts in lowpower applications, and in kilowatts or megawatts in high-power applications. In practical
applications, is also measured per centimeter or per kilometer when l is measured in
centimeters or kilometers.
In practical engineering applications, it is convenient to use decibels (dB) as a measure of
relative changes of quantities. The attenuation coefcient is then measured in decibels per
meter or decibels per kilometer when l is measured in meters or kilometers:

1
Pout
1
Pout
dB m1
, dB km1
,
10 log
10 log
Pin
Pin
lm
lkm
(3.186)
where Pin and Pout are measured in the same unit which can be watts, milliwatts, or microwatts.
In the case of a low-loss ber, the propagation length l in the ber is usually measured in
kilometers, and is conventionally given in decibels per kilometer. Comparing (3.185) with
(3.186), we nd that

dB km1 4:32 km1 and km1 0:23 dB km1 :
(3.187)
Power can also be measured in decibels and has the unit of decibel-watts (dBW), decibelmilliwatts (dBm), or decibel-microwatts (dB), dened as
PdBW 10 log PW, PdBm 10 log PmW, PdB 10 log PW:
(3.188)
132
When power is given in decibel-watts or decibel-milliwatts and the attenuation coefcient is in

decibels per kilometer, (3.185) can be expressed as

Pout dBW Pin dBW dB km1 lkm
(3.189)

Pout dBm Pin dBm dB km1 lkm:
(3.190)
or, equivalently,
A similar formula can be written for power measured in decibel-microwatts. These formulas are
convenient and useful in practical applications as they relate the input power, output power, and
attenuation in a simple arithmetic relation.
EXAMPLE 3.19
An optical ber has an attenuation coefcient of 0:4 dB km1 at 1:3 m. An optical
signal at an input power level of Pin 10 mW is transmitted through this ber over a distance
of l 100 km. What is the output power? If the attenuation coefcient is slightly reduced to
0:35 dB km1 , what is the output power?
Solution:
The input power is Pin 10 mW 10 dBm. With 0:4 dB km1 , the output power is
Pout Pin l 10 dBm 0:4 dB km1 100 km 30 dBm 103 mW 1 W:
If the attenuation coefcient is slightly reduced to 0:35 dB km1 , the output power is
Pout Pin l 10 dBm 0:35 dB km1 100 km 25 dBm 102:5 mW 3:16 W:
For a transmission distance of 100 km, the output power is increased by more than 200% when
the attenuation coefcient is reduced by only 0:05 dB km1 .
Problems
3.1.1 Explain why a TEM mode eld can exist only in an optically homogeneous space where
is a constant of space, and not in an optically inhomogeneous space where varies
in space.
3.1.2 Can a dielectric waveguide support TEM modes? Explain.
3.1.3 Can a planar optical structure support hybrid modes? Explain.
3.1.4 What types of guided modes does each of the following structure support: (a) a planar
metallic structure, (b) a planar dielectric structure, (c) a hollow cylindrical metallic
structure, and (d) a cylindrical dielectric structure?
Problems
133
3.1.5 Show that (a) the dot-product orthonormality relation of (3.20) applies to TE modes, (b)
the dot-product orthonormality relation of (3.22) applies to TM modes, and (c) both
relations apply to TEM modes.
3.2.1 The principal indices of refraction of InP, which is a cubic crystal, at the 1:3 m
wavelength are nx ny nz 3:205. Find the propagation constant and the wavelength
in the crystal for an optical wave at 1:3 m that propagates through an InP crystal
under each of the following conditions. In each case, does the polarization state change
as the wave propagates through the crystal?
(a) Linearly polarized along ^x , propagating along ^y .
(b) Linearly polarized along ^y , propagating along ^z .
(c) Linearly polarized along ^z , propagating along ^x .
(d) Circularly polarized in the xy plane, propagating along ^z .
(e) Circularly polarized in the yz plane, propagating along ^x .
3.2.2 The principal indices of refraction of LiNbO3 , which is a negative uniaxial crystal,
at the 1:3 m wavelength are nx ny no 2:222 and nz ne 2:145. Find
the propagation constant and the wavelength in the crystal for an optical wave at
1:3 m that propagates through a LiNbO3 crystal under each of the following conditions. In each case, does the polarization state change as the wave propagates through
the crystal?
3.2.3 The principal indices of refraction of KTP, which is a biaxial crystal, at the
1:3 m wavelength are nx 1:734, ny 1:742, and nz 1:822. Find the propagation
constant and the wavelength in the crystal for an optical wave at 1:3 m
that propagates through a KTP crystal under each of the following conditions.
In each case, does the polarization state change as the wave propagates through the
crystal?
3.2.4 The principal indices of refraction of LiNbO3 at 1:3 m are nx ny no 2:222
and nz ne 2:145. Design a waveplate based on LiNbO3 for rotating the polarization
direction of a linearly polarized wave at 1:3 m by 30o . Give the possible thicknesses
of the plate and the arrangement for this purpose.
3.2.5 The principal indices of refraction of LiNbO3 at 1:3 m are nx ny no 2:222
and nz ne 2:145. Design a waveplate based on LiNbO3 for converting a linearly
polarized wave into a circularly polarized wave at 1:3 m. Give the possible thicknesses of the plate and the arrangement for this purpose.
134
3.2.6 The permittivity tensor of a KDP crystal at 1 m in an arbitrarily chosen Cartesian

coordinate system is found to be
0
1
2:174
0
0:039
0@ 0
2:280
0 A:
0:039
0
2:266
3.2.7
3.2.8
3.2.9
3.2.10
(a) Is the KDP crystal birefringent or nonbirefringent? If it is birefringent, is it uniaxial

or biaxial? What are its principal indices of refraction?
(b) If it is used to make a half-wave plate at 1 m, what is the thickness of the plate?
(c) If it is used to make a quarter-wave plate at 1 m, what is the thickness of the plate?
The principal indices of refraction of quartz at the 600 nm wavelength are nx
ny 1:544 and nz 1:553.
(a) Quartz is clearly a birefringent crystal, is it positive or negative uniaxial?
(b) What kind of quartz plate can be used to rotate the polarization direction of a
linearly polarized wave by 90 to its orthogonal linear polarization? Describe the
arrangement for this function and nd the thickness of the plate.
(c) What kind of quartz plate can be used to convert a circularly polarized wave into a
linearly polarized wave? Describe the arrangement for this function and nd the
thickness of the plate. How is the direction of the output linear polarization determined?
The principal indices of refraction of BBO, which is a negative uniaxial crystal, are nx
ny no 1:677 and nz ne 1:557 at the 500 nm wavelength. Consider a
propagation direction k^ that makes an angle of 45 with respect to the x principal
axis and an angle of 60 with respect to the z principal axis.
(a) Find the polarization directions ê o and ê e , and the corresponding propagation
constants k o and ke , of the ordinary and extraordinary normal modes.
(b) Find the walk-off angle of the extraordinary wave. What is the separation of
the ordinary and extraordinary beams if an optical wave that consists of both
ordinary and extraordinary components at this wavelength propagates in this direction through a BBO crystal over a distance of 3 mm?
The principal indices of refraction of quartz, which is a positive uniaxial crystal, are
nx ny no 1:544 and nz ne 1:553 at the 600 nm wavelength. Consider a
propagation direction k^ that makes an angle of 60 with respect to the x principal
axis and an angle of 30 with respect to the z principal axis.
(a) Find the polarization directions ê o and ê e , and the corresponding propagation
constants k o and ke , of the ordinary and extraordinary normal modes.
(b) Find the walk-off angle of the extraordinary wave. What is the separation of the
ordinary and extraordinary beams if an optical wave that consists of both ordinary
and extraordinary components at this wavelength propagates in this direction
through a quartz crystal over a distance of 5 mm?
Show that there is no walk-off for an extraordinary wave when it propagates in any
direction that lies in the xy plane of a uniaxial crystal, for which the z principal axis is the
unique optical axis.
Problems
135
3.3.1 Give two examples of TEM modes that are not plane waves: (a) one example in purely
dielectric medium and (b) another example not in purely dielectric medium.
3.3.2 A fundamental Gaussian beam from an Er:ber laser at the 1:53 m wavelength exits
the ber with a spot size of w0 8 m, which is determined by the ber core radius. The
beam then propagates in free space without being collimated. Find the beam divergence
angle, the Rayleigh range, and the confocal parameter of the beam. What are the spot
sizes and the radii of curvature of the beam at the distances of 1 mm, 1 cm, 10 cm, and
1 m, respectively, from the end of the ber?
3.3.3 A Gaussian beam of an unknown wavelength in free space is found to have spot sizes of
w0 100 m at the beam waist and wz 300 m at a distance of z 15 cm from the
beam waist. Find the wavelength, the Rayleigh range, and the divergence angle of
the beam.
3.3.4 A fundamental Gaussian laser beam that has a power of P 10 W at a wavelength of
600 nm is focused to a small spot size for an intensity at the beam center of I 0
2:5 MW cm2 at its beam waist. What is the beam-waist radius w0 of the beam? What
is the divergence angle of the beam? What are its spot size and beam-center intensity at
a distance of 5 m from the beam waist? If the spot size is increased to w0 50 m
at the beam waist, what are the changes in the beam-center intensities at the beam waist
and at 5 m from the waist, respectively?
3.4.1 Consider reection and transmission of TE and TM waves at the interface of two lossless
dielectric media that have real refractive indices of n1 and n2 , respectively. Use (3.91) and
(3.95) to show the following facts.
(a) For external reection of a TE wave, the reected eld has a phase change at any
incident angle. For internal reection of a TE wave, the reected eld has no phase
change at any incident angle that is smaller than the critical angle.
(b) For external reection of a TM wave, the reected eld has no phase change at
any incident angle that is smaller than the Brewster angle, i < B , but has a phase
change at any incident angle that is larger than the Brewster angle, i > B . For
internal reection of a TM wave, the reected eld has a phase change at any
incident angle that is smaller than the Brewster angle, i < B , but has no phase
change at any incident angle that is larger than the Brewster angle and smaller than
the critical angle, B < i < c .
3.4.2 When a collimated beam of broadband white light covering the spectrum from red to
violet is incident at an oblique angle from free space on a at surface of ordinary glass,
the transmitted beam is no longer collimated. Sketch how the spectral components of the
transmitted beam spread from red to violet. Give a brief explanation why they spread in
that manner.
3.4.3 The refractive index of a glass plate is 1.5. It can be used as a reection-type polarizer
so that if a beam is incident on its surface at a proper angle, the reected beam is
always linearly polarized no matter what the polarization of the incident beam is. If the
glass plate is placed in air, what is this proper incident angle from the air? What is the
polarization of the reected beam at this incident angle?
136
3.4.4 The refractive index of diamond at 1:0 m is n 2:39. What is the reectivity
of the diamond surface at normal incidence? At a particular incident angle, a specic
linearly polarized optical wave at 1:0 m is completely transmitted through a
diamond surface exposed to air. What are this incident angle and the specic polarization
of the incident wave that make this happen?
3.4.5 The refractive index of water is 1.33. For the 600 nm wavelength, nd the parameters
of the radiation modes at the airwater interface for internal reection at the two different
incident angles of 45 and 75 , respectively. What is the penetration depth of
the evanescent tail into the air if a radiation mode is found to be a one-sided radiation
mode at a particular incident angle? What are the phase shifts on reection at the interface
for TE and TM waves, respectively?
3.4.6 At the 1:5 m wavelength, the refractive index of intrinsic GaAs is 3.38. Find the
parameters of the radiation modes at the airGaAs interface for internal reection at the
two different incident angles of 30 and 60 , respectively. What is the penetration depth
of the evanescent tail into the air if a radiation mode is found to be a one-sided radiation
mode at a particular incident angle? What are the phase shifts on reection at the interface
for TE and TM waves, respectively?
3.4.7 Consider the interface between SiO2 and silver. The refractive index of SiO2 is 1.46 in the
visible spectral region. Use the plasma frequency p 1:36 1016 rad s1 of Ag to nd
the surface plasma frequency of this interface. What are the cutoff frequency and cutoff
wavelength for the surface plasmon mode? Does the surface plasmon mode exist at the
500 nm wavelength? If it exists, nd its propagation constant and characteristic
parameters. Find the penetration depths of the mode into the SiO2 and the silver to nd
its connement at the interface.
3.4.8 Consider the interface between GaAs and silver. The refractive index of GaAs varies
with optical wavelength, increasing with decreasing wavelength. For simplicity, take the
refractive index of GaAs to be 3.51 at 1 m. Use the plasma frequency p 1:36
1016 rad s1 of Ag to nd the surface plasma frequency of this interface. What are the
cutoff frequency and cutoff wavelength for the surface plasmon mode? Does the surface
plasmon mode exist at the 500 nm and 1 m wavelengths, respectively? If it
exists, nd its propagation constant and characteristic parameters. Find the penetration
depths of the mode into the GaAs and the silver to nd its connement at the interface.
3.5.1 A step-index planar GaAs=AlGaAs waveguide has a GaAs core and AlGaAs cover
and substrate. At 900 nm, the GaAs core has n1 3:593, the AlGaAs substrate
has n2 3:409, and the AlGaAs cover of a different composition has n3 3:261. In
what range can the propagation constant of a guided mode, if it exists, be found at the
900 nm wavelength? Ignoring wavelength-dependent changes in the refractive
indices, for what wavelengths can a guided mode be found to have a propagation constant
of 2:5 107 m1 ? What happens to the answers if the AlGaAs composition for the
cover is changed so that n3 3:453?
3.5.2 A step-index planar glass waveguide has a glass core of n1 1:54, a glass substrate
of a different composition of n2 1:47, and a free-space cover of n3 1:00. The core
Problems
3.5.3
3.5.4
3.5.5
3.5.6
3.5.7
3.5.8
3.5.9
137
thickness is d 1:5 m. What is the range of optical wavelength for the waveguide
to support the TE0 mode but not the TE1 mode? What is the range of optical wavelength for the waveguide to support the TM0 mode but not the TM1 mode? What is the
range of optical wavelength for the waveguide to support the TE0 mode but not the
TM0 mode?
A step-index planar glass waveguide has a glass core of n1 1:54 and a substrate and a
cover of n2 n3 1:47. The core thickness is d 1:5 m. What is the range of optical
wavelength for the waveguide to support the TE0 mode but not the TE1 mode? What is
the range of optical wavelength for the waveguide to support the TM0 mode but not
the TM1 mode? What is the range of optical wavelength for the waveguide to support the
TE0 mode but not the TM0 mode?
What is the most outstanding difference between symmetric and asymmetric waveguides
in terms of nding guided modes?
A planar dielectric waveguide supports exactly three modes among all types of modes.
Name these modes. Which mode has the largest propagation constant? Which one has
the smallest propagation constant?
An asymmetric InGaAsP=InP waveguide has a refractive index of n1 3:432 for its
core, and indices of n2 3:354 and n3 3:166 for its two cladding layers. What is
the required core thickness for the waveguide to have one and only one guided mode at
1:55 m, including modes of all different polarizations?
A symmetric step-index planar InGaAsP=InP waveguide has the high-index InGaAsP for
its core and the low-index InP for its cladding layers. At 1:55 m, the core index is
n1 3:432 and the cladding index is n2 n3 3:166. If a single-mode waveguide
is desired, what is the required core thickness? Is the waveguide truly single-mode if
this requirement is met? Name the mode or modes.
A symmetric step-index planar InGaAsP=InP waveguide has a core index of n1 3:438
and a cladding index of n2 3:205. The core thickness is d 0:60 m.
(a) At the 1:30 m wavelength, how many guided modes are supported by the
waveguide? What are they?
(b) At what wavelengths does the waveguide support only one TE mode and one
TM mode?
A symmetric step-index planar GaAs=Al0:3 Ga0:7 As waveguide has the high-index GaAs
for its core and the low-index Al0:3 Ga0:7 As for its two cladding layers. At 1:5 m,
the core index is n1 3:38 and the cladding index is n2 3:22.
(a) If a single-mode waveguide is desired, what is the required core thickness? Is the
waveguide truly single-mode if this requirement is met? Name the mode or modes.
(b) If the core thickness is chosen to be d 2 m, how many guided modes are
supported by the waveguide? What are they?
(c) If the waveguide thickness is kept at d 2 m, but its structure is made asymmetric
by lowering the index of only one cladding layer, would existing modes start
disappearing or new modes start appearing if that index is sufciently reduced? What
is the rst mode to disappear or appear if this happens?
138
3.6.1 The effective index of refraction of a single-mode optical ber as a function of optical
wavelength around 1:3 m is found to be approximated as n 1:465 0:0114
1:3 0:004 1:33 , where is in micrometers.

(a) Characterize the phase-velocity dispersion of this ber at 1:2 m and 1:5 m,
respectively.
(b) Find and characterize the group-velocity dispersion of this ber at 1:2 m and
1:5 m, respectively.
(c) Express the group-velocity dispersion as D in the unit of ps km1 nm1 at 1:2 m
and 1:5 m, respectively.
3.6.2 The ber described in Problem 3.6.1 is used to transmit two optical pulses at 1:2 m
and 1:5 m, respectively. Each pulse has a pulse duration of t ps 5 ps and a
spectral width of ps 1 nm. Find the temporal widths of these two pulses after
propagating over a distance of 5 km in the ber.
3.6.3 How far can the pulse at each of the three wavelengths described in Example 3.17
propagate through that ber before the pulse broadening caused by group-velocity
dispersion is larger than the original pulse duration?
3.6.4 The ordinary and extraordinary indices of refraction of LiNbO3 in the wavelength range
between 1:0 and 2:0 m vary with wavelength approximately as
no 2:2158 0:002862 0:00622 ,
ne 2:1395 0:002472 0:00522 :
(3.191)
Answer each of the following questions for the ordinary and extraordinary waves,
respectively.
(a) Within this wavelength range, where does LiNbO3 have normal dispersion? Where
does it have anomalous dispersion?
(b) Within this wavelength range, where does LiNbO3 have positive group-velocity
dispersion? Where does it have negative group-velocity dispersion?
(c) Find the refractive index, the group index, and the group-velocity dispersion of
LiNbO3 at the three wavelengths of 1:0 m, 1:5 m, and 2:0 m.
(d) Express the group-velocity dispersion as D in the unit of fs cm1 nm1 .
3.6.5 An optical pulse has a pulse duration of t ps 100 fs and a spectral width of
ps 75 nm. Use the values of D obtained in Problem 3.6.4(d) for LiNbO3 to nd
the pulse broadening caused by group-velocity dispersion after the pulse propagates over
1 cm in LiNbO3 . Find also the distance that the pulse can propagate in LiNbO3 before
its pulse duration doubles. Answer both questions for the pulse polarized in the ordinary
and extraordinary axes, respectively, and for its center wavelength at 1:0 m, 1:5 m,
and 2:0 m, respectively.
^j given in (1.56)
3.7.1 By using the denition of the optical intensity I jS n^j jS S n
for a coherent wave and the equation k E 0 H given in (3.31), show that the
optical intensity of a plane-wave mode projected on the surface that is normal to its
propagation direction k^ is given by the expression in (3.183).
Bibliography
139
3.7.2 Show that under the condition that 00 0 , so that 00 0 , n00 n0 , and k 0 , the
absorption coefcient can be approximated as
k0
00
00
00
2 00
0
0
k

k
:
0
1 0
n02
n0
(3.192)
3.7.3 At the 300 nm wavelength, Si has a complex refractive index of n 5:0 i4:16, and
GaAs has n 3:73 i2:0. Find the absorption coefcients and the absorption depths of
Si and GaAs at this wavelength. What is the complex susceptibility for each material at
this wavelength?
3.7.4 The complex susceptibility of GaAs is 17:31 i3:70 at 500 nm and 12:55
i0:63 at 800 nm. Find the absorption coefcient and the absorption depth of GaAs
at these wavelengths.
3.7.5 At 800 nm, Si has an absorption depth of 1 9:8 m and a reectivity of 32:9% at
normal incidence on its surface exposed to air. Find its complex refractive index and
complex susceptibility at this wavelength.
3.7.6 An optical ber of a length l 120 km has an attenuation coefcient of 0:3 dB km1 at
1:3 m and 0:15 dB km1 at 1:55 m. If 2 mW of optical power at each

wavelength is launched into the ber, what is the output power at each wavelength?
3.7.7 An optical ber has an attenuation coefcient of 0:5 dB km1 at 1:3 m and
0:2 dB km1 at 1:55 m. If 1 mW of optical power at each wavelength is launched
into the ber and the detection limit of a detector at each wavelength is 1 W, what is the
maximum length of the ber for the power at each wavelength to be detectable by the
detector?
Bibliography
Buckman, A. B., Guided-Wave Photonics. Fort Worth, TX: Saunders College Publishing, 1992.
Ebeling, K. J., Integrated Optoelectronics: Waveguide Optics, Photonics, Semiconductors. Berlin: SpringerVerlag, 1993.
Haus, H. A., Waves and Fields in Optoelectronics. Englewood Cliffs, NJ: Prentice-Hall, 1984.
Hunsperger, R. G., Integrated Optics: Theory and Technology, 5th edn. New York: Springer-Verlag, 2002.
Iizuka, K., Elements of Photonics, Vols. I and II. New York: Wiley, 2002.
Kasap, S. O., Optoelectronics and Photonics: Principles and Practices, 2nd edn. Upper Saddle River, NJ:
Prentice-Hall, 2012.
Liu, J.M., Photonic Devices. Cambridge: Cambridge University Press, 2005.
Marcuse, D., Theory of Dielectric Optical Waveguides, 2nd edn. Boston, MA: Academic Press, 1991.
140

Nishihara, H., Haruna, M., and Suhara, T., Optical Integrated Circuits. New York: McGraw-Hill, 1989.
Pollock, C. R. and Lipson, M, Integrated Photonics. Boston, MA: Kluwer, 2003.
Syms, R. and Cozens, J., Optical Guided Waves and Devices. London: McGraw-Hill, 1992.
Yariv, A. and Yeh, P., Photonics: Optical Electronics in Modern Communications. Oxford: Oxford University
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
4 - Optical Coupling pp. 141-168
4
4.1
Optical Coupling
COUPLED-MODE THEORY
..............................................................................................................
Coupled-mode theory deals with the coupling of normal modes of propagation due to spatially
dependent perturbations. The theory has broad applicability. It applies to the coupling of spatial
modes in various optical structures, including Gaussian spatial modes in a homogeneous
medium, interface modes, and waveguide modes.
The space- and time-dependent electric and magnetic elds of a normal mode at a given
frequency are expressed in the form of (3.1) and (3.2). Because the coupled-mode theory
describes mode coupling caused by spatially dependent perturbations, no temporal changes are
involved. Therefore, the time dependence of all elds remains exp it throughout the
interaction so that it can be ignored in the expressions of the elds while =t is replaced
by i in Maxwells equations. Then, the two Maxwell equations for wave propagation can be
written in the form:
E i0 H,
(4.1)
H i E:
(4.2)
The normal modes of an unperturbed optical structure are governed by (4.1) and (4.2). They
are mutually orthogonal and are normalized through the orthonormality relation given in (3.18).
These normal modes form a basis for linear expansion of any optical eld at the frequency in
the optical structure:
X
^ x; y exp i z,
Er
A E
(4.3)
Hr
^ x; y exp i z,
A H
(4.4)
^ and H
^ are normalized mode elds; the linear expansion sums over all discrete
where E
indices of the guided modes and integrates over all continuous indices of the radiation and
evanescent modes. In the original, unperturbed structure where these modes are dened, the
normal modes do not couple because they are mutually orthogonal. Then, the expansion
coefcients A are constants that are independent of x, y, and z, as discussed in Section 3.1.
In the presence of a spatially dependent perturbation to an optical structure, the modes
dened by the original structure are not exact normal modes of the perturbed structure. For
this reason, the perturbation can cause coupling of these modes as they propagate. As a result, if
an optical eld in the perturbed structure is expanded in terms of the normal modes of the
142
Optical Coupling
unperturbed structure, the expansion coefcients are not constants of propagation but vary with
z as the optical eld propagates through the structure:
Er
Hr
^ x; y exp i z,
A zE
(4.5)
^ x; y exp i z:
A zH
(4.6)
Because the power in a normal mode is given by P jA j2 , according to (3.27), the z

dependence of A z in the above indicates that the power of a mode that is coupled to another
mode does not remain a constant of propagation. Thus, coupling of modes leads to exchange of
mode power.
4.1.1 Single-Structure Mode Coupling

We rst consider the coupling between normal modes in a single optical structure, such as a
single waveguide, that is subject to some perturbation. By single structure, we mean that the
entire optical structure is considered in dening the normal modes characterized by normalized

^ , H
^ of propagation constants . The structure can be a simple structure, such
mode elds E
as a homogeneous medium, a single interface, or a single waveguide; or it can be a compound
structure that consists of multiple interfaces or multiple waveguides. In any event, no matter
how complicated the structure might be, it is considered as a single entity and is described with
a single r to dene the normal modes.
A spatially dependent perturbation to the structure at a frequency of can be represented
by a single perturbing polarization, Pr, so that the equations in (4.1) and (4.2) are modied
as
E i0 H,
(4.7)
H i E iP:
(4.8)
Any optical eld propagating in this perturbed structure can be expanded as (4.5) and (4.6)
while its propagation is governed by these two equations with P 6 0. Meanwhile, the normal
mode elds dened by the unperturbed structure, which are dened by (4.1) and (4.2), also
satisfy these two equations with P 0.
Applying (4.7) and (4.8) to two arbitrary sets of elds, E1 ; H1 and E2 ; H2 , with respective
perturbations of P1 and P2 , we nd the Lorentz reciprocity theorem:

E1 H
2 E2 H1 i E1 P2 E2 P1 ,
(4.9)
which holds for any two sets of elds that are respectively associated with two arbitrary
perturbations. To derive the couple-mode equation, we take E1 ; H1 to be the optical eld
propagating in the perturbed structure with P1 P, which can be expanded as (4.5) and

^ , H
^ dened by the unperturbed structure
(4.6), and E2 ; H2 to be the normal mode elds E
143
with P2 0. By substituting these into (4.9) and integrating both sides of the resultant
equation over the cross section of the waveguide, we nd

Xd
i z
i z
^
^
^ Pdxdy:
^
^
E H E H ^z dxdy ie
E
A ze
dz

(4.10)
By applying the orthonormality relation given in (3.18) to (4.10), we nd the general form of
the coupled-mode equations:
dA

iei z
dz
^ Pdxdy,
E
(4.11)

where the plus sign is used when > 0 for mode to be forward propagating in the positive z
direction, and the minus sign is used when < 0 for mode to be backward propagating in the
negative z direction.
The general form of the coupled-mode equations expressed in (4.11) is applicable to mode
coupling caused by any kind of spatially dependent perturbation on any feature of the optical
structure. For example, P can be a perturbing polarization at the frequency on the elds in a
waveguide due to any of the external effects discussed in Section 2.6 or due to any nonlinear
optical susceptibility discussed in Section 2.7.
For the simple case where the perturbation can be represented by a change in the linear
polarization as
X
^ ei z ,
P E
A E
(4.12)
the coupled-mode equations can be expressed in the form:

dA X
i A ei z ,
dz
(4.13)
where

^ dxdy
^ E
E
(4.14)

is the coupling coefcient between mode and mode . This result is applicable to isotropic and
anisotropic structures. For an optical structure made of isotropic media, simply reduces to a
^ E
^ E
^ E
^ in (4.14). For a lossless optical structure, the
scalar so that E
dielectric tensor is a Hermitian matrix so that ij

ji , as discussed in Section 2.2. Consequently, mode coupling in a lossless dielectric single structure is symmetric with

:
(4.15)
144
Optical Coupling
EXAMPLE 4.1
Any physical mechanism that creates a change in the optical permittivity of a material can
possibly be a perturbation for the coupling of two modes in a waveguide. Is the mode coupling
caused by the electro-optic Pockels effect symmetric? Is that caused by optical absorption in a
semiconductor due to current injection symmetric?
Solution:
The Pockels effect mainly changes the permittivity tensor without causing additional optical
loss. The permittivity change is Hermitian: . Thus the mode coupling caused by this
effect is symmetric:

^
E^
E dxdy

A
E^ E^
dxdy

1
^
A
E^
E dxdy

0
@
1
^
A
E^
E dxdy

The permittivity change associated with optical absorption is not Hermitian: 6 . Thus
the mode coupling caused by this effect is not symmetric:

^
E^
E dxdy

@
0

@
0

6 @
1
A
E^ E^
dxdy
1
^
A
E^
E dxdy
6
:
1
^
A
E^
E dxdy

145
4.1.2 Multiple-Structure Mode Coupling

In a compound optical structure, such as a structure that consists of more than one waveguide, there
are two alternative approaches to the analysis of the characteristics of optical elds that propagate in
the structure. One approach is to treat the compound structure as a single super structure by expanding
any optical eld in terms of its normal modes, known as the super modes, which are found by solving
Maxwells equations directly with the boundary conditions dened by the entire super structure. The
alternative approach is to divide the compound structure into separate substructures, expand the elds
in terms of the normal modes of the individual substructures, and treat the problem with a coupledmode approach. The rst approach can yield exact solutions and is sometimes desirable. However, it
is not generally possible to obtain the exact super-mode solutions for complicated structures. The
coupled-mode approach yields approximate solutions, but it can be applied to most structures without
difculty. In addition, it gives an intuitive picture of how optical waves interact in a compound
structure. Here we consider the coupled-mode formulation for multiple substructures.
The concept of dividing a super structure into a combination of individual substructures is
illustrated in Fig. 4.1. In this illustration, the individual waveguides are the substructures of the
multiple-waveguide super structure. The multiple-waveguide super structure is described by
x; y, whereas the individual waveguides are described by a x; y, b x; y, c x; y, and so on.
The normal modes are solved for each individual substructure. The elds in the entire structure
can be expanded in terms of these normal modes in the same form as (4.5) and (4.6) but with the
summation over the index covering all the modes of every substructure. From the standpoint of
any substructure, the presence of other substructures is a perturbation to it. Thus, for substructure
i that is described by i x; y, the entire structure looks like i x; y plus a perturbation of
i x; y x; y i x; y:
(4.16)
The coupled-mode equations for the multiple-structure scenario can be obtained by using the
reciprocity theorem of (4.9) and then following a procedure similar to that taken above to obtain
Figure 4.1 Schematic diagram of three coupled waveguides showing the decomposition into individual
waveguides, in solid curves, plus the corresponding perturbation, in dashed curves, for each of them.
146
Optical Coupling
the coupled-mode equations for the single structure. Because the mathematics is quite involved,
only the results are given in the following without detailed derivation.
The coupled-mode equations for multiple substructures can still be written in the same form
as that of (4.13):

dA X
i A ei z ,
dz
(4.17)
where the plus sign is taken if mode is forward propagating, and the minus sign is used if it is
backward propagating. It is noted that the summation over the index runs through the modes
of every substructure, not just the modes of one single substructure. In contrast to that for
single-structure coupling discussed above, the coupling coefcients for multiple-structure
coupling have a complicated form and are best expressed in terms of matrix elements:

~ ,
(4.18)
c c1
where c 1 if mode is forward propagating and c 1 if it is backward propagating, as

~ ~ are given,
can be seen from (4.19) below. The elements of the matrices c c and
respectively, by
c
^
^
^
^
E H E H ^z dxdy c
(4.19)

and

~
^ dxdy:
^ E
E
(4.20)

Note that in (4.20) is the perturbation, dened in (4.16), to the substructure that denes the

^ , H
^ of normal mode . The coefcient c represents the overlap coefcient of
elds E

^ , H
^ , H
^ and E
^ , which can be the mode elds of different substructures in the super
E
structure. In general, c 6 0 because modes of different substructures are not necessarily
orthogonal to each other. Because the mode elds used in (4.19) are normalized, we have
c 1 or c 1, depending
on whether mode is forward or backward propagating as

mentioned above, and c 1 for any and . Note also the difference between the form of ~
expressed in (4.20) and that of the single-structure coupling coefcients given in (4.14).
As discussed above and expressed in (4.15), the coupling between modes of a single structure
is always symmetric with
if the structure is dielectric and lossless. By contrast, the
coupling between modes of different substructures in a super structure, such as those of
different individual waveguides in a multiple-waveguide structure, is generally asymmetric:
~ 6 ~
and 6
(4.21)
where and refer to modes of two different substructures. Indeed, it can be shown by using
the reciprocity theorem that
~ ~

c c
c :
2
147
(4.22)
This relation indicates that there is a direct relationship between the coupling coefcients and
the propagation constants. It has the following implications.
1. Unless or c c
0, coupling between two modes is not symmetric, i.e.,
6 , because the normal modes of different substructures are not necessarily orthogonal to each other.
2. The coupling of modes of the same order between two identical substructures is always
symmetric because , resulting in ~ ~

and .
3. The relation in (4.22) applies to modes of a single structure as well. In this situation,
~ . Therefore,
c c
0 if 6 , and
in (4.15) holds true for the
normal modes of the same structure because they are mutually orthogonal.
4. It is not possible to change the coupling between two modes without simultaneously
changing their overlap coefcient or their propagation constants.
4.2
TWO-MODE COUPLING
..............................................................................................................
The coupling between two modes is the simplest and most common situation of mode coupling.
It includes coupling between two modes of the same structure, such as mode coupling in a
single waveguide that is modulated by a grating, or coupling between modes of two substructures, such as mode coupling in a directional coupler that is formed by two parallel waveguides.
For two-mode coupling, the coupled-mode equations can be written in a simple form that can
be analytically solved. In this section, we consider the general formulation of two-mode
coupling.
We have shown that both coupling among modes of a single structure and coupling among
modes of different substructures can be described by coupled-mode equations of the same form
as given in (4.13) and (4.17). The only difference is that the coupling coefcients in (4.17) for
multiple-structure mode coupling are dened differently from those in (4.13) for singlestructure mode coupling. This commonality is convenient because the general solutions of
the coupled-mode equations can be applied to both cases. For a particular problem, we only
have to calculate the coupling coefcients that are specic to the problem under consideration.
For two-mode coupling either in a single structure or between two different substructures, the
eld expansion in (4.5) and (4.6) consists of only two modes, designated as mode a and mode b
of amplitudes A and B, respectively. Thus, coupled-mode equations of the form given in (4.13)
or (4.17) reduce to the following two coupled equations:

dA
iaa A iab Beib a z ,
dz
(4.23)
dB
ibb B iba Aeia b z :
dz
(4.24)
148
Optical Coupling
For coupling between two modes of a single structure, the coupling coefcients in these
equations are given by (4.14), which are always symmetric with ab
ba if the structure is
dielectric and lossless. For coupling between modes of two different substructures, the coupling
coefcients are given by (4.18), which can be explicitly expressed as
aa
ba
~aa cab ~ba =cbb

~ab cab ~bb =cbb
, ab
,
1 cab cba =caa cbb
1 cab cba =caa cbb
~ba cba ~aa =caa

~bb cba ~ab =caa
, bb
:
1 cab cba =caa cbb
1 cab cba =caa cbb
(4.25)
As discussed earlier and expressed in (4.21), in general ab 6

ba for coupling between modes
of two different substructures.
The iaa A and ibb B terms in the coupled equations (4.23) and (4.24) are self-coupling terms.
These terms are caused by the fact that the normal modes see in the perturbed structure an index
prole that is different from the index prole of the unperturbed original structure where the
modes are dened. They can be removed from the equations by expressing the normal-mode
expansion coefcients as
2 z
3
~ z exp 4i aa zdz5,

(4.26)
Az A
0
~ z exp 4i bb zdz5,

Bz B
(4.27)
where a plus or minus sign is chosen for a forward-propagating or backward-propagating mode,

respectively. Then (4.23) and (4.24) can be transformed into two coupled equations in terms of
~ and B
~ to remove the self-coupling terms:
A
~
dA
~ iz ,
iab zBe
dz
(4.28)
~
dB
~ iz ,
iba zAe
dz
(4.29)

where
2
z 4b z bb zdz5 4a z aa zdz5:
0
(4.30)
As shown in (4.28)(4.30), we have to consider the fact that each coupling coefcient can be
a function of z because can be a function of z but the integration in (4.14) and (4.20) is
carried out only over x and y. In the case when ab z and ba z are arbitrary functions of z, the
coupled-mode equations cannot be analytically solved. In this situation, there is no need to
further simplify the coupled-mode equations because they can only be numerically solved.
However, for optical structures of practical interest that are designed for two-mode coupling,
149
is usually either independent of z or periodic in z. Then, the coupling coefcients are either
independent of z or periodic in z. In either case, (4.28) and (4.29) can be reduced to the
~ and B
~ with ab and ba being constants that are independfollowing general form in terms of A
ent of z:
~
dA
~ i2z ,
iab Be
dz
(4.31)
~
dB
~ i2z :
iba Ae
dz
(4.32)

The parameter 2 is the phase mismatch between the two modes. Perfectly phase-matched
coupling of two modes with 0 is always symmetric with ab
ba irrespective of whether
these two modes belong to the same structure or two different substructures.
The general form of (4.31) and (4.32) applies to both cases of uniform and periodic
perturbations, but the details of the parameters vary between the two cases.
4.2.1 Uniform Perturbation

In this case, is only a function of x and y but is not a function of z. Then all of the coupling
coefcients aa , bb , ab , and ba in (4.28)(4.30) are constants that are independent of z. We
then nd that
~ zeiaa z ,
Az A
~ zeibb z ,
Bz B
(4.33)
and 2z z b bb a aa z for (4.30) so that

2 b bb a aa :
(4.34)
The choice of sign in each in (4.33) and (4.34) is consistent with that in (4.26) and (4.27)
discussed above. The physical meaning of the self-coupling coefcients, aa and bb , is a
change in the propagation constant of each normal mode. While the propagation constants of
the normal modes in the original unperturbed structure are a and b , their values are changed
by the perturbation characterized by . These modes now propagate with the modied
propagation constants a aa and b bb , respectively, which take into account the effect
of the perturbation on the structure. In addition, they couple to each other through ab and ba .
With the simple transformation of (4.33) and the phase mismatch 2 given in (4.34), twomode coupling due to a uniform perturbation is described by the general form of (4.31) and
(4.32) with constant values of ab and ba . A good example of two-mode coupling due to a
uniform perturbation is that in a two-channel directional coupler, which consists of two parallel
single-mode waveguides, as shown schematically in Fig. 4.2. This is the case of multiplestructure coupling. If the two waveguides are not identical, the directional coupler is not
symmetric. Then, in general ba 6
ab , as discussed in Section 4.1. Furthermore, 2 6 0 except
for a certain possible phase-matched optical frequency because aa 6 bb and a 6 b in
general. If the two waveguides are identical, the directional coupler is symmetric. Then,
ba
ab , aa bb , and a b so that 2 0 for all frequencies.
150
Optical Coupling
Figure 4.2 Schematic diagram of (a) a two-channel directional coupler of a length l consisting of two parallel
waveguides and (b) its index prole assuming two step-index waveguides on the same substrate. The coupler is
symmetric if na nb n1 and d a d b d.
4.2.2 Periodic Perturbation

In this case, is a periodic function of z, and so are the coupling coefcients aa z, bb z, ab z,
and ba z in (4.28)(4.30). The periodic perturbation has a period of and a wavenumber of
2
:
(4.35)
Each coupling coefcient, being periodic in z with a period of , can be expanded in a Fourier series:
X
X
q exp iqKz
q exp iqKz
(4.36)
z
K
where q represents the order of coupling, the summation over q runs through all integers, and
1
q
z exp iqKzdz:
(4.37)
Using (4.36) for ab z and ba z, (4.28) and (4.29) can be expressed as

X
~
dA
~ iziqKz ,
ab qBe
i
dz
q
(4.38)
X
~
dB
~ iziqKz :
ba qAe
i
dz
q
(4.39)

For aa z and bb z, we nd that

z
X q

zdz 0z
eiqKz 1 :
iqK
q60
0
(4.40)
151
The 0 term represents a possible uniform perturbation that might exist due to a uniform
bias in the periodic . It can be removed by redening or by considering it separately. In
any event, for Kz 1,

X q

iqKz
1 Kz:
(4.41)
e

q60 iqK

Therefore, the contributions of the q 6 0 terms of aa z and bb z to the z-dependent phases in
(4.38) and (4.39) are negligible so that
2
3 2
3
z
z
z qKz 4b z bb zdz5 4a z aa zdz5 qKz
(4.42)
0
fb bb 0 a aa 0 qK gz:
With this approximation, the coupled-mode equations in the case of a periodic perturbation can
be expressed as
X
~
dA
~ iziqKz
iab qBe
~ i2z ,
ab qBe
i
dz
q
(4.43)
X
~
dB
~ iziqKz
iba qAe
~ i2z ,
ba qAe
i
dz
q
(4.44)

where
2 b bb 0 a aa 0 qK:
(4.45)
Note that only one term in the Fourier series that yields a minimum value for jj is kept in each
of the two coupled-mode equations expressed in (4.43) and (4.44) because only this term will
effectively couple the two modes. Thus, the coupled-mode equations in (4.43) and (4.44) have
the general form of (4.31) and (4.32) with ab ab q and ba ba q being constants that
are independent of z, where q is the integer chosen to minimize the phase mismatch given
in (4.45).
The most common periodic perturbations are gratings. The simplest gratings are onedimensional gratings. For our purpose, such one-dimensional gratings are structures that are
periodic only in the longitudinal direction, which is taken to be the z direction. Grating
waveguide couplers have many useful applications and are one of the most important kinds
of waveguide couplers. They consist of periodic ne structures that form gratings in waveguides. The grating in a waveguide can take the form of either periodic index modulation or
periodic structural corrugation. Periodic index modulation can be permanently written in a
waveguide by periodically modulating the doping concentration in the waveguide medium, for
example, or it can be created by an electro-optic, acousto-optic, or nonlinear optical effect.
Figure 4.3 shows some examples of planar grating waveguide couplers in single waveguides. In
these examples, there is no uniform perturbation apart from the periodic perturbation; therefore,
aa 0 bb 0 0 in (4.45) for these single-waveguide grating couplers.
152
Optical Coupling
Figure 4.3 Structures of planar grating waveguide couplers with (a) and (b) periodic index modulation, (c), (d),
(e), and (f) periodic structural corrugation.
EXAMPLE 4.2
Find the qth-order coupling coefcient q for a sinusoidal grating that has a period of , as
shown in Fig. 4.3(c), such that z a cos Kz, where K 2=. Find it for a squarefunction grating that has a period of and a duty factor of , as shown in Fig. 4.3(d), such
that z a for 0 < z < and z a for < z < within each period. In each
case, which orders are useful for mode coupling?
Solution:
For the sinusoidal grating, we nd by using (4.37) that
1
q
z exp iqKzdz
1
a cos Kz exp iqKzdz
exp iKz iqKz exp iKz iqKz

dz:
2

a
q, 1 q, 1 ,
2
where q, 1 and q, 1 are the Kronecker delta functions. Therefore, only the order q 1 and
q 1 the order are useful for mode coupling because only these two orders have a nonzero
coupling coefcient of 1 1 a=2.
153
For the square-function grating, we nd by using (4.37) that
1
q
z exp iqKzdz
1
1
a exp iqKzdz
a exp iqKzdz:
2a
sin q iq
:
e
q
We nd that q for a given value of q can be made nonzero by an appropriate choice of the
duty factor . Therefore,
any order can be used if the value of is properly chosen to maximize

the value of q for a given q. However, it is possible to have q 0 for certain
combinations of the values
of q and , such as q 2 and 1=2, or q 3 and 1=3, etc.
The largest value of q appears when q 1 or q 1 while 1=2 so that

q 2a=.
A grating can also be used in a multiple-structure coupler. Figure 4.4 shows an example of a
grating placed in a dual-channel coupler that consists of two waveguides. The two waveguides
can be either identical, as in a symmetric structure, or nonidentical, as in an asymmetric
structure. In both cases, the phase mismatch of this dual-channel coupler with a grating is that
given in (4.45) with aa 0 6 0 and bb 0 6 0 due to the uniform perturbation on one
waveguide by the other waveguide, as in the directional coupler shown in Fig. 4.2.
EXAMPLE 4.3
Find the grating period for perfect phase matching of two modes a and b.
Solution:
For perfect phase matching, the phase mismatch given in (4.45) between two modes a and b of
propagation constants a and b has to be made zero by the perturbation of a grating:
2 b bb 0 a aa 0 qK 0
)
)
)
qK a aa 0 b bb 0
2
a aa 0 b bb 0
2q
q
,
a aa 0 b bb 0
q
where a aa 0 b bb 0 is the total phase mismatch including all uniform perturbations on the structure, and the sign of q is chosen to be the sign of a aa 0 b bb 0
so that the grating period has a positive value.
154
Optical Coupling
Figure 4.4 Dual-channel directional coupler with a
grating of period .
With the above general considerations, (4.31) and (4.32) represent the most general coupled
equations for two-mode coupling in structures of practical interest. They can be analytically
solved; their solutions apply to various two-mode coupling problems.
4.3
CODIRECTIONAL COUPLING
..............................................................................................................
First, we consider the coupling of two modes that propagate in the same direction, taken to be
the positive z direction, over a length of l, as is shown in Fig. 4.5. In this case, a > 0 and
b > 0. The coupled equations are
~
dA
~ i2z ,
iab Be
dz
(4.46)
~
dB
~ i2z :
iba Ae
dz
(4.47)
The equations for codirectional coupling are generally solved as an initial-value problem with
~ z and B
~ z0 and B
~ z0 at z z0 to nd the values of A
~ z at any other
given initial values of A
location z. The general solution can be expressed in the matrix form:
"
#
"
#
~ z
~ z0
A
A
Fz; z0
,
(4.48)
~ z
~ z0
B
B
where the forward-coupling matrix Fz; z0 relates the eld amplitudes at the location z0 to
those at the location z. It has the form:
2
3
c cos c zz0 i sin c zz0 izz0
iab
izz0
e
sin
zz
e
0
c
6
7
c
c
7
Fz;z0 6
4
iba
c cos c zz0 i sin c zz0 izz0 5
izz0
sin c zz0 e
e
c
c
(4.49)
where

1=2
:
c ab ba 2
(4.50)
We consider a simple case when power is launched only into mode a at z 0. Then the initial
~ 0 6 0 and B
~ 0 0. By applying these conditions to (4.48) and taking z0 0 in
values are A
(4.49), we nd that
4.3 Codirectional Coupling
155
Figure 4.5 Codirectional coupling between two modes of propagation constants a and b (a) in the same
waveguide and (b) in two parallel waveguides. A perturbation is required for codirectional coupling in the same
waveguide but is not required for codirectional coupling between two waveguides.
Figure 4.6 Periodic power exchange between two codirectionally coupled modes for (a) the phase-mismatched
condition 6 0 and (b) the phase-matched condition 0. The solid curves represent Pa z=Pa 0, and the
dashed curves represent Pb z=Pa 0.
i
~
~
A z A 0 cos c z
sin c z eiz ,
c

iba
~
~
sin c z eiz :
B z B 0
c
(4.51)
(4.52)
The power in the two modes varies with z as

~ z 2 ab ba
Pa z A
2
2

cos
,

c
~ 0
Pa 0
2c
2c
A
(4.53)

~ z 2 jba j2
Pb z B

sin2 c z:
2
~ 0
Pa 0 A
c
(4.54)
The coupling efciency for codirectional coupling over a length of l is
Pb l jba j2 2
2 sin c l:
Pa 0
c
(4.55)
156
Optical Coupling
Thus, power is exchanged periodically between two modes with a coupling length of
lc
,
2c
(4.56)
where maximum power transfer occurs. Figure 4.6 shows the periodic power exchange between
the two coupled modes as a function of z. As can be seen from Fig. 4.6, complete power transfer
can occur only in the phase-matched condition when 0.
EXAMPLE 4.4
Find the maximum coupling efciency for codirectional coupling and the length of a codirectional coupler that reaches this efciency. What happens if the phase mismatch is large such
that 2 > ab ba ?
Solution:
From (4.55), the maximum efciency for codirectional coupling is
max
jba j2
jba j2
,
2c
ab ba 2
which is reached when sin2 c l 1. Because sin2 c l is periodic, sin2 c l 1 has many
solutions. The length to reach the maximum efciency is any of
lmax 2m 1
2m 1lc for m 0, 1, 2, . . .
2c
The formulas obtained above remain valid for 2 > ab ba . There are no qualitative changes,
but only quantitative changes, when the phase mismatch is large such that 2 > ab ba . The
maximum coupling efciency decreases with increasing phase mismatch because c increases
with 2 . The length lmax to reach the maximum efciency also decreases with increasing phase
mismatch because the coupling length lc decreases with increasing c .
4.4
CONTRADIRECTIONAL COUPLING
..............................................................................................................
We now consider the coupling of two modes that propagate in opposite directions over a length
of l, as is shown in Fig. 4.7 where mode a is forward propagating in the positive z direction and
mode b is backward propagating in the negative z direction. In this case, a > 0 and b < 0.
Thus, the coupled equations are
~
dA
~ i2z ,
iab Be
dz
(4.57)
~
dB
~ i2z :
iba Ae
dz
(4.58)
4.4 Contradirectional Coupling
157
Figure 4.7 Contradirectional coupling between two modes of propagation constants a and b (a) in the same
waveguide and (b) in two parallel waveguides. A signicant perturbation is required for contradirectional
coupling in both cases.
The equations for contradirectional coupling are generally solved as a boundary value problem with
~ 0 at one end and B
~ z and B
~ l at the other end to nd the values of A
~ z
given boundary values of A
at any location z between the two ends. The general solution can be expressed in the matrix form:
"
#
"
#
~ z
~ 0
A
A
Rz; 0; l
(4.59)
~ z
~ l
B
B
~ 0 at z 0 and B
~ l
where the reverse-coupling matrix Rz; 0; l relates the eld amplitudes A
at z l to those at any location z. It has the form:
2
3
c cosh c l z i sinh c l z iz
iab sinh c z
ilz
e
e
6
7
c cosh c l i sinh c l
6
7
Rz; 0; l 6
7
4
iba sinh c l z
c cosh c z i sinh c z ilz 5
iz
e
e
(4.60)
where

1=2
:
c ab ba 2
(4.61)
We consider a simple case when power is launched only into mode a at z 0 but not into
~ 0 6 0 and B
~ l 0. By applying these
mode b at z l. Then the boundary values are A
conditions to (4.59), we nd that
~ z A
~ 0 c cosh c l z i sinh c l z eiz ,
A
~ 0
~ z A
B
iba sinh c l z
eiz :
The power in the two contradirectionally coupled modes varies with z as

~ z 2 cosh2 c l z 2 =ab ba
Pa z A

,
~ 0
Pa 0 A
cosh2 c l 2 =ab ba

~ z 2
Pb z B
sinh2 c l z
ba

:
~ 0
ab cosh2 c l 2 =ab ba
Pa 0
A
(4.62)
(4.63)
(4.64)
(4.65)
Because mode b is propagating backward with no input at z l but with an output at z 0, the
coupling efciency for contradirectional coupling over a length of l is
158
Optical Coupling
Figure 4.8 Power exchange between two contradirectionally coupled modes for (a) the phase-mismatched
condition 6 0 and (b) the phase-matched condition 0. The solid curves represent Pa z=Pa 0, and the
dashed curves represent Pb z=Pa 0.

~ 02
Pb 0 B
sinh2 c l
ba
:

~ 0
Pa 0
ab cosh2 c l 2 =ab ba
A
(4.66)
Figure 4.8 shows the power exchange between the two contradirectionally coupled modes as a
2
function of z. Power transfer approaches 100% as l ! if ab
ba and < ab ba .
~ 0 6 0 and B
~ l 0, as considered above, contradirectional coupling can
In the case when A
~ 0 at z 0 with a reection coefcient of
be viewed as reection of the eld amplitude A
~ 0
iba sinh c l
B
r jr jei
:
(4.67)
~ 0 c cosh c l i sinh c l
A
The reectivity is R jr j2 as is given in (4.66). The phase shift is

1
1
ba tan
tanh c l PM tan
tanh c l ,
2
c
c
(4.68)
where ba is the phase angle of ba , and PM =2 ba is the phase shift at the phasematched point where 0.
EXAMPLE 4.5
Find the maximum coupling efciency for contradirectional coupling and the length of a
contradirectional coupler that reaches this efciency. What happens if the phase mismatch is
large such that 2 > ab ba ?
Solution:
In the case when 2 < ab ba , the parameter c given in (4.61) has a real, positive value. Then,
sinh c l and cosh c l are both monotonic functions with sinh c l ! 1 and cosh c l ! 1 as
l ! . From (4.66), the maximum efciency for contradirectional coupling in the case when
2 < ab ba is
4.5 Conservation of Power
max
159
ba
,
ab
which can only be asymptotically reached when l ! . Therefore, lmax when 2 < ab ba .
In the case when 2 > ab ba , we nd that the parameter c given in (4.61) becomes purely
imaginary:

1=2

1=2
ic with c 2 ab ba
:
c ab ba 2
Then the coupling efciency given in (4.66) becomes
sin2 c l
ba
:
ab 2 =ab ba cos2 c l
We nd that varies with l periodically. By taking d=dc l 0, the maximum value of is

found when 2c l 2m 1. Thus, it takes place when sin2 c l 1 and cos2 c l 0 with
max
jba j2
:
2
The length to reach this maximum efciency is any of

lmax 2m 1
2m 1

2c 2 2 ab ba 1=2
for m 0, 1, 2, . . .
For contradirectional coupling, there is a qualitative change in the coupling efciency when the
phase mismatch becomes large so that 2 > ab ba .
4.5
CONSERVATION OF POWER
..............................................................................................................
Conservation of power requires that in a lossless structure the net power owing across any
cross section of the structure be a constant that does not vary along the longitudinal direction of
the structure. For codirectional coupling between two modes with the power initially launched
into only one mode such that Pa 0 6 0 but Pb 0 0, this requirement suggests that the sum
of power in the two waveguides, Pa z Pb z, be a constant independent of z because the
power in the two modes ows in the same direction. For contradirectional coupling with the
power launched into only one mode such that Pa 0 6 0 and Pb l 0, this requirement
suggests that Pa z Pb z be a constant independent of z because the power in mode b ows
in the backward direction while that in mode a ows in the forward direction. These conclusions are correct for mode coupling in a single structure, such as a single waveguide, but they
do not generally hold for coupling between modes of two different substructures, such as two
separate waveguides.
It can be seen from (4.53) and (4.54) that Pa z Pb z is not a constant of z for codirectional
coupling unless ab
ba . Similarly, from (4.64) and (4.65), it is also found that Pa z Pb z
160
Optical Coupling
is not a constant of z for contradirectional coupling when ab 6

ba . It seems that the total
power is not conserved in a lossless structure in the case of asymmetric coupling with
ab 6
ba . A close examination reveals that because cab 6 0 in the case of asymmetric
coupling, the two interacting modes are not orthogonal to each other. For this reason, the total
power ow cannot be fully accounted for by gathering the power in each individual mode as if
the modes were mutually orthogonal. Indeed, by expanding the total electric and magnetic
elds in the structure as a linear superposition of the two modes in the form of (4.5) and (4.6) to
calculate the power of the entire structure, we nd that the total power as a function of space is

Pz caa jAzj2 cbb jBzj2 2Re cab A zBzeiz
(4.69)
caa Pa z cbb Pb z Pab z,

where Pab z 2Re cab A zBzeiz can be considered as the power residing between the
two nonorthogonal modes of the two different substructures. As dened in Section 4.1, c 1
if mode is forward propagating and c 1 if mode is backward propagating. It can be
shown, using (4.53) and (4.54) for the case of codirectional coupling and using (4.64) and
(4.65) for the case of contradirectional coupling, that Pz given in (4.69) is a constant
independent of z no matter whether ab

ba or ab 6 ba . Therefore, conservation of power
holds as expected.
It can be shown simply by applying conservation of power that the coupling is symmetric
with ab
ba when Pab z 0. Conversely, if the coupling is symmetric, Pab z always
vanishes even when mode a and mode b are not orthogonal to each other. Two conclusions
can thus be made.
1. If mode a and mode b are orthogonal to each other with cab 0, then Pab z 0 and
ab
ba even when the two modes are not phase matched so that 6 0.
2. If mode a and mode b are phase matched with 0, then Pab z 0 and ab
ba even
when the two modes are not orthogonal to each other with cab 6 0.
Consequently, coupling between two modes a and b is symmetric with ab
ba if these two
modes are orthogonal to each other or if they are phase matched.
4.6
PHASE MATCHING
..............................................................................................................
As can be seen from Figs. 4.6 and 4.8, power transfer is most efcient when 0. The
parameter is a measure of phase mismatch between the two modes being coupled. For the
simple case when 2 b a , the phase-matching condition 0 is achieved when
a b . Then, the two modes are synchronized to have the same phase velocity. In the case
when includes a contribution from additional structural perturbation, such as a periodic
grating, phase matching of the two modes being coupled can be accomplished by compensating
for the difference b a with a perturbation phase factor to make 0, as can be seen
in (4.34) for a uniform perturbation and in (4.45) for a periodic perturbation. When considering
phase matching between two modes, it is important to always include all sources of contribution to the phase-mismatch parameter . When all contributions to the phase mismatch are
4.6 Phase Matching
161
considered and their effects on the coupling coefcients are accounted for, the coupling
coefcients and the phase mismatch have a relation similar to (4.22):

ab
(4.70)
ba cab cba cab 2:
Phase-matched coupling is always symmetric because ab
ba whenever 0, as seen in
(4.70). This statement is true even when cab 6 0 and a 6 b . However, symmetric coupling
does not necessarily imply a phase-matched condition because symmetric coupling can be
accomplished by having cab 0 when 6 0, as also seen in (4.70). Therefore, though 0
always implies ab
ba , the converse is not true; it is possible to have ab ba when 6 0.
The clearest example of this situation is the coupling between two phase-mismatched modes in
the same waveguide.
4.6.1 Phase-Matched Coupling

When perfect phase matching is accomplished so that 0, we can take
i
ab
ba jje :
(4.71)
c c jj:
(4.72)
Because 0, we nd that
With these relations, the matrix Fz; z0 for codirectional coupling is reduced to

iei sin jjz z0
cos jjz z0
FPM z; z0
,
iei sin jjz z0
cos jjz z0
and the matrix Rz; 0; l for contradirectional coupling is reduced to
2
3
cosh jjl z
i sinh jjz
ie
6
cosh jjl
cosh jjl 7
6
7
RPM z; 0; l 6
7:
4 i sinh jjl z
cosh jjz 5
ie
cosh jjl
cosh jjl
(4.73)
(4.74)
For perfectly phase-matched codirectional coupling, the coupling efciency is

PM sin2 jjl,
(4.75)
as shown in Fig. 4.9(a), and the coupling length is

lPM
c
:
2jj
(4.76)
PM
By choosing the interaction length to be l lPM
c , or any odd multiple of lc , 100% power
transfer from one mode to the other with PM 1 can be accomplished.
For perfectly phase-matched contradirectional coupling, the coupling efciency is
PM tanh2 jjl,
(4.77)
162
Optical Coupling
Figure 4.9 Coupling efciency PM as a function of the normalized coupling length jjl for (a) perfectly phasematched codirectional coupling and (b) perfectly phase-matched contradirectional coupling.
as shown in Fig. 4.9(b). For an interaction length of l lPM

dened in (4.76), phase-matched
c
contradirectional coupling has a coupling efciency of PM 84%. Although complete power
transfer with 100% efciency cannot be accomplished for contradirectional coupling, most
power is transferred in a length comparable to the coupling length of codirectional coupling if
perfect phase matching is accomplished.
EXAMPLE 4.6
The coupling efciency of a contradirectional coupler never reaches 100% but only approaches
100% as the length of the coupler approaches innity: ! 1 as l ! . For a practical
application, 99% might be as good. Find the length of a perfectly phase-matched contradirectional coupler that has 99%.
Solution:
The length for a perfectly phase-matched contradirectional coupler that has 99% is
found as
2
99% tanh jjl99% 0:99 )
l99%
p 3:0
1
1
tanh
0:99
:
j j
j j
EXAMPLE 4.7
A 3-dB coupler is one that has a coupling efciency of 50%. Consider a 3-dB codirectional
coupler and a 3-dB contradirectional coupler. Both have perfect phase matching and have the
same coupling coefcient of . Find the length l3dB of each phase-matched 3-dB coupler in
terms of jj?
4.6 Phase Matching
163
Solution:
Using (4.75), the length of a phase-matched 3-dB codirectional coupler is found to be one of the
many values:

1
1
1
1
2
1
3dB sin jjl3dB
p m
) l3dB
sin
for m 0, 1, 2, . . .
2
2 2jj
j j
2
Using (4.77), the length of a phase-matched 3-dB contradirectional coupler is found to have
only one value:
3dB tanh2 jjl3dB
1
2
l3dB
1
1
0:88
tanh1 p
:
j j
j j
2
The values of l3dB found above for codirectional and contradirectional coupling can be seen in
Figs. 4.9(a) and (b), respectively.
4.6.2 Phase-Mismatched Coupling

In the presence of phase mismatch with 6 0, symmetric coupling with ab
ba is still true
for coupling between two modes in the same structure but is not necessarily true for coupling
between two different substructures. Nevertheless, to illustrate the effect of phase mismatch on
the coupling efciency between two modes, we consider the simple case that ab
ba , as
expressed in (4.71).
For codirectional coupling with a phase mismatch of , the coupling efciency obtained in
(4.55) can be written in terms of jjl and j=j as
q
1
2
(4.78)
sin jjl 1 j=j2 :
1 j=j2
The maximum efciency is
max
1
1 j=j2
(4.79)
at a coupling length of
lPM
c
:
lc q
2
1 j=j
(4.80)
The maximum coupling efciency is clearly less than unity when 6 0. As shown in
Fig. 4.10(a), both lc and max decrease as j=j increases. If the interaction length is xed at
l lPM
c , the efciency drops quickly as j=j increases, as shown in Fig. 4.10(b).
For contradirectional coupling with a phase mismatch of , the coupling efciency can be
expressed in terms of jjl and j=j as
164
Optical Coupling
Figure 4.10 Effect of phase mismatch on codirectional coupling showing, as a function of j=j, (a) the
coupling length lc , normalized as lc =lPM
c , and the maximum coupling efciency max and (b) the coupling
PM
efciency for xed interaction lengths of l lPM
, 3lPM
c
c ,5lc .
Figure 4.11 Effect of phase mismatch on contradirectional
coupling showing the coupling efciency for a few different
values of jjl as a function of j=j.
sinh jjl 1 j=j2

q
:
2
2
2
cosh jjl 1 j=j j=j
2
(4.81)
The coupling efciency decreases as phase mismatch increases, as seen in Fig. 4.11. It
decreases monotonically with increasing j=j for j=j < 1; it decreases nonmonotonically
but oscillatorily for j=j > 1.
In summary, to accomplish efcient coupling between two waveguide modes, the following
three parameters have to be considered.
1. Coupling coefcient: The coupling coefcient has to exist and be sufciently large.
Problems
165
2. Phase matching: The phase mismatch has to be minimized so that j=j is made as small as
possible. Ideally, perfect phase matching with 0 is desired.
3. Interaction length: For codirectional coupling, because the efciency oscillates with
interaction length, the length has to be properly chosen. An overly large length is neither
required nor benecial. For contradirectional coupling, because the efciency monotonically
increases with the interaction length, the length has to be sufciently large but does not have
to be critically chosen. A very large length is not necessary, either.
Problems
4.1.1 Is the mode coupling caused by introducing an optical gain to a single waveguide
symmetric? Is the mode coupling caused by a slight structural change in the waveguide
symmetric?
4.1.2 Show that the general formulation for multiple-structure mode coupling is applicable to
the coupling of modes in a single waveguide.
4.2.1 Show that symmetric mode coupling in a single waveguide remains symmetric when a
lossless grating is introduced for phase matching.
4.2.2 Find the qth-order coupling coefcient q for a saw-tooth grating, as shown in
Fig. 4.3(f), that has a period of and a duty factor of such that
8
>
> 2z a,
for 0 < z < ;
<
z 1 2z
(4.82)
>
>
a, for < z < ;
:
1
with K 2=. Which orders are useful for mode coupling?

4.2.3 A single-mode GaAs/AlGaAs waveguide supports a mode that has a propagation constant of 2:5 107 m1 at 900 nm. To make a waveguide reector, the forwardpropagating wave in this mode has to be coupled to the backward-propagating wave of
the same mode. A grating is incorporated into the waveguide for phase matching. Ignore
any zeroth-order effect of the grating. Find the rst-order grating period and the secondorder grating period for this purpose.
4.2.4 A dual-channel directional coupler consists of two parallel InGaAsP/InP waveguides for
the two channels. A grating is fabricated in the space between the two channels to phase
match the waveguide modes of the two channels, as shown in Fig. 4.4. At 1:55 m,
the modes have effective indices of na 3:40 and nb 3:35, respectively. Ignore any
zeroth-order effect of the grating. Find the rst-order grating period and the second-order
grating period for phase matching the modes of the two channels in the same direction.
Find those values for phase matching the modes in the two channels for them to
propagate in opposite directions.
4.3.1 Find the length of a codirectional coupler that has a coupling efciency of half of the
maximum possible efciency for given coupling coefcients of ab and ba and phase
166
Optical Coupling
mismatch of between two modes in the case when the phase mismatch is small such that
2 < ab ba . What happens if the phase mismatch is large such that 2 > ab ba ?
4.3.2 Find the length of a codirectional coupler that has a coupling efciency of 25% of the
2 < ab ba . What happens if the phase mismatch is large such that 2 > ab ba ?
4.4.1 Find the length of a contradirectional coupler that has a coupling efciency of half of the
2 < ab ba .
4.4.2 Find the length of a contradirectional coupler that has a coupling efciency of half of the
mismatch of between two modes in the case when the phase mismatch is large such that
2 > ab ba .
4.5.1 Show that in the case of symmetric coupling with ab
ba , the powers of the two
codirectionally coupled modes given in (4.53) and (4.54) for the condition of Pa 0 6 0
and Pb 0 0 satisfy the power conservation relation Pz Pa z Pb z Pa 0
with Pab z 0.
4.5.2 Show that in the case of symmetric coupling with ab
ba , the powers of the two
contradirectionally coupled modes given in (4.64) and (4.65) for the condition of Pa 0
6 0 and Pb l 0 satisfy the power conservation relation Pz Pa z Pb z
Pa 0 Pb 0 with Pab z 0. Show also that Pa l Pb 0 Pa 0 for the total
power to be conserved.
4.6.1 Two optical waves of exactly the same wavelength and the same power are respectively
launched into the two input ports of a perfectly phase-matched 3-dB directional coupler at
the same time, as shown in Fig. 4.12. What are the possible power ratios between the two
output ports? What factor determines this ratio?
4.6.2 If the length of the coupler shown in Fig. 4.12 is doubled so that it becomes a coupler of
100% efciency, what are the possible power ratios between the two output ports? What
factor determines this ratio?
Figure 4.12 3-dB directional coupler.
Problems
167
4.6.3 A waveguide distributed Bragg reector (DBR) has a grating of square corrugation as
shown in Fig. 4.3(d). The period of the grating is , and its duty factor is . It is found that
the propagation constant of the fundamental TE0 mode of the waveguide at the
1:0 m optical wavelength is 1:0 107 m1 . It is also found that the maximum
absolute value of the coupling coefcient of this grating is jjmax 1:0 104 m1 ,
which is obtained when the parameters of the grating are properly chosen. Assume that
the waveguide structural parameters and the grating depth are xed. Only the period
and the duty factor of the grating are varied.
(a) What are the optimal choices of the period and the duty factor for the grating to
have the maximum coupling coefcient jjmax ? What is the length of the DBR if 50%
reectivity is desired?
(b) If a second-order grating has to be used, what are the best choices of its period and
its duty factor for the highest efciency? What is the length of the DBR if 50%
reectivity is desired in this case?
4.6.4 A waveguide Bragg reector is fabricated with a grating of a period in a symmetric
planar semiconductor waveguide, which has a core index of 3.25 and a cladding index of
3.20 for the wavelength of 1:55 m.
(a) Estimate the required grating period for a rst-order grating and that for a secondorder grating.
(b) Between the sinusoidal and the square gratings, choose a combination of shape and
duty factor for a rst-order grating that has a maximized coupling efciency for a
given modulation depth.
(c) If the grating chosen in (b) has a coupling coefcient of jj 1:0 104 m1 , what is
the required length of the grating for the Bragg reector to have a 90% reectivity?
4.6.5 A ber-optic frequency lter is made of two single-mode bers of different mode propagation constants. They are placed in close contact over a length of l, as shown in
Fig. 4.13. At the 1:55 m optical wavelength, the effective indices for the two ber
modes are a 5:959 106 m1 and b 5:849 106 m1 , respectively, and the
coupling coefcient between the two ber modes is ab
ba 2 103 m1 .
A grating that has a period of is built into the bers in the coupling section. The input
port of the device is port 1. The device is to function as an optical lter for separating the
1:55 m wavelength from other wavelengths.
(a) If the device is to direct all of the optical power at the 1:55 m wavelength to port
4 and to dump all other wavelengths to port 3, what is the maximum possible
coupling efciency for the 1:55 m wavelength without the grating?
(b) With a rst-order grating, what are the values of and l that have to be selected to
obtain the best efciency for directing the power at the 1:55 m wavelength to port 4?
What is the maximum efciency if the parameters of the grating are properly chosen?
168
Optical Coupling
(c) If the device is to direct the power at the 1:55 m wavelength to port 2, what is the
maximum possible coupling efciency without the grating?
(d) With a rst-order grating, what should the choice of the grating period be in order
to get the highest efciency for directing the power at the 1:55 m wavelength to
port 2? In this case, if the length l of the coupler remains the same as that found in (b),
what is the efciency of directing the 1:55 m light from port 1 to port 2?
Figure 4.13 Fiber-optic frequency lter
consisting of two single-mode bers and a
grating.
4.6.6 In designing an efcient waveguide coupler of any geometry, what are the three major
parameters that have to be considered in order to have a good efciency? In what order of
priority do they have to be considered?
Bibliography
Chuang, S. L., Physics of Photonic Devices, 2nd edn. New York: Wiley, 2009.
Marcuse, D., Theory of Dielectric Optical Waveguides, 2nd edn. Boston, MA: Academic Press, 1991.
Pollock, C. R. and Lipson, M., Integrated Photonics. Boston, MA: Kluwer, 2003.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
5 - Optical Interference pp. 169-203
5
5.1
Optical Interference
OPTICAL INTERFERENCE
..............................................................................................................
An optical eld is a sinusoidal wave that has a space- and time-varying phase. The complex
electric eld of an optical wave that propagates in a homogeneous medium can be generally
expressed in the form of (1.81):
Er; t E r; t exp ik r it ê jE r; t jeiE r;t exp ik r it ,
(5.1)
which has a total space- and time-dependent phase as given in (1.83):

r; t k r t E r; t :
(5.2)
For a waveguide mode that propagates along the longitudinal waveguide axis, taken to be the z
axis, the complex electric eld takes the form of (3.1):
E r; t E r; t exp i z it ê jE r; t jeiE z;t exp i z it ,
(5.3)
which has a total space- and time-dependent phase of

z; t z t E z; t:
(5.4)
The wave nature of an optical eld is fully characterized by its total space- and time-dependent
phase factor. Because z; t in (5.4) for a waveguide mode is mathematically a special form of
r; t in (5.2), by taking k to be ^z and E r; t to be E z; t , in the following discussion we
consider only optical waves in a homogeneous medium. The general concept applies equally to
waveguide modes. Unless otherwise specied, we also consider a lossless medium for simplicity so that the propagation constant k has a real value.
One phenomenon that clearly demonstrates the wave nature of optical elds is optical
interference of two or more elds of different phases. In this section, we consider the interference of two elds that are superimposed only once. In Section 5.2, the concept of an optical
grating based on the interference of multiple waves that emerge from a periodic optical
structure is discussed. Multiple interference leading to optical resonance and optical ltering
is discussed in Section 5.3.
Consider the superposition of two optical elds, E1 and E2 . The total eld is the linear vector
sum of the two:
E E1 E2 ê 1 jE 1 jei1 ê 2 jE 2 jei2 ,
(5.5)
where 1 k1 r 1 t E 1 and 2 k2 r 2 t E 2 are the total phases of the two elds

E1 and E2 , respectively. According to (3.183), the intensity of an optical eld is proportional to
170
jE j2 . Though (3.183) is strictly only applicable to a plane-wave normal mode that has a unique
wavevector of k and a unique frequency of while the composite eld E in (5.5) might not be a
normal mode because k1 and k2 might not be the same and 1 and 2 might not be the same, it
is clear that the intensity of the composite eld E is not simply the sum of the intensities of the
component elds E1 and E2 because
jEj2 jE1 j2 jE2 j2 E1 E

2 E1 E2

i1 2
jE 1 j2 jE 2 j2 2jE 1 jjE 2 jRe ê 1 ê
:
2e
(5.6)
The interference between the two elds E1 and E2 arises from the term

i1 2
2jE 1 jjE 2 jRe ê 1 ê
in (5.6). Clearly, interference does not exist between two orthog2e
onally polarized elds for which ê 1 ê
2 0. Note that the orthogonality between two optical
elds is dened by ê 1 ê
0,
as
given
in (1.80), but not by ê 1 ê 2 0. This is important for
2
circularly polarized or elliptically polarized elds, which have complex unit polarization
vectors. Interference occurs only when two elds are not orthogonally polarized so that
ê 1 ê
2 6 0.
Using the time-averaged
Poynting vector S dened in (1.53) and the denition of the light

intensity I S n^j while assuming that the angle between k1 and k2 is small, the intensity of
the total eld can be expressed as
I I 1 I 2 I 12 cos 1 2 ê 1 ê 2

I 1 I 2 I 12 cos k1 k2 r 1 2 t E 1 E 2 ê 1 ê 2 ,
(5.7)
where I 1 2k 1 jE 1 j2 =1 0 and I 2 2k 2 jE 2 j2 =2 0 are respectively the intensities of E1 and

E2 alone,

k1
k2

I 12 2
(5.8)
jE 1 E 2 jê 1 ê
2 0
1 0 2 0
is the intensity magnitude of the interference between the two elds, ê 1 ê 2 is the phase of
ê 1 ê
is the time average of cos 1 2 ê 1 e^2 over one
e
e1 ^
2 , and cos 1 2 ^
2
wave cycle, as dened in (1.53) for the time-averaged Poynting vector S. The phase factor
ê 1 ê 2 matters only when the two polarizations ê 1 and ê 2 are not mutually orthogonal and at
least one of them is not linearly polarized because ê 1 ê 2 0 when ê 1 ê

e 1 and
2 0 or both ^
ê 2 are real vectors. With this understanding, in the following we consider for simplicity only the
case when the two component elds have the same polarization, i.e., ê 1 ê 2 , so that
ê 1 ê
0. Then,
e
e1 ^
2 1 and ^
2
I I 1 I 2 I 12 cos 1 2

I 1 I 2 I 12 cos k1 k2 r 1 2 t E 1 E 2 ,
(5.9)
and

I 12

k1
k2
2
jE 1 E 2 j > 0:
1 0 2 0
(5.10)
171
As seen from (5.9), I 1 I 2 I 12 I I 1 I 2 I 12 . Depending on the total phase difference 1 2 , the total intensity I of the composite eld can be higher or lower than, or equal to,
the sum of the intensities I 1 and I 2 of the individual component elds. Because I 12 I 1 I 2 ,
maximum interference takes place when the two component elds have the same polarization
and the same amplitude so that I 12 I 1 I 2 .
1. Constructive interference occurs when the phase difference 1 2 is such that the total
intensity I is higher than the sum of the intensities I 1 and I 2 of the individual component
elds: I 1 I 2 < I I 1 I 2 I 12 . Complete constructive interference happens when the
two component elds are in phase, i.e., 1 2 2q, where q is an integer, so that
I I 1 I 2 I 12 . Partial constructive interference happens when the phase difference is
such that 2q =2 < 1 2 < 2q =2 but 1 2 6 2q so that I 1 I 2 < I < I 1
I 2 I 12 . These concepts of constructive interference are illustrated in Fig. 5.1 for the case
when the two component elds have the same frequency.
2. Destructive interference occurs when the phase difference 1 2 is such that the total
intensity I is lower than the sum of the intensities I 1 and I 2 of the individual component
elds: 0 I 1 I 2 I 12 I < I 1 I 2 . Complete destructive interference happens when
Figure 5.1 Constructive interference between two elds of the same frequency but of different amplitudes
showing the individual elds (dashed curves) and the composite eld (solid curve). The two component elds
have amplitudes of jE 1 j E 0 and jE 2 j 0:8E 0 in this example. (a) Complete constructive interference for
1 2 0. In this case, I 1 I 0 , I 2 0:64I 0 , and I 3:24I 0 > I 1 I 2 because the amplitude of the
composite eld is jE j 1:8E 0 . (b) Partial constructive interference for 1 2 =4 as an example. In this
case, I 1 I 0 , I 2 0:64I 0 , and I 2:77I 0 > I 1 I 2 because the amplitude of the composite eld is
jE j 1:665E 0 .
172
the two component elds are completely out of phase, i.e., 1 2 2q 1, and they
have the same amplitude to completely cancel each other so that I I 1 I 2 I 12 0.
Partial destructive interference happens when the two elds cancel each other only partially
but not completely so that I 6 0 but 0 < I < I 1 I 2 . Partial destructive interference occurs
under one of the two following different situations. The two elds are completely out of
phase, 1 2 2q 1, but they do not have the same amplitude, jE 1 j 6 jE 2 j, so
that I 12 < I 1 I 2 ; or the phase difference is such that 2q 1 =2 < 1 2 <
2q 1 =2 but 1 2 6 2q 1. These concepts of destructive interference are
illustrated in Fig. 5.2 for the case when the two component elds have the same frequency.
Figure 5.2 Destructive interference between two elds of the same frequency showing the elds and
intensities of the individual elds (dashed curves) and the composite eld (solid curve). (a) Complete
destructive interference for 1 2 and jE 1 j jE 2 j so that I 0. (b) Partial destructive interference
for 1 2 but jE 1 j 6 jE 2 j so that I 6 0 but 0 < I < I 1 I 2 . (c) Partial destructive interference
for 1 2 3=4 and jE 1 j jE 2 j so that I 6 0 but 0 < I < I 1 I 2 .
173
Interference between two optical elds can create intensity patterns that vary in space or time,
or both, because the phase difference 1 2 can be a function of space or time, or both. As
seen in (5.9), the phase difference 1 2 k1 k2 r 1 2 t E 1 E 2 has three
components.
1. When k1
6 k2 , the spatially varying phase factor k1 k2 r creates periodic spatial interference fringes that have a period of 2=jk1 k2 j along the k1 k2 direction. These
interference fringes disappear when k1 k2 : When 1 2 and E 1 E 2 is time independent, these interference fringes in space are stationary patterns that do not vary with time.
Figure 5.3 shows the stationary periodic fringes produced by the interference between two
elds of the same polarization, same amplitude, and same frequency, but different wavevectors.
2. When 1 6 2 , the temporally varying phase factor 1 2 t causes periodic temporal
beats that have a frequency of f j1 2 j=2. In the case when j1 2 j 1 and
j1 2 j 2 , these beats create a detectable temporal intensity variation at the frequency
f . This periodic temporal intensity variation disappears when 1 2 . When k1 k2 and
E 1 E 2 is space independent, these periodic beats in time are spatially uniform patterns
that do not vary in space. Figure 5.4 shows the periodic temporal beats produced by the
interference between two elds of the same polarization, same amplitude, and same wavevector, but different frequencies.
3. The phase factor E 1 E 2 depends on the phases of the two optical elds E 1 and E 2 . It denes
the coherence between the two elds. The two elds are temporally coherent with each other if
E 1 E 2 is a constant of time; they are spatially coherent if E 1 E 2 is a constant of space.
The two elds are temporally incoherent if E 1 E 2 varies randomly with time on the scale of
the optical cycle; they are spatially incoherent if E 1 E 2 varies randomly with space on the
scale of the optical wavelength. Between the extremes of complete coherence and complete
incoherence, the two elds can be partially coherent to different degrees in time, space, or both.
Figure 5.3 Stationary periodic fringes produced by the interference between two optical elds of the same
polarization, same amplitude, and same frequency, but different wavevectors.
174
Figure 5.4 Periodic temporal beats (sold curve) produced by the interference between two elds (dashed
curves) of the same polarization, same amplitude, and same wavevector, but different frequencies. The
envelope of the beat notes is shown in dashed gray curves.

The time average cos k1 k2 r 1 2 t E 1 E 2 depends strongly on the
degree of coherence. When the two elds are coherent, E 1 E 2 does not vary on the
time scale of the optical cycle or on the space scale of the optical wavelength, but it can still

vary in time or space slowly so that cos k1 k2 r 1 2 t E 1 E 2

cos k1 k2 r 1 2 t E 1 E 2 : The phase factor E 1 E 2 is a constant of both
space and time when the phases of the eld amplitudes E 1 and E 2 are constants or vary in the
same manner with space and time. It varies with space or time when the phases of the two eld
amplitudes vary differently with space or time; it varies with both space and time when the phases
of the eld amplitudes have different spatial variations and different temporal variations. Thus, a
modulation on the total intensity I in space or time, or both, can be accomplished by properly
modulating this phase factor. The principles of most interferometers are based on this concept.
EXAMPLE 5.1
A glass wedge of a refractive index n has a small wedge angle of as shown in Fig. 5.5. It has a
length of l in the x direction and a height of h in the y direction. A monochromatic plane optical
wave at the wavelength vertically illuminates the wedge from above. If the optical wave is
coherent, nd the locations of the bright and dark fringe lines when viewed from above. What is
the period of the fringes? How many periods of interference fringes appear on the top surface of
the wedge? What happens to the fringes if the light is not completely coherent?
Solution:
The incident wave propagates in the negative y direction with a wavevector of ki k^y . When
viewed from above, there are two reected waves, from the two surfaces of the glass wedge,
respectively. The rst is reected from the top wedge surface; it has a wavevector of
k1 k sin 2^x k cos 2^y at an angle of 2 from the y direction. The second is reected
from the bottom wedge surface; it has a wavevector of k2 k^y in the y direction. Thus,
k1 k2 k sin 2^x k cos 2 1^y :
175
Figure 5.5 Interference

fringes formed by reected
waves from the two
surfaces of a glass wedge.
Because the two reected waves are from the same source, they have the same frequency:
1 2 . However, the two reected waves have different phases because the top reection is
external reection at nearly normal incidence with a phase change of for the electric eld,
whereas the bottom reection is internal reection at normal incidence with no phase change.
If the incident optical wave is coherent, the phase of the two reected waves does not vary
with time so that E 1 E 2 . Then,

cos k1 k2 r 1 2 t E 1 E 2 cos 2kx sin cos 2kx sin :
Therefore,
I I 1 I 2 I 12 cos 2kx sin :
Bright fringe lines appear at the locations where cos 2kx sin 1 so that I I 1 I 2 I 12 ;
dark fringe lines appear where cos 2kx sin 1 so that I I 1 I 2 I 12 . We nd that a dark
fringe line appears at the tip of the wedge at x 0. Therefore, the dark and bright fringe lines
appear, respectively, at the locations:
l
m
m
,
m 0, 1, 2 . . .
k sin
2n sin
2nh

1
1 l
b
xm m
m
m
,
m 0, 1, 2 . . .
2 k sin
2 2n sin
2 2nh
xdm m
where we take sin h=l for a small angle of . The period of the fringes is found for
2k sin 2:

:
k sin 2n sin 2nh
The number of periods over the length is

M
l 2nl sin 2nh

:
176
If the incident optical wave is not coherent, then E 1 E 2 is not a constant of time. Because
the two reected waves are from the same source, whether they will create interference fringes
or not depends on the coherence time of the incident wave, i.e., the degree of coherence or
incoherence of the wave. The difference in the optical path lengths between the two reected
waves depends on the location of the fringe. It is y 2nh for the last fringe located at the end
of the wedge at x l, and it is
8
m,
for the mth dark fringe,
<

xm
ym y
1
: m , for the mth bright fringe:
l
2
The corresponding time difference of the two waves for the last fringe located at the end of the
wedge is
t
y 2nh M
,
c
c
and it is
8m
> ,
ym <

t m
1 1
>
c
: m
,
2
for the mth dark fringe,

for the mth bright fringe:
For the mth fringe to appear, the coherence time coh of the incident optical wave has to be such
that coh > tm , which means that coh is longer than m optical cycles for the mth dark fringe and
longer than m 1=2 cycles for the mth bright fringe. If the coherence time is sufciently long
such that coh > t, then all fringes on the surface of the wedge appear.
5.1.1 Double-Slit Interference

Youngs double-slit experiment established the wave nature of light. Figure 5.6 illustrates the
double-slit interference. We consider a monochromatic plane wave of a frequency and a
wavevector ki k^x , which is normally incident on two identical slits separated at a spacing of
in the z direction. The observation point is in the direction that makes an angle of with
respect to the x axis and is on a plane at a distance of l from the plane of the slits. The optical
path lengths from the two slits to the observation point are r 1 and r 2 , respectively. In the limit
that l , the path difference is
r 2 r 1 sin :
(5.11)
Because the incoming wave is normally incident on the plane of the slits, the elds that emerge
from the two slits have the same phase at the exit plane of the slits. Because the two slits have
the same geometrical dimensions, these elds have the same polarization and the same
amplitude such that E 1 E 2 ê E 0 . The total eld at the observation point is the linear
superposition of the elds from the two slits:
177
Figure 5.6 Double-slit interference.

E E 1 eikr1 it E 2 eikr2 it ê E 0 eikr1 it 1 ei ,
(5.12)
k r 2 r 1 k sin
(5.13)
where
is the phase difference at the observation point between the two elds that come from the two
slits. The intensity at the observation point is
I 4I 0 cos2
,
2
(5.14)
where I 0 / jE 0 j2 is the intensity contributed by a single slit alone. This result can be obtained
from (5.9) because I 1 I 2 I 0 , I 12 I 1 I 2 2I 0 , and 1 2 .
EXAMPLE 5.2
Find the angles at which the double-slit interference from normal incidence of a plane wave
shows bright interference fringes. Find the locations of the bright fringes on a screen that is at a
distance of l from the slits.
Solution:
The intensity pattern of the double-slit interference from normal incidence of a plane wave is
that given in (5.14). A bright interference fringe appears when
cos 2
1
2
2q for q 0, 1, 2, . . .
178
Using (5.13), the qth-order bright interference fringe appears at the angles q :
k sin q 2q
sin q
2q q
k
n
q sin1
q
,
n
where n is the refractive index of the medium. On a screen that is located at a distance of l from
the slits, the qth-order bright fringe is found at a distance of
zq l sin q q
l
n
from the zeroth-order bright fringe, which is located at z 0.
5.1.2 Optical Interferometers

Optical interference has been developed into many advanced concepts and applications. One
important application is interferometry, which uses optical interference to interrogate the
characteristics, including the polarization state, the wavevector, the frequency, and the phase,
of an optical wave with respect to a reference wave. Many types of interferometers have been
developed. The most important ones for photonics applications include the Michelson interferometer, MachZehnder interferometer, and FabryProt interferometer. The Michelson
interferometer and the MachZehnder interferometer are illustrated below. The FabryProt
interferometer is discussed in Section 5.3.
Michelson Interferometer
The Michelson interferometer was used in the historical MichelsonMorley experiment.
Figure 5.7 shows its basic structure. The single beam splitter in this structure denes four
optical paths. The two paths that are respectively on the left of and below the beam splitter
Figure 5.7 Michelson
interferometer. BS, beam splitter.
179
dene two ports, each of which serves as a port for both input and output. Input light can be sent
into either port or into both ports, but usually only one input is supplied, as shown in Fig. 5.7
where only port 1 receives an input of an intensity I in while port 2 receives no input. By
contrast, both ports always function as output ports with output intensities of I out, 1 and I out, 2 ,
respectively, though the output intensity at a port can be zero when totally destructive interference occurs at the port.
The input wave is split by the beam splitter into two waves, each of which enters one of the
two internal paths that are respectively above and on the right of the beam splitter. Each internal
path ends with a totally reective mirror, which reects the light back to the beam splitter. The
beam splitter again divides each returning wave into one reected wave and one transmitted
wave for the two output ports. Each output eld is the combination of one reected eld from
one internal path and one transmitted eld from the other internal path: The output eld at port
1 is the linear superposition of the reected eld from the vertical internal path and the
transmitted eld from the horizontal internal path, whereas the output eld at port 2 is the
linear superposition of the transmitted eld from the vertical internal path and the reected eld
from the horizontal internal path.
Though the two component elds of each output eld come from different internal paths, they
have the same polarization, the same frequency, and the same wavevector because they both
originate from the same input eld and they propagate in the same direction. Their phase
difference depends only on the optical length difference of the two internal paths and the phase
change caused by reection or transmission at the beam splitter. Because the phase change at
the beam splitter has a xed value, the output intensity at a port can be varied by varying the
optical length difference of the two internal paths. Note that what matters is not the physical
length difference of the paths but the optical length difference. The optical length difference can
be varied by varying the physical length difference, through moving one or both mirrors, or by
varying the refractive index along one or both paths, through modulating the medium using any
of the effects discussed in Sections 2.6 and 2.7.
The beam splitter is partially reective and partially transmissive. In practice, it has negligible
absorption so that R T 1. The beam splitter can have any reectance/transmittance ratio,
but complete destructive interference is possible only when it is a 50/50 beam splitter so that the
reected eld and the transmitted eld have the same magnitude though possibly different
phases. Conservation of energy requires that I out, 1 I out, 2 I in when there is no loss in the
system. Clearly, I out, 1 0 and I out, 2 I in when complete destructive interference occurs
at port 1, whereas I out, 2 0 and I out, 1 I in when complete destructive interference occurs at
port 2. Thus, complete constructive interference occurs at one output port when complete
destructive interference occurs at the other output port. This condition is clearly required by
conservation of energy, but it is not trivial if we take a closer look. It implies that the two
component elds for the total output eld at port 1 are completely in phase when those at
port 2 are completely out of phase. This seems puzzling: each output eld is the combination of
one reected eld and one transmitted eld through the beam splitter, but one combination is
constructive while the other is destructive at the same time.
To resolve this puzzle, we have to pay attention to two key properties of the functioning of an
optical beam splitter. (1) An optical beam splitter always has a layer of properly designed and
180
accurately implemented coating on one of its two surfaces to accomplish the desired reectance/transmittance ratio. The other surface is often antireection coated to eliminate unwanted
reection. In any event, reection takes place on only one surface of the beam splitter. Because
the two waves returning from the two different internal paths reach the beam splitter from
different sides, one undergoes external reection while the other undergoes internal reection.
(2) For any polarization, a transmitted eld through a lossless dielectric interface has no phase
change with respect to the incident eld. A reected eld may have either no phase change or a
phase change of , depending on its polarization, its incident angle, and whether it undergoes
external reection or internal reection; in any case, the phase difference between external
reection and internal reection for a given polarization at a given incident angle is always .
(See Problem 3.4.1.) Considering the above two characteristics, it is clear that the phase
difference between the two eld components at one output port is always different by a phase
factor of from that at the other output port because the reected eld component for one
output port comes from external reection and that for the other output port is from internal
reection. For this reason, constructive interference happens at one output port when destructive interference takes place at the other output port, ensuring conservation of energy.
Assume that the beam splitter has the reective surface on the left side. Then, reection on
the left side of the beam splitter is external reection with a phase change of and reection on
the right side of the beam splitter is internal reection with no phase change. If the beam splitter
is a 50/50 splitter, the output intensities of the two output ports are
I out, 1 I in cos2
,
2
I out, 2 I in sin2
,
2
(5.15)
where is the phase difference of the two optical paths. In the case when the two paths are
lled with the same uniform medium, 2kla lb , where la and lb are respectively the
lengths of the two arms, and the factor 2 accounts for the fact that the wave in each arm travels
through the arm twice before returning to the beam splitter.
MachZehnder Interferometer
Figure 5.8 shows the basic structure of the MachZehnder interferometer. With two beam
splitters, this structure is different from that of the Michelson interferometer in two basic
features: The output ports are separate from the input ports, and light propagates through each
of the two separate internal paths only once. Despite these differences, the fundamental concepts
discussed above for the Michelson interferometer are applicable to the MachZehnder interferometer. The output intensity at a given port can be varied by varying the difference of the optical
path lengths between the two paths, which can be accomplished by varying the physical length
difference between the two paths or by varying the refractive index in the medium along one or
both paths. When constructive interference occurs at one output port, destructive interference
happens at the other output port. Thus, I out, 1 I out, 2 I in for a lossless system.
Assume that each beam splitter has the reective surface on the left side. Then, reection on
the left side of each beam splitter is external reection with a phase change of and reection
on the right side of each beam splitter is internal reection with no phase change. If both beam
splitters are 50/50 splitters, the output intensities of the two output ports are
181
Figure 5.8 MachZehnder interferometer. BS, beam splitter.
Figure 5.9 MachZehnder interferometers in the waveguide form using (a) two Y-junction waveguides
and (b) two directional couplers. Only one input is supplied in this illustration. In general, the lengths of the
two arms are not identical.
I out, 1 I in sin2
,
2
I out, 2 I in cos2
,
2
(5.16)
where is the phase difference of the two optical paths. In the case when the two paths are
lled with the same uniform medium, k la lb , where la and lb are respectively the
lengths of the two arms, and the wave in each arm travels through the arm only once before
reaching the output beam splitter.
The MachZehnder interferometer can be implemented in various waveguide forms.
Figure 5.9 shows two common forms using (a) Y-junctions and (b) directional couplers for
182
the beam-splitting function. In the case when the Y-junctions and the directional couplers are all
3-dB couplers such that 1=2, we nd that
T cos2
(5.17)
for the interferometer using 3-dB Y-junctions shown in Fig. 5.9(a), and
T sin2
(5.18)
for the interferometer using 3-dB directional couplers shown in Fig. 5.9(b), where a b
is the phase difference of the two optical paths.
5.1.3 Standing Wave

In the analysis and discussion presented above following (5.7), we have assumed that the angle
between the two wavevectors k1 and k2 of the interfering waves is small, or zero as in the case
of the interferometers. In the case when the angle between k1 and k2 is large, the principle of
linear superposition expressed as (5.5) is still valid and interference between two elds still
occurs, but the intensity of the combined eld expressed as (5.7) is not valid. Here we consider
the special case when two waves have the same polarization, ê 2 ê 1 ê , the same amplitude,
E 2 E 1 E, and the same frequency, 2 1 , but they propagate in opposite directions, k2 k1 k, so that E1 ê 1 E 1 eik1 ri1 t ê Eeik rit and E2 ê 2 E 2 eik2 ri2 t
ê Eeik rit . The linear superposition of these two elds yields
Er; t E1 r; t E2 r; t ê Eeik rit ê Eeik rit 2ê Eeit cos k r:
(5.19)
For simplicity of discussion without loss of generality, we assume linear polarization and a
eld amplitude of E jEj by taking its phase to be zero. Then the real eld of the combined
eld can be expressed as
Er; t Er; t E
r; t 4ê jEj cos t cos k r:
(5.20)
The spatial variation of this eld is decoupled from the temporal variation. We nd that
Er; t vanishes for all times at the xed locations, known as nodes, where k r 2q 1=2
for integers q so that cos k r 0, as shown in Fig. 5.10. The nodes are periodically
distributed along the line dened by k^ at a spacing of =k =2n, where =n is the
wavelength of the optical eld in the medium of a refractive index n. At the locations where
k r 2q so that cos k r 1, we nd that Er; t 4ê jE j cos t ; such locations are
known as antinodes. The antinodes are also periodically distributed along the line dened
by k^ at a spacing of =k =2n. An antinode is found at the midpoint between two
neighboring nodes.
Because the nodes and antinodes are xed in space, the eld given in (5.20) appears to stand
still in space. It does not travel but only oscillates in time. Therefore, the interference of the two
contrapropagating waves of the same polarization, the same frequency, and the same amplitude
results in a standing wave.
183
Figure 5.10 Standing wave. Nodes, labeled with N, are periodically distributed along the line dened by k^ at a
spacing of =k =2n. Each antinode, labeled with A, is located at the midpoint between two neighboring
nodes. A standing wave oscillates in time but appears to stand still in space.
5.2
OPTICAL GRATINGS
..............................................................................................................
An optical grating is a periodic optical structure. Some waveguide grating structures are
illustrated in Fig. 4.3. The functioning of an optical grating can be understood from the
viewpoint of phase matching, as discussed in Chapter 4, or from the viewpoint of optical
interference. In this section, we make the connection between these two viewpoints.
The concept of double-slit interference discussed in the preceding section can be extended to
equally spaced multiple slits of identical geometrical parameters, which form a periodic
structure of a period in the z direction, as shown in Fig. 5.11. The slits are on the yz plane,
which is normal to the x axis. Being a periodic optical structure, this multiple-slit structure can
be considered a grating. Indeed, it functions as a transmissive diffraction grating, also called a
transmission grating.
5.2.1 Normal Incidence

We rst consider normal incidence of a monochromatic plane wave of a frequency and a
wavevector ki k^x on the periodic multiple-slit structure, as shown in Fig. 5.11. Because the
incoming plane wave is normally incident on the plane of the slits, the elds that emerge
from all of the slits have the same phase at the exit plane of the slits, which is perpendicular
to ki . They also have the same polarization and the same amplitude because the slits have the
same geometrical dimensions. Therefore, on the exit plane of the slits, E 1 E 2
E N ê E 0 :
As seen in Fig. 5.11, at a distant point in the direction at an angle of with respect to the x
axis, the phases of the rays coming from different slits increase between successive slits by the
amount of k sin given in (5.13). Following the same reasoning for the double slits, the
total eld at the distant point in this direction is the linear superposition of the elds coming
from all slits to the point:
E E 1 eikr1 it E 2 eikr2 it E N eikrN it

ê E 0 eikr1 it 1 ei eiN1
ê E 0 eikr1 it
iN
1e
:
1 ei
(5.21)
184
Figure 5.11 Normal incidence of a monochromatic plane wave on a periodic multiple-slit structure.
The intensity at the distant observation point is

I I0
sin2 N=2
,
sin2 =2
(5.22)
where I 0 / jE 0 j2 is the intensity contributed by a single slit alone. Using the mathematical
relations
lim
x!0
sin2 Nx
N 2 and
sin2 x
sin2 N x q sin2 Nx
for q 0, 1, 2, ,
sin2 x q
sin2 x
(5.23)
we nd that the intensity I has maxima of the value N 2 I 0 when

k sin q 2q,
(5.24)
where q is an integer that represents the order of diffraction.

Figure 5.12 shows the intensity distribution given in (5.22) for the multiple-slit structure as
a function of the phase factor k sin , which varies with the angle for xed values of k
and . The primary maxima that have the peak intensity of N 2 I 0 appear at the angles that satisfy
the condition given in (5.24). Secondary maxima of lower peak intensities exist between
primary maxima. As the number N of the periods in the structure increases, the peak intensity
of each primary maximum increases quadratically as N 2 while the width decreases linearly as
N 1 ; meanwhile, the peak intensities of all secondary maxima decrease.
The periodic multiple-slit structure functions as a transmissive diffraction grating that has a
wavenumber of
K
(5.25)
185
Figure 5.12 Intensity distribution as a function of the phase factor k sin for a multiple-slit structure
functioning as a transmission grating. As the number N of periods increases, the primary maxima representing
the diffraction orders have peak intensities increasing as N 2 and widths decreasing as N 1 while the peak
intensities of all secondary maxima decrease.
in the z direction. Each primary maximum in the spatial intensity distribution of the transmitted
light represents a diffraction order. The qth-order diffracted beam has a wavevector of
kq k cos q ^x k sin q ^z :
(5.26)
Using the relations in (5.24) and (5.25), we nd that

k sin q qK:
(5.27)
Because the wavevector of the incident wave is ki k^x , there is a phase mismatch of

(5.28)
kq kq ki k cos q 1 ^x k sin q ^z k cos q 1 ^x qK^z
between the qth-order diffracted beam and the incident wave.
A phase mismatch between two waves is a momentum difference between two photons of the
two waves. Clearly from (5.28), except for the zeroth order, an incident photon acquires
momentum changes in both x and z directions in the process to exit as a diffracted photon.
Because of conservation of momentum, any momentum change of a photon has to be compensated by an opposite momentum change of another physical object. The momentum change

kq, x k cos q 1 in the x direction is easily compensated by an opposite momentum
186
change of the entire multiple-slit structure in the x direction because it is in the direction normal
to the plane of the structure and it is negligibly small for the mass of the structure. In the z
direction, however, no such momentum compensation is possible if the slits are absent from the
structure because no force in the direction parallel to the plane of the structure can be exerted on
the structure. In the presence of the periodic slits, the periodicity along the z direction provides
the necessary compensation for the momentum change of kq, z k sin q in the z direction
when the phase-matching condition k sin q qK of (5.27) is satised. Therefore, constructive
interference for a diffracted beam is equivalent to phase matching for the beam.
5.2.2 Oblique Incidence

In the above, we considered a monochromatic plane wave that is normally incident on the
multiple-slit structure at an incident angle of i 0 so that ki k^x . The equivalent concepts of
constructive interference and phase matching for the diffraction orders can also be applied to
oblique incidence at a nonzero incident angle of i 6 0 so that ki k i, x ^x k i, z^z k cos i ^x
k sin i^z 6 k^x , as shown in Fig. 5.13. With an incident wave of this wavevector, the eld
emerging from the slits has a phase shift of k i, z k sin i from one slit to the next in
the z direction so that E 1 ê E 0 , E 2 E 1 eiki, z ê E 0 eik sin i , . . . , E N E 1 eiN1ki, z
ê E 0 eiN1k sin i . Applying these relations to (5.21), we nd that the phase factor k sin
for normal incidence at i 0 is generalized to
k sin k sin i
(5.29)
for oblique incidence at i 6 0. Therefore, the condition given in (5.24) for nding the maxima
of the diffracted intensity distribution is generalized to the condition:
k sin q k sin i 2q,
where q is an integer.
Figure 5.13 Oblique incidence of a monochromatic plane wave on a periodic multiple-slit structure.
(5.30)
187
From the phase-matching point of view, the condition in (5.30) can be easily obtained from
the condition for phase matching assisted by the grating of a wavenumber K in the z direction:
kz k q, z k i, z qK,
(5.31)
which is identical to (5.30) in the form of

k sin q k sin i qK:
(5.32)
As discussed above for the case of normal incidence, it is also true for oblique incidence that
phase matching in the x direction normal to the plane of the grating structure does not set a
required condition because it is automatically satised by a compensating momentum change of
the massive structure. Note that the zeroth order takes place at 0 i .
EXAMPLE 5.3
A monochromatic plane wave at the 651 nm wavelength is normally incident on the plane
of an array of equally spaced slits. The 20th-order diffraction peak is found at the angle of
20 10 . Find the spacing between neighboring slits. If a plane wave at 488 nm is
normally incident on the slits, what is the diffraction angle of the 20th-order diffraction peak? If
it is obliquely incident for the 20th-order diffraction peak to appear at 20 10 , what is the
required incident angle?
Solution:
For normal incidence with 651 nm and 20 10 , (5.27) requires that
k sin q qK
sin q q
q
20 651 109
m 75 m:
sin 10
sin q
For 488 nm at normal incidence, the 20th-order diffraction peak appears at

k sin q qK
q sin1
20 sin1
20 488 109
7:48 :
6
75 10
For oblique incidence, the incident angle is found using (5.32) as

q
1
sin q
k sin q k sin i qK ) i sin
:
For the 20th-order diffraction peak of 488 nm to appear at 20 10 , we nd

20 488 109
1

2:49 :
sin 10
i sin
75 106
5.2.3 Grating at an Interface

When an optical wave is incident on a grating at an interface between two different optical
media, as shown in Fig. 5.14, diffraction orders in reection and in transmission can both
188
Figure 5.14 (a) Optical grating at an interface. (b) Phase-matching conditions for the reective and
transmissive diffraction orders.
appear. Assuming that the incident wave comes from medium 1, which has a refractive index of
n1 , at an incident angle of i with respect to the normal of the interface, the diffraction orders on
both sides of the interface are determined by the phase-matching conditions:
k1 sin 1q k1 sin i qK
(5.33)
for the reective diffraction orders in medium 1, and

k2 sin 2q k2 sin i qK
(5.34)
for the transmissive diffraction orders in medium 2. Note that for the zeroth order, 10 i in
reection and n2 sin 20 n1 sin i in transmission, which are just those required by Snells law
for a at surface when the grating does not exist. Here we only consider the phase-matching
conditions that determine the direction of each diffraction order; whether a diffraction order
appears or not also depends on the shape and the geometrical parameters of the grating, as
discussed in Example 4.2.
EXAMPLE 5.4
A grating that has a period of 2 m is fabricated on the surface of a glass plate, which has a
refractive index of 1:5. It is exposed to air. A laser beam at the wavelength of 850 nm is
normally incident on the grating from the air side. How many diffraction orders are possible on
each side? What is the diffraction angle of each order?
Solution:
For normal incidence, i 0 . Thus, the phase-matching conditions in (5.33) and (5.34) reduce
to k1 sin 1q qK and k2 sin 2q qK, which can be expressed as
sin 1q
q
q
and sin 2q
:
n1
n2
189
Every diffraction angle is required to be within the range between 90 and 90 , i.e., 1
sin 1q 1 and 1 sin 2q 1.
On the air side, n1 1; thus
1 sin 1q
q
1
n1
0 jqj
n1 1 2 106
2:35:
850 109
There are ve diffraction orders on the air side for q 2,1, 0, 1, 2. The diffraction angles
with respect to the surface normal are
1q sin1
q
q 850 109
sin1
n1
1 2 106
1q 58:21 , 25:15 , 0 , 25:15 , 58:21 :
On the glass side, n2 1:5; thus

1 sin 2q
q
1
n2
0 jqj
n2 1:5 2 106
3:52:
850 109
There are seven diffraction orders on the glass side for q 3, 2, 1, 0, 1, 2, 3. The
diffraction angles with respect to the surface normal are
2q sin1
)
q
q 850 109
sin1
n2
1:5 2 106
1q 58:21 , 34:52 , 16:46 , 0 , 16:46 , 34:52 , 58:21 :
5.2.4 Surface GratingWaveguide Coupling

A grating fabricated on the surface of a waveguide can couple a radiation eld that propagates
in the homogeneous space on one side of the waveguide into a waveguide mode. In reverse
operation, it can also couple a waveguide mode into a radiation eld from the surface of the
waveguide. These concepts are illustrated in Fig. 5.15.
For this purpose, it is necessary to phase match the radiation eld with the waveguide mode
in the longitudinal direction of the waveguide, which is taken to be the z direction. For coupling
Figure 5.15 Surface
grating for (a) input
coupling and (b) output
coupling of a
waveguide mode.
190
with a waveguide mode that has a propagation constant of , the incident optical wave has to
satisfy the phase-matching condition:
k2 sin 2q qK
(5.35)
if the wave is incident from the substrate side of a refractive index n2 at an incident angle of
2q , or
k3 sin 3q qK
(5.36)
if the wave is incident from the cover side of a refractive index n3 at an incident angle of 3q .
The same phase-matching conditions are used to determine the directions of output coupling.
Note that the phase-matching conditions given in (5.35) and (5.36) only determine the
directions of the radiation elds that can be coupled into or out from a waveguide mode, but
they do not tell us the efciency of the coupling. The coupling efciency is determined by the
coupling coefcient, which depends on the shape, the depth, and other geometrical parameters
of the grating, as discussed in Example 4.2.
EXAMPLE 5.5
A sinusoidal grating that can only serve as a rst-order grating is fabricated on the surface of a
GaAs slab waveguide as shown in Fig. 5.15. The cover of the waveguide is simply air so that
n3 1. At the wavelength of 1:3 m, the propagation constant of the TE0 mode of this
waveguide is 1:62 107 nm, corresponding to an effective index of n 3:35. If it is
desired that a laser beam at this wavelength be coupled into this guided mode through the
surface grating at an incident angle of i 45 , what is the required period of the grating?
Solution:
Because a sinusoidal grating can be used only as a rst-order grating, it is necessary that the
phase-matching condition is satised for q 1 or q 1. Because the wave is incident from
the cover side, the condition is that from (5.36) with q 1:
k 3 sin 31 K
n3
1 n
sin 31
n n3 sin 31
With 1:3 m, n 3:35, n3 1, and 31 i 45 , the required grating period is
n n3 sin 31
1:3 106
m 492 nm:
3:35 1 sin 45
5.2.5 Flat Interface

The phase-matching concept can be applied to reection and refraction at a at, smooth
interface between two media of different indices n1 and n2 to obtain Snells law discussed in
Section 3.4. For a smooth surface that is not modied by any periodic structure, we can take the
191
limit of an innitely large period, ! , thus a zero wavenumber of K 0. Then, by applying

(5.31) with K 0 to the reected and transmitted waves, we obtain the following phasematching condition:
k r, z ki, z k t, z ki, z 0,
(5.37)
which yields the condition of ki sin i k r sin r k t sin t given in (3.88) and Snells law
expressed in (3.89) and (3.90).
5.3
FABRYPROT INTERFEROMETER
..............................................................................................................
The basic principle of the FabryProt interferometer is the interference of multiple reections from
two partially reective parallel surfaces. The desired reectivity for each of these two surfaces can
be obtained by proper coating. The basic structure of the FabryProt interferometer takes two
different forms. The rst form shown in Fig. 5.16(a) consists of two partially reective mirrors on
the parallel inner surfaces of two dielectric plates; the outer surfaces of the plates are antireection
coated and often wedged to prevent unwanted reection from these surfaces. In the second form
shown in Fig. 5.16(b), the two partially reective surfaces are the parallel surfaces of a transparent
dielectric plate; a FabryProt interferometer of this form is usually called a FabryProt etalon.
The two structures shown in Figs. 5.16(a) and (b) have the same interferometric characteristics despite the differences in their detailed structures. For both structures, we consider a
physical spacing of l that is lled with a medium of a refractive index n between the two
partially reective surfaces, as shown in Fig. 5.16. The direction normal to the reective
surfaces is taken to be the z direction. We consider for generality oblique incidence of a
Figure 5.16 (a) FabryProt interferometer. The outer surfaces of the wedged plates are antireection coated.
(b) FabryProt etalon.
192
monochromatic plane wave of a frequency and a wavelength . The wavevector of the wave
that is transmitted through the rst partially reective surface makes an angle of with respect
to the normal of the reective surface; this angle is not necessarily the same as the incident
angle of the wave coming from outside because the refractive index of the outside medium is
not necessary the same as that inside the interferometer. The eld-amplitude reection coefcients r 1 and r2 of the left and right mirrors, respectively, can be expressed as
1=2
1=2
r1 R1 ei1 ,
r 2 R2 ei2 ,
(5.38)
where R1 and R2 are the intensity reectivities of the left and right reective surfaces,
respectively, and 1 and 2 are the phase changes of the optical elds upon reection on these
surfaces. As discussed in Section 3.4, the reection coefcients r 1 and r 2 are functions of the
incident angle and the polarization of the optical eld.
Multiple partial reections inside the interferometer take place at the two partially reective
surfaces, as seen in Fig. 5.16. Between the two reective surfaces, all forward-propagating
waves have the same wavevector at an angle of with respect to the z direction so that
k z k cos , and all backward-propagating waves have the same wavevector at an angle of
with respect to the z direction so that k z k cos k cos . Each forward or backward pass through the spacing of a length l causes a phase shift of kl cos . Each time a wave
reaches a reective surface, part of it is transmitted and the rest of it is reected; multiple
reections by the reective surfaces produce multiple transmitted waves. At a given location
on the outside of the interferometer, each successive transmitted eld is related to the
preceding transmitted eld by a factor of
1=2 1=2
1=2 1=2
r 1 r 2 ei2kl cos R1 R2 ei2kl cos 1 2 R1 R2 eiRT ,
(5.39)
where
RT 2kl cos 1 2 4
nl
nl
cos 1 2 4 cos 1 2
c
(5.40)
is the total phase shift caused by a round-trip passage between the two reective surfaces. This
phase shift includes the phase shift of 2kl cos from the double passes through the medium in
the spacing and the localized phase shifts of 1 and 2 from reections at the two reective
surfaces.
The interferometer has two output ports: one in the forward direction for the total transmitted
eld and the other in the backward direction for the total reected eld. The total transmitted
eld through the interferometer at the forward output port is the linear sum of all transmitted
elds through the second reective surface:
Etout E 0 eit E 1 eit E 2 eit

2
1=2 1=2 iRT
1=2 1=2 iRT
it
E0e
1 R1 R2 e
R1 R2 e

E 0 eit
1
1=2 1=2
1 R1 R2 eiRT
(5.41)
193
where E 0 is the transmitted eld that directly passes through the two reective surfaces, E 1 is
the transmitted eld after one reection by each reective surface, and E 2 is the transmitted
eld after two reections by each reective surface, and
2
. so forth.

1=2 1=2 iRT
t
From (5.41), the total transmitted intensity is I out I 0 1 R1 R2 e , where the intensity
I 0 of the directly transmitted eld E 0 is related to the input intensity as I 0 1 R1 1 R2 I in .
Therefore, the transmittance of a lossless FabryProt interferometer for the forward output
port is
T FP
I tout
1 R1 1 R2
1 R1 1 R2
:

2
2

I in
1=2 1=2
1=2 1=2
1=2 1=2 i
2
RT
R
4R
R
sin
=2
1

R
R
e
1

R

RT
1
2
1
2
1
2
(5.42)
The reectance of the FabryProt interferometer for the backward output port is
RFP
I rout
1 T FP :
I in
(5.43)
The maximum transmittance of the FabryProt interferometer is

1 R1 1 R2
T max
2 :
FP
1=2 1=2
1 R1 R2
(5.44)
The maximum transmittance is T max

FP 1 for a lossless symmetric FabryProt interferometer
max
that has R1 R2 , but T FP < 1 for an asymmetric FabryProt interferometer that has R1 6 R2 .
We can dene a normalized transmittance as
T FP
T^ FP max
T FP
where
1
2
4F
1 2 sin2 RT =2
(5.45)
1=4 1=4
R1 R2
1=2 1=2
1 R1 R2
(5.46)
is the nesse of the lossless FabryProt interferometer. As expressed in (5.46) and plotted in
Fig. 5.17, the nesse of a lossless FabryProt interferometer is a nonlinear function of the
product, R1 R2 , of the reectivities of the two reective surfaces that form the interferometer.
The normalized transmittance T^ FP of a lossless FabryProt interferometer expressed in
(5.45) is plotted in Fig. 5.18 as a function of the round-trip phase shift RT for a few values
of the nesse of the interferometer. The strong dependence of T^ FP on RT is the consequence of
the interference of the multiple reections between the two reective surfaces. The transmittance peaks appear at

2
c
RT 2q, q q 1
,
(5.47)
2
2nl cos
where q is an integer so that all transmitted elds resulting from multiple reections in the
interferometer are in phase for constructive interference. The separation between two
194
Figure 5.17 Finesse, F, of a lossless FabryProt interferometer as a function of the product, R1 R2 , of the
reectivities of the two reective surfaces of the interferometer.
Figure. 5.18 Normalized transmittance T^ FP of a lossless FabryProt interferometer as a function of the roundtrip phase shift RT for a few values of the nesse of the interferometer.
neighboring peaks in the spectrum is called the free spectral range, which has a round-trip
phase difference of FSR and a frequency difference of FSR :
c
FSR 2, FSR
:
(5.48)
2nl cos
Away from the peaks, the transmittance is low because the transmitted elds are out of phase,
resulting in destructive interference. Each transmittance peak has a nite FWHM linewidth,
line , measured in terms of the shift in the round-trip phase, or line , measured in terms of the
optical frequency. Actually, the nesse is dened as the ratio of the free spectral range to the
linewidth:
F
FSR FSR
:
line
line
(5.49)
195
The relation given in (5.46) for a lossless FabryProt interferometer is a valid approximation
for F 1: Therefore, the linewidth decreases with increasing nesse, which in turn increases
nonlinearly with the value of R1 R2 .
As seen in (5.40), the round-trip phase RT is a function of the wavelength of the optical
wave, the physical spacing l of the interferometer, the refractive index n of the medium between
the two reective surfaces, and the angle at which the wave propagates inside the interferometer and is incident on the reective surfaces. The transmittance of a FabryProt interferometer can be varied by varying any of these physical parameters. The strong dependence of
the transmittance on the optical wavelength, thus on the optical frequency, allows a high-nesse
FabryProt interferometer to be used as an optical spectrum analyzer. A high nesse leads to a
narrow linewidth for the transmittance peaks, thus a high resolution for the optical spectrum
analyzer. Further detailed characteristics of the FabryProt interferometer used as an optical
resonator are discussed in Chapter 6.
EXAMPLE 5.6
What happens to the maximum transmittance T max
FP , the nesse F, the frequencies q at which
the peak transmittance occurs, the free-spectral range FSR , and the spectral linewidth line of
a FabryProt interferometer in each of the following situations? (a) The reectivity R1 or R2 is
increased, or both are increased. (b) The spacing l is increased. (c) The index n of the medium
between the reective surfaces is increased. (d) The angle at which the wave propagates
between the reective surfaces is increased.
Solution:
The transmittance of a FabryProt interferometer is a direct function of only three parameters,
R1 , R2 , and RT , as seen in (5.42); however, RT is a function of the parameters l, n, , and the
optical frequency . Each of the other characteristics of the FabryProt interferometer depends
on some of these parameters but is independent of the other parameters.
(a) The reectivity R1 or R2 is increased, or both are increased. From (5.44), we nd that T max
FP
does not monotonically vary with R1 or R2 . Indeed, we nd
max
max
dT FP
dT FP
signR2 R1 and sign
signR1 R2 :
sign
dR1
dR2
Therefore, T max
FP increases with increasing R1 if R1 < R2 , but it decreases with R1 if
max
R1 R2 , including when R1 R2 because T max
FP reaches its largest value of T FP 1 when
R1 R2 . Similarly, T max
FP increases with increasing R2 if R1 > R2 , but it decreases with R2 if
R1 R2 , including when R1 R2 . From (5.46), we nd that the nesse F monotonically
increases with the product R1 R2 ; therefore, it increases when R1 R2 is increased through
increasing either R1 or R2 , or both. From (5.47) and (5.48), we nd that both q and FSR
do not vary with R1 or R2 . From (5.49), we nd that line decreases when the product R1 R2
is increased because line FSR =F.
(b) The spacing l is increased. From (5.44) and (5.46), we nd that both T max
FP and F are
independent of the spacing l; they do not change as l is increased. From (5.47) and (5.48),
196
we nd that both q and FSR decrease when the spacing l is increased. From (5.49), we
nd that line decreases with increasing l because line FSR =F.
(c) The index n of the medium between the reective surfaces is increased. From (5.40), (5.47),
and (5.48), we nd that the index n and the spacing l always appear together in the form of
their product nl. Indeed what counts is the optical path length nl, rather than the physical
length. Therefore, increasing the index n has exactly the same consequences as increasing
the spacing l discussed in (b).
(d) The angle at which the wave propagates between the reective surfaces is increased.
From (5.40), (5.47), and (5.48), we nd that actually the angle always appears together
with the index n and the spacing l in the form of nl cos . Increasing reduces the effective
optical path length nl cos . Therefore, increasing is equivalent to reducing the spacing l
or the refractive index n: Both T max
FP and F do not change with ; both q and FSR increase
with increasing ; line increases with increasing .
5.3.1 Optical Thin Films

Optical thin lms are thin layers of optical materials that have thicknesses on the order of the
optical wavelength. An optical thin lm can be either a free-standing layer in a homogeneous
medium, such as the lm of a soap bubble in air, or a layer deposited on a substrate of a
different optical property, such as a thin SiO2 layer on a silicon substrate. A sophisticated thinlm structure can be composed of multiple thin layers of different optical properties. Figure 5.19
shows some examples of optical thin lms.
Figure 5.19 Examples of optical thin lms.
197
A single optical thin lm has the structure, thus the basic optical property, of a FabryProt
interferometer in the etalon form. The two surfaces of the thin lm act as the two partially
reective surfaces of the interferometer. Multiple reections take place in the thin lm between
these two surfaces. Therefore, the reectance and transmittance of an optical thin lm are
functions of the optical wavelength, the incident angle, the thickness and refractive index of the
thin lm, and the refractive indices of the media on the two sides of the thin lm. An optical
thin lm often exhibits a color because of the strong wavelength dependence of its reectance
and transmittance. A thin lm that has a spatially varying thickness can produce a spectrum of
spatially distributed colors, as often seen in soap bubbles or oil slicks.
EXAMPLE 5.7
An oil lm of a uniform thickness l 100 nm oats on water. The refractive index of the oil
lm is noil 1:40 and that of water is nw 1:33. When it is illuminated by white light at
normal incidence, which wavelength in the visible spectral range shows the highest reection?
What color does it appear to be? If the same lm is coated on a glass surface of a refractive
index ng 1:50, does it show the same high reection?
Solution:
For the oil lm on water, we nd that noil > nw > nair . Therefore, for the wave inside the oil
lm as an interferometer, the reection at the airlm interface and that at the lmwater
interface are both internal reection with no phase changes so that 1 2 0. Then,
according to (5.47), for normal incidence the peak transmittance for dark reection occurs at

2
c
c
c
2noil l
) dark
q q 1
q
,
2
2nl
2noil l
q
q
and the minimum transmittance for bright reection occurs at

1 1 2 c
1
c
c
4noil l
q1=2 q
) bright
q
:
2
2
2nl
2 2noil l
q1=2 2q 1
With noil 1:40 and l 100 nm, we nd that the only bright that falls within the 400 to 700 nm
visible spectral range is found for q 1 at
bright
4noil l
4noil l 4 1:4 100 nm 560 nm:
2q 1
The next bright reection takes place for q 2 at 186:7 nm, which is in the deep UV.
Therefore, the lm appears to be green.
If the same lm is coated on a glass surface of a refractive index ng 1:50, then
ng > noil > nair . In this situation, the reection at the airlm interface is still internal reection
with 1 0, but that at the lmglass interface is external reection with 1 . Then,
according to (5.47), for normal incidence the peak transmittance for dark reection occurs at

1 2
c
1
c
c
4noil l
q q
) dark
q
,
2
2nl
2 2noil l
q 2q 1
198
and the minimum transmittance for bright reection occurs at

1 1 2 c
c
c
2noil l
q1=2 q
) bright
q 1
:
2
2
2nl
2noil l
q1=2 q 1
With noil 1:40 and l 100 nm, we nd that no bright falls within the 400 to 700 nm visible
spectral range because the largest value for bright is found for q 2 at 280 nm, which is in the
UV. Therefore, this lm appears to be colorless on glass.
A thin lm on an optical surface can dramatically change the reection and transmission
properties of the surface. Thin-lm coating is an important technology for designing and
achieving desired reection and transmission properties of an optical surface, and thin-lm
optics has been developed into an important eld in optics. Sophisticated thin lms consisting
of multiple layers of different thicknesses and different refractive indices are used for advanced
optical coatings. A desired reection property, such as broadband antireection, broadband
total reection, narrowband antireection, or narrowband high reection, can be obtained by
coating an optical surface with a properly designed thin-lm structure. Applications of thin-lm
optical coatings range from high-precision coatings for optical lters and laser mirrors to lowemission glass panes for house windows.
EXAMPLE 5.8
A uniform thin lm of MgF2 , which has a refractive index of nf 1:38 is deposited on the
surface of a glass lens, which has a refractive index of ng 1:50, to serve as an antireective
coating at the wavelength of 552 nm. What is the minimum thickness of the thin lm?
What other thicknesses can be chosen? How effective is this thin lm as an antireective
coating? How can the thin-lm material be chosen to further increase the effectiveness of the
antireective coating?
Solution:
There are two interfaces: the airMgF2 interface and the MgF2glass interface. Because the
refractive index increases from one medium to the next with nair 1, nf 1:38, and ng 1:50,
for the wave inside the thin lm as an interferometer, the reection at the airlm interface is
internal reection with no phase change and that at the lmglass interface is external reection
with a phase change of ; thus 1 0 and 2 . For the lm to serve as an antireective
coating, it is desired that T FP T max
FP , which takes place at the optical frequencies q given
in (5.47):

1 2
c
1 c
q q
q
2
2nl cos
2 2nf l
for normal incidence. With the given wavelength at c= 552 nm, the acceptable thicknesses are

lq
199

1
c
1
1
552
1
q
q
q
nm 200 q
nm:
2 2nf
2 2nf
2 2 1:38
2
Therefore, the minimum thickness is lmin 100 nm for q 1, and any thickness that is larger
than the minimum thickness by an integral multiple of 200 nm, such that l 1002m 1 nm,
also works.
Without the coating, the reectivity at the airglass interface is

nair ng 2 1 1:52

0:04:
R

nair ng
1 1:5
With the thin-lm coating, the reectivities at the two interfaces are

nair nf 2 1 1:382
nf ng 2 1:38 1:52
3

R1
1 1:38 0:0255, R2 nf ng 1:38 1:5 1:736 10 :
nair nf
The reectivity of the coated surface is
1 R1 1 R2
RFP 1 T max
2 1 0:986 0:014:
FP 1
1=2 1=2
1 R1 R2
Therefore, the thin-lm coating cuts the reectivity by 65% from 0:04 to 0:014.
To increase the effectiveness of the antireective coating, the material of the thin lm has to
be chosen so that R1 and R2 have closer values. The coating results in total antireection with
RFP 0 when R1 R2 so that T max
FP 1. This can be accomplished by choosing the refractive
p
index of the thin lm
to be nf nair ng . For this thin lm to be totally antireective, a material
p
of an index nf 1 1:5 1:225 has to be chosen for the lm.
5.3.2 Interference Filters

A high-nesse FabryProt interferometer can be used as an interference lter to selectively
transmit a desired wavelength. The wavelength selectivity of the lter is determined by its free
spectral range; a larger free spectral range allows fewer transmission wavelengths within a given
spectrum. For a desired transmission wavelength , the largest spectral range for an interference
lter is FSR c=, which is achieved when the optical path length of the interferometer is
half the optical wavelength: nl =2. For such a lter, the next transmission peak occurs at the
second harmonic frequency, 2, of the desired transmission frequency , i.e., at the wavelength =2
that is half the desired transmission wavelength , if the dispersion of the refractive index n is
negligible between and 2. The pass band around the transmission frequency is determined by
the linewidth of the interferometer. As discussed above, for a given free spectral range the
linewidth can be reduced by increasing the nesse through increasing the product R1 R2 of the
reectivities of the reective surfaces. By properly coating the two reective surfaces for high
reectivities, an interference lter of a narrow linewidth on the order of a nanometer or an
angstrom can be obtained. Such a highly selective, narrow-linewidth lter is also called a line lter.
200
Problems
5.1.1 Show that in the case when the angles between the wavevectors k1 and k2 of two optical
elds is small, the intensity of the combined optical eld projected on a plane that is
normal to k1 k2 is approximately that given in (5.7).
5.1.2 A glass wedge of a refractive index n 1:5 as shown in Fig. 5.5 has a length of l 5 cm
and a height of h 1 mm. It is vertically illuminated with coherent light at the
600 nm wavelength. What is the period of the interference fringes? How many dark and
bright interference fringes appear on the surface of the wedge?
5.1.3 If the incident light in Problem 5.1.2 is not completely coherent, what is the minimum
coherence time of the wave for all of the interference fringes to appear on the wedge? If
1000 periods of interference fringes appear, what is the coherence time of the incident light?
5.1.4 An air wedge is formed between two at glass plates by making them in contact at one
end but separated by the thickness of a piece of paper at the other end. When it is
vertically illuminated with monochromatic coherent light at the 500 nm wavelength,
exactly 400 periods of interference fringes are seen. What is the thickness of the paper?
5.1.5 A laser beam at the 532 nm wavelength is normally incident on two slits that are
spaced at 200 m. What is the angle between the two bright interference fringes of
the diffraction orders q 10? On a screen that is at a distance of l 2 m from the slits,
what is the separation of these two fringes?
5.1.6 Two slits separated by 100 m are illuminated with a laser beam at normal incidence. On a screen that is at a distance of l 2:5 m from the slits, it is found that the
separation between two neighboring dark fringes is 12:2 mm, what is the wavelength of
the laser light?
5.1.7 A laser beam is sent into a Michelson interferometer that is constructed in free space, as
shown in Fig. 5.7.
(a) When the mirror of one arm is moved to increase the length of the arm by 0:5 mm
while the other arm is xed, the intensity pattern at each output port repeats itself
1880 times. Find the wavelength of the laser beam.
(b) The two arms are adjusted such that I out, 1 I in and I out, 2 0. Then, a thin glass
plate that has a refractive index of n 1:46 and a thickness of d 1 mm is inserted
perpendicularly to the beam path into one of the two arms without changing the
optical alignment. What are the output intensities I out, 1 and I out, 2 now?
5.1.8 A laser beam is sent into a MachZehnder interferometer that is constructed in free space,
as shown in Fig. 5.8.
(a) When the mirror of one arm is moved to increase the length of the arm by 0:5 mm
while the other arm is xed, the intensity pattern at each output port repeats itself 940
times. Find the wavelength of the laser beam.
(b) The two arms are adjusted such that I out, 1 I in and I out, 2 0. Then, a thin glass
plate that has a refractive index of n 1:46 and a thickness of d 1 mm is inserted
Problems
201
perpendicularly to the beam path into one of the two arms without changing the
optical alignment. What are the output intensities I out, 1 and I out, 2 now?
5.1.9 A waveguide MachZehnder interferometer uses Y-junction couplers for its input and
output ports, as shown in Fig. 5.9(a). It has a symmetric structure with an equal length of
la lb l for the two arms. The two Y-junctions are both 3-dB couplers. Thus, 0,
and the transmittance is T 1. By changing the refractive index of the medium in one
arm with respect to the other through the Pockels effect, for example, the phase shifts
through the two arms can be made different for 6 0 so that T 6 1. Find the minimum
necessary index difference n between the two arms for T 0 at an optical wavelength
of . At 1 m, what is the minimum value of n for an equal arm length of
l 1 mm? If the MachZehnder interferometer has a symmetric structure with la lb
l using two 3-dB directional couplers, as shown in Fig. 5.9(b), the transmittance is T 0
with 0. Then, what is the minimum necessary index difference n between the two
arms for T 1 at an optical wavelength of ? At 1 m, what is the minimum value of
n for an equal arm length of l 1 mm?
5.2.1 Identical slits in an array are equally spaced at 20 m. A plane wave at the
532 nm wavelength is normally incident on the slits. How many diffraction peaks can be
found in transmission within the range of angles between 30 and 30 ? If the wave is
obliquely incident at an angle of i 15 , how many diffraction peaks can be found in
transmission within the range of angles between 30 and 30 ?
5.2.2 Three perfectly aligned plane optical waves at 1 450 nm, 2 550 nm, and 3 650 nm
are normally incident at the same time on an array of identical slits that are equally spaced at
. The diffraction peaks in transmission are examined. It is clear that the zeroth-order peaks
for all three wavelengths completely overlap at q 0 for q1 q2 q3 0.
(a) What are the lowest nonzero diffraction orders q1 and q2 for 1 and 2 , respectively,
that have exactly overlapped peaks? What is the minimum slit spacing for this to be
possible?
(b) Answer the questions in (a) for 2 and 3 .
(c) Answer the questions in (a) for 1 and 3 .
(d) What are the nonzero diffraction orders q1 , q2 , q3 for 1 , 2 , 3 , respectively, that
have exactly overlapped peaks? What is the smallest slit spacing for this to be
possible?
5.2.3 A grating on the surface of a glass plate has a period of 800 nm. The glass plate has a
refractive index of 1:5. A laser beam is normally incident on the grating from the air.
Only two nonzero diffraction orders, for q 1 and q 1, are allowed on the glass side,
but no nonzero diffraction orders are allowed on the air side. What is the possible
wavelength of the incident laser light?
5.2.4 A collimated laser beam at 800 nm is incident on a grating at an airglass interface
from the air side. The refractive index of this glass is 1.5. At normal incidence, three
diffraction peaks for q 1, 0, and 1 are found on the glass side. By carefully varying
the incident angle of the laser beam, it is found that the q 1 diffraction peak just
disappears when the incident angle is i 12:1 . Find the grating period. How many
202
diffraction peaks can be found at an incident angle of i 10 from the air and glass
sides, respectively? At what angles are these diffraction peaks found?
5.2.5 Consider the waveguide and the grating of a period 492 nm found in Example 5.5.
The waveguide supports the TE0 mode at the 1:55 m wavelength. The effective
index of this mode at this wavelength is n 3:33. Find the incident angle for a laser
beam at 1:55 m to be coupled into this guided mode.
5.2.6 A surface grating that has a period of 300 nm is fabricated on the surface of a GaAs/
AlGaAs slab waveguide as shown in Fig. 5.15. The cover of the waveguide is simply air
with n3 1. At the wavelength of 900 nm, the GaAs core has n1 3:59 and the
AlGaAs substrate has n2 3:39. The waveguide supports only the TE0 mode of an
unknown propagation constant. If it is found that a laser beam at 900 nm can be
coupled into this guided mode through the surface grating at an incident angle of
i 30 , what is the propagation constant of the mode? What grating period will allow
coupling of this laser beam into this waveguide mode at normal incidence with i 0 ?
5.3.1 A laser beam is sent at normal incidence into a FabryProt interferometer that is constructed in free space with R1 R2 0:5.
(a) When one reective surface is xed in location but the other is moved to increase the
spacing between them by 0:5 mm, the transmitted intensity pattern repeats itself 1880
times. Find the wavelength of the laser beam.
(b) The interferometer is adjusted such that T FP 1. Then, a thin glass plate that has a
refractive index of n 1:46 and a thickness of d 1 mm is inserted perpendicularly
to the beam path into the spacing without changing the optical alignment. What is the
transmittance of the interferometer now?
5.3.2 A lossless FabryProt interferometer consists of two highly reective surfaces with
R1 95% and R2 90%, which are separated by a spacing of l in free space. What are
the maximum transmittance and the nesse of this interferometer? It is used as an optical
spectrum analyzer. If a spectral resolution with a linewidth of line 0:1 nm at the
500 nm wavelength is desired, what is the required spacing l of the interferometer? What
is the wavelength separation FSR between neighboring transmission peaks? If a higher
resolution is needed, how should the spacing be changed in order to reduce the spectral
linewidth by half to line 0:05 nm?
5.3.3 A FabryProt etalon consists of a thin glass plate that has a refractive index of n 1:50
and a thickness of l 100 m. Its surfaces are coated such that its peak transmittance is
100% and it has a spectral linewidth of line 5 GHz for high spectral resolution. Find
the values of R1 and R2 that allow the etalon to have these properties.
5.3.4 An oil lm that has a refractive index of noil 1:40 oats on a smooth water surface,
which has nw 1:33. It reects most strongly at the 672 nm red wavelength and appears
to have no reection at the 504 nm blue wavelength. What is the thickness of the oil lm?
5.3.5 A material that has a refractive index of nf 1:25 is used for the thin lm discussed in
Example 5.8, which is deposited on the surface of a glass lens that has a refractive index
of ng 1:50. To serve as an antireective coating at the wavelength of 552 nm, what
Bibliography
203
is the minimum thickness required for the thin lm? What other thicknesses can be
chosen? How effective is this thin lm as an antireective coating?
5.3.6 The refractive index of Si at the 1:0 m wavelength is nSi 3:61. If an antireective
thin lm is to be coated on a smoothly polished Si surface, how should the refractive
index of the thin-lm material be chosen so that the coated surface is totally antireective
when exposed to air? What should the refractive index of the thin lm be chosen if the
surface is to become totally antireective in water, which has a refractive index of
nw 1:33?
Bibliography
Serway, R. A. and Jewett, J. W., Physics for Scientists and Engineers, 9th edn. Boston, MA: Brooks Cole,
2013.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
6 - Optical Resonance pp. 204-223
6
6.1
Optical Resonance
OPTICAL RESONATOR
..............................................................................................................
As discussed in Section 5.3, multiple reections take place between the two reective surfaces
of a FabryProt interferometer, resulting in multiple transmitted elds. A transmittance peak
occurs when the round-trip phase shift RT between the two reective surfaces is an integral
multiple of 2 so that all of the transmitted elds are in phase. From the viewpoint of the eld
inside the interferometer, this condition results in optical resonance between the two reective
surfaces. Thus a FabryProt interferometer behaves as an optical resonator, also called a
resonant optical cavity. At resonance, the eld amplitude inside an optical resonator reaches a
peak value due to constructive interference of multiple reections. The optical energy stored in
an optical cavity peaks at its resonance frequencies.
An optical cavity can take a variety of forms. Figure 6.1 shows the schematic structures of a
few different forms of optical cavities. Though an optical cavity has a clearly dened longitudinal axis, the axis can lie on a straight line, as in Fig. 6.1(a), or it can be dened by a folded
path, as in Figs. 6.1(b), (c), and (d). A linear cavity dened by two end mirrors, as in Fig. 6.1(a),
is known as a FabryProt cavity because it takes the form of the FabryProt interferometer.
A folded cavity can simply be a folded FabryProt cavity that supports a standing intracavity
eld, as in Fig. 6.1(b). A folded cavity can also be a non-FabryProt ring cavity that supports
two independent, contrapropagating intracavity elds, as in Figs. 6.1(c) and (d).
An optical cavity provides optical feedback to the optical eld in the cavity. Optical
resonance occurs when the optical feedback is in phase with the intracavity optical eld. The
optical feedback in a FabryProt cavity is provided simply by the two end mirrors that have
the reective surfaces perpendicular to the longitudinal axis, as in Figs. 6.1(a) and (b). In a ring
cavity, it is provided by the circulation of the laser eld along a ring path dened by mirrors, as
in Fig. 6.1(c), or a ring path dened by an optical ber, as in Fig. 6.1(d). The cavity can also be
constructed with an optical waveguide, as in the case of a semiconductor laser or a ber laser. In
the following discussion, we take the coordinate dened by the longitudinal axis to be the z
coordinate, and the transverse coordinates that are perpendicular to the longitudinal axis to be
the x and y coordinates. In a folded cavity, the z axis is thus also folded along with the
longitudinal optical path. Sophisticated optical cavities can use gratings to provide distributed
feedback; such advanced cavities are not shown in Fig. 6.1 and are not discussed in this chapter.
In a ring cavity, an intracavity eld completes one round trip by circulating inside the
cavity in only one direction. The two contrapropagating elds that circulate in a ring cavity
in opposite directions are independent of each other even when they have the same frequency.
In a FabryProt cavity, an intracavity eld has to travel the length of the cavity twice in
6.1 Optical Resonator
205
Figure 6.1 Schematics of a few different forms of optical cavities: (a) linear FabryProt cavity with end mirrors;
(b) folded FabryProt cavity with end mirrors; (c) three-mirror ring cavity with two independent, contrapropagating
elds; and (d) ring cavity with two independent, contrapropagating elds guided by an optical-ber waveguide.
opposite directions to complete a round trip. The time it takes for an intracavity eld to
complete one round trip in the cavity is called the round-trip time,
T
round-trip optical path length lRT

,
c
c
(6.1)
206
Optical Resonance
Figure 6.2 Passive laser cavities with a gain lling factor under optical injection: (a) a FabryPerot cavity and (b)
a ring cavity. The refractive index of the gain medium is n, while that of the background medium in the cavity is n0 .
A laser cavity is simply a passive optical cavity when its gain medium is absent or is present but not pumped.
where the round-trip optical path length lRT takes into account the refractive index of the
medium inside the cavity.
The space inside an optical cavity can be lled with a variety of optical media of different
properties. For example, a laser cavity contains at least a gain medium. The gain medium may ll
up the entire length of the cavity, or it may occupy a fraction of the cavity length. For a laser cavity
of a length l that contains a gain medium of a length lg , as shown in Fig. 6.2, we can dene an
overlap factor between the gain medium and the intensity distribution of the laser mode as the ratio
jEj2 dxdydz
V gain
lg
gain
:

(6.2)

V mode
l
2
jEj dxdydz
cavity
This ratio is commonly known as the gain lling factor for a gain medium that takes up only a
fraction of the length of the laser cavity, whereas it is related to the mode connement factor in
a waveguide laser, such as a ber laser or a semiconductor laser. When the gain medium lls up
an optical cavity and covers the entire intracavity eld distribution, 1; otherwise, < 1.
Take the refractive index of the gain medium to be n and that of the intracavity medium
excluding the gain medium to be n0 ; then, the round-trip optical path length can be expressed as

2nl 1 n0 l 2nl, for a linear cavity;
(6.3)
lRT
for a ring cavity;
nl 1 n0 l nl,
where n n 1 n0 is the weighted average index of refraction throughout the laser
cavity. When an optical cavity contains optical elements other than a gain medium, n is still the
weighted average index throughout the cavity with n0 being the weighted average index of the
background medium and these optical elements.
207
Consider an intracavity eld, Ec z, at any location z along the longitudinal axis inside an
optical cavity. When this eld completes a round trip in the cavity and returns back to the
location z, it is amplied or attenuated by a factor a to become aEc z. The complex
amplication or attenuation factor a can be generally expressed as
a GeiRT ,
(6.4)
where G is the round-trip gain factor for the eld amplitude, equivalent to the power gain in a
single pass through a linear FabryProt cavity, and RT is the round-trip phase shift for the
intracavity eld. Both G and RT have real values, and G 0. For a cavity that has a net optical
gain, G > 1, and the intracavity eld is amplied. For a cavity that has a net optical loss, G < 1,
and the intracavity eld is attenuated.
EXAMPLE 6.1
Consider a linear cavity, as shown in Fig. 6.1(a), and a ring cavity, as shown in Fig. 6.1(c). The
linear cavity has two mirrors with R1 R2 0:9, which are separated at l 1:5 m. The ring
cavity has three mirrors with R1 R2 0:9 and R3 1, which are separated at l12 0:7 m
and l23 l31 0:4 m. Find the physical length, the round-trip length lRT , the round-trip time T,
and the round-trip gain factor G of each cavity.
Solution:
For the linear cavity, the physical length is simply l 1:5 m dened by the separation of the
two mirrors. The round-trip length and the round-trip time are, respectively,
llinear
2l 3 m, T linear RT 10 ns:
c
In a round trip through the linear cavity, the intracavity intensity changes by a factor of R1 R2
because the intracavity light is reected once by each of the two mirrors in each round trip.
Therefore, the round-trip gain factor for the eld amplitude is
p
Glinear R1 R2 0:9:
llinear
RT
For the ring cavity, the physical length is simply l l12 l23 l31 1:5 m dened by the ring
length. The round-trip length and the round-trip time are, respectively,
lring
l l12 l23 l31 1:5 m, T ring RT 5 ns:
c
In a round trip through the ring cavity, the intracavity intensity changes by a factor of R1 R2 R3
because the intracavity light is reected once by each of the three mirrors in each round trip.
Therefore, the round-trip gain factor for the eld amplitude is
p
Gring R1 R2 R3 0:9:
lring
RT
6.2
LONGITUDINAL MODES
..............................................................................................................
We rst consider the resonant characteristics of a passive optical cavity. A passive cavity
cannot generate or amplify an optical eld; thus G < 1. In order to maintain an intracavity eld
208
Optical Resonance
in such a cavity, it is necessary to constantly inject an input optical eld, Ein , into the cavity. As
shown in Fig. 6.2, the forward-traveling component of the intracavity eld at the location z1 just
inside the cavity next to the injection point is the sum of the transmitted input eld and the
fraction of the intracavity eld that returns after one round trip through the cavity:
Ec z1 t in Ein aEc z1 ,
(6.5)
where t in is the complex transmission coefcient for the input eld. We nd that
t in
(6.6)
Ec z1
Ein :
1a
The transmitted output eld, Eout , is proportional to the intracavity eld: Eout / Ec z1 . Therefore, the output intensity is proportional to the input intensity through the following relationship,
I out /
I in
j1 aj
I in
2
1 G 4G sin2 RT =2
(6.7)
The proportionality constant of this relationship depends on the transmittance of the output mirror
and the intracavity attenuation over the distance from the point at z1 to the output point. The
transmittance of the cavity is T c I out =I in , which is scaled by the value of this proportionality
constant. For our discussion in the following, this proportionality constant is irrelevant. Therefore,
we only have to consider the normalized transmittance of the passive cavity:
T^ c
1
1
h
i
h
i
,
2
2
2
1 4G=1 G sin RT =2 1 4=G=1 1=G sin2 RT =2
(6.8)
which is obtained by normalizing T c to its peak value.

Clearly, T^ c has a peak value of unity, as expected for a normalized quantity. In Fig. 6.3, T^ c is
plotted as a function of the round-trip phase shift RT for a few different values of G. We nd that
Figure 6.3 Normalized transmittance of an optical cavity as a function of the round-trip phase shift in the
cavity. In a resonator that has a xed, frequency-independent optical path length, the round-trip phase shift
is directly proportional to the optical frequency. The longitudinal mode frequencies are dened by the
frequencies corresponding to the resonance peaks. The spectral shape for a gain factor of G is the same as that
for a gain factor of 1=G. Thus, the curve for G 0:1 is the same as that for G 10, that for G 0:5 is the
same as that for G 2, and so on.
209
the spectral shape for a gain factor of G is the same as that for a gain factor of 1=G. Therefore, a
passive cavity that has a gain factor of Gp G < 1 has the same spectral characteristics as an
active cavity that has a gain factor of Ga 1=G > 1. Note that the characteristics of T^ c shown in
Fig. 6.3 are the same as those of T^ FP shown in Fig. 5.18 because a FabryProt interferometer
can be considered as an optical cavity. Clearly, T^ FP given in (5.45) for a FabryProt interferometer can be identied with T^ c in (6.8) for a general optical cavity by properly relating the
nesse F of a cavity to the gain factor G, as is given below in (6.12).
At a given input eld intensity, the intracavity eld intensity of a passive cavity is proportional to T^ c because the transmitted output eld intensity is directly proportional to the
intracavity eld intensity while it is also proportional to T^ c . Therefore, resonances of the cavity
occur at the peaks of T^ c , where the intracavity intensity reaches its maximum level with respect
to a constant input eld intensity. As can be seen from Fig. 6.3, the resonance condition of the
cavity is that the round-trip phase shift is an integral multiple of 2:
RT 2q,
q 1, 2, . . . :
(6.9)
From (6.9) and Fig. 6.3, we nd that the separation between two neighboring resonance peaks
of T^ c is
L 2
(6.10)
and that the FWHM of each resonance peak is

1G
:
G1=2
The nesse, F, of the cavity is the ratio of the separation to the FWHM of the peaks:
c 2
(6.11)
L G1=2
:
(6.12)
c 1 G
In the simplest situation that the optical eld is a plane wave at a frequency of , the roundtrip phase shift can be generally expressed as
F
RT
lRT local ,
c
(6.13)
where the rst term on the right-hand side is the phase shift contributed by the propagation of
the optical eld over an optical path length of lRT , and the second term, local , is the sum of all
the localized, and usually xed, phase shifts such as those caused by reection from the mirrors
of a cavity. In the case when the frequency of the input eld is xed, the resonance condition
given in (6.9) can be satised by varying the optical path length lRT of the cavity, either by
varying the physical length of the cavity or by varying the refractive index of the intracavity
medium, or both. The optical cavity then functions as an optical interferometer, which is used to
accurately measure the frequency and the spectral width of an optical wave.
When both the optical path length and the localized phase shifts are xed, as is typically the
case for a laser resonator, the resonance condition of RT 2q is satised only if the optical
frequency satises the condition:
q
c
lRT
2q local ,
(6.14)
210
Optical Resonance
or
q
c
lRT
q
local
:
2
(6.15)
These discrete resonance frequencies are the longitudinal mode frequencies of the optical
resonator because they are dened by the resonance condition of the round-trip phase shift
along the longitudinal axis of the cavity. The frequency spacing, L , between two neighboring
longitudinal modes is known as the free spectral range, also called the longitudinal mode
frequency spacing, of the optical resonator. The FWHM of a longitudinal mode spectral peak is
c , which is known as the longitudinal mode width of the cavity. If the values of lRT and local
are independent of frequency, then L / L and c / c . Therefore, the nesse of an
optical resonator is the ratio of its free spectral range to its longitudinal mode width:
F
L L
:
c c
(6.16)
From (6.15), we nd that the longitudinal mode frequency spacing is related to the round-trip
time as
L q1 q
c
1
:
lRT T
(6.17)
The longitudinal mode width of the cavity can be expressed as

c
L 1 G
L :
F
G1=2
(6.18)
EXAMPLE 6.2
Find the nesse F, the longitudinal mode frequency spacing L , and the longitudinal mode
width c of the linear and ring cavities that are considered in Example 6.1.
Solution:
For the linear cavity, the nesse is
1=2
F linear
Glinear
0:91=2
29:8:
1 Glinear
1 0:9
The longitudinal mode frequency spacing is
linear
L
1
T linear
1
Hz 100 MHz:
10 109
The longitudinal mode width is

linear
linear
100
L
MHz 3:36 MHz:

F linear
29:8
211
For the ring cavity, the nesse is

1=2
F ring
Gring
1 Gring
0:91=2
29:8:
1 0:9

ring
L
1
T ring
1
Hz 200 MHz:
5 109

ring
6.3
ring
200
L
MHz 6:71 MHz:

F ring
29:8
TRANSVERSE MODES
..............................................................................................................
Any realistic optical cavity has a nite transverse cross-sectional area. Therefore, the resonant
optical eld inside a realistic optical cavity cannot be a plane wave. Indeed, there exist certain
normal modes for the transverse eld distribution in a given optical cavity. Such transverse eld
patterns are known as the transverse modes of a cavity. A transverse mode of an optical cavity
is a stable transverse eld pattern that reproduces itself after each round-trip pass in the cavity,
except that it might be amplied or attenuated in magnitude and shifted in phase.
The transverse modes of an optical cavity are dened by the transverse boundary conditions
that are imposed by the transverse cross-sectional index prole of the cavity. For a cavity that
utilizes an optical waveguide for lateral connement of the optical eld, the transverse modes
are the waveguide modes, such as the TE and TM modes of a slab waveguide or the TE, TM,
HE, and EH modes of a cylindrical ber waveguide. For a nonwaveguiding cavity, the
transverse modes are TEM elds determined by the shapes and sizes of the end mirrors of
the cavity, as well as by the properties of the medium and any other optical components inside
the cavity. The Gaussian modes discussed in Section 3.3 are an important set of such unguided
TEM modes.
In an optical cavity that supports multiple transverse modes, the round-trip phase shift is
generally a function of the transverse mode indices m and n. Therefore, the resonance condition
can be explicitly written as
RT
mn 2q:
(6.19)
As a result, the resonance frequencies of the cavity, mnq or mnq , are dependent on both
longitudinal and transverse mode indices. When the frequency spacing between neighboring
transverse modes is smaller than that between neighboring longitudinal modes, multiple
resonance frequencies of different transverse modes can exist for each longitudinal mode, as
illustrated schematically in Fig. 6.4.
212
Optical Resonance
Figure 6.4 Cavity resonance frequencies associated with different longitudinal and transverse modes. For
clarity, the heights of the transverse modes are made arbitrarily decreasing.
In a cavity that consists of an optical waveguide, the propagation constant mn is a

function of the waveguide mode. If the physical length of the waveguide cavity is l, the
effective round-trip optical path length of a waveguide mode is
8 c
>
< 2 mn l, for a linear cavity;
RT
(6.20)
lmn c
>
: mn l,
for a ring cavity:
The round-trip optical path length lRT

mn generally varies from one mode to another due to the
modal dispersion of the waveguide. In addition, the localized phase shift can also be mode
dependent. Therefore, instead of the resonance frequencies q given by (6.14) for a plane wave,
the resonance frequencies mnq of a waveguide cavity are found by solving, for integral values
of q, the following resonance condition,
RT
mn
RT
l local
mn 2q:
c mn
(6.21)
In a nonwaveguiding cavity, the propagation constant, k, is a property of only the medium and
is not mode dependent. Nevertheless, a mode-dependent on-axis phase variation mn z does
exist, which is given in (3.76) for a HermiteGaussian mode as discussed in Section 3.3. The
total on-axis phase variation of the TEMmn Gaussian mode is mn z kz mn z, which
includes the mode-independent phase shift kz and the mode-dependent phase shift mn z.
Consequently, the cavity resonance condition for a Gaussian mode is a modication of that for
a plane wave made by adding the round-trip contribution of the mode-dependent phase shift:
RT
mn
local
lRT RT
mn mn 2q,
c
(6.22)
where the localized phase shift can, in general, also be mode dependent.
It is clear from the above discussion that the qth longitudinal mode frequency of a given
longitudinal mode index q varies among different transverse modes, as illustrated in Fig. 6.4.
For transverse modes dened by a waveguide structure, the longitudinal mode frequency
spacing Lmn mnq1 mnq between two neighboring longitudinal modes, q and q 1,
of the same transverse mode mn varies slightly among different transverse modes, as illustrated
in Example 6.3. Because a higher-order transverse waveguide mode has a smaller propagation
constant, thus a smaller effective index of refraction, Lmn is generally larger for a higher-order
213
transverse mode. By comparison, the longitudinal mode frequency spacing Lmn stays constant
for different transverse Gaussian modes dened in free space because all Gaussian modes are
TEM modes of the same propagation constant. The mode-dependent phase shift mn z only
changes the mode frequency mnq but not the difference Lmn between two neighboring
longitudinal modes mnq and mnq 1.
EXAMPLE 6.3
A GaAs/AlGaAs semiconductor optical cavity has the longitudinal structure of a linear Fabry
Prot cavity and the transverse structure of a slab waveguide. The cavity has a physical length
of l 500 m. The GaAs/AlGaAs slab waveguide supports three TE modes at the 870 nm
wavelength, with propagation constants of TE0 2:61 107 m1 , TE1 2:58 107 m1 ,
and TE2 2:53 107 m1 for the TE0 , TE1 , and TE2 modes, respectively. The end surfaces of
the cavity are not coated. Find the effective round-trip optical path length lRT
m , the round-trip
L
time T m , the longitudinal mode frequency spacing m , and the longitudinal mode width cm
for each transverse mode.
Solution:
For the linear cavity, the effective round-trip optical path length of each transverse waveguide
mode is found using (6.20):
c
l
RT
RT
m l m
) lRT
TE0 3614 m, lTE1 3572 m, lTE2 3503 m:
The round-trip time of the cavity for each transverse waveguide mode is
lRT
m 2
lRT
m
) T TE0 12:05 ps, T TE1 11:91 ps, T TE2 11:68 ps:
c
The longitudinal mode frequency spacing for each transverse waveguide mode is
Tm
Lm
1
Tm
LTE0 83:0 GHz, LTE1 84:0 GHz, LTE2 85:6 GHz:
To nd cm , it is necessary to nd the nesse. The effective refractive index for each mode is
found, which is used to nd the reectivities of the cavity and the nesse:
nm
R1, m R2, m RTEm
1=4
Fm
m
) nTE0 3:61, nTE1 3:57, nTE2 3:50;
2

1 nm 2
) RTE 32:1%, RTE 31:6%, RTE 30:9%;

0
1
2
1 n m
1=4
R1, m R2, m
1=2
1=2
1 R1, m R2, m
1=2
RTEm
1 RTEm
F TE0 2:62, F TE1 2:58, F TE2 2:53:
The longitudinal mode width cm for each transverse waveguide mode is

cm
Lm
Fm
cTE0 31:7 GHz, cTE1 32:6 GHz, cTE2 33:8 GHz:
214
Optical Resonance
6.4
CAVITY LIFETIME AND QUALITY FACTOR
..............................................................................................................
Here we consider some important parameters of a passive optical cavity of zero optical gain so
that res 0, thus g 0. Such a passive optical cavity is known as a cold cavity. To be specic,
we identify the round-trip gain factor for the eld amplitude in a cold cavity as Gc , or as Gcmn for
the transverse mode mn.
Because there is no optical gain in a cold cavity, the cavity has a net loss from nite
transmission through the end mirrors and various passive loss mechanisms so that Gc < 1.
Any optical eld that initially exists in the cavity gradually decays as it circulates inside the
cavity. Because the eld amplitude is attenuated by a factor of Gc per round trip, the intensity
and thus the number of intracavity photons are attenuated by a factor of G2c per round trip. We
can dene a photon lifetime, also called cavity lifetime, c , and a cavity decay rate, c , for a cold
cavity through the relation:
G2c eT=c ec T :
(6.23)
Therefore, the cavity lifetime is found as

c
T
:
2 ln Gc
(6.24)
The cavity decay rate is the decay rate of the optical energy stored in a cavity and is given by
c
1
2
ln Gc :
c
T
(6.25)
In general, the value of Gc for a given cavity is mode dependent. Usually, the fundamental
transverse mode has the lowest loss because its eld distribution is transversely most concentrated toward the center along the longitudinal axis of the cavity. As the order of a mode
increases, its loss in the cavity increases due to the increased diffraction loss caused by the
transverse spreading of its eld distribution. Consequently, both c and c are also mode
dependent: cmnq and cmnq . Unless a specic mode-discriminating mechanism is introduced in
a cavity, either intentionally or unintentionally, the fundamental mode generally has the largest
c and, correspondingly, the lowest c .
The quality factor, Q, of a resonator is generally dened as the ratio of the resonance
frequency, res , to the energy decay rate, , of the resonator:

energy stored in the resonator
res
:
(6.26)
Q res
average power dissipation

Therefore, the quality factor of a cold cavity is
Q
q
q c ,
c
(6.27)
where q is the longitudinal mode frequency. For a low-loss, high-Q cavity, Gc is not much less
than unity; then, it can be shown by using (6.17), (6.18), and (6.23) that
c
c
2 c 2
(6.28)
6.4 Cavity Lifetime and Quality Factor
215
and
Q
q
:
c
(6.29)
Note that though it is not explicitly spelled out in (6.27) and (6.29), the quality factor is a
function of not only the longitudinal-mode index q but also the transverse-mode indices m and
n: Q Qmnq . To be precise, (6.27) should be written as
Qmnq
mnq
mnq c :
c
(6.30)
For an optical cavity, the dependence of Qmnq on the longitudinal-mode index q is generally
insignicant because q is a very large number except in the case of a very short microcavity. By
comparison, the dependence of Qmnq on the transverse-mode indices m and n cannot be ignored.
Indeed, Q00q for the fundamental transverse mode is generally larger than Qmnq for any highorder transverse mode because the fundamental transverse mode generally has the lowest loss.
EXAMPLE 6.4
Find the photon lifetime c , the cavity decay rate c , and the quality factor Q at the 500 nm
wavelength of the linear and ring cavities that are considered in Example 6.1.
Solution:
For the linear cavity, the photon lifetime is

linear
c
T linear
10

ns 47:5 ns:
linear
2 ln 0:9
2 ln Gc
The cavity decay rate is
linear
c
1
linear
c
1
s1 2:1 107 s1 :
47:5 109
The quality factor Q at 500 nm is

2c linear 2 3 108
47:5 109 1:79 108 :
c
500 109
For the ring cavity, the photon lifetime is
Qlinear linear
ring

c
T ring
5

ns 23:7 ns:
ring
2 ln 0:9
2 ln Gc

ring
1
ring
c
1
s1 4:2 107 s1 :
9
23:7 10
The quality factor Q at 500 nm is

Qring ring
2c ring 2 3 108
23:7 109 8:93 107 :

c
500 109
216
Optical Resonance
6.5
FABRYPROT CAVITY
..............................................................................................................
The most common type of optical cavity is the FabryProt cavity, which consists of two end
mirrors in the form of the FabryProt interferometer and, in the case when it is used as a laser
cavity, an optical gain medium, as shown in Fig. 6.5. The radii of curvature of the left and right
mirrors are R1 and R2 , respectively. The sign of the radius of curvature is taken to be positive
for a concave mirror and negative for a convex mirror. For example, the cavity shown in
Fig. 6.5 has R1 > 0 and R2 > 0 because it is formed with two concave mirrors.
6.5.1 Stability Criterion

Most of the important features of a nonwaveguiding FabryProt cavity can be obtained by
applying the following simple concept. For the cavity to be a stable cavity in which a Gaussian
mode can be established, the radii of curvature of both end mirrors have to match the wavefront
curvatures of the Gaussian mode at the surfaces of the mirrors: Rz1 R1 and Rz2 R2 ,
where z1 and z2 are, respectively, the coordinates of the left and right mirrors measured from the
location of the Gaussian beam waist. Based on this concept, we have from (3.71) two relations:
z1
z2R
z2
R1 and z2 R R2 :
z1
z2
(6.31)
From these relations, we nd that

z2R
lR1 lR2 lR1 R2 l

R1 R2 2l2
(6.32)
where l z2 z1 is the length of the cavity dened by the separation between the two end
mirrors.
Given the values of R1 , R2 , and l, stable Gaussian modes exist for the cavity if both relations
in (6.31) can be satised with a real and positive parameter of zR > 0 from (6.32) for a nite,
positive beam -waist spot size w0 according to (3.69). Then the cavity is stable. If the relations
in (6.31) cannot be simultaneously satised with a real and positive value for zR , then the cavity
is unstable because no stable Gaussian mode can be established in the cavity. Application of
this concept yields the stability criterion for a FabryProt cavity:
Figure 6.5 FabryProt cavity
containing an optical gain
medium with a lling factor .
Changes of Gaussian beam
divergence at the boundaries of
the gain medium are ignored in
this plot.

0
l
1
R1

l
1
1:
R2
217
(6.33)
In a stable FabryProt cavity, the mode-dependent on-axis phase shift in a single pass through
the cavity from the left mirror to the right mirror is simply mn z2 mn z1 for the TEMmn
HermiteGaussian mode. Therefore, the round-trip mode-dependent on-axis phase shift is
RT
mn 2 mn z2 mn z1 :
(6.34)
With proper modications, the above concept can be used to nd the characteristics and
stability criterion of a cavity that has multiple mirrors, such as a folded FabryProt cavity or a
ring cavity.
EXAMPLE 6.5
A two-mirror FabryProt cavity as shown in Fig. 6.5 has a cavity length of l 1 m. One
mirror has a radius of curvature of R1 2 m. Find the condition that the radius of curvature R2
of the other mirror has to satisfy in order for the cavity to be stable. Choose a proper value for
R2 so that the cavity is stable and is most symmetric. Find the beam spot size w0 at the beam
waist for a Gaussian beam at 600 nm that is stably established in the cavity. Where is the
beam waist located?
Solution:
With l 1 m and R1 2 m, the stability condition in (6.33) requires that

l
l
1
l
0 1
1
1 ) 0
1 ) jR2 j l 1 m:
1
R1
R2
2
R2
Under this condition, R2 can be either positive or negative but its magnitude has to be larger
than 1 m. For the cavity to be stable and most symmetric, we can choose R2 R1 2 m.
Then, using (6.32), we nd the Rayleigh range:
p
3
lR1 lR2 lR1 R2 l 3 2
2
m:
m ) zR
zR
2
2
4
R1 R2 2l
The spot size at the beam waist is

w0
zR
1=2
p1=2

600 109 3
m 407 m:
2
Because R2 R1 , by symmetry the beam waist must be located right at the center of the
cavity.
6.5.2 Characteristic Parameters

We consider a cavity that contains an isotropic gain medium with a lling factor of . The
surfaces of the gain medium are antireection coated so that there is no reection inside the
218
Optical Resonance
cavity other than the reection at the two end mirrors. If the gain medium lls up the entire
cavity, we simply make 1 in the results obtained below. The FabryProt cavity has a
physical length of l between the two end mirrors. The eld reection coefcients are r 1 and r 2
for the left and right mirrors, respectively. They are generally complex to account for the phase
changes on reection, 1 and 2 , respectively, and can be expressed as
1=2
r 1 R1 ei1 ,
1=2
r 2 R2 ei2 ,
(6.35)
where R1 and R2 are the reectivities of the left and right mirrors, respectively.
The dielectric property of the intracavity gain medium includes the permittivity of the
background material and a resonant susceptibility res that characterizes the laser transition. To clearly identify the effect of each contribution, it is instructive to explicitly express
the permittivity of the gain medium, including the contribution of the resonant laser transition, as
res 0 res ,
(6.36)
where 0 n2 is the background permittivity of the gain medium excluding the resonant
susceptibility. Because res 0 for a cold cavity, the weighted average of the propagation
constant for the intracavity eld in a cold cavity is
k
n
k 1 k0 ,
c
(6.37)
where k n=c is the propagation constant in the gain medium and k0 n0 =c is that in the
surrounding medium. The round-trip optical path length in this cavity is lRT 2nl.
Usually there is an intracavity background loss contributed by a variety of mechanisms that
are irrelevant to the laser transition, such as scattering or absorption. In addition, modedependent diffraction losses exist for the intracavity optical eld due to the nite sizes of the
end mirrors. The combined effect of these losses can be accounted for by taking a spatially
averaged, mode-dependent loss coefcient, mn , so that the effective propagation constant is
complex with a mode-dependent imaginary part: k i mn =2. This loss is known as the
distributed loss of the cavity mode. In general, mn k for a practical optical cavity.
By following a mode eld through one round trip in the cavity, we nd that

a r 1 r 2 exp i2kl mn l i RT
mn
(6.38)
for the TEMmn HermiteGaussian mode. Therefore, by using (6.4) and (6.35), we nd that both
the round-trip gain factor and the round-trip phase shift are mode dependent:
1=2 1=2
Gcmn R1 R2 e mn l
(6.39)
RT
RT
mn 2kl mn 1 2 :
(6.40)
and
Using (6.40) for the resonance condition given in (6.19), we nd the resonance frequencies of
the cold FabryProt cavity:
cmnq
2q RT
mn 1 2 ,
2nl
cmnq

c
RT
mn 1 2
q
,
2
2
2nl
cmnq
219
(6.41)
where the superscript c indicates the fact that the frequencies are those for a cold cavity with
res 0. These frequencies are clearly functions of the transverse-mode indices because of the
RT
mode-dependent phase shift RT
mn . However, because mn is not a function of the longitudinalmode index q, the frequency separation between two neighboring longitudinal modes of the
same transverse mode group is a mode-independent constant:
L cmn, q1 cmnq
c
1
:
2nl T
(6.42)
Here we assume that the background optical property of the medium is not very dispersive so
that the background refractive index n can be considered a constant that is independent of
optical frequency in the narrow range between neighboring modes of interest.
Using (6.12) and (6.39), the nesse of the lossy FabryProt cavity is
1=4 1=4
R1 R2 e mn l=2
1=2 1=2
1 R1 R2 e mn l
(6.43)
which is mode dependent due to the mode-dependent loss mn . The longitudinal mode width,
c L =F, is also mode dependent for the same reason. For a cavity that has a negligible
loss, we can take mn 0; then, (6.43) reduces to the familiar formula for the nesse of a
lossless FabryProt interferometer as given in (5.46):
1=4 1=4
R1 R2
1=2 1=2
1 R1 R2
(6.44)
Therefore, for a nondispersive, lossless FabryProt cavity, L , F, and c are all independent
of the longitudinal and transverse mode indices though the mode frequency mnq is a function of
all three mode indices.
Using (6.24) and (6.39), the mode-dependent photon lifetime of the FabryProt cavity can
be expressed as
cmnq
nl
p ,
cmn l ln R1 R2
and the mode-dependent cavity decay rate can be expressed as

c
1 p
c
mn ln R1 R2 :
mnq
n
l
(6.45)
(6.46)
Clearly, both cmnq and cmnq are also mode dependent due to the mode-dependent distributed loss
mn . However, they are independent of the longitudinal mode index q under the assumption that
the background refractive index n, the loss mn , and the mirror reectivities R1 and R2 are not
sensitive to the frequency differences among different longitudinal modes. If any of these
parameters vary signicantly within the range of the longitudinal modes of interest, then the
dependence of cmnq and cmnq on the index q cannot be ignored.
220
Optical Resonance
A FabryProt cavity that is used as a laser cavity has a Q value ranging from the order of 103
for a cavity of a high-gain laser that has low mirror reectivities to the order of 108 for a cavity
of a low-gain laser that has high mirror reectivities. A FabryProt cavity that is used as a
high-resolution optical spectrum analyzer can have an even higher Q value.
EXAMPLE 6.6
The FabryProt cavity of a high-gain InGaAsP/InP semiconductor laser emitting at the 1.3 m
wavelength has an effective average refractive index of n n 3:5 dened by the InGaAsP/
InP waveguide mode, a physical length of l 300 m, and mirror reectivities of R1
R2 0:3. The structure supports only one transverse mode. Assume a negligibly small for
simplicity. Find the round-trip time, the longitudinal mode frequency spacing, the nesse, the
longitudinal mode width, the photon lifetime, the cavity decay rate, and the quality factor of this
cavity as a cold cavity.
Solution:
The round-trip time of the cavity is
T
2nl 2 3:5 300 106

s 7 ps:
c
3 108

L
1
1
Hz 142:9 GHz:
T 7 1012
Assuming no distributed loss, the nesse of the cavity is

1=4 1=4
R1 R2
1=2 1=2
1 R1 R2
0:31=4 0:31=4
2:46:
1 0:31=2 0:31=2

c
L 142:9
GHz 58:1 GHz:
2:46
F
The photon lifetime is

nl
3:5 300 106
p
s 2:9 ps:
p

c
c ln R1 R2
3 108 ln 0:3 0:3
c
1
1
s1 3:4 1011 s1 :

c 2:9 1012
To nd the quality factor, we note that the frequency is found using 2c= for the given
optical wavelength of 1:3 m. Thus, using (6.27), we nd the quality factor of this cavity to
be
Problems
Q c
221
2c
2 3 108
2:9 1012 4:2 103 :
c
1:3 106
The approximate relation (6.29) yields a slightly smaller value of Q 4:0 103 . A Q value on
the order of 103 is relatively low for a laser cavity. Even so, the difference between (6.27) and
(6.29) is only about 5%.
Problems
6.1.1 A folded FabryProt cavity as shown in Fig. 6.1(b) has two end mirrors with R1 R2
0:8 and a middle mirror with Rm 0:9 for folding the cavity, which is separated from the
two end mirrors at l1m 0:8 m and l2m 0:3 m, respectively. A glass rod that has a
length of lg 0:2 m and a refractive index of ng 1:5 is placed along the beam path
between the two mirrors of R1 and Rm . Find the physical length, the round-trip length lRT ,
the round-trip time T, and the round-trip gain factor G of the cavity.
6.1.2 A ring cavity as shown in Fig. 6.1(c) has three mirrors with R1 R2 0:8 and R3 0:9,
which are separated at l12 0:5 m and l23 l31 0:3 m. A glass rod that has a length of
lg 0:2 m and a refractive index of ng 1:5 is placed along the beam path between the
two mirrors of R1 and R2 . Find the physical length, the round-trip length lRT , the roundtrip time T, and the round-trip gain factor G of the cavity.
6.1.3 An optical-ber ring cavity as shown in Fig. 6.1(d) has one inputoutput coupler that has
a coupling efciency of 20%. The ber loop has a length of l 2 m, and the
effective index of the ber mode is n 1:47. Find the physical length, the round-trip
length lRT , the round-trip time T, and the round-trip gain factor G of the cavity.
6.2.1 Find the nesse F, the longitudinal mode frequency spacing L , and the longitudinal
mode width c of the folded FabryProt cavity considered in Problem 6.1.1.
mode width c of the ring cavity considered in Problem 6.1.2.
mode width c of the ber ring cavity considered in Problem 6.1.3.
6.3.1 An InP/InGaAsP semiconductor optical cavity has the longitudinal structure of a linear
FabryProt cavity and the transverse structure of a slab waveguide. The cavity has a
physical length of l 400 m. The slab waveguide supports two TE and two TM modes
at the 1:3 m wavelength, with propagation constants of TE0 1:67 107 m1 ,
TM0 1:65 107 m1 , TE1 1:57 107 m1 , and TM1 1:56 107 m1 for the
TE0 , TM0 , TE1 , and TM1 modes, respectively. The end surfaces of the cavity are not
coated. Find the effective round-trip optical path length lRT , the round-trip time T, the
longitudinal mode frequency spacing L , and the longitudinal mode width c for each
transverse mode.
222
Optical Resonance
6.4.1 Find the photon lifetime c , the cavity decay rate c , and the quality factor Q at the
850 nm wavelength of the folded FabryProt cavity considered in Problems 6.1.1 and 6.2.1.
850 nm wavelength of the ring cavity considered in Problems 6.1.2 and 6.2.2.
850 nm wavelength of the ber ring cavity considered in Problems 6.1.3 and 6.2.3.
6.4.4 An optical cavity has two characteristic time constants: the round-trip time T and the
photon lifetime c . Once they are known, most of the other characteristic parameters of
the cavity can be found. Find the cold-cavity eld-amplitude gain factor Gc , the nesse F,
the longitudinal mode frequency spacing L , the longitudinal mode width c , the cavity
decay rate c , and the quality factor Q at the 1:3 m wavelength for an optical cavity
that has T 1 ns and c 20 ns.
6.4.5 An optical cavity has two characteristic spectral parameters: the longitudinal mode
frequency spacing L and the longitudinal mode width c . Once they are known, most
of the other characteristic parameters of the cavity can be found. Find the nesse F, the
cold-cavity eld-amplitude gain factor Gc , the round-trip time T, the photon lifetime c ,
the cavity decay rate c , and the quality factor Q at the 1:064 m wavelength for an
optical cavity that has L 150 MHz and c 5 MHz.
6.4.6 An optical cavity has two characteristic quality factors: the nesse F and the quality factor
Q at a specic resonance frequency. Once they are known, most of the other characteristic
parameters of the cavity can be found. Find the cold-cavity eld-amplitude gain factor Gc ,
the photon lifetime c , the cavity decay rate c , the round-trip time T, the longitudinal mode
frequency spacing L , and the longitudinal mode width c for an optical cavity that has a
nesse of F 100 and a quality factor of Q 2 108 at the 532 nm wavelength.
6.5.1 Show for a linear FabryProt cavity of a length l as shown in Fig. 6.5 that the locations
of the left and right end mirrors measured from the beam waist are, respectively,
lR2 l
lR1 l
, z2
,
(6.47)
R1 R2 2l
R1 R2 2l
where R1 and R2 are the radii of curvature of the left and right mirrors, respectively.
Show also that the Rayleigh range of a stable Gaussian beam dened by the cavity is that
given by (6.32).
z1
6.5.2 A linear FabryProt cavity in free space has a concave left mirror that has a radius of
curvature of R1 2 m and a convex right mirror that has a radius of curvature of
R2 1 m. The cavity length is l 1:5 m. Is the cavity stable? If it is stable, where
is the Gaussian beam waist located? What is the beam waist spot size?
6.5.3 A symmetric linear FabryProt cavity in free space has a cavity length of l and two
mirrors of the same radius of curvature of R1 R2 R 1 m.
(a) In what range can the cavity length be chosen to make the cavity stable?
(b) For different choices of the cavity length, where is the location of the beam waist of
the Gaussian beam that is dened by the cavity?
(c) Find the cavity length that maximizes the waist spot size of the Gaussian beam? What
is this spot size for an optical wavelength of 1:064 m?
Bibliography
223
(d) For a beam waist spot size of w0 350 m, what is the cavity length that has to be
chosen?
(e) If the cavity length is chosen to be l 1:5 m, is the cavity stable? If it is stable, what
is the beam waist spot size?
6.5.4 The length of the InGaAsP/InP FabryProt cavity described in Example 6.6 is doubled
to l 600 m. At the 1:3 m wavelength, the effective index of n n 3:5 and
the mirror reectivities of R1 R2 0:3 remain unchanged, while the distributed loss is
still negligible. Find the round-trip time, the longitudinal mode frequency spacing, the
nesse, the longitudinal mode width, the photon lifetime, the cavity decay rate, and the
quality factor of this cavity. How are these parameters changed as compared to those
found in Example 6.6?
6.5.5 The length of the InGaAsP/InP FabryProt cavity described in Example 6.6 remains
l 300 m. At the 1:3 m wavelength, the effective index of n n 3:5 and the
mirror reectivities of R1 R2 0:3 remain unchanged, but the cavity now has a small
distributed loss of 10 cm1 . Find the round-trip time, the longitudinal mode frequency spacing, the nesse, the longitudinal mode width, the photon lifetime, the cavity
decay rate, and the quality factor of this cavity. How are these parameters changed as
compared to those found in Example 6.6?
6.5.6 An optical-ber FabryPerot cavity has a physical length of l 20 m, an averaged
intracavity refractive index of n 1:45, a distributed loss of 0:005 m1 , and mirror
reectivities of R1 R2 80%.
(a) What are the round-trip optical path length, the round-trip time, and the longitudinal
mode frequency spacing of this cavity?
(b) Find the free spectral range, the nesse, and the longitudinal mode width of this cavity.
(c) What are the cavity decay rate, the photon lifetime, and the quality factor for
1:3 m?
Bibliography
Milonni, P. W. and Eberly, J. H., Laser Physics. New York: Wiley, 2010.
Siegman, A. E., Lasers. Mill Valley, CA: University Science Books, 1986.
Silfvest, W. T., Laser Fundamentals. Cambridge: Cambridge University Press, 1996.
Svelto, O., Principles of Lasers, 5th edn. New York: Springer, 2010.
Verdeyen, J. T., Laser Electronics, 3rd edn. Englewood Cliffs, NJ: Prentice-Hall, 1995.
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
7 - Optical Absorption and Emission pp. 224-248
7
7.1
Optical Absorption and Emission
OPTICAL TRANSITIONS
..............................................................................................................
Optical absorption and emission occur through the interaction of optical radiation with electrons in a material system that denes the energy levels of the electrons. Depending on the
properties of a given material, electrons that interact with optical radiation can be either those
bound to individual atoms or those residing in the energy-band structures of a material such as a
semiconductor. In any event, the absorption or emission of a photon by an electron is associated
with a resonant transition of the electron between a lower energy level j1i of energy E1 and an
upper energy level j2i of energy E 2 , as illustrated in Fig. 7.1. The resonance frequency, 21 , of
the transition is determined by the separation between the energy levels:
v21
E2 E1
:
h
(7.1)
In an atomic or molecular system, a given energy level usually consists of a number of

degenerate quantum mechanical states that have the same energy. The degeneracy factors g1
and g2 account for the degeneracies in the energy levels j1i and j2i, respectively.
There are three basic types of processes associated with resonant optical transitions of electrons
between two energy levels: absorption, stimulated emission, and spontaneous emission, which are
illustrated in Figs. 7.1(a), (b), and (c), respectively. Absorption and stimulated emission of a photon
are both associated with induced transitions between two energy levels caused by the interaction of
an electron with existing optical radiation. An electron that is initially in the lower level j1i can
absorb a photon to make a transition to the upper level j2i. An electron that is initially in the upper
level j2i can be stimulated by the optical radiation to emit a photon while making a downward
transition to the lower level j1i. By contrast, spontaneous emission is not induced. Irrespective of
the presence or absence of existing optical radiation, an electron initially in the upper level j2i can
spontaneously relax to the lower level j1i by emitting a spontaneous photon.
Figure 7.1 (a) Absorption, (b) stimulated emission, and (c) spontaneous emission of photons resulting from
resonant transitions of electrons in a material.
225
A photon that is emitted through stimulated emission has the same frequency, phase,
polarization, and propagation direction as the optical radiation that induces the process. By
contrast, spontaneously emitted photons are random in phase and polarization, and they are
emitted in all directions, though their frequencies are still dictated by the separation between the
two energy levels, subject to a degree of uncertainty determined by the linewidth of the
transition. Therefore, stimulated emission results in the amplication of an optical signal,
whereas spontaneous emission merely adds noise to an optical signal. Absorption simply leads
to the attenuation of an optical signal.
7.1.1 Spectral Lineshape

A resonant transition is selective of the frequency of the interacting optical eld because the
process is associated with the absorption or emission of a photon that has a frequency
determined by the energy change of the electron making the transition, as indicated in (7.1).
The spectral characteristic of a resonant transition is never innitely sharp, however. The nite
spectral width of a resonant transition is dictated by the uncertainty principle of quantum
mechanics, but it can be intuitively understood using the reasoning in Section 2.3. One
important conclusion learned from the discussion in Section 2.3 is that any response that has
a nite relaxation time in the time domain must have a nite spectral width in the frequency
domain. As we shall see later, the induced transition rates of both absorption and stimulated
emission between two energy levels in a given system are directly proportional to the spontaneous emission rate from the upper to the lower of the two levels. Therefore, it is a basic law of
physics that any allowed resonant transition between two energy levels has a nite relaxation
time because at least the upper level has a nite lifetime due to spontaneous emission.
Consequently, every optical process associated with a resonant transition between two specic
energy levels is characterized by a lineshape function, g^v or g^ , of a nite linewidth. The
lineshape function is generally normalized as
g^vdv g^d 1, where g^v 2^

g :
(7.2)
7.1.2 Homogeneous Broadening

If all of the atoms in a material that participate in a resonant interaction associated with the
energy levels j1i and j2i are indistinguishable, their responses to an electromagnetic eld are
characterized by the same transition resonance frequency 21 and the same relaxation rate 21 .
Note that 21 is the phase relaxation rate of the resonant interaction between the electromagnetic eld and the two energy levels. In such a homogeneous system, the physical mechanisms
that broaden the linewidth of the transition affect all atoms equally. Spectral broadening caused
by such mechanisms is called homogeneous broadening.
From the discussion in Section 2.3, the spectral characteristics of a damped response that is
characterized by a single resonance frequency and a single relaxation rate, such as that of a
226
resonant interaction in a homogeneously broadened system, are described by the resonant

susceptibility given in (2.26), with its real and imaginary parts given in (2.27). As discussed
later in Section 7.2, in the interaction of an optical eld with a material, the absorption and
emission of optical energy are characterized by the imaginary part 00res of the resonant susceptibility of the material. Therefore, the spectral characteristics of resonant optical absorption and
emission in a homogeneously broadened medium are described by the Lorentzian lineshape
function of 00res given in (2.27). The normalized Lorentzian lineshape function, which is
normalized using (7.2), for the resonant transitions between j1i and j2i has the form:
1
21
g^
,
21 2 221
(7.3)
which has a FWHM of h 221 , or

g^ v
vh
2v v21 2 vh =22
(7.4)
where
vh
21
(7.5)
is the FWHM of g^v. We see that the spectrum has a nite width that is determined by the
relaxation rate 21 .
The fundamental mechanism for homogeneous broadening is lifetime broadening due to the
nite lifetimes, 1 and 2 , respectively, of the energy levels, j1i and j2i, that are involved in the
resonant transitions. The population in an energy level can relax through both radiative and
nonradiative transitions to lower levels. Radiative relaxation is associated with population
relaxation through spontaneous emission of radiation. The radiative relaxation rate of the
transition from level j2i to level j1i is characterized by a rate constant A21 , known as the
Einstein A coefcient, which denes a time constant sp 1=A21 , known as the spontaneous
radiative lifetime, between j2i and j1i. Both A21 and sp are discussed in further detail later.
The total radiative relaxation rate, rad
all radiative
2 , of level j2i is the sum of the rates ofX
rad
spontaneous transitions from j2i to all levels of lower energies: 2
A . The
i 2i
nonrad
nonradiative relaxation rate, 2
, accounts for all other population relaxation mechanisms
that do not result in the emission of photons. The total relaxation rate, 2 , of level j2i is the sum
of its radiative and nonradiative relaxation rates, and the lifetime of the energy level has both
radiative and nonradiative contributions:
nonrad
2 rad
,
2 2
1
1
1
rad nonrad ,
2 2
2
(7.6)
rad
nonrad
1=nonrad
. This concept can be applied to level j1i
where 2 1=2 , rad
2 1=2 , and 2
2
to obtain similar relations for 1 and 1 .
Though 2 has contributions of both radiative and nonradiative relaxations, the uorescence
due to spontaneous emission from level j2i decays in time at the total relaxation rate 2 because
its strength is proportional to the population in level j2i, which relaxes at the total relaxation
227
rate. Therefore, the decay time constant of the uorescent emission from level j2i is 2 , not rad
2 .
For this reason, the total lifetimes 1 and 2 are known as the uorescence lifetimes of energy
levels j1i and j2i, respectively. The contributions of various relaxation rates to the radiative and
nonradiative lifetimes, and to the uorescence lifetimes, of the upper and lower energy levels
are summarized in Fig. 7.2.
The nonradiative relaxation rate of an energy level is a function of extrinsic factors, such as
collisions and thermal vibrations. It can therefore be changed by varying the conditions of the
surrounding environment. The minimum broadening is called natural broadening, which is
caused only by radiative relaxation when all nonradiative processes are eliminated. The linewidth due to natural broadening is determined by the radiative phase relaxation rate caused by
radiative decays of the two energy levels:

1 rad
1 1
1
natural
rad
rad
21
(7.7)
21 1 2
rad :
2
2 rad
2
1
The total phase relaxation rate that characterizes lifetime broadening of the linewidth accounts for
the lifetimes of the two energy levels due to both radiative and nonradiative relaxation processes:

1
1 1
1
life
:
(7.8)
21 1 2
natural
21
2
2 1 2
and life
The contributions to natural
21
21 are also summarized in Fig. 7.2. Note that the linewidth is
determined by the lifetimes of both upper and lower levels. In the case when the lower level j1i
is the ground level of an atomic system, we have 1 0 and 1 . Then, the linewidth due to
lifetime broadening is solely determined by the lifetime of the upper level, 2 .
Other mechanisms that affect all atoms equally can further increase the homogeneous linewidth without changing the uorescence lifetime of either the upper or the lower level. One
Figure 7.2 Contributions of various relaxation rates to the radiative and nonradiative lifetimes, and to the
uorescence lifetimes, of the upper and lower energy levels. The homogeneous natural linewidth is determined
by the radiative lifetimes, whereas the lifetime-broadened linewidth is determined by the uorescence lifetimes.
228
important mechanism is collision-induced phase randomization of the emitted radiation. Collisions among atoms in a gas or liquid and collisions between atoms and phonons in a solid
normally have two possible effects. One effect is to reduce the uorescence lifetimes of the
upper and lower levels by increasing the nonradiative relaxation rates. Such a process increases
life
nonrad
lifetime broadening; its effect is included in life
and
21 through the dependence of 21 on 1
nonrad
contained in 1 and 2 , respectively. Collisions can also increase a homogeneous line2
width without reducing the uorescence lifetimes by simply interrupting the phase of the
radiation emitted through radiative relaxation. This dephasing process, quantied by a
linewidth-broadening factor dephase
, is often more important than the lifetime-reduction pro21
cess, resulting in a homogeneous linewidth that is signicantly broader than the linewidth due
to lifetime broadening. Therefore, the homogeneous linewidth can increase with both pressure
and temperature in a gas medium, and with active-ion concentration and temperature in a liquid
or solid medium. In general, the homogeneous linewidth including the contributions of
such extrinsic mechanisms is a function of pressure, P, active-ion concentration, N, and
temperature, T:
dephase
natural
21 P, N, T life
life
:
21 21
21 21
(7.9)
EXAMPLE 7.1
The energy levels of Nd:YAG are shown in Fig. 7.3. The highest level 4 F3=2 of the active Nd3
ion relaxes to four lower levels at different radiative relaxation rates characterized by the
Einstein A coefcients shown for different emission wavelengths. The lowest level 4 I9=2 is
the ground level, which does not relax to any other level. The dominant transition of this system
is that associated with the well-known Nd:YAG emission wavelength of 1:064 m, which
takes place between the upper level 4 F3=2 , labeled j2i, and the lower level 4 I11=2 , labeled j1i.
The upper level 4 F3=2 has a lifetime of 2 240 s predominantly due to radiative relaxation;
Figure 7.3 Energy levels of Nd:YAG.
229
the lower level 4 I11=2 has a lifetime of 1 200 ps purely from nonradiative relaxation. (a) Find
the radiative, nonradiative, and total relaxation rates for the upper and lower levels, j2i and j1i,
respectively. (b) Find the natural linewidth and the lifetime-broadened linewidth for the
1:064 m emission line. If no other mechanisms further broaden this line, what is its lineshape
and linewidth? (c) At room temperature, dephasing due to phonon collisions contributes a
dephasing rate of dephase
3:75 1011 s1 to the linewidth. What is the homogeneous line21
width of this emission line at room temperature?
Solution:
All of the processes considered here cause homogeneous broadening because they are
common to all Nd3 ions. Inhomogeneous broadening mechanisms are not considered in
this example.
(a) The upper level j2i relaxes both radiatively and nonradiatively to four lower levels, but the
lower level j1i relaxes only nonradiatively to the ground level. The total relaxation rates of
the two levels are, respectively,
2
1
1
s1 4167 s1 ,

2 240 106
1
1
s1 5 109 s1 :

1 200 1012
The radiative relaxation rates of the two levels are, respectively,

X
A2i 3868 s1 , rad
rad
2
1 0:
i
The nonradiative relaxation rates of the two levels are, respectively,

1
nonrad
2 rad
2
2 299 s ,
9 1
nonrad
1 rad
1
1 5 10 s :
(b) Using the results from (a), we nd that

1
1
rad
1
rad
1934 s1 ,
natural
21
1 2 0 3868 s
2
2
1
1
9
1
2:5 109 s1 :
life
21 1 2 5 10 4167 s
2
2
The natural linewidth and the lifetime-broadened linewidth are, respectively,
natural
natural
21
616 Hz,
life
life
21
796 MHz:
If no other mechanisms further broaden this line, this emission line has a Lorentzian
lineshape that has a homogeneously broadened linewidth of h life 796 MHz:
(c) With a dephasing rate of dephase
3:75 1011 s1 , the total phase relaxation rate is
21
dephase
3:775 1011 s1 :
21 life
21 21
230
Thus, the homogeneous linewidth is broadened to

h
21
120 GHz:
The linewidth is further broadened by inhomogeneous mechanisms discussed below. For the
1:064 m line of Nd:YAG, the total linewidth varies with temperature and with the
quality of the YAG crystal. Increasing temperature increases the homogeneous linewidth,
whereas a poorer crystal quality leads to a larger inhomogeneous linewidth. In any event, this
emission line of Nd:YAG is predominantly homogeneously broadened at room temperature.
7.1.3 Inhomogeneous Broadening

A resonant transition can be further broadened by inhomogeneous broadening if certain
physical mechanisms exist that do not equally affect all atoms, causing energy levels j1i or
j2i, or both, to shift differently among different groups of atoms. The resulting inhomogeneous
shifts of the transition resonance frequency cause inhomogeneous broadening of the transition
spectrum on top of the original homogeneous broadening.
If we express the homogeneous lineshape function given in (7.4) as g^h , 21 to explicitly
indicate that its transition resonance frequency is 21 , the homogeneously broadened spectrum
of a group of atoms whose resonance frequency is shifted from 21 to k is g^h , k . The
distribution of atoms in an inhomogeneous system can be described by a probability density
function pk with
pk dk 1:
(7.10)
The probability that the resonance frequency of a given atom falls in the range between k and
k dk is pk dk . Then, the overall spectral lineshape of the inhomogeneously broadened
transition is
g^ pk ^
g h , k dk :
(7.11)
The overall lineshape function obtained from (7.11) depends on the degree of inhomogeneous
broadening in comparison to homogeneous broadening. Mathematically, it depends on the
spread of the distribution function pk in comparison to the homogeneous linewidth.
One possibility for inhomogeneous broadening is the existence of different isotopes, which
have slightly different resonance frequencies for a given resonant transition. In this situation,
pk dk represents the percentage of each isotope group among all atoms and (7.11) becomes
simply the weighted sum of the isotope groups.
Other mechanisms for inhomogeneous broadening include the Doppler effect in a gaseous
medium at a low pressure and the random distribution of active impurity atoms doped in a solid
231
host. The inhomogeneous frequency shifts caused by these mechanisms are usually randomly
distributed, resulting in a Gaussian functional distribution for pk . In an extremely inhomogeneously broadened system, the spread of this distribution dominates the homogeneous linewidth. Then, the transition is characterized by a normalized Gaussian lineshape:
"
#
2ln 21=2
0 2
g^ 1=2
exp 4 ln 2
,
(7.12)
inh
2inh
where 0 is the center frequency and inh is the FWHM of the inhomogeneously broadened
spectral distribution. In terms of the angular frequency, the normalized Gaussian lineshape is
"
#
2ln 21=2
0 2
g^ 1=2
exp 4 ln 2
,
(7.13)
inh
2inh
where 0 20 and inh 2inh .
Whether a medium is homogeneously or inhomogeneously broadened is often a function of
pressure and temperature. In a gas at a low pressure, the velocity distribution of the gas
molecules in thermal equilibrium is characterized by the Maxwellian velocity distribution,
which is a Gaussian function. This velocity distribution leads to a Gaussian distribution of
Doppler frequency shifts with a linewidth of D given by
3=2
D 2
ln 2
1=2

k B T 1=2 23=2 ln 21=2 k B T 1=2
Mc2
M
(7.14)
where is the emission wavelength, kB is the Boltzmann constant, T is the temperature in kelvin, and
M is the mass of the atom or molecule that emits the radiation. When this Doppler-broadening effect
dominates, the Gaussian lineshape has an inhomogeneous linewidth of inh D .
Figure 7.4 Normalized Lorentzian (solid curves) and Gaussian (dashed curves) lineshape functions of the same
FWHM with (a) a normalized area as g^ is dened and (b) a normalized peak value. For the Lorentzian
lineshape, 0 21 and h . For the Gaussian lineshape, inh .
232
The normalized Lorentzian lineshape function and the normalized Gaussian lineshape function of the same FWHM are compared in Fig. 7.4. In Fig. 7.4(a), we show g^ as expressed in
(7.4) for the Lorentzian lineshape and in (7.12) for the Gaussian lineshape, both with a
normalized area as dened in (7.2). In Fig. 7.4(b), the lineshapes are normalized to have the
same peak value.
EXAMPLE 7.2
The transition for the well-known HeNe emission wavelength of 632:8 nm takes place
between the 3s2 level, which is the upper level j2i, and the 2p4 level, which is the lower
level j1i, of the Ne atom. The upper and lower levels for this emission both relax
20
rad
radiatively, with 2 rad
and
2 30 ns and 1 1 10 ns. Two Ne isotopes, Ne
22
20
Ne , contribute to this emission, with more than 90% due to Ne . For simplicity, we
take the atomic mass number of Ne to be 20. The typical HeNe laser medium operates at a
temperature of T 400 K and a low gas pressure of P 2:5 torr. (a) Find the radiative,
nonradiative, and total relaxation rates for the upper and lower levels, j2i and j1i,
respectively. (b) Find the natural linewidth and the lifetime-broadened linewidth of the
emission line. (c) Find the linewidth caused by Doppler broadening. (d) What is the
lineshape and linewidth of this emission line?
Solution:
Natural broadening and lifetime broadening are homogeneous broadening mechanisms,
whereas Doppler broadening is an inhomogeneous broadening mechanism. Pressure-induced
broadening is a homogeneous mechanism, but it can be ignored in this problem because of the
low gas pressure of P 2:5 torr.
(a) Both the upper level j2i and the lower level j1i relax radiatively. For each level, the total
relaxation rate is the same as the radiative relaxation rate:
2 rad
2
1
1
s1 3:3 107 s1 ,

2 30 109
1 rad
1
1
1
s1 1 108 s1 :

1 10 109
The nonradiative relaxation rates of the two levels are both zero:
nonrad
nonrad
2 rad
1 rad
2
2 0, 1
1 0:
(b) Using the results from (a), we nd that

1
1
rad
8
7
1
rad
6:7 107 s1 ,
natural
21
1 1 1 10 3:3 10 s
2
2
1
1
8
7
1
life
6:7 107 s1 :
21 1 2 1 10 3:3 10 s
2
2
233
The natural linewidth and the lifetime-broadened linewidth are the same:
life
life
natural
21
natural 21
21:2 MHz:
If no other mechanisms further broaden this line, this emission line has a Lorentzian
lineshape that has a homogeneously broadened linewidth of h life 21:2 MHz.
(c) The mass of a Ne atom is M 20 1:66 1027 kg 3:32 1026 kg for a mass
number of 20. Therefore, the Doppler-broadened linewidth at T 400 K is

1=2

23=2 ln 21=2 k B T 1=2
23=2 ln 21=2 1:38 1023 400
Hz 1:5 GHz:
D
M
632:8 109
3:32 1026
(d) Because D life , the homogeneous lifetime broadening is completely dominated by
the inhomogeneous Doppler broadening. Therefore, the lineshape of this emission line is
Gaussian with a linewidth of inh D 1:5 GHz:
7.1.4 Mixed Broadening

When the pressure of a gaseous medium is increased, frequent collisions among the gas
molecules shorten the lifetimes of the excited states of the molecules. This effect reduces 2 ,
and it can also reduce 1 if the lower level is not the ground level. The resulting pressureinduced lifetime broadening causes the homogeneous linewidth to increase. At a certain
pressure, the homogeneous linewidth h nally dominates the Doppler linewidth D . Then
the medium becomes predominantly homogeneously broadened.
Another good example is the linewidth associated with the impurity ions doped in a solid
host, such as Nd:YAG or Nd:glass. At a low temperature, the homogeneous linewidth of
the Nd3 ions is narrow. The lineshape is dominated by inhomogeneous shifts of the
resonance frequency due to variations in the local environment of individual Nd3 ions.
As a result, the lineshape function is inhomogeneously broadened. As the temperature
increases, the homogeneous linewidth increases because of increased collisions of phonons
with the ions. At room temperature, the spectral line of Nd:YAG at 1.064 m has a total
linewidth of 120 to 180 GHz with an inhomogeneous component of only about
6 to 30 GHz. Therefore, as illustrated in Example 7.1, Nd:YAG is pretty much homogeneously broadened at room temperature. In comparison, Nd:glass has a much larger
inhomogeneous linewidth than Nd:YAG because the glass host provides a larger range of
local variations than the YAG crystal. At room temperature, the same spectral line of Nd:
glass appears at 1.054 m with a total linewidth of 5 to 7 THz, which is almost all
inhomogeneously broadened.
Clearly a lineshape can be neither Lorentzian nor Gaussian when the homogeneously
broadened linewidth h and the inhomogeneously broadened linewidth inh of an emission
line are on the same order of magnitude. In this situation, the line prole is a convolution of the
234
Lorentzian prole of a width h and the Gaussian prole of a width inh . The result is a Voigt
lineshape that has a linewidth of
0:5346h 0:21662h 2inh 1=2 :
7.2
(7.15)
TRANSITION RATES
..............................................................................................................
The probability per unit time for a resonant optical process to occur is measured by the
transition rate of the process. Because of the resonant nature of the interaction, the transition
rate of an induced process is a function of both the spectral distribution of the optical radiation
and the spectral characteristics of the resonant transition.
The spectral distribution of an optical eld is characterized by its spectral energy density,
u, which is the energy density of the optical radiation per unit frequency interval at the
optical frequency . The total energy density of the radiation is
u ud:
(7.16)
The spectral intensity distribution, I, of the radiation is related to u by the relation

c
I u,
n
(7.17)
where n is the refractive index of the medium, and the total intensity is simply
I Id:
(7.18)
Because an induced transition is stimulated by optical radiation, its transition rate is proportional to the energy density of the optical radiation within the spectral response range of the
transition. The transition rate for the upward transition from j1i to j2i, associated with
absorption, in the frequency range between and d is
W 12 d B12 u^
g d,
(7.19)
whereas that for the induced downward transition from j2i to j1i, associated with stimulated
emission, in the frequency range between and d is
W 21 d B21 u^
g d:
(7.20)
Because the spontaneous emission rate is independent of the energy density of the radiation, the
spontaneous emission spectrum is determined solely by the lineshape function of the transition:
W sp d A21 g^d:
(7.21)
The A and B constants dened above are known as the Einstein A and B coefcients,
respectively. The rates associated with the transitions between two atomic levels j1i and j2i
235
Figure 7.5 Resonant transitions in the interaction of a radiation eld with two atomic levels j1i and j2i of
population densities N 1 and N 2 , respectively.
in the interaction with a radiation eld of an energy density u are summarized in Fig. 7.5.
The total induced transition rates are
W 12 W 12 d B12 u^
g d
and
W 21 W 21 d B21 u^
g d:
0
(7.22)
(7.23)
The total spontaneous emission rate is
W sp W sp d A21 :
(7.24)
The induced and spontaneous transition rates of a given system are not independent of each
other but are directly proportional to each other. Their relationship was rst obtained by
Einstein by considering the interaction of blackbody radiation with an ensemble of identical
atomic systems in thermal equilibrium. The spectral energy density of blackbody radiation at a
temperature T is given by Plancks formula:
u
8n3 h3
1
,
3
h=k
T 1
B
c
e
(7.25)
where k B is the Boltzmann constant.

As shown in Fig. 7.5, the population densities per unit volume of the atoms in levels j2i and
j1i are N 2 and N 1 , respectively. The number of atoms per unit volume making the downward
transition per unit time accompanied by the emission of radiation in a frequency range from to
d is N 2 W 21 W sp d, and the number of atoms per unit volume making the upward
transition per unit time through the absorption of radiation in the same frequency range is
N 1 W 12 d. In thermal equilibrium, both the spectral density of blackbody radiation and the
atomic population density in each energy level reach a steady state, meaning that
N 2 W 21 W sp N 1 W 12 :
(7.26)
236
This relation spells out the principle of detailed balance in thermal equilibrium. Therefore, the
steady-state population distribution in thermal equilibrium satises
N2
W 12
B12 u
:
N 1 W 21 W sp B21 u A21
(7.27)
In thermal equilibrium at a temperature T, however, the population ratio of the atoms in the
upper and the lower levels follows the Boltzmann distribution. Taking into account the
degeneracy factors, g2 and g1 , of these energy levels, we have
N 2 g2 hv=kB T
e
N 1 g1
(7.28)
for the population densities associated with a transition energy of h. Combining (7.27) and
(7.28), we have
u
A21 =B21
:
g1 B12 =g2 B21 ehv=kB T 1
(7.29)
Identifying (7.29) with (7.25), we nd that

A21 8n3 h3
B21
c3
(7.30)
g1 B12 g2 B21 :
(7.31)
and
The spontaneous radiative lifetime of the atoms in level j2i associated with the radiative
spontaneous transition from j2i to j1i is
sp
1
1
:
W sp A21
(7.32)
The spectral dependence of the spontaneous emission rate can be expressed as

W sp
1
g^ :
sp
(7.33)
According to the relations in (7.30) and (7.31), the transition rates of both of the induced
processes of absorption and stimulated emission are directly proportional to the spontaneous
emission rate. In terms of sp , the spectral dependence of the stimulated-emission transition
from j2i to j1i can be generally expressed as
W 21
c3
c2
u^
g
I^
g ,
8n3 hv3 sp
8n2 hv3 sp
(7.34)
and that for the absorption transition from j1i to j2i can be found as
W 12
g2
W 21 :
g1
(7.35)
237
Because W is the transition rate per unit frequency according to the denition in (7.19)
(7.21), we have Wd Wd. Therefore, W sp 2W sp , W 21 2W 21 ,
and W 12 2W 12 .
EXAMPLE 7.3
A cylindrical Nd:YAG rod has a length of l 5 cm and a diameter of d 6 mm. The Nd3
ions are doped in the YAG host at 1.2% atomic concentration for a total concentration of
N t 1:66 1020 cm3 . The rod is uniformly pumped such that 1% of the Nd3 ions are
excited to the 4 F3=2 level and then left to relax spontaneously. Use the parameters given in
Fig. 7.3 for the energy levels of Nd:YAG to answer the following questions regarding the
emission at the two lines of 1:064 m and 1:34 m. (a) Find the spontaneous
radiative lifetimes for the transitions of the two emission lines, respectively. (b) What are
the decay times of the spontaneous emission at the two emission lines, respectively? (c)
What are the optical energies of the spontaneous emission at the two wavelengths,
respectively? (d) What are the powers of the spontaneous emission at the two wavelengths,
respectively?
Solution:
The Nd:YAG rod has a volume of
V d=22 l 6 103 =22 5 102 m3 1:41 106 m3 :
It is pumped to have a concentration in the upper level j2i of
N 2 1%N t 1:66 1018 cm3 1:66 1024 m3 :
(a) The spontaneous radiative lifetime of each transition is determined by the A coefcient of
the transition. From Fig. 7.3, we nd A1:064 1940 s1 and A1:34 493 s1 . Therefore, the
spontaneous radiative lifetimes are, respectively,
sp
1:064
1
1
s 515 s,
A1:064 1940
sp
1:34
1
1
s 2:03 ms:
A1:34 493
(b) Because the spontaneous emission at both emission lines results from the population in level
j2i, the number density S1:064 of the spontaneous photons that are emitted at 1:064 m and
the number density S1:34 of the spontaneous photons emitted at 1:34 m are both proportional to N 2 . Therefore, the uorescence at both wavelengths decays at the same rate as that of
N 2 . The uorescence time is the same for both wavelengths and is the lifetime 2 240 s of
level j2i, given in Fig. 7.3.
(c) Though the number densities S1:064 and S1:34 of the spontaneous photons emitted at
1:064 m and 1:34 m, respectively, are both proportional to N 2 and both decay at the
same decay time, their magnitudes are respectively proportional to the spontaneous radiative relaxation rates, A1:064 and A1:34 , of their transitions:
238
S1:064
A1:064
N 2 A1:064 2 N 2 1940 240 106 1:66 1024 m3 7:73 1023 m3 ,
2
S1:34
A1:34
N 2 A1:34 2 N 2 493 240 106 1:66 1024 m3 1:96 1023 m3 :
2
The photon energies at the two wavelengths are, respectively,

hv1:064
1:2398
1:2398
eV, hv1:34
eV:
1:064
1:34
The spontaneous optical energies emitted at the two wavelengths are, respectively,
U 1:064 hv1:064 S1:064 V
U 1:34 hv1:34 S1:34 V
1:2398
1:6 1019 7:73 1023 1:41 106 J 203 mJ;
1:064
1:2398
1:6 1019 1:96 1023 1:41 106 J 41 mJ:
1:34
Because these optical energies both decay at the uorescence time of 2 240 s,
P1:064
P1:34
U 1:064 203 103
W 846 W,
2
240 106
U 1:34
41 103
W 17 W:
2
240 106
7.2.1 Transition Cross Section

It is often useful to express the transition probability of an atom in its interaction with optical
radiation at a frequency of in terms of the transition cross section, . For transitions
between energy levels j1i and j2i, the transition cross sections 21 and 12 are dened
through the following relations to the transition rates,
W 21
I
21
h
(7.36)
W 12
I
12 :
h
(7.37)
and
The transition cross section 21 , which is associated with stimulated emission, is also called
the emission cross section, e , whereas 12 , which is associated with absorption, is also
called the absorption cross section, a . From (7.34), we nd that
e 21
c2
g^ :
8n2 2 sp
(7.38)
239
According to (7.35), we nd that g1 12 g2 21 . Therefore,

a 12
g2
g
21 2 e :
g1
g1
(7.39)
The transition cross sections have the unit of area in square meters but are often quoted in
square centimeters. Note that because is simply dened as the value of the
transition cross section at the frequency rather than as that per unit frequency, but W
2W and g^ 2^
g . Therefore, in terms of ,
e 21
2 c2
g^
n2 2 sp
and a
g2
e :
g1
(7.40)
For the ideal Lorentzian and Gaussian lineshapes expressed in (7.4) and (7.12), respectively,
the peak value of g^ occurs at the center of the spectrum and is a function of the linewidth
only. By applying this fact to (7.38), the peak value of the emission cross section at the center
wavelength of the spectrum can be expressed as
he
2
4 2 n2 h sp
(7.41)
for a homogeneously broadened medium that has an ideal Lorentzian lineshape, and as
inh
e
ln 21=2 2
4 3=2 n2 inh sp
(7.42)
for an inhomogeneously broadened medium that has an ideal Gaussian lineshape. In

practice, the experimentally measured peak emission cross section usually differs from
that calculated using these formulas because the spectral lineshape of a realistic gain
medium is generally determined by a combination of many different mechanisms and,
consequently, is rarely ideal Lorentzian or ideal Gaussian. Nevertheless, these formulas
provide a good estimate for the peak value of the emission cross section. They also clearly
indicate that the emission cross section varies quadratically with the emission wavelength
but is inversely proportional to both the emission linewidth and the spontaneous radiative
lifetime of the transition.
The characteristics of some representative laser materials are listed in Table 7.1. As seen in
Table 7.1, the parameters vary over a wide range among different types of optical gain
media. For example, the peak value of the emission cross section varies from 6 1025 m2
for Er:ber to 2:5 1016 m2 for the Ar-ion laser, whereas the spontaneous emission linewidth varies from 60 MHz for CO2 to 100 THz for Ti:sapphire. The uorescence lifetime
varies from the order of 1 ns for a semiconductor gain medium to the order of 10 ms for
Er:ber.
240
Table 7.1 Characteristics of some laser materials
Gain medium
Wavelength System
(m)
Cross section
e (m2)
Spontaneous
linewidthc
Lifetimesd
(nm)
sp
Index
n
HeNe
0.6328
I,4
3.0 1017
1.5 GHz
0.002
300 ns 30 ns
Ar ion
0.488
I,4
2.5 1016
2.7 GHz
0.004
13 ns
10 ns
CO2
10.6
I,4
3.0 1022
60 MHz
0.02
4s
1 s
Copper vapor
0.5105
I,3
8.6 1018
2.3 GHz
0.002
500 ns 500 ns
KrF excimer
0.248
H,3
2.6 1020
10 THz
10 ns
8 ns
R6G dye
0.570.65
H/I,Q2
2.3 1020
30 THz
33
6 ns
4 ns
1.4
Rubye
0.6943
H,3
1.252.5 1024 330 GHz
0.53
3 ms
3 ms
1.76
Nd:YAG
1.064
H,4
210 1023
150 GHz
0.56
515 s 240 s
1.82
Nd:glass
1.054
I,4
4.0 1024
6 THz
22
330 s 330 s
1.53
Er:ber
1.53
H/I,3
6.0 1025
5 THz
40
10 ms
10 ms
1.46
0.661.1
H,Q2
3.4 1023
100 THz
180
3.9 s 3.2 s
1.76
0.781.01
H,Q2
4.8 1024
83 THz
200
67 s
67 s
1.4
H/I,Q2
15 1020
1020 THz 20100
1 ns
1 ns
34
Ti:sapphire
Cr:LiSAF
Semiconductor 0.371.65
a
H, homogeneously broadened; I, inhomogeneously broadened; Q2, quasi-two-level system; 3, three-level system;

4, four-level system.
b
Both the absorption and emission cross sections depend on the optical frequency. The absorption and emission
cross sections generally have different peak values and different spectral dependences. Listed is the peak value of
the emission cross section.
c
The spontaneous linewidth determines the gain bandwidth of a medium when population inversion is achieved.
d
The spontaneous lifetime sp is related to the transition rate, whereas the uorescence lifetime 2 is related to the
upper-level population relaxation. The uorescence lifetime of a gaseous medium varies with temperature and
pressure; that of a liquid or solid medium varies with temperature, the host material, and the concentration of the
active ions or molecules. For example, 2 of CO2 varies from 100 ns to 1 ms depending on temperature and pressure.
e
Ruby is sapphire (Al2O3) doped with Cr3+ ions. The sapphire crystal is uniaxial. For ruby, the value of e for
emission with Ec, which is listed, is larger than that for Ekc.
f
For Ti:sapphire, the value of e for Ekc, which is listed, is larger than that for Ec.
EXAMPLE 7.4
The 1:064 m emission line of Nd:YAG considered in Example 7.1 has a predominantly
homogeneously broadened total linewidth of 150 GHz and a spontaneous radiative relaxation
rate of A 1940 s1 . The refractive index of the YAG crystal is n 1:82. The 632:8 nm
241
emission line of HeNe considered in Example 7.2 has a predominantly inhomogeneously

broadened total linewidth of 1.5 GHz and a spontaneous radiative lifetime of sp 300 ns. The
refractive index of the low-pressure HeNe gas is n 1. Find the peak emission cross sections
for these two lines.
Solution:
For the 1:064 m emission line of Nd:YAG, we take h 150 GHz to be the homogeneous linewidth as an approximation because this line is predominantly homogeneously
broadened. The spontaneous radiative lifetime is sp 1=A 515 s. Then, using (7.41), the
emission cross section is found to be
he
2
1:064 106 2
m2 1:12 1022 m2 ,
4 2 n2 h sp 4 2 1:822 150 109 515 106
which is slightly larger than, but consistent with, the value listed in Table 7.1.
For the 632:8 nm emission line of HeNe, we take inh 1:5 GHz to be the inhomogeneous linewidth as an approximation because this line is predominantly inhomogeneously
broadened. With a spontaneous radiative lifetime of sp 300 ns, the emission cross section is
found using (7.42) to be
inh
e
ln 21=2 2
ln 21=2 632:8 109 2
m2 3:33 1017 m2 ,
4 3=2 n2 inh sp 4 3=2 12 1:5 109 300 109
which is slightly larger than, but consistent with, the value listed in Table 7.1.
7.3
ATTENUATION AND AMPLIFICATION OF OPTICAL FIELDS
..............................................................................................................
Optical absorption results in the attenuation of an optical eld, whereas stimulated emission
leads to the amplication of an optical eld. To quantify the net effect of a resonant transition
on the attenuation or amplication of an optical eld, we consider the interaction of a
monochromatic plane optical eld at a frequency of with a material that consists of electronic
or atomic systems with population densities N 1 and N 2 in energy levels j1i and j2i, respectively. Because the spectral intensity distribution of the monochromatic plane optical eld that
has an intensity of I is simply I I0 , the total induced transition rates between
energy levels j1i and j2i in this interaction are
I
I
(7.43)
e and W 12 a :
h
h
The net power that is transferred from the optical eld to the material is the difference
between that absorbed by the material and that emitted due to stimulated emission:
W 21
W p hW 12 N 1 hW 21 N 2 N 1 a N 2 e I:
(7.44)
242
In the case when W p > 0, there is net power absorption by the medium from the optical eld
due to resonant transitions between energy levels j1i and j2i. The absorption coefcient, also
called attenuation coefcient, is

g1
N 1 a N 2 e N 1 N 2 a :
(7.45)
g2
In the case when W p < 0, net power is transferred from the medium to the optical eld,
resulting in the amplication of the optical eld. The gain coefcient, also called the
amplication coefcient, is

g2
g N 2 e N 1 a N 2 N 1 e :
(7.46)
g1
The coefcients and g have the unit of per meter, also often quoted per centimeter. Note that
and g g because . Note also that g because a
negative gain is a positive loss, and vice versa.
According to (7.43), both e and a have positive values because W 21 0 and W 12 0
by denition. Therefore, > 0 and g < 0 if N 1 > g1 =g2 N 2 , whereas g > 0 and
< 0 if N 2 > g2 =g1 N 1 . A material in its normal state in thermal equilibrium absorbs
optical energy because the lower energy level is more populated than the upper energy level. In
order to provide a net optical gain to the optical eld, a material has to be in a nonequilibrium
state of population inversion for the upper level to be more populated than the lower level.
EXAMPLE 7.5
The 1:064 m emission line of Nd:YAG has 2 240 s for the upper level j2i and
1 200 ps for the lower level j1i, as shown in Fig. 7.3. We consider here the Nd:YAG rod
in Example 7.3, which is doped with Nd3 ions at 1.2% atomic concentration for a total
concentration of N t 1:66 1020 cm3 . If it is not pumped, what is its absorption coefcient
at 1:064 m at T 300 K? If the rod is uniformly pumped such that 1% of the total Nd3
ions are excited to level j2i, what is the absorption or gain coefcient at 1:064 m?
Solution:
The lower level j1i is not the ground level. From Fig. 7.3, we nd that its energy above the
ground level is
E 10
1:2398
1:2398
eV
eV 0:21 eV:
0:9
1:064
At T 300 K, k B T 25:9 meV. Thus, the population density of Nd3 ions in this level is
approximately

0:21
E 10 =kB T
N 1 N te
3 104 N t
N t exp
25:9 103
which is negligibly small because level j1i lies sufciently high above the ground level.
Therefore, the absorption coefcient at 1:064 m is negligibly small: 0.
243
When 1% of the total Nd3 ions are excited to level j2i, we have
N 2 1%N t 1:66 1018 cm3 1:66 1024 m3 :
In this situation, the excited ions can relax from level j2i to level j1i, but any ion reaching level j1i
quickly relaxes to the ground level because level j1i has a short lifetime of 1 200 ps
2 .
Therefore, level j1i remains almost empty, N 1 0, as compared to level j2i. The emission cross
section of the 1:064 m line found in Example 7.4 is e 1:12 1022 m2 . Consequently,
at 1:064 m the Nd:YAG rod has a gain coefcient of
g N 2 e N 1 a N 2 e 1:66 1024 1:12 1022 m1 186 m1 :
This is a very large gain coefcient even though only 1% of the total Nd3 ions are excited. In
practice, depending on the design of the laser cavity, only a smaller percentage of ions has to be
excited for laser action.
7.3.1 Resonant Optical Susceptibility

The macroscopic optical properties of a medium are characterized by its electric susceptibility.
As seen in Section 2.3, resonances in an optical medium contribute to the dispersion in the
susceptibility of the medium. Clearly, the optical properties of a medium are functions of the
resonant optical transitions between the energy levels of the electrons in the medium.
From the viewpoint of the macroscopic optical properties of a medium, the interaction
between an optical eld and a medium is characterized by the polarization induced by the
optical eld in the medium. The power exchange between the optical eld and the medium is
given by (1.34). For the resonant interaction of an isotropic medium with a monochromatic
plane optical eld at a frequency of 2, we have Et Eeit E eit and
it
Pres t 0 res Eeit
res E e , where Pres is the polarization contributed by the
resonant transitions and res is the resonant susceptibility. Using (1.34), we nd that the timeaveraged power density absorbed by the medium is
W p 2 0 00res jEj2 00res I:

(7.47)
nc
By identifying (7.47) with (7.44), we nd that the imaginary part of the susceptibility contributed by the resonant transitions between energy levels j1i and j2i is
00res
nc
N 1 a N 2 e :
(7.48)
The real part 0res of the resonant susceptibility can be found through the KramersKronig
relations given in (2.53).
As discussed in Sections 2.1 and 2.3, a medium causes an optical loss if 00 > 0, and it
provides an optical gain if 00 < 0. It is also clear from (7.47) that there is a net power loss from
the optical eld due to absorption by the medium if 00res > 0, but there is a net power gain for
the optical eld if 00res < 0. By comparing (7.48) with (7.45) and (7.46), we nd that the
medium has an absorption coefcient given by
244
00

nc res
(7.49)
in the case of normal population distribution when 00res > 0, whereas it has a gain coefcient
given by
g
00

nc res
(7.50)
in the case of population inversion so that 00res < 0.

Note that the material susceptibility characterizes the response of a material to the excitation
of an electromagnetic eld. Therefore, the magnitude of the resonant susceptibility 00res only
accounts for the contributions from the induced processes of absorption and stimulated
emission, and not that from the process of spontaneous emission. Spontaneous emission causes
natural broadening of the spectral width of 00res , as discussed in Section 7.1. The resonant
susceptibility contributed by the induced transitions between two energy levels is proportional
to the population difference between the two levels, but the power density of the optical
radiation due to spontaneous emission is a function of the population density in the upper
energy level alone.
The coefcients and g respectively characterize the attenuation and growth of the optical
intensity per unit length traveled by the optical wave in a medium. The intensity of a
monochromatic plane wave at the resonance frequency varies with distance along its propagation direction, taken to be the z direction, as
dI
I
dz
(7.51)
in the case of optical attenuation when 00res > 0, and

dI
gI
dz
(7.52)
in the case of optical amplication when 00res < 0.

EXAMPLE 7.6
What is the imaginary part 00res of the resonant susceptibility, at 1:064 m, of the pumped
Nd:YAG rod considered in Example 7.5? The refractive index of Nd:YAG is n 1:82. The rod
has a length of l 5 cm. If a beam at 1:064 m that has a power of Pin 1 mW is sent into
one end of the Nd:YAG rod uniformly over the cross-sectional area of the rod, what is the
optical power coming out at the other end?
Solution:
From Example 7.5, the gain coefcient at 1:064 m for the pumped Nd:YAG rod is
g 186 m1 . Using (7.50), we nd the imaginary part of the resonant susceptibility at
1:064 m:
Problems
00res
245
nc
n
1:82 1:064 106
186 5:73 105 :
g g
2
For uniform illumination, (7.52) can be written in terms of the optical power to nd the output
power as
2
dP
gP ) Pout Pin egl 1 103 e186510 W 10:9 W:
dz
Problems
7.1.1 A ruby laser rod is a sapphire crystal doped with active Cr3 ions. The upper level j2i of
the transition for the ruby emission wavelength of 694:3 nm is the E level of the Cr3
ion, and the lower level j1i is the 4 A2 ground level. The population in the E level relaxes
only to the 4 A2 ground level, and the relaxation is purely radiative. The upper level
lifetime is 2 3 ms. At room temperature, this emission line has a predominantly
homogeneous linewidth of 330 GHz.
(a) Find the radiative, nonradiative, and total relaxation rates for the upper and lower
levels, j2i and j1i, respectively.
(b) Find the natural linewidth and the lifetime-broadened linewidth for the 694:3 nm
emission line. If no other mechanisms further broaden this line, what are its lineshape
and linewidth?
(c) The homogeneous broadening at room temperature is contributed by dephasing due
to phonon collisions. What is the dephasing rate dephase
?
21
7.1.2 Ti:sapphire and Cr:LiSAF are solid-state laser media. Ti:sapphire contains active Ti3
ions doped in a sapphire crystal, and Cr:LiSAF contains active Cr3 ions doped in a
LiSAF crystal. The uorescence lifetime of Ti:sapphire is 2 3:2 s, and that of Cr:
LiSAF is 2 67 s. For both systems, the lower level j1i is the ground level. Both
media have very broad spontaneous linewidths that are predominantly homogeneously
broadened, with 100 THz for Ti:sapphire and 83 THz for Cr:LiSAF. What
are the expected lifetime-broadened homogeneous linewidths of these two media?
Explain why these two media have such broad homogeneous linewidths.
7.1.3 The CO2 laser gain medium contains the gas mixture of CO2 , N2 , and He with about the
same fractional ratio of CO2 and N2 , and somewhat more He. The 10:6 m emission
takes place between two vibrational levels of the CO2 molecule. The upper level j2i has a
radiative lifetime of rad
2 4 s, and the lower level j1i has a radiative lifetime of
rad
1 200 ms. The N2 molecules help to pump the CO2 molecules to the upper level
j2i, while the He atoms help to de-excite the N2 and CO2 molecules back to their
respective ground levels. The collisions of the CO2 molecules with the N2 molecules
and the He atoms change the lifetimes 2 of the upper level and 1 of the lower level by
inducing nonradiative relaxations from these levels. As a result, 2 and 1 depend on the
246
pressure and temperature of the gas mixture. The working temperature of a CO2 laser
ranges from 400 K to 700 K. The working gas pressure varies from below 50 torr to
760 torr for different CO2 lasers.
(a) Find the radiative relaxation rates for the upper and lower levels, j2i and j1i,
respectively. What is the natural linewidth of the emission line?
(b) The molecular mass number of CO2 is 44. Find the range of the Doppler-broadened
linewidth for the CO2 lasers.
(c) Consider a CO2 laser medium of a relatively low pressure working at T 400 K,
which has 2 10 s and 1 1 s. Find the nonradiative and total relaxation rates
for the upper and lower levels, j2i and j1i, respectively. What are the homogeneously and inhomogeneously broadened linewidths of the emission line? What are
the lineshape and the total linewidth? Is it homogeneously or inhomogeneously
broadened?
(d) Consider a CO2 laser medium of a high pressure working at T 700 K, which
has 2 100 ns and 1 1 ns. Find the nonradiative and total relaxation rates for
the upper and lower levels, j2i and j1i, respectively. What are the homogeneously and inhomogeneously broadened linewidths of the emission line? What
are the lineshape and the total linewidth? Is it homogeneously or inhomogeneously broadened?
7.1.4 The argon-ion laser has two emission lines at 488 nm and 514:5 nm. Both lines are
almost entirely broadened by Doppler broadening at the typical operating temperature of
T 1200 C. The Ar atom has an atomic mass number of 40. Find the linewidths and the
lineshapes of the two emission lines, respectively.
7.2.1 A cylindrical ruby rod, which is a sapphire crystal doped with active Cr3 ions, has a
length of l 6 cm and a diameter of d 5 mm. The Cr3 ions has a total concentration
of N t 1:58 1019 cm3 . The upper level j2i of the transition for the ruby emission
wavelength of 694:3 nm relaxes only radiatively through this emission line with a
3
lifetime of 2 rad
ions
2 3 ms. The rod is uniformly pumped such that 50% of the Cr
are excited to the upper level and then left to relax spontaneously.
(a) Find the spontaneous radiative lifetime for the transition of this emission line. What is
the decay time of the spontaneous emission?
(b) What are the optical energy and the power of the spontaneous emission?
7.2.2 Two emission lines have exactly the same wavelength and the same linewidth, but one
has a Lorentzian lineshape while the other has a Gaussian lineshape. If the optical
transitions for both emission lines have the same spontaneous lifetime and the two media
have the same refractive index, do they have the same peak emission cross section? If
they do not have the same peak emission cross section, which one has a larger cross
section? What is the difference?
7.2.3 Two emission lines have exactly the same center wavelength, the same linewidth, the
same peak emission cross section, and they take place in two media that have the same
refractive index, but one has a Lorentzian lineshape and the other has a Gaussian lineshape. What is the possible parameter that has different values for these two transitions?
Problems
247
7.2.4 Are the emission cross section and the absorption cross section of the same spectral line
associated with the transitions between the same pair of energy levels necessarily the
same? Explain.
7.2.5 The upper level j2i of the transition for the ruby emission wavelength of 694:3 nm
is the E level of the active Cr 3 ions doped in the ruby crystal, which has a degeneracy
of g2 2, and the lower level j1i is the 4 A2 ground level, which has a degeneracy of
g1 4. The population in the E level relaxes radiatively only through this emission line
to the 4 A2 ground level with 2 rad
2 3 ms. At room temperature, this emission line
has a homogeneous linewidth of 330 GHz. The refractive index of the ruby
crystal is n 1:76. Find the peak emission and absorption cross sections for this
spectral line.
7.2.6 The 510:5 nm emission line of the copper vapor laser has a linewidth of 2:3 GHz,
which is almost entirely caused by Doppler broadening, and a spontaneous radiative
lifetime of sp 500 ns. The refractive index of the low-pressure gaseous medium is
n 1. Find the peak emission cross section of this line.
7.3.1 A large absorption cross section of Ti:sapphire appears at the wavelength of a 490 nm
with a a 6:4 1024 m2 , while e a 3 1028 m2 . The peak emission cross
section appears at the wavelength of e 795 nm with e e 3:4 1023 m2 , while
a e 8 1026 m2 . The lower level is the ground level. A Ti:sapphire rod that is not
pumped is found to have an absorption coefcient of a 200 m1 at a 490 nm.
(a) Find the total doping concentration N t of the active Ti3 ions in this rod.
(b) If a gain coefcient of ge 20 m1 is desired at e 795 nm, what percent of the
Ti3 ions have to be excited to the upper level?
7.3.2 Ti:sapphire has a refractive index of n 1:76. A Ti:sapphire rod has a length of
l 10 cm.
(a) When it is not pumped, it has an absorption coefcient of a 200 m1 at
a 490 nm. Find the imaginary part 00res of the resonant susceptibility at this
wavelength. If a beam that has a power of Pin a 1 W at a 490 nm is sent into
the rod from one end, what is the output power at the other end? How much of the
power is absorbed?
(b) It is pumped so that it has a gain coefcient of ge 20 m1 at e 795 nm. Find
the imaginary part 00res of the resonant susceptibility at this wavelength. If a beam that
has a power of Pin e 1 mW at e 795 nm is sent into the rod from one end,
what is the output power at the other end? How much of the power is emitted through
stimulated emission?
7.3.3 Because the lower level of the HeNe emission line at 632:8 nm is not the ground
level, an unexcited Ne atom does not absorb light at this wavelength. The emission cross
section of this emission line is e 3 1017 m2 . An optical beam at 632:8 nm is
sent through a uniformly pumped HeNe tube that has a length of l 1 m. If the output
power is 120% of the input power, what is the population density of the excited Ne atoms
in the upper level of the emission line?
248
7.3.4 An Er:ber is doped with an Er3 ion concentration of N t 2:2 1024 m3 . It is found
to have an absorption cross section of a 5:7 1025 m2 and an emission cross section
of e 7:9 1025 m2 at the 1:53 m wavelength. The lower level is the ground
level. Assume uniform pumping throughout the ber. Assume also that all Er3 ions are
distributed only between the two levels of the 1:53 m transition.
(a) What is its intrinsic absorption coefcient 0 at this wavelength when the Er:ber is
not pumped?
(b) What percent of the Er3 ions have to be pumped to the upper level for the ber to be
transparent with g 0?
(c) What percent of the Er3 ions have to be pumped to the upper level for a gain
coefcient of g 0:2 m1 ?
(d) What percent of the Er3 ions have to be pumped to the upper level for a gain
coefcient of g 0 ?
(e) What is the maximum gain coefcient g max when all Er3 ions are pumped to the
upper level? Compare it to the intrinsic absorption coefcient 0 , which is the
maximum value of the absorption coefcient.
Bibliography
Rosencher, E. and Vinter, B., Optoelectronics. Cambridge: Cambridge University Press, 2002.
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
8 - Optical Amplification pp. 249-273
8
8.1
Optical Amplication
POPULATION RATE EQUATIONS
..............................................................................................................
From the discussion in the preceding chapter, it is clear that population inversion is the basic
condition for an optical gain. For any system in its normal state in thermal equilibrium, a lowenergy level is always more populated than a high-energy level, hence there is no population
inversion. Population inversion in a system can only be accomplished through a process called
pumping by actively exciting the atoms in a low-energy level to a high-energy level. If left
alone, the atoms in a system relax to thermal equilibrium. Therefore, population inversion is a
nonequilibrium state that cannot be sustained without active pumping. To keep a constant
optical gain, continuous pumping is required to maintain population inversion. This condition is
clearly consistent with the law of conservation of energy: amplication of an optical wave leads
to an increase in optical energy, which is possible only if the required energy is supplied by a
source.
Pumping is the process that supplies energy to the gain medium for the amplication of an
optical wave. There are many different pumping techniques, including optical excitation,
electric current injection, electric discharge, chemical reaction, and excitation with particle
beams. The use of a specic pumping technique depends on the properties of the gain medium
being pumped. The lasers and optical ampliers of particular interest in photonic systems are
made of either dielectric solid-state media doped with active ions, such as Nd:YAG and Er:
glass ber, or direct-gap semiconductors, such as GaAs and InP. For a dielectric gain medium,
the most commonly used pumping technique is optical pumping using either an incoherent light
source, such as a ashlamp or a light-emitting diode, or a coherent light source from another
laser. A semiconductor gain medium can also be optically pumped, but it is usually pumped by
electric current injection. In this section, we consider the general conditions for pumping to
achieve population inversion. Detailed pumping mechanisms and physical setups are not
addressed here because they depend on the specic gain medium used in a particular
application.
The net rate of increase of the population density in a given energy level is described by a rate
equation. As we shall see below, pumping for population inversion in any practical gain
medium always requires the participation of more than two energy levels. In general, a rate
equation has to be written for each energy level that is involved in the process. For simplicity
but without loss of validity, however, we shall explicitly write down only the rate equations for
the two energy levels, j2i and j1i, that are directly associated with the resonant transition of
interest. We are not interested in the population densities of other energy levels but only in how
they affect N 2 and N 1 .
250
Optical Amplication
In the presence of a monochromatic optical wave that has an intensity of I at a frequency of v,

the rate equations that govern the temporal evolution of N 2 and N 1 are
dN 2
N2
I
N 2 e N 1 a ,
R2
dt
2 hv
(8.1)
dN 1
N1 N2
I
R1
N 2 e N 1 a ,
dt
1 21 hv
(8.2)
where R2 and R1 are the total rates of pumping into energy levels j2i and j1i, respectively, and
2 and 1 are the uorescence lifetimes of levels j2i and j1i, respectively. The total rate of
population relaxation, including radiative and nonradiative spontaneous relaxations, from level
j2i to level j1i is 1
21 . Because it is possible for the population in level j2i to also relax to other
1
energy levels, the total population relaxation rate of level j2i is 1
2 21 . Therefore, in
general, we have
2 21 sp :
(8.3)
1
Note that 1
21 is not the same as 21 dened in (7.9): 21 is purely the rate of population
relaxation from level j2i to level j1i, whereas 21 is the rate of phase relaxation of the
polarization associated with the transition between these two levels. For an optical gain
medium, level j2i is known as the upper laser level, and level j1i is known as the lower laser
level. The uorescence lifetime 2 of the upper laser level is an important parameter that
determines the effectiveness of a gain medium. Generally speaking, for a gain medium to be
useful, the upper laser level has to be a metastable state that has a relatively large 2 .
To account for the difference between the emission cross section and the absorption cross
section, the effective population inversion can be more accurately dened as
N N2
a
N 1:
e
(8.4)
With this denition for the effective population inversion, the gain coefcient is simply
g e N :
(8.5)
This relation is also valid for nding the absorption coefcient. A positive gain coefcient
g > 0 is found when the system reaches effective population inversion so that N > 0; it has a
negative gain coefcient, i.e., a positive absorption coefcient, g > 0 when effective
population inversion is not accomplished so that N < 0.
For the different systems discussed in the following section, the two rate equations given in
(8.1) and (8.2) for N 2 and N 1 can be combined into one equation for the effective population
inversion N:
dN
N
I
R e N,
dt
2
hv
(8.6)
where R is the effective pumping rate for population inversion and

1
a
1
e
(8.7)
251
is the bottleneck factor that characterizes the effectiveness of pumping a system for population
inversion. It is more difcult to reach population inversion in a system that has a larger value of
. Note that the detailed form of the effective pumping rate R depends on the pumping
mechanism and the pumping scheme. It can be a function of the effective population inversion
N, as in the situation when the gain medium contains a xed density of active atoms or
molecules. In this case, the pumping rate R cannot be generally taken as an independent
external parameter. However, it is possible in a different situation that the pumping rate can
be taken as an independent external parameter, such as in the case of a semiconductor gain
medium that is pumped by current injection where the pumping rate is determined by the
injection current. In the following section, we consider the case when a gain medium contains a
xed, nite concentration of active atoms or molecules so that the pumping rate R is a function
of the effective population inversion N.
EXAMPLE 8.1
A Nd:YAG crystal is doped with 1 at.% of Nd3 ions for a concentration of N t
1:38 1026 m3 . For its 1:064 m laser line, the emission cross section is found to be e
4:5 1023 m2 and the absorption cross section is a 0 because the lower laser level of this
laser line is effectively empty all the time. A ruby crystal is doped with 0.05 wt.% of Cr3 ions for
a concentration of N t 1:58 1025 m3 . For its 694:3 nm laser line, the emission cross
section is found to be e 1:34 1024 m2 and the absorption cross section is
a 1:25 1024 m2 . The variations in the measured emission and absorption cross sections
of these gain media are caused by the population ratios in the degenerate states of each laser level,
which vary with doping and temperature. Find the bottleneck factors for these two laser media.
Solution:
The bottleneck factor of this Nd:YAG crystal at 1:064 m is
1
a
0
1
1:
e
4:5 1023
The bottleneck factor of this ruby crystal at 694:3 nm is

1
a
1:25 1024
1
1:93:
e
1:34 1024
The 1:064 m laser line of Nd:YAG has the smallest possible bottleneck factor of 1
because a 0. The 694:3 nm laser line of ruby has a bottleneck factor of 1:93, which
is close to 2, because a is comparable to e .
8.2
POPULATION INVERSION
..............................................................................................................
Population inversion between the upper laser level j2i of a degeneracy g2 and the lower laser
level j1i of a degeneracy g1 in a medium is generally dened as
252
Optical Amplication
N2 N1
g
g
>
so that N 1 < 1 N 2 and N 2 > 2 N 1 :
g2
g1
g2
g1
(8.8)
According to (7.45) and (7.46), this condition makes v < 0 and g v > 0 so that the medium
shows a positive optical gain. However, in many systems, the degenerate states in level j1i or
j2i, or both, are split into closely spaced sublevels to form small energy bands. When the
energy spread of the sublevels in a laser level is sufciently large, the population in the level
can be distributed unevenly so that (7.39) is not valid, i.e., a v 6 g2 =g1 e v. In this
situation, the second equal sign in (7.45) and (7.46) is not valid though the rst equal sign is
still valid:

g1
(8.9)
v N 1 a v N 2 e v 6 N 1 N 2 a v
g2
and

g v N 2 e v N 1 a v 6

g2
N 2 N 1 e v:
g1
(8.10)
For this reason, when the condition for population inversion given in (8.8) is achieved in a
medium, we might nd a v g2 =g1 e v for an optical gain at an optical frequency v while
at the same time we might nd a v0 > g2 =g1 e v0 for an optical loss at another frequency
v0 . Therefore, the population inversion condition in (8.8) does not guarantee an optical gain at a
particular optical frequency v in the case when the population in level j1i or j2i is distributed
unevenly among its sublevels so that a v 6 g2 =g1 e v.
What really matters to an optical wave at a given frequency is the optical gain at that specic
frequency. For this reason, in the following discussion, we shall consider, instead of the
condition in (8.8), the condition that guarantees an optical gain at the frequency v,
g v N 2 e v N 1 a v N e v > 0,
(8.11)
as the effective condition for population inversion as far as an optical signal at the frequency v is
concerned. Clearly, by dening the effective population inversion N as in (8.4), the effective
condition for population inversion is simply N > 0. This population inversion condition can be
reached even when N 2 < N 1 in the case when a < e . On the other hand, if a > e , it is
possible that N 2 > N 1 but N 2 is not sufciently large so that N < 0 and effective population
inversion for an optical gain is not reached.
The pumping requirement for the condition in (8.11) to be satised depends on the properties of
a medium. For atomic and molecular media, there are three different basic systems. Each has a
different pumping requirement to reach effective population inversion for an optical gain. The
pumping requirement can be found by solving the coupled rate equations given in (8.1) and (8.2).
EXAMPLE 8.2
Use the parameters given in Example 8.1 to nd the effective population inversion required to
have a gain coefcient of g 10 m1 for the 1:064 m laser line of Nd:YAG and that
required for the 694:3 nm laser line of ruby.
253
Solution:
For the 1:064 m laser line of Nd:YAG, e 4:5 1023 m2 . Therefore, the required
effective population inversion is
g
10
m3 2:22 1023 m3 :

23
e 4:5 10
For the 694:3 nm laser line of ruby, e 1:34 1024 m2 . Therefore, the required effective population inversion is
N
g
10
m3 7:46 1024 m3 :

e 1:34 1024
For the same gain coefcient, the population inversion required for the ruby laser line is about
34 times that required for the Nd:YAG laser line because the emission cross section of the
Nd:YAG laser line is about 34 times that of the ruby laser line.
N
8.2.1 Two-Level System

When the only energy levels involved in the pumping and the relaxation processes are the upper
and lower laser levels, j2i and j1i, the system can be considered as a two-level system, as
shown in Fig. 8.1. In such a system, level j1i is the ground level, which has 1 , and level
j2i relaxes only to level j1i, so that 21 2 . The total population density is N t N 1 N 2 .
While a pumping mechanism excites atoms from the lower laser level to the upper laser level
of a two-level system, the same pump also stimulates atoms in the upper laser level to relax to
the lower laser level. Therefore, irrespective of the specic pumping technique used, it is
always true that R2 R1 W p12 N 1 W p21 N 2 , where W p12 and W p21 are the pumping transition
probability rates, or simply the pumping rates, from j1i to j2i and from j2i to j1i, respectively.
Under these conditions, (8.1) and (8.2) are equivalent to each other. The upward and downward
pumping rates are not independent of each other but are directly proportional to each other
because both are associated with the interaction between the same pump source and a given pair
of energy levels. We take the upward pumping rate to be W p12 W p and the downward
Figure 8.1 (a) Pumping scheme of a true two-level system. (b) Pumping scheme of a quasi-two-level system.
254
Optical Amplication
pumping rate to be W p21 pW p , where p is a constant that depends on the detailed characteristics of the two-level atomic system and the pump source. In the steady state when
dN 2 =dt dN 1 =dt 0 , we then nd that
g N 2e N 1a
W p 2 e p a a
Nt:
1 1 pW p 2 I 2 =hv e a
Using the relation in (7.43), we nd that, in the case of optical pumping,

W p21 pe e p
p p p ,
W 12 a a p
(8.12)
(8.13)
where pa and pe are the absorption and emission cross sections, respectively, at the pump
wavelength.
In a true two-level system, shown in Fig. 8.1(a), the energy levels j2i and j1i can respectively
be degenerate with degeneracies g2 and g1 , but the population density in each level is evenly
distributed among the degenerate states in the level. In this situation, p pe = pa
g1 =g2 e = a . Then, we nd from (8.12) that
g N 2e N 1a
a

N t < 0:
1 I 2 =hv W p 2 = a e a
(8.14)
No matter how a true two-level system is pumped, it is clearly not possible to achieve population
inversion for an optical gain in the steady state. This situation can be understood by considering
the fact that the pump for a two-level system has to be in resonance with the transition between the
two levels, thus simultaneously inducing downward and upward transitions. In the steady state,
the two-level system reaches thermal equilibrium with the pump at a nite temperature, resulting
in a Boltzmann population distribution of the form given in (7.28) without population inversion.
As discussed above and illustrated in Fig. 8.1(b), however, in many systems an energy level
is actually split into a band of closely spaced, but not exactly degenerate, sublevels with its
population density unevenly distributed among these sublevels. This type of system is not a true
two-level system, but is known as a quasi-two-level system, if either or both of the two levels
are split in such a manner. By properly pumping a quasi-two-level system, it is possible to reach
the needed population inversion in the steady state for an optical gain at a particular laser
frequency v because the ratio p pe = pa at the pump frequency vp can now be made different
from the ratio e = a at the laser frequency v due to the uneven population distribution among
the sublevels within an energy level. From (8.12), we nd that the pumping requirements for a
quasi-two-level system to have a steady-state optical gain are
p
pe e
a
and W p >
:
p <
2 e p a
a a
(8.15)
Because the absorption spectrum is generally shifted to the short-wavelength side of the
emission spectrum, these conditions can be satised by pumping sufciently strongly at a
higher transition energy than the photon energy at the peak of the emission spectrum. In the
case of optical pumping, this condition means that the pump wavelength has to be shorter than
the emission wavelength. Figure 8.1(b) illustrates such a pumping scheme for a quasi-two-level
255
system. Indeed, many laser gain media, including laser dyes, semiconductor gain media, and
vibronic solid-state gain media, are often pumped as a quasi-two-level system.
8.2.2 Three-Level System

Population inversion in the steady state is possible for a system that has three energy levels
involved in the process. Figure 8.2 shows the energy-level diagram of an idealized three-level
system. The lower laser level j1i is the ground level, E 1 E 0 , or is very close to the ground
level, within an energy separation of E10 E 1 E0 k B T from the ground level, so that it is
initially populated. The atoms are pumped to an energy level j3i above the upper laser level j2i.
An effective three-level system satises the following conditions.
1. Population relaxation from level j3i to level j2i is very fast and efcient, ideally
2 32 3 , so that the atoms excited by the pump quickly end up in level j2i.
2. Level j3i lies sufciently high above level j2i with E32 E 3 E2 k B T so that the
population in level j2i cannot be thermally excited back to level j3i.
3. The lower laser level j1i is the ground level, or its population relaxes very slowly if it is not the
ground level, so that 1 . Furthermore, level j2i relaxes mostly to level j1i so that 21 2 .
Under these conditions, R2 W p N 1 , R1 W p N 1 , and N 1 N 2 N t . The parameter W p is
the effective pumping rate of exciting atoms in the ground level to eventually reach the upper
laser level. It is proportional to the pump power.
In the steady state under constant pumping, W p is a constant and dN 2 =dt dN 1 =dt 0.
With these conditions, we nd that
g N 2 e N 1 a
W p2e a
Nt:
1 W p 2 I 2 =hv e a
(8.16)
Therefore, the pumping requirement for a positive optical gain under steady-state population
inversion is
Wp >
a
:
2e
(8.17)
This condition sets the minimum pumping requirement for a three-level system to have a
positive optical gain. This requirement can be understood by considering the fact that almost
Figure 8.2 Energy levels of a three-level
system.
256
Optical Amplication
all of the population initially resides in the lower laser level j1i. To achieve effective population
inversion, the pump has to be strong enough to sufciently depopulate level j1i while the
system has to be able to keep the excited atoms in level j2i. In the case when a e , for a
bottleneck factor of 2, no population inversion occurs before at least half of the total
population is transferred from level j1i to level j2i. This is the bottleneck effect that limits the
energy conversion efciency of a three-level laser system as compared to a quasi-two-level or
four-level system.
8.2.3 Four-Level System

A four-level system, shown schematically in Fig. 8.3, is more efcient than a three-level
system. A four-level system differs from a three-level system in that the lower laser level j1i
lies sufciently high above the ground level j0i with E10 E 1 E0 k B T so that in thermal
equilibrium the population in level j1i is negligibly small compared to that in level j0i.
Pumping takes place from level j0i to level j3i.
An effective four-level system also has to satisfy the conditions concerning levels j3i and j2i
discussed above for an effective three-level system. In addition, it has to satisfy the condition
that the population in level j1i relaxes very quickly to the ground level, ideally 1 10 2 ,
so that level j1i remains relatively unpopulated in comparison to level j2i when the system is
pumped. Under these conditions, N 1 0 and R2 W p N t N 2 , where the effective pumping
rate W p is again proportional to the pump power. Because N 1 0, (8.2) can be ignored. For a
four-level system, we can also take a 0, for a bottleneck factor of 1, because its
absorption coefcient at the laser wavelength is zero even when it is not pumped.
In the steady state with a constant W p , we nd by taking dN 2 =dt 0 for (8.1) and taking
a 0 that
g N 2e
W p2e
Nt:
1 W p 2 e I=hv
(8.18)
This result indicates that there is no minimum pumping requirement for an ideal four-level
system that satises the conditions discussed above. No bottleneck effect limits an ideal fourlevel system because level j1i is initially empty in such a system. Real systems are rarely ideal,
but a practical four-level system is still much more efcient than a three-level system.
Figure 8.3 Energy levels of a four-level
system.
257
8.2.4 Transparency
When the gain coefcient is zero, g 0, the medium becomes transparent, or bleached, to the
optical signal, neither absorbing nor amplifying it. An ideal four-level system is transparent at
no pumping. A quasi-two-level or three-level system reaches transparency, or the bleached
condition, at the transparency pumping rate:
W trp
a
1
,
2 e p a 2 1 p 1
(8.19)
where is the bottleneck factor dened in (8.7). This relation is valid for all systems though it is
obtained for a two-level or three-level system. For a four-level system, we simply nd from
(8.19) that W trp 0 because a 0 and 1 for the system. For a system to have an optical
gain, the pumping rate has to be higher than the transparency pumping rate: W p > W trp . For a
four-level system, any pumping leads to a gain because it is always true that W p > W trp 0 as
long as the system is pumped. For a two-level or three-level system, which has a 6 0 so that
> 1, it is possible for the system to have no optical gain but optical attenuation when it is not
sufciently pumped such that W trp > W p > 0.
The relation in (8.19) gives the necessary pumping effort for a system to reach transparency
and then an optical gain above it. Another useful measure is the population density N 2 that has
to be pumped to the upper laser level in order for a system to have an optical gain. For a twolevel or three-level system, N 1 N 2 N t . By simultaneously solving N 1 N 2 N t and
N 2 e N 1 a g, the population of the upper laser level is found:

aN t g
1
N
N2
1 Nt :
(8.20)
e a
Though this relation is obtained by using N 1 N 2 N t , which is not valid for a four-level
system, the relation is still valid for a four-level system because it reduces to N 2 g= e in the
case of a four-level system, for which a 0. Therefore, this relation is valid for all systems.
The relation given in (8.20) is valid for any valid value of g, which can be positive, zero, or
negative. In the case of a four-level system, it is always true that g 0. In the case of a quasi-twolevel or three-level system, g < 0 when the medium is not sufciently pumped to reach
transparency. Because the maximum value of the absorption coefcient for a two-level or threelevel system is 0 a N t while 0 g 0 e = a , we nd from (8.20) that N 2 0 for
any values of g, including g < 0 when the system has a positive absorption coefcient of
g > 0 for optical attenuation, g 0 when the system neither attenuates nor amplies the optical
signal, and g > 0 when the system has a positive gain coefcient for optical amplication.
Because g 0 and N 0 at transparency, the transparency population density for the upper
laser level is obtained from (8.20) as

a
1
tr
N2
N t 1 Nt:
(8.21)
e a
Population inversion with N > 0 for a positive optical gain of g > 0 is reached when N 2 > N tr2
so that the system is above transparency. Clearly, the bottleneck factor gives a measure of the
ease or difculty in reaching the transparency point. For a four-level system, such as the
258
Optical Amplication
Nd:YAG laser, 1 because a 0; thus N tr2 0. In this situation, any population density N 2
pumped to the upper laser level contributes to an optical gain even when most of the active
atoms remain in the ground level, which is not the lower laser level of the system. For a twolevel or three-level system, > 1; thus N tr2 > 0. In this situation, a population density of
N 2 > N tr2 > 0 in the upper laser level is required for the system to have an optical gain, and
it increases with the value of . In many three-level systems, such as the ruby laser, the value of
is close to 2; in this situation, about half of all active atoms have to be pumped to the upper
laser level before the system can have any optical gain. In some quasi-two-level systems,
however, the value of is close to 1 though larger than 1; then it is relatively easy, though not
as easy as for a four-level system, for the system to reach population inversion for a positive
optical gain.
EXAMPLE 8.3
Consider the Nd:YAG and ruby crystals that have the parameters given in Example 8.1. Find the
population density of the upper laser level required for the Nd:YAG crystal to reach transparency at its 1:064 m laser line and that required for the ruby crystal to reach transparency at
its 694:3 nm laser line. What percent of all active ions are excited in each case?
Solution:
For the Nd:YAG crystal, we have 1 and N t 1:38 1026 m3 from Example 8.1. The
population density of the upper laser level required for the Nd:YAG crystal to reach transparency at its 1:064 m laser line is found using (8.21) to be

1
1
tr
N2 1 Nt 1
1:38 1026 m3 0:
1
The percentage of all active ions that are excited to the upper laser level is 0%.
For the ruby crystal, we have 1:93 and N t 1:58 1025 m3 from Example 8.1. The
population density of the upper laser level required for the ruby crystal to reach transparency at
its 694:3 nm laser line is found using (8.21) to be

1
1
tr
N2 1 Nt 1
1:58 1025 m3 7:61 1024 m3 :
1:93
The percentage of all active ions that are excited to the upper laser level is
N tr2 7:61 1024
48%:
N t 1:58 1025
We nd that no active ions have to be excited for the Nd:YAG crystal to reach the
transparency point because it is a four-level system that has a bottleneck factor of 1. By
comparison, as many as 48% of all active ions have to be excited to the upper laser level for the
ruby crystal to reach transparency because it is a three-level system that has a large bottleneck
factor of 1:93.
8.3 Optical Gain
8.3
259
OPTICAL GAIN
..............................................................................................................
When the condition in (8.11) is satised for a system, an optical gain coefcient at a specic
optical frequency v can be found as g v N 2 e v N 1 a v. The optical gain coefcient is a
function of the optical signal intensity, I, as a result of the dependence of N 2 and N 1 on I due to
stimulated emission, which changes the population densities by causing downward transitions
from level j2i to level j1i. This effect causes saturation of the optical gain coefcient by the
optical signal. For all three basic systems discussed above, the optical gain coefcient can be
expressed as a function of the optical signal intensity I:
g
g0
,
1 I=I sat
(8.22)
where g 0 is the unsaturated gain coefcient, which is independent of the optical signal
intensity, and I sat is the saturation intensity of a medium, which can be generally
expressed as
I sat
hv
:
se
(8.23)
The time constant s is an effective saturation lifetime of the population inversion. It can be
considered as an effective decay time constant for the optical gain coefcient through the
relaxation of the effective population inversion. Both g 0 and s are functions of the intrinsic
properties of a gain medium, as well as of the pumping rate. They can be found from (8.12),
(8.16), and (8.18) for the quasi-two-level, three-level, and four-level systems, respectively. The
results are summarized below.
Quasi-two-level system:

g0 W pse a N t,
s 2
Three-level system:
1 a = e
:
1 1 pW p 2

g0 W pse a N t,
s 2
1 a = e
:
1 W p2
(8.24)
(8.25)
(8.26)
(8.27)
Four-level system:
g0 W pseN t,
(8.28)
2
:
1 W p2
(8.29)
260
Optical Amplication
The minimum pumping requirement for a medium to have an optical gain is clearly g 0 > 0.
This is the condition for reaching transparency discussed in Section 8.2. For a desired unsaturated gain coefcient of g 0 , the required pumping rate can be found by solving (8.24) and (8.25)
for a quasi-two-level system, (8.26) and (8.27) for a three-level system, and (8.28) and (8.29)
for a four-level system. The results are summarized below.
Quasi-two-level system:
Wp
1
aN t g0
:
2 e p a N t 1 pg 0
(8.30)
Wp
1 aN t g0
:
2 eN t g0
(8.31)
Wp
1
g0
:
2 eN t g0
(8.32)
Three-level system:
Four-level system:
The different forms of unsaturated gain coefcient g 0 and saturation lifetime s found above
for different systems can be expressed in a general form for all systems by using the parameter p
and the bottleneck factor to account for the differences among the systems. Meanwhile, the
required pumping rate for an unsaturated gain coefcient of g 0 can be found expressed in a
general form for all systems. They are given below.
General forms for all systems:

g0 W ps 1 eN t,
(8.33)
,
1 1 pW p 2
(8.34)
1
1 e N t g 0
:
2 1 p 1 e N t 1 pg 0
(8.35)
s 2
Wp
For a quasi-two-level system, p 0 and 1. When using a specic quasi-two-level system, it

is desirable to make p as small as possible by properly choosing the pumping parameters and it
is desirable to make as close to unity as possible by properly choosing the laser emission
wavelength. For a three-level system, p 0 and > 1; the value of is usually close to 2 for
the typical three-level system, but it can be less than 2 or sometimes greater than 2. The large
bottleneck factor makes a three-level system inefcient, as discussed in Section 8.2. For a
four-level system, p 0 and 1, making the system most efcient in pumping for an
optical gain.
In the limit when p ! 0, a quasi-two-level system is identical to a three-level system. In the
limit when p ! 0 and a ! 0 ! 1, a quasi-two-level system behaves like a four-level
system. In the limit when a ! 0 ! 1, a three-level system behaves like a four-level
system.
8.3 Optical Gain
261
EXAMPLE 8.4
The Nd:YAG laser crystal described in Example 8.1 has 2 240 s for its 1:064 m laser line.
The ruby laser crystal described in Example 8.1 has 2 3 ms for its 694:3 nm laser line. (a)
Find the pumping rates for the 1:064 m Nd:YAG laser line to reach transparency and to have an
unsaturated gain coefcient of g 0 10 m1 , respectively. What are the saturation lifetime and the
saturation intensity in each case? (b) Answer the same questions for the 694:3 nm ruby laser line.
Solution:
The two laser media belong to different systems and have different parameters.
(a) The Nd:YAG at 1:064 m is a four-level system with e 4:5 1023 m2 and a 0.
The doping density is N t 1:38 1026 m3 . The photon energy is
hv
1:2398
eV 1:165 eV:
1:064
Using (8.32), (8.29), and (8.23) for a four-level system, we nd the pumping rate, the
saturation lifetime, and the saturation intensity for g 0 0 at transparency to be
W trp 0,
trs 2 240 s,
I trsat
hv
1:165 1:6 1019
W m2 17:3 MW m2 :

6
23
tr
s e 240 10 4:5 10
The parameters for an unsaturated gain coefcient of g 0 10 m1 are

Wp
1
g0
1
10
s1 6:72 s1 ,

2 e N t g 0 240 106 4:5 1023 1:38 1026 10
s
I sat
2
240 106
s 239:6 s,
1 W p 2 1 6:72 240 106
hv
1:165 1:6 1019
W m2 17:3 MW m2 :

s e 239:6 106 4:5 1023
(b) The ruby at 694:3 nm is a three-level system with e 1:34 1024 m2 and a
1:25 1024 m2 . The doping density is N t 1:58 1025 m3 . The photon energy is
hv
1239:8
eV 1:786 eV:
694:3
Using (8.31), (8.27), and (8.23) for a three-level system, we nd the pumping rate, the
saturation lifetime, and the saturation intensity for g 0 0 at transparency to be
W trp
1 a
1
1:25 1024 1

s 311 s1 ,
2 e 3 103 1:34 1024
262
Optical Amplication
trs 2
I trsat
1 a = e
2 3 ms,
1 W trp 2
hv
1:786 1:6 1019
W m2 71:1 MW m2 :

trs e 3 103 1:34 1024
The parameters for an unsaturated gain coefcient of g 0 10 m1 are

Wp
1 aN t g0
1
1:25 1024 1:58 1025 10 1

s 888 s1 ,
2 e N t g 0 3 103 1:34 1024 1:58 1025 10
s 2
I sat
1 a = e
1 1:25=1:34
3 103
s 1:58 ms,
1 W p2
1 888 3 103
hv
1:786 1:6 1019
W m2 139:4 MW m2 :

s e 1:58 103 1:34 1024
EXAMPLE 8.5
The Nd:YAG crystal considered in Example 8.4 can be optically pumped with an absorption
cross section of pa 3:0 1024 m2 at the p 808 nm pump wavelength, whereas the ruby
crystal considered in Example 8.4 can be optically pumped with an absorption cross section of
pa 2:0 1023 m2 at the p 554 nm pump wavelength. Assume a 100% pump quantum
efciency for the following questions. (a) Find the required pump intensities at p 808 nm to
pump the 1:064 m Nd:YAG laser line to transparency and to have an unsaturated gain
coefcient of g 0 10 m1 , respectively. (b) Find the required pump intensities at p 554 nm
to pump the 694:3 nm ruby laser line to transparency and to have an unsaturated gain
coefcient of g 0 10 m1 , respectively.
Solution:
The pumping transition probability rate W p determines the number per second of active atoms
excited by the pump to the upper laser level. If the pump has a pump quantum efciency of p
when N p pump photons are absorbed, only p N p atoms are excited. Thus, the required pump
intensity for a pumping transition probability rate of W p is
Ip
1 hvp
W p:
p pa
With p 1 assumed in this example, we have

Ip
hvp W p
:
pa
(a) For the Nd:YAG crystal, p 808 nm and pa 3:0 1024 m2 . The pump photon energy is
hvp
1239:8
eV:
808
8.3 Optical Gain
263
From Example 8.4, the transparency pumping rate is W trp 0 and the pumping rate for
g 0 10 m1 is W p 6:72 s1 . Therefore, the required pump intensity for transparency is
I trp
hvp W trp
pa
0,
and that for g 0 10 m1 is

Ip
hvp W p 1239:8
6:72
W m2 550 kW m2 :

1:6 1019
p
24
808
a
3:0 10
(b) For the ruby crystal, p 554 nm and pa 2:0 1023 m2 . The pump photon energy is
hvp
1239:8
eV:
554
From Example 8.4, the transparency pumping rate is W trp 311 s1 and the pumping rate for
g 0 10 m1 is W p 888 s1 . Therefore, the required pump intensity for transparency is
I trp
hvp W trp
pa
1239:8
311
W m2 5:57 MW m2 ,
1:6 1019
554
2:0 1023
and that for g 0 10 m1 is

Ip
hvp W p 1239:8
888
W m2 15:9 MW m2 :

1:6 1019
p
554
a
2:0 1023
8.3.1 Unsaturated Gain

The unsaturated gain coefcient g 0 is also known as the small-signal gain coefcient because it
is the gain coefcient of a weak optical signal that does not saturate the gain medium. At
transparency, g 0 0 because g 0. For a four-level system, g 0 > 0 as long as the medium is
pumped because there is no minimum pumping requirement for transparency. For a quasi-twolevel or three-level system, g 0 > 0 only when the pumping level exceeds its minimum pumping
requirement for transparency; below that, the medium has absorption because g 0 < 0.
It can be seen from (8.24)(8.29) that for any system, g 0 increases with pump power less than
linearly because s decreases with the pump power though W p is linearly proportional to the
pump power. This dependence of s on the pump power is caused by the fact that as the pump
excites atoms from the ground state to any excited state to eventually reach the upper laser
level, it depletes the population in the ground level. Consequently, as the pump power
increases, fewer atoms remain available for excitation in the ground level, thus reducing the
differential increase of the effective population inversion with respect to the increase of the
pump power.
264
Optical Amplication
8.3.2 Gain Saturation

The optical gain coefcient is a function of the intensity of the optical wave that propagates in
the gain medium; it decreases as the optical signal intensity increases. According to (8.22), the
optical gain coefcient g is reduced to half of the unsaturated gain coefcient g 0 when the
optical signal intensity reaches the saturation intensity such that I I sat . The smaller the value
of I sat , the easier it is for the gain to be saturated. For a quasi-two-level system,
s 2 1 p a = e at transparency. For a three-level or four-level system, s 2 at transparency. For all three systems, s < 2 when the gain medium is pumped above transparency for a
positive gain coefcient. Therefore, I sat increases as the gain medium is pumped harder for a
larger unsaturated gain coefcient.
EXAMPLE 8.6
The Nd:YAG laser crystal considered in Example 8.4 has a saturation intensity of
I sat 17:3 MW m2 when it is pumped to have an unsaturated gain coefcient of
g 0 10 m1 at 1:064 m. The ruby laser crystal also considered in Example 8.4 has a
saturation intensity of I sat 139:4 MW m2 when it is pumped to have an unsaturated gain
coefcient of g 0 10 m1 at 694:3 nm. Two Gaussian laser beams of the same power
of P 1:5 W at these two wavelengths are both collimated to have the same spot size of
w0 300 m in each crystal. Find the saturated gain coefcient for each crystal when the
beam at the respective wavelength is sent through each crystal.
Solution:
Each Gaussian beam has a cross-sectional area of

2
w20 300 106
A
m2 1:4 107 m2 :
2
2
The peak intensity of each beam is
I
P
1:5
W m2 10:7 MW m2 :
A 1:4 107
For the Nd:YAG laser crystal, the saturated gain coefcient is

g
g0
1 I=I sat
10
m1 6:18 m1 :
10:7
1
17:3
For the ruby laser crystal, the saturated gain coefcient is

g
g0
1 I=I sat
10
m1 9:29 m1 :
10:7
1
139:4
The gain coefcient of the Nd:YAG laser line is more saturated than that of the ruby laser line
because the saturation intensity of the Nd:YAG laser line is lower than that of the ruby laser line.
8.4 Optical Amplication
8.4
265
OPTICAL AMPLIFICATION
..............................................................................................................
Any medium that has an optical gain can be used to amplify an optical signal. Depending on the
physical mechanism that is responsible for the optical gain, there are two different categories of
optical ampliers: nonlinear optical ampliers and laser ampliers. The optical gain of a
nonlinear optical amplier originates from a nonlinear optical process in a nonlinear medium,
whereas the gain of a laser amplier results from the population inversion in a gain medium as
discussed in the preceding section.
Ignoring the effect of noise, the amplication of the intensity, I s , of an optical signal
propagating in the z direction through a laser amplier can be described by
dI s
g 0 z
I s,
gI s
dz
1 I s =I sat
(8.36)
where g 0 z is the unsaturated gain coefcient and I sat is the saturation intensity of the gain
medium, both dened in the preceding section. Here we assume transverse uniformity but consider
the possibility of longitudinal nonuniformity by taking the unsaturated gain coefcient g 0 z to be a
function of z. Such a longitudinally nonuniform gain distribution is a common scenario for an
amplier under longitudinal optical pumping because of pump absorption by the gain medium.
In the following discussion, we assume for simplicity that the signal beam is collimated
throughout the length of the amplier such that its divergence is negligible. This assumption
allows us to express (8.36) in terms of the power, Ps , of the optical signal as
dPs
g 0 z
gPs
Ps ,
dz
1 Ps =Psat
(8.37)
where Psat is the saturation power obtained by integrating I sat over the cross-sectional area of
the signal beam. By integrating (8.37), the following relation is obtained:

z
Ps z
Ps z Ps 0
exp
(8.38)
exp g 0 zdz,
Ps 0
Psat
0
where Ps 0 is the power of the signal beam at z 0. When Ps Psat , the power of the optical
signal grows exponentially with distance. The growth slows down as Ps approaches the value of
Psat . Eventually, the signal grows only linearly with distance when Ps Psat .
The power gain of a signal is dened as
Pout
s
,
(8.39)
Pin
s
out
where Pin
s and Ps are the input and output powers of the signal, respectively. By using the
relation in (8.38) while identifying Pout
and Pin
s
s with Ps l and Ps 0, respectively, for an
amplier that has a length of l, an implicit relation is found for the power gain of the signal:

Pin
s
G G0 exp 1 G
,
(8.40)
Psat
G
where G0 is the unsaturated power gain, or the small-signal power gain. For a single pass
through the amplier, G0 is given by
266
Optical Amplication
Figure 8.4 Gain, normalized to the unsaturated gain as G=G0 , of a laser amplier as a function of the input signal
power, normalized to the saturation power as Pin
s =Psat , for different values of the unsaturated power gain G0 .
l
G0 exp g 0 zdz:
(8.41)
Note that, according to (8.40), G0 G > 1 because g 0 > 0 for an amplier. For a small optical
out
signal such that Pin
s < Ps Psat , the power gain is simply the small-signal power gain so that
G G0 . If the signal power approaches or even exceeds the saturation power of the amplier,
the relation in (8.40) clearly indicates that G < G0 because of gain saturation. In this situation,
the overall gain G can be found by solving (8.40) when the values of Pin
s and Psat , as well as that
of G0 , are given. Figure 8.4 shows the amplier gain as a function of the input signal power for
a few different values of the unsaturated power gain G0 .
EXAMPLE 8.7
A Nd:YAG laser rod and a ruby laser rod with the properties described in the preceding
examples both have a length of l 10 cm and a cross-sectional diameter of d 6 mm. The
refractive index of Nd:YAG is 1.82, and that of ruby is 1.76. Each is uniformly pumped to
have an unsaturated gain coefcient of g 0 10 m1 at its laser wavelength, 1:064 m
for Nd:YAG and 694:3 nm for ruby. The saturation intensities at g 0 10 m1 are
found in Example 8.4 to be I YAG
1:73 MW m2 for the Nd:YAG laser line and
sat
2
I ruby
for the ruby laser line. Two collimated Gaussian signal beams at the
sat 139:4 MW m
two laser wavelengths that have the same spot size of w0 400 m in the rod and the same
power of Pin
s 5 W are respectively sent through the Nd:YAG and ruby rods for amplication.
What are the output signal powers from the Nd:YAG and ruby ampliers, respectively?
Solution:
The primary difference between the Nd:YAG amplier and the ruby amplier is their different
saturation intensities. Because their signal wavelengths are different, the two Gaussian beams
have different Rayleigh ranges when their spot sizes are the same. With w0 400 m, the
Rayleigh ranges of the two beams are
267

2
nw20 1:82 400 106
zR
m 86 cm for 1:064 m,
1:064 106

2
nw20 1:76 400 106
m 1:27 m for 694:3 nm:

zR
694:3 109
Both Rayleigh ranges are much larger than the l 10 cm length of each rod, and the spot size
of each beam is much smaller than the cross-sectional diameter of each rod. Therefore, each
Gaussian beam can be considered to be collimated throughout each rod with an approximate
beam cross-sectional area of

2
w20 400 106
A
m2 2:51 107 m2 :
2
2
Then, the saturation powers are
6
7
I YAG
W 4:34 W for the Nd:YAG amplifier,
PYAG
sat
sat A 17:3 10 2:51 10
ruby
6
7
W 35 W for the ruby amplifier:
Pruby
sat I sat A 139:4 10 2:51 10
With l 10 cm and a uniform unsaturated gain coefcient of g 0 10 m1 for both rods, both
ampliers have the same unsaturated power gain of
G0 exp g 0 l e1:0 :
Using (8.40), the power gain for an input signal power of Pin
s 5 W can be found for each
amplier:

Pin
5
1:0
s
GYAG G0 exp 1 GYAG YAG e exp 1 GYAG
) GYAG 1:51,
4:34
Psat
"
#

Pin

5
1:0
s
Gruby G0 exp 1 Gruby ruby e exp 1 Gruby
) Gruby 2:27:
35
Psat
Thus, the output signal powers are
in
Pout
s, YAG GYAG Ps 1:51 5 W 7:55 W for the Nd:YAG amplifier,
in
Pout
s, ruby Gruby Ps 2:27 5 W 11:35 W for the ruby amplifier:
8.5
SPONTANEOUS EMISSION
..............................................................................................................
Spontaneous emission occurs whenever the upper laser level of a system is populated, irrespective of the lower-level population. The population of the upper laser level for any system is
given in (8.20):
268
Optical Amplication
N2
aN t g
,
e a
(8.42)
where g > 0 when the system is above transparency with an optical gain, g 0 when the
system is at transparency, and g < 0 when the system is below transparency with an optical
attenuation coefcient of g.
According to the discussion in Section 7.1, the spontaneous emission power is proportional to
N 2 but is independent of N 1 . Therefore, regardless of whether the medium has a gain or a loss,
the spontaneous emission power density, which is dened as the spontaneous emission power
per unit volume of the medium in watts per cubic meter, is
^ sp hv N 2 hv
a N t g ,
P
sp
sp e a
(8.43)
where g can be positive for a medium pumped above transparency, zero for a system at
transparency, or negative for a medium below transparency. For a gain volume of V, the
spontaneous emission power is
^ sp V:
Psp P
(8.44)
The spontaneous emission power density at transparency, which is known as the critical
uorescence power density, is
^ trsp hv N 2 hv
a N t :
P
sp
sp e a
(8.45)
The critical uorescence power for a gain volume of V is

^ trsp V:
Ptrsp P
(8.46)
^ trsp 0 and Ptrsp 0 because a 0 so that it is transparent

For an ideal four-level system, P
^ trsp 6 0 and Ptrsp 6 0
without pumping. For a quasi-two-level system or a three-level system, P
^ trsp and Ptrsp are
because a 6 0. A practical quasi-two-level system usually has a e so that P
^ sp and Psp when the medium is pumped for a positive gain of
respectively much smaller than P
^ sp and Psp
^ trsp and Ptrsp are often respectively comparable to P
g > 0. For a three-level system, P
when the medium is pumped for a positive gain of g > 0 because a and e are of the same
order of magnitude.
When an optical medium is pumped below transparency, it can still emit light through
spontaneous emission as long as N 2 > 0 though N 2 < N tr2 in this situation. Even when an
optical medium is pumped above transparency, spontaneous emission still occurs, and the
power of spontaneous emission can still dominate that of stimulated emission before laser
action takes place. Such spontaneous emission power is the basis of incoherent luminescent
light sources. For example, light-emitting diodes are solid-state light sources that emit spontaneous emission generated by electroluminescence through radiative relaxation of electronhole
pairs that are injected by an electric current.
269
In a laser amplier that amplies an optical signal through stimulated emission, the spontaneous emission is also amplied, resulting in amplied spontaneous emission. Amplied
spontaneous emission is the major source of optical noise for a laser amplier. It is also the
major source of optical noise for a laser oscillator.
EXAMPLE 8.8
Consider the Nd:YAG and ruby crystals that have the characteristics described in the preceding
examples. As found in Example 8.3, the population density of the upper laser level required for
the 1:064 m Nd:YAG laser line to reach transparency is N tr2 0, whereas that required for
the 694:3 nm ruby laser line to reach transparency is N tr2 7:61 1024 m3 . The spontaneous lifetimes are sp 515 s for the Nd:YAG laser line and sp 3 ms for the ruby laser line.
A Nd:YAG laser rod and a ruby laser rod both have a length of l 10 cm and a cross-sectional
diameter of d 6 mm. Find the critical uorescence power density and the critical uorescence
power for each rod.
Solution:
The volume of each rod is

2
2
d
6 103
V
l
10 102 m3 2:83 106 m3 :
2
2
For the Nd:YAG rod, because N tr2 0, both the critical uorescence power density and the
critical uorescence power are zero:
^ trsp 0 and Ptrsp 0:
P
For the ruby rod, N tr2 7:61 1024 m3 , sp 3 ms, and the photon energy is
hv
1239:8
eV 1:786 eV:
694:3
Therefore, the critical uorescence power density and the critical uorescence power for the
ruby rod are, respectively,
hv tr 1:786 1:6 1019
tr
^
N
7:61 1024 W m3 725 MW m3
P sp
3
sp 2
3 10
and
^ trsp V 725 106 2:83 106 W 2:05 kW:
Ptrsp P
270
Optical Amplication
Problems
8.1.1 Show that the rate equation given in (8.6) for the effective population inversion is valid
for all systems if the differences among the systems are accounted for by using the
bottleneck factor dened in (8.7). Show also that the effective pumping rate is
R R2 1
Nt
:
2
(8.47)
Hint: Use (8.20) directly for the relation between the population density of the upper laser
level and the gain coefcient dened in (8.5).
8.1.2 A Ti:sapphire crystal is doped with 0.024 wt.% of Ti2 O3 for a Ti3 ion concentration of
N t 7:9 1024 m3 . At the 800 nm wavelength, it has an emission cross section of
e 3:4 1023 m2 and an absorption cross section of a 8 1026 m2 . Find its
bottleneck factor at this laser wavelength.
8.1.3 An Er:ber is doped with an Er3 ion concentration of N t 2:2 1024 m3 . It has an
absorption cross section of a 5:7 1025 m2 and an emission cross section of e
7:9 1025 m2 at the 1:53 m wavelength. Find its bottleneck factor at this laser
wavelength. What is the effective population inversion for a gain coefcient of g
0:3 m1 at 1:53 m?
8.2.1 Verify the relation given in (8.20) for the population density of the upper laser level for a
gain coefcient of g at an effective population inversion of N.
8.2.2 A Nd:YAG crystal is doped with 1 at.% of Nd3 ions for a concentration of
N t 1:38 1026 m3 . For its 1:064 m laser line, the emission cross section is found
to be e 4:5 1023 m2 and the absorption cross section is a 0 because the lower
laser level of this laser line is effectively empty all the time. A ruby crystal is doped with
0.05 wt.% of Cr3 ions for a concentration of N t 1:58 1025 m3 . For its 694:3 nm
laser line, the emission cross section is found to be e 1:34 1024 m2 and the absorption cross section is a 1:25 1024 m2 . Find the effective population inversion and the
population density of the upper laser level required for the 1:064 m Nd:YAG laser
line to have a gain coefcient of g 6 m1 . Find those values required for the
694:3 nm ruby laser line to have a gain coefcient of g 6 m1 . What percent of
all active ions are excited in each case? Explain the difference between the two media.
e 3:4 1023 m2 and an absorption cross section of a 8 1026 m2 .
(a) Find the population density of the upper laser level required for this Ti:sapphire crystal
to reach transparency at 800 nm. What percent of all active ions are excited?
(b) What is the effective population inversion for a gain coefcient of g 15 m1 at
800 nm? What is the population density of the upper laser level for this effective
population inversion? What percent of all active ions are excited? What percent of the
excited ions effectively contribute to the population inversion?
Problems
271
8.2.4 An Er:ber is doped with an Er3 ion concentration of N t 2:2 1024 m3 . It has an
absorption cross section of a 5:7 1025 m2 and an emission cross section of e
7:9 1015 m2 at the 1:53 m wavelength.

(a) Find the population density of the upper laser level required for this Er:ber to reach
transparency at 1:53 m. What percent of all active ions are excited?
(b) What is the effective population inversion required for a gain coefcient of g
0:3 m1 at 1:53 m? What is the population density of the upper laser level for
this effective population inversion? What percent of all active ions are excited? What
percent of the excited ions effectively contribute to the population inversion?
8.3.1 With a constant upward pumping transition probability rate of W p into the upper laser
level j2i by depleting the population in the lower laser level j1i, and a constant downward
pumping transition probability rate of pW p that depletes the population in the upper level,
the total pumping rate to the upper laser level is R2 W p N 1 pN 2 . Show by using
N 1 N 2 N t and (8.20) that the effective pumping rate found in Problem 8.1.1 can be
expressed in terms of the total population N t and the effective population inversion N as

1
N t 1 pW p N:
(8.48)
R 1 1p W p
2
Use this pumping rate and the rate equation given in (8.6) for the effective population
inversion to show that in the steady state the gain coefcient can be expressed in the form
of (8.22) with the saturation intensity I sat taking the form of (8.23), the unsaturated gain
coefcient g 0 having the form of (8.33), and the saturation lifetime s having the form of
(8.34).
8.3.2 By using (8.33) and (8.34), show that the required pumping probability rate for an
unsaturated gain coefcient of g 0 is that given in (8.35).
8.3.3 By using the general expression in (8.34), nd the saturation lifetime at the transparency
point for all systems.
e 3:4 1023 m2 and an absorption cross section of a 8 1026 m2 . It has an
upper laser level lifetime of 2 3:2 s. It can be optically pumped at the pump wavelength of p 532 nm, where the absorption cross section is pa 7:4 1024 m2 and the
emission cross section is pe 3 1026 m2 . The pump quantum efciency is p 0:9.
(a) Find the pumping rates for this Ti:sapphire to reach transparency and to have an
unsaturated gain coefcient of g 0 15 m1 at 800 nm, respectively. What are
the saturation lifetime and the saturation intensity in each case?
(b) Find the required pump intensities at p 532 nm to pump this Ti:sapphire to
transparency and to have an unsaturated gain coefcient of g 0 15 m1 , respectively.
(c) When this Ti:sapphire is pumped to have an unsaturated gain coefcient of g 0
15 m1 at 800 nm, a collimated Gaussian laser beam at this wavelength that has a
power of P 1 W and a spot size of w0 200 m is sent through this crystal. Find
the saturated gain coefcient.
272
Optical Amplication
8.3.5 An Er:ber is doped with an Er3 ion concentration of N t 2:2 1024 m3 in its core.
This ber is a cylindrical waveguide that has a core radius of a 4:5 m. At the
1:53 m wavelength, the Er:ber has an absorption cross section of a 5:7 1025 m2 ,
an emission cross section of e 7:9 1025 m2 , and an upper laser level lifetime of
2 10 ms. It can be optically pumped as a three-level system at the pump wavelength of
p 980 nm, where the absorption cross section is pa 2:58 1025 m2 . At the signal
wavelength of 1:53 m and the pump wavelength of p 980 nm, the guided signal
and pump waves respectively have effective mode radii of 4:1 m and p 3:3 m
for their intensity proles. The fractions of the signal and pump intensities that overlap
with the core doped with active ions are determined by the connement factors, which are
0:70 and p 0:72, respectively. The pump quantum efciency is p 0:8.
(a) Find the pumping rates for this Er:ber to reach transparency and to have an
unsaturated gain coefcient of g 0 0:3 m1 , respectively, at 1:53 m. What
are the saturation lifetime and the saturation intensity in each case?
(b) Find the required pump intensities at p 980 nm to pump this Er:ber to transparency and to have an unsaturated gain coefcient of g 0 0:3 m1 , respectively.
(c) Find the required pump powers for transparency and for g 0 0:3 m1 by accounting
for the overlap between the guided pump beam and the active core.
(d) When this Er:ber is pumped to have an unsaturated gain coefcient of g 0 0:3 m1
at 1:53 m, a guided laser beam at this wavelength that has a power of P
1 mW is sent through this ber. Find the saturated gain coefcient by accounting for
the overlap between the guided signal beam and the active core.
8.4.1 If the spot sizes of both beams in Example 8.6 are increased to w0 800 m, what is the
output power from each amplier?
8.4.2 A Ti:sapphire laser rod of the characteristics described in Problem 8.3.4 has a length of
l 4 cm and a cross-sectional diameter of d 3 mm. The refractive index of sapphire is
1.76. The laser rod is uniformly pumped to have an unsaturated gain coefcient of g 0
15 m1 at the wavelength of 800 nm. The saturation intensity at g 0 15 m1 is
I sat > 2 GW m2 . A collimated Gaussian signal beam at 800 nm that has a spot size
of w0 300 m in the rod and a power of Pin
s 1 W is sent through the Ti:sapphire
amplier. What is the output signal power from this Ti:sapphire amplier?
8.4.3 An Er:ber amplier of the characteristics described in Problem 8.3.5 has a length of
l 10 m. It is uniformly pumped to have an unsaturated gain coefcient of g 0 0:3 m1 at
its laser wavelength of 1:53 m. After accounting for the overlap between the guided
signal beam and the active core, the saturation power at g 0 0:3 m1 is Psat 1:49 mW. If
a guided signal beam at 1:53 m that has a power of Pin
s 10 W is sent through the
Er:ber amplier, what is the amplied output signal power? What is the output signal
power if the input signal power is increased to Pin
s 1 mW?
8.5.1 A Nd:YAG crystal is doped with a Nd3 concentration of N t 1:38 1026 m3 . For its
1:064 m laser line, the emission cross section is e 4:5 1023 m2 , the absorption cross section is a 0, and the spontaneous lifetime is sp 515 s. A ruby crystal
is doped with a Cr3 concentration of N t 1:58 1025 m3 . For its 694:3 nm laser
Bibliography
273
line, the emission cross section is e 1:34 1024 m2 , the absorption cross section is
a 1:25 1024 m2 , and the spontaneous lifetime is sp 3 ms. The refractive index
of Nd:YAG is 1.82, and that of ruby is 1.76. A Nd:YAG laser rod and a ruby laser rod
both have a length of l 10 cm and a cross-sectional diameter of d 6 mm. Find the
spontaneous emission power density and the spontaneous emission power of each rod
when each is uniformly pumped to have an unsaturated gain coefcient of g 0 10 m1 .
8.5.2 A Ti:sapphire laser rod has a length of l 4 cm and a cross-sectional diameter of
d 3 mm. It is doped with a Ti3 ion concentration of N t 1:0 1025 m3 . At the
800 nm wavelength, it has an emission cross section of e 3:4 1023 m2 and an
absorption cross section of a 8 1026 m2 . Its upper laser level for the 800 nm
emission has a total lifetime of 2 3:2 s and a spontaneous lifetime of sp 3:9 s.
(a) Find the critical uorescence power density and the critical uorescence power of
the rod.
(b) Find the spontaneous emission power density and the spontaneous emission power of
the rod when it is uniformly pumped to have an unsaturated gain coefcient of g 0
15 m1 at 800 nm.
8.5.3 An Er:ber that has a length of l 10 m is doped with an Er3 ion concentration of N t
2:2 1024 m3 in its core, which has a radius of a 4:5 m. It has an absorption cross
section of a 5:7 1025 m2 and an emission cross section of e 7:9 1025 m2 at
the 1:53 m wavelength. Its upper laser level for the 1:53 m emission has the
same total lifetime and spontaneous lifetime of 2 sp 10 ms.
(a) Find the critical uorescence power density and the critical uorescence power of
the ber.
(b) Find the spontaneous emission power density and the spontaneous emission power
of the ber when it is uniformly pumped to have an unsaturated gain coefcient of
g 0 0:3 m1 at 1:53 m.
Bibliography
Iizuka, K., Elements of Photonics for Fiber and Integrated Optics, Vol. II. New York: Wiley, 2002.
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
9 - Laser Oscillation pp. 274-296
9
9.1
Laser Oscillation
CONDITIONS FOR LASER OSCILLATION
..............................................................................................................
The word laser is the acronym of light amplication by stimulated emission of radiation.
A medium that is pumped to population inversion has an optical gain to amplify an optical
eld through stimulated emission. Besides optical amplication, however, positive optical
feedback is normally required for laser oscillation. This requirement is fullled by placing
the gain medium in an optical resonator. One major characteristic of laser light is that it is
highly collimated and is spatially and temporally coherent. The directionality of laser light is a
direct consequence of the fact that laser oscillation takes place only along a longitudinal axis
dened by the optical resonator. The spatial and temporal coherence results from the fact that a
photon emitted by stimulated emission is coherent with the photon that induces the emission.
The gain medium emits spontaneous photons in all directions, but only the radiation that
propagates along the longitudinal axis within a small divergence angle dened by the resonator
obtains sufcient regenerative amplication through stimulated emission to reach the threshold
for oscillation. In order for the oscillating laser eld to be most efciently amplied in the
longitudinal direction, any spontaneous photons emitted in a direction outside of that small
angular range must not be allowed to compete for the gain. For this reason, a functional laser
oscillator is necessarily an open cavity that provides optical feedback only along the longitudinal axis. Most of the randomly directed spontaneous photons quickly escape from the cavity
through the open sides. Only a very small fraction of them that happen to be emitted within the
divergence angle of the laser eld mix with the coherent oscillating laser eld to become the
major incoherent noise source of the laser.
A laser is basically a coherent optical oscillator, and the basic function of an oscillator is to
generate a coherent signal through resonant oscillation without an input signal. No external
optical eld is injected into the optical cavity for laser oscillation. The intracavity optical eld
has to grow from the eld that is generated by spontaneous emission from the intracavity gain
medium. When steady-state oscillation is reached, the coherent laser eld at any given location
inside the cavity has to be a constant of time in both phase and magnitude. In the model shown
in Fig. 9.1, the situation of steady-state laser oscillation requires that Ein 0 while Ec z 6 0 at
any intracavity location z does not change with time. By applying this concept to (6.5) while
using (6.4), we nd the condition for steady-state laser oscillation:
a G exp iRT 1,
(9.1)
where a is the round-trip complex amplication factor for the intracavity eld, G is the roundtrip gain factor for the intracavity eld amplitude, and RT is the round-trip phase shift for the
9.1 Conditions for Laser Oscillation
275
Figure 9.1 FabryProt laser.
intracavity eld, as dened in (6.4). This general condition for laser oscillation applies to lasers
of various cavity structures that use different feedback mechanisms, including FabryProt
lasers, ring lasers, and distributed-feedback lasers. To illustrate the implications of this condition, we consider in the following the simple FabryProt laser shown in Fig. 9.1 that contains
an isotropic gain medium with a lling factor of .
The total permittivity of the gain medium, including the contribution of the resonant laser
transition, is res 0 res , as given in (6.36). Therefore, the total complex propagation
constant of the gain medium, including the contribution from the resonant transition, is
g
1=2
(9.2)
kg 0 0 res 1=2 k kres i ,
2
where
0res
0
(9.3)
,
2
2n
2nc res
00
00res :
(9.4)
g k res
2
n
nc
Here g is the gain coefcient of the laser medium, which is identied in (7.50), and kres is the
corresponding change in the propagation constant caused by the change in the refractive index
of the gain medium due to the changes in the population densities of the laser levels. When
population inversion is achieved, 00res < 0 so that the gain coefcient g has a positive value.
By replacing k for a cold medium with k g for a pumped gain medium, we nd that k given in (6.38)
for a cold cavity has to be replaced with k k res ig=2 when an actively pumped laser cavity is
considered. We then nd for an active laser cavity the mode-dependent round-trip gain factor,
kres k
1=2 1=2
Gmn R1 R2
exp mn g mn l,
(9.5)
and the mode-dependent round-trip phase shift,

RT
RT
mn 2k k res l mn 1 2 :
(9.6)
Because both Gmn and RT

mn are real parameters, the oscillation condition given in (9.1) can be
satised for a given laser mode to oscillate only if the gain condition
Gmn 1
(9.7)
and the phase condition

RT
mn 2q,
q 1, 2, . . .
are simultaneously fullled. Note that both Gmn and RT

mn are frequency dependent.
(9.8)
276
Laser Oscillation
9.1.1 Laser Threshold

The condition in (9.7) implies that there exist a threshold gain and a corresponding threshold
pumping level for laser oscillation. For the FabryProt laser shown in Fig. 9.1, which has a
length of l and contains a gain medium of a length lg for a lling factor of lg =l, the
threshold gain coefcient, g th
mn , of the TEMmn mode is given by
1 p
ln R1 R2 ,
l
(9.9)
p
g th
mn lg mn l ln R1 R2 :
(9.10)
g th
mn mn
or
Because the distributed loss mn is mode dependent, the threshold gain coefcient g th
mn varies
from one transverse mode to another. In addition, the effective gain coefcient can be different
for different transverse modes because different transverse modes have different eld distribution patterns and thus overlap with the gain volume differently. The transverse mode that has
the lowest loss and the largest effective gain at any given pumping level reaches threshold rst
and starts oscillating at the lowest pumping level. In the typical laser, the transverse mode that
reaches threshold rst is normally the fundamental TEM00 mode.
Unless a frequency-selecting mechanism is placed in a laser to create a frequencydependent loss that varies from one longitudinal mode to another, the threshold gain coefcient g th
mn varies little among the mnq longitudinal modes of different q values that share the
common mn transverse mode pattern. It is possible, however, to introduce a frequencyselecting device to a laser cavity to make mn and, consequently, g th
mn of a given mn transverse
mode highly frequency dependent for the purpose of selecting or tuning the oscillating laser
frequency.
The power required to pump a laser to reach its threshold is called the threshold pump
power, Pth
p . Because the threshold gain coefcient is mode dependent and frequency
dependent, the threshold pump power is also mode dependent and frequency dependent.
The threshold pump power of a laser mode can be found by calculating the power required
for the gain medium to have an unsaturated gain coefcient equal to the threshold gain
coefcient of the mode: g 0 g th
mn mnq , assuming uniform pumping throughout the gain
medium. For a quasi-two-level or three-level laser, there is also a transparency pump power,
Ptrp , for g 0 0, assuming uniform pumping. In the situation of nonuniform pumping, these
conditions for reaching threshold and transparency have to be modied. Clearly, Ptrp < Pth
p by
denition.
EXAMPLE 9.1
A Nd:YAG laser for the 1:064 m laser wavelength consists of a Nd:YAG laser rod of a
length lg 3 cm as a gain medium in a FabryProt cavity, which is formed by two mirrors of
reectivities R1 90% and R2 100% at a physical spacing of l 10 cm. The surfaces of the
laser rod are antireection coated to eliminate losses and undesirable effects. The crosssectional area of the laser rod is larger than that of the TEM00 Gaussian laser mode. This laser
9.2 Mode-Pulling Effect
277
mode has a distributed optical loss of 0:1 m1 . Find the threshold gain coefcient of this
laser mode.
Solution:
Using (9.10), we nd with the given parameters that the threshold gain coefcient of the TEM00
Gaussian laser mode is
g th
9.2
p
p
1
1
l ln R1 R2
0:1 0:1 ln 0:9 1 m1 2:09 m1 :
lg
0:03
MODE-PULLING EFFECT
..............................................................................................................
Comparing (9.6) for an active FabryProt laser with (6.40) for its cold cavity, we nd that,
through its dependence on kres , the round-trip phase shift of a eld in a laser cavity is a
function of 0res . Consequently, the longitudinal mode frequencies mnq at which a laser
oscillates are not exactly the same as the longitudinal mode frequencies cmnq given in (6.41)
for the cold FabryProt cavity.
Using (9.6) and (9.8), we nd that the longitudinal mode frequencies of a FabryProt laser
are related to those of its cold cavity by the relation:
mnq
cmnq

0res 1
0res
c
:
1
mnq 1
2nn
2nn
(9.11)
Clearly, the laser mode frequencies mnq differ from the cold-cavity mode frequencies because they
vary with the resonant susceptibility, which depends on the level of population inversion in the gain
medium. This dependence of the laser mode frequencies on the population inversion in the gain
medium is caused by the fact that the refractive index and the gain of the medium are directly connected
to each other, as is dictated by the KramersKronig relation. This effect causes a frequency shift of
mnq mnq cmnq
0res c
2nn mnq
(9.12)
for the oscillation frequency of mode mnq. Because of the frequency dependence of 0res , the
dependence of this frequency shift on 0res results in the mode-pulling effect demonstrated in
Fig. 9.2. Near the transition resonance frequency, 21 , of the gain medium, 0res is highly dispersive.
When a medium is pumped to have population inversion for a transition that has a resonance
frequency of 21 , 00res < 0 for either < 21 or > 21 , but 0res < 0 for < 21 and
0res > 0 for > 21 . As a result, mnq > cmnq for cmnq < 21 , whereas mnq < cmnq for
cmnq > 21 . Therefore, in comparison to the resonance frequencies of the cold cavity, the mode
frequencies of a laser are pulled toward the transition resonance frequency of the gain medium. In
278
Laser Oscillation
Figure 9.2 Frequency-pulling effect for laser modes. Compared to the resonance frequencies of the cold cavity
shown as dotted lines, the mode frequencies of an active laser shown as solid lines are pulled toward the
transition resonance frequency of the gain medium in the situation of population inversion. The real and
imaginary parts of the gain susceptibility as a function of optical frequency are shown.
addition, the longitudinal modes belonging to a common transverse mode are no longer equally
spaced in frequency. In a laser of a relatively high gain and a large dispersion, such as a
semiconductor laser, this effect can result in a large variation in the frequency spacing between
neighboring laser modes.
Because of the frequency dependence of the gain coefcient g due to the frequency
dependence of 00res , different longitudinal modes not only experience different values of
refractive index but also see different values of gain coefcient, as also illustrated in Fig. 9.2.
A longitudinal mode that has a frequency close to the gain peak at the transition resonance
frequency has a higher gain than one that has a frequency far away from the gain peak.
EXAMPLE 9.2
A Nd:YAG laser contains a Nd:YAG rod described in Example 8.1 in a cavity described in
Example 9.1. The refractive index of the Nd:YAG crystal is n 1:82. Find the largest
frequency shift of the longitudinal mode frequencies of the Nd:YAG laser due to the modepulling effect. How large is this frequency shift compared to the longitudinal mode frequency
spacing?
Solution:
From Example 9.1, we nd that the gain coefcient is g g th 2:09 m1 when the TEM00
laser mode is pumped to its threshold. The overlap factor is lg =l 0:3; thus, the weighted
average refractive index seen by the laser mode is
n 0:3 1:82 1 0:3 1 1:246:
With 1:064 m at the transition frequency 21 , we nd that the maximum value of the
imaginary part of the resonant susceptibility associated with this laser transition is
00res 21
279
nc
n
1:82 1:064 106
g g
2:09 6:44 107 ,
2
21
2
which appears at the line center. Because this laser transition is a discrete atomic transition, the
real part 0res has the largest absolute value at two frequencies. With 00res 21 < 0, 0res has the
largest negative value of 0res 00res 21 =2 at the frequency 21 and the largest
positive value of 0res 00res 21 =2 at 21 , as seen in Figs. 2.3 and 9.2. Thus,
j 0res jmax j 00res 21 =2j 3:22 107 :
For a Nd:YAG laser at 1:064 m, =21 2 104 because the gain linewidth is about
vg = 120 GHz, whereas the laser frequency is v21 21 =2 c= 283 THz. Therefore, we can take the approximation that c 21 21 for (9.12) to nd the
absolute value of the largest frequency shift caused by mode pulling:
jvjmax
jjmax j 0res jmax

3:22 107

21
283 1012 Hz 20:1 MHz:
2
2nn
2 1:82 1:246
This is the largest amount of frequency shift, which occurs for a longitudinal mode that has a coldcavity mode frequency at either the positive or negative half-width points vc, v21 vg =2. As
shown in Fig. 9.2, the mode that is closest to the lower frequency, vc, v21 vg =2, is pulled
up by an amount of approximately jvjmax , whereas the mode that is closest to the higher
frequency, c, v21 g =2, is pulled down by an amount of approximately jvjmax .
L
c
3 108
Hz 1:204 GHz:
2nl 2 1:246 10 102
Thus, the percentage of the maximum mode-pulling frequency shift is

jjmax
20:1 106
1:67%:
L
1:204 109
This frequency shift is appreciable though small. It is small because the dispersive effect of the
optical gain is small in the Nd:YAG medium. It can be much larger in a highly dispersive gain
medium, such as a semiconductor laser gain medium.
9.3
OSCILLATING LASER MODES
..............................................................................................................
Because the gain coefcient is a function of frequency, the net gain coefcient, g g th
mn , of a
laser mode is always frequency dependent and varies among different transverse modes and
among different longitudinal modes no matter whether the threshold gain coefcient g th
mn of a
transverse mode is frequency dependent or not. At a low pumping level before the laser starts
280
Laser Oscillation
oscillating, the net gain is negative for all laser modes. As the pumping level increases, the
mode that rst reaches its threshold starts to oscillate.
Once a laser starts oscillating in one mode, whether any other longitudinal or transverse modes
have the opportunity to oscillate through further increase of the pumping level is a complicated
issue of mode interaction and competition that depends on a variety of factors, including the
properties of the gain medium, the structure of the laser, the pumping geometry, the nonlinearity
in the system, and the operating condition of the laser. Here we only discuss some basic concepts
in the situation of steady-state oscillation of a CW laser. Interaction and competition among laser
modes are more complicated when a laser is pulsed than when it is in CW operation. Therefore,
some of the conclusions obtained below may not be valid for a pulsed laser.
The gain condition in (9.7) implies that once a given laser mode is oscillating in the steady state,
the gain that is available to this mode does not increase with increased pumping above the threshold
pumping level because Gmn has to be kept at unity for the steady-state oscillation of a laser mode.
Thus the effective gain coefcient of an oscillating mode is clamped at the threshold level of the
mode as long as the pumping level is kept at or above threshold. The mechanism for holding down
the gain coefcient at the threshold level is the effect of gain saturation discussed in Section 8.3. An
increase in the pumping level above threshold only increases the eld intensity of the oscillating
mode in the cavity, but the gain coefcient is saturated at the threshold value by the high intensity of
the intracavity laser eld. The fact that the gain of a laser mode oscillating in the steady state is
saturated at the threshold value has a signicant effect on the mode characteristics of a CW laser.
9.3.1 Homogeneously Broadened Lasers

When the gain medium of a laser is homogeneously broadened, all modes that occupy the same
spatial gain region compete for the gain from the population inversion in the same group of active
atoms. As the mode that rst reaches threshold starts oscillating, the entire gain curve supported by
this group of atoms saturates. Because this oscillating mode is normally the one that has a
longitudinal mode frequency closest to the gain peak and a transverse mode pattern of the lowest
loss, the gain curve is saturated in such a manner that its value at this longitudinal mode frequency is
clamped at the threshold value of the transverse mode that has the lowest threshold gain coefcient
among all transverse modes. If the gain peak does not happen to coincide with this mode frequency,
it still lies above the threshold when the gain curve is saturated, as shown in Fig. 9.3. Nevertheless,
all other longitudinal modes belonging to this transverse mode have frequencies away from the gain
peak. Therefore, even with increased pumping, they do not have sufcient gain to reach threshold
because the entire gain curve shared by these modes is saturated, as illustrated in Fig. 9.3. Other
transverse modes that are supported solely by this group of saturated, homogeneously broadened
atoms do not have the opportunity to oscillate either, because the gain curve is saturated below their
respective threshold levels. Nevertheless, because different transverse modes have different spatial
eld distributions, a high-order transverse mode may draw its gain from a gain region outside of the
region that is saturated by a low-order transverse mode. Therefore, when the pumping level is
increased, a high-order transverse mode may still reach its relatively high threshold for oscillation if
a low-order transverse mode of a low threshold is already oscillating.
Consequently, for a homogeneously broadened CW laser in steady-state oscillation, only one
among all of the longitudinal modes belonging to a particular transverse mode will oscillate, but
281
Figure 9.3 Gain saturation in a homogeneously broadened laser. Only one longitudinal mode whose frequency
is closest to the gain peak oscillates. The entire gain curve is saturated such that the gain at the single oscillating
frequency remains at the loss level.
it is possible for more than one transverse mode to oscillate simultaneously at a high pumping
level. Note that this conclusion does not hold true for a pulsed laser. It is possible for multiple
longitudinal modes belonging to the same transverse mode to oscillate simultaneously in a
pulsed laser even when its gain medium is homogeneously broadened.
EXAMPLE 9.3
The Nd:YAG laser described in Examples 9.1 and 9.2 has a Lorentzian gain lineshape that has a
bandwidth of g 0:45 nm for the laser line at 0:064 m. It is pumped at a level such that
the peak unsaturated gain coefcient is twice the threshold gain coefcient: g max
2g th . How
0
many longitudinal modes have their unsaturated gain coefcients pumped above the threshold?
How many longitudinal modes oscillate?
Solution:
The gain bandwidth in terms of frequency is

g g

:
With g 0:45 nm and 1:064 m,
c
3 108
0:45 109 Hz 119:25 GHz:
g g 2 g
1:064 106 2
2g th , the two frequencies at the two ends of the
When the laser is pumped such that g max
0
FWHM of the gain bandwidth have an unsaturated gain coefcient of g 0 g th . Therefore,
every mode that has a frequency within the FWHM, g 119:25 GHz, of the gain bandwidth
has an unsaturated gain coefcient above the threshold value. From Example 9.2, the longitudinal mode frequency spacing is
L
c
3 108
Hz 1:204 GHz:
2nl 2 1:246 10 102
Then,
282
Laser Oscillation
g 119:25
99:04:
L
1:204
Therefore, depending on where the longitudinal mode frequencies are located with respect to
the gain peak, 99 or 100 longitudinal modes have unsaturated gain coefcients that are above
the threshold value.
Because the gain spectrum has a Lorentzian lineshape, the laser is homogeneously broadened.
Therefore, ideally only one longitudinal mode oscillates. Though 99 or 100 longitudinal modes are
each pumped to have an unsaturated gain coefcient above the threshold value, all of them except
the oscillating mode are saturated below the threshold by the oscillating mode, which reaches the
threshold rst. In practice, however, we often nd that a Nd:YAG laser oscillates steadily in more
than one mode because it is not completely homogeneously broadened though it is predominantly
so. The degree of inhomogeneous broadening determines the number of oscillating modes.
9.3.2 Inhomogeneously Broadened Lasers

In a laser that has an inhomogeneously broadened gain medium, there are different groups of active
atoms in the same spatial gain region. Each group saturates independently. Two modes occupying
the same spatial gain region do not compete for the same group of atoms if the separation of their
frequencies is larger than the homogeneous linewidth of each group of atoms. When one longitudinal mode reaches threshold and oscillates, the gain coefcient is saturated only within the spectral
range of a homogeneous linewidth around its frequency, while the gain coefcient at frequencies
outside this small range continues to increase with increased pumping. As the pumping level
increases, other longitudinal modes can successively reach threshold and oscillate. As a result, at a
sufciently high pumping level, multiple longitudinal modes belonging to the same transverse
mode can oscillate simultaneously. The saturation of the gain coefcient in a small spectral range
within a homogeneous linewidth around each of the frequencies of these oscillating modes, but not
across the entire gain curve, creates the effect of spectral hole burning in the gain curve of an
inhomogeneously broadened laser medium, as illustrated in Fig. 9.4. Different transverse modes
Figure 9.4 Spectral hole burning effect in the gain saturation of an inhomogeneously broadened laser. Multiple
longitudinal modes oscillate simultaneously at a sufciently high pumping level. The gain at each oscillating
frequency is saturated at the loss level. The mode-pulling effect is ignored in this illustration.
283
also saturate independently in an inhomogeneously broadened medium if their frequencies are

sufciently separated. Therefore, an inhomogeneously broadened laser can also oscillate in
multiple transverse modes.
EXAMPLE 9.4
A HeNe laser has a Doppler-broadened gain bandwidth of g 1:5 GHz at its laser
wavelength of 632:8 nm. The laser has a cavity length of l 32 cm. It is pumped at a
level such that the peak unsaturated gain coefcient is twice the threshold gain coefcient:
g max
2g th . How many longitudinal modes have their unsaturated gain coefcients pumped
0
above the threshold? How many longitudinal modes oscillate?
Solution:
When the laser is pumped such that g max
2g th , the two frequencies at the two end of the
0
FWHM vg of the gain bandwidth have an unsaturated gain coefcient of g 0 g th . Therefore,
the laser has a bandwidth of v vg 1:5 GHz. Every mode that has a frequency within this
bandwidth has an unsaturated gain coefcient above the threshold value. With l 32 cm and
n 1 for the gaseous HeNe laser gain medium, the longitudinal mode frequency spacing is
L
c
3 108
Hz 468:75 MHz:
2nl 2 1 32 102
Then,
1:5 109
3:2:
L 468:75 106
Therefore, three or four longitudinal modes have unsaturated gain coefcients that are above
the threshold value, depending on where the longitudinal mode frequencies are located with
respect to the gain peak. Because the gain spectrum is Doppler broadened, the laser is
inhomogeneously broadened. All longitudinal modes above threshold oscillate.
9.3.3 Laser Linewidth

The linewidth of an oscillating laser mode is still described by (6.18):
mnq
1 Gmnq L
mn ,
Gmnq
(9.13)
where the longitudinal mode frequency spacing Lmn might vary for different transverse modes.
From this relation, we see that in practice the round-trip eld gain factor Gmnq of a laser mode in
steady-state oscillation cannot be exactly equal to unity because the laser linewidth cannot be
zero, due to the existence of spontaneous emission. In reality, in steady-state oscillation the
value of Gmnq is slightly less than unity, with the small difference made up by spontaneous
emission. Clearly, the linewidth of an oscillating laser mode is determined by the amount of
284
Laser Oscillation
spontaneous emission that is channeled into the laser mode. Therefore, (9.13) is not very useful
for calculating the linewidth of a laser mode in steady-state oscillation without knowing the
exact value of Gmnq in the presence of spontaneous emission.
A detailed analysis taking into account spontaneous emission yields the SchawlowTownes
relation for the linewidth of a laser mode in terms of the laser parameters:
ST
mnq
2hvcmnq 2
Pout
mnq
N sp
hv
2 cmnq 2 Pout
mnq
N sp ,
(9.14)
where cmnq and cmnq are respectively the cold-cavity linewidth and the photon lifetime of the
oscillating mnq mode, Pout
mnq is the output power of the oscillating laser mode, and
N sp
eN 2
eN 2 N 2
eN 2 aN 1
g
N
(9.15)
is the spontaneous emission factor that measures the degree of the effective population inversion
in the gain medium. The effective population inversion dened as N g= e in (8.5) is the
population density that is able to contribute to the coherent stimulate emission, which does not
broaden the laser linewidth, whereas all of the upper level population N 2 contributes to the
incoherent spontaneous emission, which broadens the laser linewidth. The effect of spontaneous
emission on the linewidth of an oscillating laser mode enters the relation in (9.14) through the
population densities of the laser levels in the form of the spontaneous emission factor.
Because N sp 1, the ultimate lower limit of the laser linewidth, which is known as the
SchawlowTownes limit, is that given in (9.14) for N sp 1. It can also be seen that the
linewidth of a laser mode decreases as the laser power increases. This phenomenon is easily
understood. Because the gain of an oscillating laser mode is clamped at its threshold level,
increased pumping above threshold does not increase the population inversion, and thus does
not increase the spontaneous emission, which is proportional to the population of the upper
laser level. When the power of an oscillating laser mode increases with increased pumping, the
coherent stimulated emission increases proportionally but the incoherent spontaneous emission
is clamped at its threshold level. As a result, the linewidth of the laser mode decreases with
increasing laser power.
EXAMPLE 9.5
Find the minimum possible linewidth that is set by the SchawlowTownes limit for the
oscillating laser mode of the Nd:YAG laser described in Examples 9.1 and 9.2 when the laser
is pumped sufciently above the threshold so that the output power of the mode at
1:064 m is 100 mW.
Solution:
The Nd:YAG laser described in Examples 9.1 and 9.2 has a FabryProt cavity that has a
length of l 10 cm, a weighted average index of n 1:246, a distributed loss of 0:1 m1 ,
and mirror reectivities of R1 90% and R2 100%. Therefore, from (6.45), the cold-cavity
photon lifetime of the laser mode is
9.4 Laser Power
285
nl
1:246 10 102
p s 6:63 ns:
p
cl ln R1 R2 3 108 0:1 10 102 ln 0:9 1
Because Nd:YAG is a four-level system which has a 0, it has N sp 1 as can be seen from
(9.15). The photon energy at the 1:064 m laser wavelength is
hv
1:2398
eV 1:165 eV:
1:064
For an oscillating laser mode that has an output power of Pout 100 mW, the minimum
possible linewidth set by the SchawlowTownes limit is found using (9.14):
vST
hv
1:165 1:6 1019
N
1 Hz 6:7 mHz:
sp
2 2c Pout
2 6:63 109 2 100 103
This minimum possible oscillating laser mode linewidth is nine orders of magnitude smaller
than the cold-cavity longitudinal linewidth of vc 2 c 1 27:9 MHz. The signicant line
narrowing is caused by the coherent stimulated emission. However, the SchawlowTownes
linewidth found above is only the fundamental lower bound limited by the spontaneous
emission noise, which can be approached if all other noise sources are eliminated in the ideal
condition. In practice, the linewidth of an oscillating laser mode is much larger than the
SchawlowTownes linewidth, though generally much smaller than the cold-cavity linewidth,
because it is broadened by many mechanisms such as the noise from pump power uctuations,
mechanical vibrations, and temperature uctuations of the laser.
9.4
LASER POWER
..............................................................................................................
In this section, we consider the output power of a laser. Because the situation of a multimode
laser can be quite complicated due to mode competition, we consider for simplicity only a CW
laser that oscillates in a single longitudinal and transverse mode. The parameters mentioned in
this section are not labeled with mode indices because all of them are clearly associated with the
only oscillating mode. The simple case of a FabryProt cavity that contains an isotropic gain
medium with a lling factor of as shown in Fig. 9.1 is considered. To illustrate the general
concepts, we consider the situation when the gain medium is uniformly pumped so that the
entire gain medium has a spatially independent gain coefcient.
For the single oscillating mode of the FabryProt laser considered here, the round-trip gain
factor G is that given by (9.5), and the cavity decay rate c dened by (6.23) is that given by
(6.46). Therefore,
G2 exp 2gl c T:
(9.16)
Because G2 is the net amplication factor of the intracavity eld energy, which is proportional
to the intracavity photon number, in a round-trip time T of the laser cavity, we can dene an
286
Laser Oscillation
intracavity energy growth rate, or intracavity photon growth rate, g, for the oscillating laser
mode through the relation
G2 exp g c T:
(9.17)
We nd, by comparing (9.17) with (9.16), the gain parameter of the gain medium:
g
2gl cg
:
n
T
(9.18)
By comparing (6.46) with (9.9), we nd that

c
2g th l
cg
th :
T
n
(9.19)
Note that while the unit of g and g th is per meter, the unit of g and c is per second.
The relation in (9.18) translates the gain coefcient that characterizes spatially dependent
amplication through the gain medium of a propagating intracavity laser eld into an intracavity energy growth rate that characterizes the temporal growth of the energy in a laser mode.
The relation in (9.19) clearly indicates that the threshold intracavity energy growth rate for laser
oscillation is the cavity decay rate:
gth c :
(9.20)
This relation can also be obtained by applying the threshold condition of G 1 to the relation in
(9.17). It is easy to understand because for a laser mode to oscillate, the growth of intracavity
photons in that mode through amplication by the gain medium has to completely compensate for
the decay of photons caused by all the loss mechanisms. Therefore, we shall call the energy growth
rate g and the cavity decay rate c , both of which are specic to a laser mode, the gain parameter
and the loss parameter, respectively, of the laser mode. Note that the gain parameter g of the laser
mode is reduced by the lling factor from the gain parameter g of the gain medium.
By using temporal growth and decay rates instead of spatial gain and loss coefcients to
describe the characteristics of a laser, we are in effect moving from a spatially distributed
description of the laser to a lumped-device description. In the lumped-device description, a laser
mode is considered an integral entity with its spatial characteristics effectively integrated into
the parameters g and c . The detailed spatial characteristics of the mode are irrelevant and are
lost in this description. Therefore, instead of the intensity of the oscillating laser eld, we have
to consider the intracavity photon density, S, of the oscillating laser mode. For a FabryProt
laser that contains a gain medium of a lling factor so that the average refractive index inside
the cavity is n n 1 n0 as dened in (6.3), the average intracavity photon density of
the laser mode is
S
nI
,
chv
(9.21)
where I is the spatially averaged intracavity intensity and hv is the photon energy of the
oscillating laser mode.
Because the gain parameter g is directly proportional to the gain coefcient g of the gain
medium, the relation between the unsaturated gain parameter g0 and the saturated gain
9.4 Laser Power
287
parameter g of a laser mode in the lumped-device description can be obtained by converting

the relation between g 0 and g discussed in Section 9.3 through the relation in (9.18). Therefore,
for the gain parameter of a laser mode, we have
g
g0
g0
and g
1 S=Ssat
1 S=Ssat
(9.22)
cg 0
n
(9.23)
where
g0
is the unsaturated gain parameter of the gain medium and

Ssat
nI sat
n
chv c s e
(9.24)
is the saturation photon density of the laser mode.

When a CW laser oscillates in the steady state, the value of g for the oscillating mode is
clamped at its threshold value of c , just as the value of g is clamped at g th . Therefore, by setting
g to equal c and using (9.22), we nd that the intracavity photon density of a CW laser mode
in steady-state oscillation is

g0
1 Ssat r 1Ssat , for r 1:
(9.25)
S
c
The dimensionless pumping ratio r represents that a laser is pumped at r times its threshold. It is
dened as
r
g0 g 0
:
c
g th
(9.26)
Assuming that the pumping efciency is the same at transparency, at threshold, and at the
operating point, the pumping ratio can be expressed in terms of the pump power as
r
Pp Ptrp
tr
Pth
p Pp
(9.27)
where Ptrp is the pump power for the gain medium to reach transparency, Pth
p is that for the laser to
reach its threshold, and Pp is the pump power at the operating point. Note that (9.25) is valid only
for r 1 when the laser oscillates because only then is the laser gain saturated. For r < 1, the laser
does not reach threshold. The laser cavity is then lled with spontaneous photons at a density that
is small in comparison to the high density of coherent photons when the laser oscillates at r 1.
From the intracavity photon density of the oscillating laser mode, we can easily nd the total
intracavity energy contained in this mode:
U mode hvV mode S,
(9.28)
where V mode is the volume of the oscillating mode. The mode volume can be found by
integrating the normalized intensity distribution of the mode over the three-dimensional
288
Laser Oscillation
space dened by the laser cavity; it is usually a fraction of the volume of the cavity. The
output power of the laser is simply the coherent optical energy emitted from the laser per
second. Therefore, it is simply the product of the mode energy and the output-coupling rate,
out , of the cavity:
Pout out U mode out hvV mode S r 1out hvV mode Ssat :
(9.29)
The output-coupling rate is also called the output-coupling loss parameter because it contributes to the total loss of a laser cavity; it is a fraction of the total loss parameter c . One can
indeed write c i out , where i is the internal loss of the laser that does not contribute to
the output coupling of the laser power.
As an example, for the FabryProt laser that has c given by

c
1 p
(9.30)
ln R1 R2
c
n
l
as expressed in (6.46), we have the internal loss given by i c=n and the output-coupling
loss given by
out
c p
c p c p
ln R1 R2 ln R1 ln R2 out, 1 out, 2 ,
nl
nl
nl
where
out;1
c p
ln R1
nl
and
out, 2
c p
ln R2
nl
(9.31)
(9.32)
are the output-coupling losses of mirror 1 and mirror 2, respectively. In this case, out is the total
output-coupling loss through both mirrors. Therefore, Pout given in (9.29) is the total output
power emitted through both mirrors. For the output power emitted through each mirror, we nd
that
Pout;1 U mode out, 1
out, 1
Pout and Pout, 2 U mode out, 2 out, 2 Pout :

out
out
(9.33)
It is convenient to dene the saturation output power as

Psat
out out hvV mode Ssat :
(9.34)
Using the denition of Ssat in (9.24), it can be shown that

p
Psat
out Psat ln R1 R2 ,
(9.35)
where Psat is the saturation power of the gain medium found by integrating I sat over the crosssectional area of the gain medium. Combining (9.29) with (9.34), we can express the output
laser power in terms of Psat
out as
Pout r 1Psat
out :
(9.36)
Note that Psat

out is not the level at which the output power of a laser saturates. Its physical
meaning can be easily seen from (9.35) and (9.36). From (9.35), we nd that the output power
of a laser is Psat
out when the intracavity laser power is at the level Psat of the gain medium. From
9.4 Laser Power
289
sat
(9.36), we nd that Pout Psat
out when r 2; in other words, a laser has an output power of Pout
when it is pumped at twice its threshold level.
EXAMPLE 9.6
The Nd:YAG gain medium of the laser described in Examples 9.1 and 9.2 has a saturation
intensity of I sat 17:3 MW m2 , which stays almost constant for an unsaturated gain coefcient g 0 over the range from 0 to 10 m1. With a cavity length of l 10 cm, the two cavity
mirrors are chosen such that at the 1:064 m laser wavelength, the TEM00 Gaussian mode
has a beam waist spot size of w0 500 m located at the center of the Nd:YAG rod, which has
a length of lg 3 cm. (a) Find the pumping ratio r and the corresponding unsaturated gain
coefcient g 0 required for the laser mode to have an output power of 100 mW. (b) If the laser is
pumped at a level for an unsaturated gain coefcient of g 0 10 m1 , what is the pumping ratio
and the output power of the laser mode?
Solution:
For the TEM00 Gaussian mode that has a beam waist spot size of w0 500 m in the Nd:YAG
rod, the Rayleigh range, from (3.69), is
zR
nw20 1:82 500 106 2
m 1:34 m:
1:064 106
Because zR l > lg , the beam spot stays constant throughout the cavity. Therefore, the mode
volume of the oscillating laser mode is
V mode
w20
500 106 2
Al
l
10 102 m3 3:93 108 m3 :
2
2
The weighted average refractive index of the laser mode is n 1:246, from Example 9.2. The
photon energy for 1:064 m is hv 1:165 eV, from Example 9.5. With a saturation
intensity of I sat 17:3 MW m2 , the saturation photon density of the oscillating laser mode is
Ssat
nI sat
1:246 17:3 106
m3 3:85 1017 m3 :

chv 3 108 1:165 1:6 1019
The output coupling rate is

out
p 1
c p
3 108

ln
0:9 1 s 1:27 108 s1 :
ln R1 R2
nl
1:246 10 102
The saturation output power is found using (9.34):

Psat
out out hvV mode Ssat 358 mW:
(a) For an output power of Pout 100 mW, we nd by using (9.36) that the required pumping
ratio is
r 1
Pout
100
1:28:
sat 1
Pout
358
290
Laser Oscillation
From Example 9.1, the threshold gain coefcient is g th 2:09 m1 . Therefore, by (9.26),
the unsaturated gain coefcient at this pumping ratio is
g 0 rg th 1:28 2:09 m1 2:68 m1 :
(b) When the laser is pumped to have an unsaturated gain coefcient of g 0 10 m1 , by (9.26)
the pumping ratio is
r
g0
10
4:78:
g th 2:09
Therefore, from (9.36), the output laser power is

3
W 1:35 W:
Pout r 1Psat
out 4:78 1 358 10
To explicitly express the output laser power as a function of the pump power, it is necessary
to specify the pumping mechanism and the pumping geometry. Irrespective of the pumping
details, it is generally true that a laser has zero coherent output power but only uorescence
before it reaches threshold, whereas its coherent output power grows linearly with the pump
power above threshold before nonlinearity occurs at a high pump power. Upon reaching the
threshold, the output laser eld also shows dramatic spectral narrowing that accompanies the
start of laser oscillation. According to (9.14) and the discussion following it, the linewidth of an
oscillating laser mode continues to narrow with increasing laser power as the laser is pumped
higher above threshold. The reason is that above threshold the coherent stimulated emission
increases with the pumping ratio, whereas the spontaneous emission, which is proportional to
the population of the upper laser level, is clamped at its threshold value. These are the unique
characteristics that distinguish a laser from other types of light sources, such as uorescent light
emitters and luminescent light sources. However, a real laser does not have such exact ideal
characteristics, mainly because of the presence of spontaneous emission and nonlinearities in
the gain medium.
Figure 9.5 shows the typical characteristics of the output power Pout of a single-mode laser as
a function of the pump power Pp . The linear relation between Pout and Pp is a consequence of
applying the linear relation between g 0 and Pp to (9.26) for (9.27). As discussed in Section 8.3,
the linear relation between g 0 and Pp is itself an approximation near the transparency point of a
gain medium. As the pump power increases to a sufciently high level, the unsaturated gain
coefcient of a medium cannot continue to increase linearly with the pump power because of
the depletion of the ground-level population. Therefore, we should expect that the output power
of a laser will not continue its linear increase with the pump power but will increase less than
linearly with the pump power at high pumping levels. On the other hand, once the gain medium
of a laser is pumped so that its upper laser level begins to be populated, it emits spontaneous
photons regardless of whether the laser is oscillating or not. Clearly, the output power of a laser
that is pumped below threshold is not exactly zero because uorescence from spontaneous
emission is already emitted from the laser before the laser reaches threshold. Though this
9.4 Laser Power
291
Figure 9.5 Typical characteristics of the output

power of a single-mode laser as a function of the
pump power.
uorescence is incoherent and its power is generally small for a practical laser, it is signicant
for a laser below and right at threshold. Above threshold, it is the major source of incoherent
noise for the coherent eld of the laser output.
The overall efciency of a laser, known as the power conversion efciency, is
c
Pout
:
Pp
(9.37)
The approximately linear dependence of the laser output power on the pump power above
threshold leads to the concept of the differential power conversion efciency, also known as the
slope efciency, of a laser, dened as
s
dPout
:
dPp
(9.38)
Referring to the laser power characteristics shown in Fig. 9.5, the threshold of a laser can usually
be lowered by increasing the nesse of the laser cavity, thus lowering the values of c and out , but
only at the expense of reducing the differential power conversion efciency of the laser. In the
linear region of the laser power characteristics, s is clearly a constant that is independent of the
operating point of the laser. By contrast, c increases with the pump power, but c is always
smaller than s in the linear region. At high pumping levels where the laser output power does not
increase linearly with the pump power because of nonlinearity, s is no longer independent of the
operating point. It can even become smaller than c in certain unfavorable situations.
EXAMPLE 9.7
The Nd:YAG laser considered in Example 9.5 is optically pumped from two sides of the laser
rod with two diode laser arrays at the 808 nm pump wavelength. Because the Nd:YAG laser is a
four-level system, its transparency pump power is zero, Ptrp 0. Furthermore, the pumping
ratio is approximately proportional to the pump power: r / Pp . It is found that the pump power
292
Laser Oscillation
required to reach the pumping ratio for an unsaturated gain coefcient of g 0 10 m1 is
Pp 16:5 W. Use the data obtained in Example 9.6 to answer the following questions. (a) Find
the threshold pump power. (b) Find the conversion efciency and the slope efciency when the
laser has an output power of Pout 100 mW as in Example 9.6(a). (c) Find the conversion
efciency and the slope efciency when the laser has an unsaturated gain coefcient of
g 0 10 m1 as in Example 9.6(b).
Solution:
From Example 9.6(b), r = 4.78 for g 0 10 m1 . Therefore, r 4:78 for Pp 16:5 W.
Because Nd:YAG is a four-level system, it is transparent without pumping. Therefore,
Ptrp 0. From (9.27), we have
r
Pp Ptrp
Pth
p
Ptrp
Pp
,
Pth
p
and
dr
r
4:78 1
W 0:29 W1 :
dPp Pp 16:5
(a) The laser reaches its threshold when the pumping ratio is r th 1. Therefore, the threshold
pump power is
Pth
p
rth
1
W
W 3:45 W:
0:29
0:29
(b) From Example 9.6(a), we nd that r 1:28 for Pout 100 mW. At this pumping ratio,
Pp rPth
p 1:28 3:45 W 4:42 W:
Therefore, from (9.37), the power conversion efciency is
c
Pout 100 103
2:26%:
Pp
4:42
From Example 9.6, we have Psat

out 358 mW. Using (9.38) and (9.36), we nd that the
slope efciency is
s
dPout
dr sat
P 0:29 358 103 10:4%:

dPp
dPp out
(c) When the laser is pumped with a pump power of Pp 16:5 W to give an unsaturated gain
coefcient of g 0 10 m1 , we nd r = 4.78 and Pout 1:35 W from Example 9.6(b).
Therefore, from (9.37), the power conversion efciency is
Problems
293
Pout 1:35
8:18%:
Pp
16:5
The slope efciency is the same as that found in (b):

s
dPout
dr sat
P 0:29 358 103 10:4%:

dPp
dPp out
Problems
9.1.1 A HeNe laser has a FabryProt cavity formed by two mirrors of reectivities R1
95% and R2 100% at its laser wavelength of 632:8 nm. The cavity length is
l 32 cm. The effective refractive index of the HeNe gas is n 1. The TEM00
Gaussian laser mode has a distributed optical loss of 0:05 m1 . Find the threshold
gain coefcient of this laser mode.
9.1.2 An optical-ber laser emitting at 1:53 m has a ring cavity as shown in Fig. 6.1(d). It
has one inputoutput coupler that has a coupling efciency of 10%. The ber loop
has a total length of l 10 m, which contains a gain section of a length lg 1 m. The
effective index of the ber laser mode is n 1:47 and the distributed loss is
10 dB km1 . What is the threshold gain coefcient of this laser mode?

9.1.3 A GaAs/AlGaAs semiconductor laser emitting at 860 nm has a FabryProt cavity
formed by two at, cleaved surfaces of reectivities R1 R2 32% for the TE0 mode of
the GaAs/AlGaAs waveguide. The gain region is the GaAs waveguide core, which is
pumped uniformly throughout the cavity length such that the cavity and the gain medium
have the same length of l lg 350 m. The laser oscillates in the single transverse TE0
waveguide mode, which has a connement factor of 0:3 dened by the overlap
factor of the TE0 mode intensity prole with the waveguide core gain region. The
distributed loss is 25 cm1 . Find the threshold gain coefcient of this laser mode.
If one of the cleaved cavity surfaces is optically coated for 100% reectivity, what is the
threshold gain coefcient?
9.2.1 The optical gain of a homogeneously broadened laser is contributed by a discrete optical
transition between two atomic energy levels at a transition resonance frequency of 21 .
A longitudinal mode q of the laser has its cold-cavity frequency tuned to the transition
resonance frequency such that cq 21 . When the laser is pumped above the threshold
for this mode to oscillate, what is the oscillating frequency of the laser? How much is the
frequency shift due to mode pulling?
9.2.2 The optical gain in a semiconductor laser medium is contributed by excess electrons and
holes in the conduction and valence bands, respectively, of the semiconductor. The gain
is determined by the excess carrier concentration N, which is the density of the
294
Laser Oscillation
electronhole pairs in excess of the thermal-equilibrium concentrations of electrons and

holes. As a result, the relationship between the real and imaginary parts of the resonant
susceptibility is not simply the Lorentzian function that characterizes a discrete atomic
transition. Nevertheless, an optical gain still causes a change in the refractive index of the
medium. This effect is usually described by an experimentally measured antiguidance
factor, also known as the linewidth enhancement factor, dened as
b
n0 =N
2 n0 =N
4 n0 =N

,
n00 =N
c g=N
g=N
(9.39)
where n0 and n00 are, respectively, the real and imaginary parts of the refractive index
of the medium, and g is the gain coefcient. A GaAs/AlGaAs semiconductor laser
emitting at 850 nm has a FabryProt cavity, which is pumped uniformly so
that the cavity and the gain medium have the same length of l lg 300 m. The
gain medium has an antiguidance factor of b 3:5. The effective refractive index is
n 3:65 when the laser medium is pumped to transparency at 850 nm. The laser
is pumped to give a gain coefcient of g 5 104 m1 . Besides shifting the frequency of each longitudinal mode, the mode-pulling effect caused by the antiguidance
factor changes the longitudinal mode frequency spacing. Find the frequency shift of a
longitudinal mode at the 850 nm laser wavelength. Find the change in the longitudinal mode frequency spacing.
9.3.1 A GaAs/AlGaAs vertical-cavity surface-emitting semiconductor laser emitting at
850 nm has a very short cavity. Its gain region is composed of a few thin quantum wells,
and its reective mirrors are distributed Bragg reectors of periodic structures. For the
longitudinal mode frequencies, the effective physical length of the cavity is leff 1:2 m
and the effective refractive index is neff 3:52. The laser is pumped to give a gain
bandwidth of g 48 nm above the laser threshold. How many longitudinal modes
oscillate?
9.3.2 A HeNe laser has a Doppler-broadened gain bandwidth of g 1:5 GHz at its laser
wavelength of 632:8 nm. The laser has a cavity length of l 32 cm.
(a) It is pumped at a level such that the peak unsaturated gain coefcient is four times the
threshold gain coefcient: g max
4g th . How many longitudinal modes have their
0
unsaturated gain coefcients pumped above the threshold? How many longitudinal
modes oscillate?
(b) If a longitudinal mode frequency is tuned to the frequency of the gain peak, what is
the value of g max
for the laser to oscillate only in this mode?
0
9.3.3 An Er:ber laser emitting at 1:53 m has a cold-cavity linewidth of c 520 kHz.
It is doped with an Er3 ion concentration of N t 2:2 1024 m3 . At 1:53 m, the
absorption cross section is a 5:7 1025 m2 , and the emission cross section is
e 7:9 1025 m2 . The gain coefcients of its oscillating modes are saturated at
g 0:25 m1 . The population density of the upper laser level for this gain coefcient
can be found using (8.42). What is the minimum possible linewidth set by the Schawlow
Townes limit for an oscillating laser mode that has an output power of Pout 1 mW?
Problems
295
9.4.1 A Ti:sapphire laser consists of a Ti:sapphire crystal of a length lg 2 cm in a Fabry

Prot cavity, which has a physical length of l 16 cm dened by two mirrors of
reectivities R1 100% and R2 95% at the laser emission wavelength of
800 nm. The TEM00 Gaussian mode dened by the cavity has a beam waist spot
size of w0 150 m located at the center of the Ti:sapphire crystal, which has a
refractive index of n 1:76. The end surfaces of the crystal are antireection coated to
eliminate undesirable losses. At the 800 nm laser wavelength, the Ti:sapphire crystal
has an emission cross section of e 3:4 1023 m2 and an absorption cross section of
a 8 1026 m2 . Over the range of laser operation considered here, the saturation
lifetime can be taken as s 2 3:2 s. The distributed loss of the laser cavity,
including the absorption of the Ti:sapphire crystal at 800 nm, is 0:1 m1 . The
laser is optically pumped at the pump wavelength of p 532 nm.
(a) Find the threshold gain coefcient of this laser.
(b) Find the saturation output power of this laser.
(c) What are the pumping ratio and the unsaturated gain coefcient required for the laser
to have an output power of Pout 1 W?
(d) The transparency pump power of the laser is Ptrp 1:4 W, and the threshold pump
power is Pth
p 5:0 W. What is the pump power that is required for Pout 1 W?
(e) What are the power conversion efciency and the slope efciency when the laser has
an output power of Pout = 1 W?
9.4.2 The Ti:sapphire laser described in Problem 9.4.1 is pumped to have an unsaturated gain
coefcient of g 0 5 m1 .
(a) What are the pumping ratio and the pump power?
(b) Find the output laser power at this pumping level.
(c) What are the power conversion efciency and the slope efciency at this
operating point?
9.4.3 A semiconductor laser is pumped by current injection. The injected current generates
excess electronhole pairs in the active region of the laser. The excess electronhole pairs
act as the source of the optical gain. When the details of the laser structure and the
parameters of the gain medium are known, the power and efciency of a semiconductor
laser can be analyzed as discussed in Section 9.4. Alternatively and equivalently, the
output power of a semiconductor laser can be found by considering that one photon is
emitted when an electronhole pair recombines radiatively. Thus, for a semiconductor
laser,
Pout inj
out hv
I I th ,
c e
(9.40)
where inj is the current injection efciency, out is the output coupling rate, c is the
cavity decay rate, hv is the laser photon energy, e is the electronic charge, I is the
injection current, and I th is the threshold injection current for the laser to start oscillating.
The injection efciency inj is the fraction of the total injection current that actually
296
Laser Oscillation
contributes to the generation of useful electronhole pairs in the active region of the laser.
If the bias voltage of the laser is V, the power conversion efciency is

Pout Pout
out hv
I th
c
inj
,
(9.41)
1
Pp
VI
c eV
I
and the slope efciency is
s
dPout dPout
hv
inj out
:
dPp
VdI
c eV
(9.42)
Now, consider the GaAs/AlGaAs laser described in Problem 9.1.3 but with R1 1 and
R2 0:32. The effective refractive index of the laser mode is n 3:63. The injection
efciency is inj 0:7, the threshold current is I th 20 mA, and the bias voltage is
V 2 V.
(a) Find the output laser power for an injection current of I 40 mA.
(b) What are the power conversion efciency and the slope efciency at this
operating point?
Bibliography
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
10 - Optical Modulation pp. 297-361
10
Optical Modulation
10.1 TYPES OF OPTICAL MODULATION
..............................................................................................................
Optical modulation allows one to control an optical wave or to encode information on a carrier
optical wave. The inverse process that recovers the encoded information is demodulation. There
are many types of optical modulation, which can be categorized in several different ways.
1. According to the particular optical-eld parameter being modulated, optical modulation can be
categorized into different modulation schemes: phase modulation, frequency modulation,
polarization modulation, amplitude modulation, spatial modulation, and diffraction modulation.
2. Depending on whether the information is encoded in the analog or digital form, optical
modulation can be either analog modulation or digital modulation.
3. Optical modulation can be categorized as direct modulation or external modulation. Direct
modulation is directly performed on an optical source, which is usually a light-emitting
diode (LED) or a laser, without using a separate optical modulator. External modulation is
performed on an optical wave using a separate optical modulator to change one or more
characteristics of the wave.
4. Optical modulation is accomplished by varying the optical susceptibility of the modulator
material. Depending on whether the real or imaginary part of the susceptibility is responsible
for the functioning of the modulator, optical modulation can be categorized as refractive
modulation or absorptive modulation. Refractive modulation is performed by varying the
real part of the susceptibility, thus varying the refractive index of the material; absorptive
modulation is performed by varying the imaginary part of the susceptibility, thus varying the
absorption coefcient of the material.
5. Optical modulation can be categorized according to the physical mechanism behind the
change of the optical susceptibility, such as electro-optic modulation, acousto-optic modulation, magneto-optic modulation, all-optical modulation, and so forth.
6. Depending on the geometric relation between the modulating signal and the modulated
optical wave, optical modulation can be transverse modulation or longitudinal modulation.
In transverse modulation, the signal is applied in a direction perpendicular to the propagation
direction of the optical wave. In longitudinal modulation, the signal is applied along the
propagation direction of the optical wave.
7. Optical modulation can be performed on unguided or guided optical waves. Correspondingly, the structure of an optical modulator can take the form of a bulk or waveguide device.
A bulk modulator is used to modulate an unguided optical wave. A waveguide modulator is
used to modulate a guided optical wave.
298
Optical Modulation
Optical switching is a special form of optical modulation. Generally speaking, optical

switching is large-signal digital optical modulation that results in the switching between two
or more discrete values of an optical parameter or between two or more optical modes. It can be
performed on any type of optical modulation. The characteristic of the optical wave being
switched can be its phase, frequency, amplitude, polarization, propagation direction, or spatial
pattern. Optical switching can also be performed between two or more normal modes in a
waveguide structure.
10.2 MODULATION SCHEMES
..............................................................................................................
As discussed in Section 1.7, an unguided optical eld is characterized by its polarization ê ,
magnitude jE j, phase E , wavevector k, and frequency :
Er, t Er, t exp ik r it
ê jEr, tjeiE r, t exp ik r it:
(10.1)
The total phase of this eld is that given in (1.83):

r; t k r t E r; t:
(10.2)
As described in (3.25), a guided optical eld propagating along the z direction can be expressed
as a linear superposition of normal modes:
X
Er, t
A z, tE^ x, y exp i z it
X
(10.3)
E^ x, yjA z, tjeiA z, t exp i z it:
The eld in a mode is also characterized by ve eld parameters: the vectorial mode eld
pattern E^v x; y, the magnitude jAv z; tj of the complex mode amplitude Av z; t, the phase
Av z; t of the complex mode amplitude Av z; t , the mode propagation constant v , and the
frequency . The total phase of the eld in mode v is
v z; t v z t Av z; t :
(10.4)
Optical modulation can be performed on any of the eld parameters. Therefore, there exist
many modulation techniques based on different schemes. Each modulation scheme has been
further developed into many advanced modulation formats.
In general, the concept of a modulation scheme or format that is developed for an
electromagnetic carrier wave at a low frequency, such as a radio frequency, can be adapted
and applied to optical modulation. Also common to low-frequency carriers and optical
carriers is that the modulation signal can be either analog or digital. The three basic modulation schemes for all carrier frequencies are phase modulation (PM), frequency modulation
(FM), and amplitude modulation (AM) for analog modulation, which take the forms of phaseshift keying (PSK), frequency-shift keying (FSK), and amplitude-shift keying (ASK) for
digital modulation.
299
Due to the differences between optical waves and low-frequency electromagnetic waves
regarding the eld characteristics and the material properties in their respective spectral regions,
some schemes and certain considerations are specic to optical modulation. In addition to the
three basic modulation schemes of phase modulation, frequency modulation, and amplitude
modulation, optical modulation can also be performed on the polarization ê of the eld for
polarization modulation, on the spatial distribution jE r; t j of the eld for spatial modulation,
and on the direction k^ of wave propagation for diffraction modulation.
Because of the dispersive nature and the intrinsic coupling between the real and imaginary
parts of the optical susceptibility, as well as its tensorial nature in the case of an anisotropic
crystal, a modulation signal often affects more than one parameter of the modulated optical
eld. For example, amplitude modulation that is carried out by varying the absorption or
amplication coefcient, through varying 00 , of the material in a modulator is usually accompanied by a variation in 0 , thus varying the refractive index and resulting in a modulation on
the phase of the optical wave. This is the case for direct modulation discussed in Section 10.3.
As another example, phase modulation using a modulator made of an anisotropic crystal can
sometimes be accompanied by a polarization change of the optical eld. In any event, a
modulation scheme is chosen based on the eld parameter on which we intend to code the
information. The accompanying modulation on other eld parameters is a side effect that has to
be avoided or suppressed as much as possible, if it is unavoidable.
Phase modulation is the most fundamental of all modulation schemes. By controlling the
optical phase while properly manipulating the optical wave, a desired modulation on any other
eld parameter can be accomplished. On the other hand, certain eld parameters can be directly
modulated without changing the optical phase. The concepts of basic optical modulation
schemes are described in the following. The techniques and the physical mechanisms that
can be used for these modulation schemes are discussed in later sections.
10.2.1 Phase Modulation

A phase-modulated optical eld at a xed location, taken to be r 0 for simplicity of
expression, is a function of time of the form:
E0; t ê jE j exp iE t it,
(10.5)
where the time-varying phase E t carries the encoded information, whereas ê , jE j, and do
not vary with time. In analog phase modulation, E t is a continuous function of time; in
digital phase modulation, i.e., PSK, E t changes stepwise with time. The temporal characteristics of the optical eld under analog and digital phase modulation are shown in Figs. 10.1(a)
and (b), respectively. The magnitude and frequency of the carrier eld stay constant under
phase modulation because only the phase varies with time.
In phase modulation, the largest meaningful phase change is 2 because phase is periodic
with a period of 2; therefore, the range of phase modulation is usually chosen to be from 0 to
2 or from to . In PSK, the 2 phase range is equally divided into discrete levels
representing different digital values. The phase shifts from one discrete level to another discrete
300
Optical Modulation
Figure 10.1 (a) Analog phase modulation with an analog signal. (b) Digital phase modulation using two
discrete phases separated by for BPSK. The eld magnitude and the carrier frequency stay constant while the
phase varies with time.
level. In binary PSK (BPSK), two discrete phases separated by , such as f0; g or
f=2; 3=2g, are used to respectively represent the two binary bits of 0 and 1, as shown in
Fig. 10.1(b). In quadrature PSK (QPSK), four discrete phases that are equally spaced at an
interval of =2, such as f0; =2; ; 3=2g or f=4; 3=4; 5=4; 7=4g, are used to represent
the four possible two-bit combinations of f00; 01; 10; 11g by encoding two bits with each phase.
Optical phase modulation is normally accomplished through refractive modulation. By
modulating the refractive index of a material through which an optical wave propagates, the
phase of the wave can be modulated. The physical mechanisms that can be used for this purpose
are discussed in Section 10.4.
10.2.2 Frequency Modulation

A frequency-modulated optical eld has a time-varying frequency of t that carries the
encoded information:
E0; t ê jE j exp iE itt ,
(10.6)
where ê , jE j, and E do not vary with time. In analog frequency modulation, t varies
continuously with time; in digital frequency modulation, i.e., FSK, t shifts abruptly from
one frequency to another. In binary FSK (BFSK), two different frequencies are used to
represent the two binary bits of 0 and 1 for a digital signal. More than two frequencies can
be used to digitize a signal in multiple symbols; for example, in quadrature FSK (QFSK), four
frequencies are used to represent the four possible two-bit combinations of f00; 01; 10; 11g by
encoding two bits with each frequency.
Figures 10.2(a) and (b) show the temporal characteristics of the optical eld under analog
frequency modulation and BFSK, respectively. The magnitude of the carrier eld stays constant
301
Figure 10.2 (a) Analog frequency modulation. (b) Digital frequency modulation using two different
frequencies for BFSK. The eld magnitude stays constant while the carrier frequency varies with time.
while the frequency varies with time. Note the ne differences in the characteristics of the
modulated waveforms between frequency modulation and phase modulation by comparing
Fig. 10.2 to Fig. 10.1.
Frequency modulation can be achieved by phase modulation over a large phase range
because, from (1.87),
t
E:
t
t
(10.7)
In contrast to the case for phase modulation discussed above, however, the modulated phase
change for frequency modulation is not limited to a range of 2. Instead, the range of phase
change is a function of the magnitude and the duration of the frequency shift from the original,
unshifted carrier frequency. For example, for BFSK that shifts the frequency between and 0 ,
a time-varying phase of E t 0 t t 0 has to be maintained from the time t 0 when
the frequency is shifted from to 0 until the time when the frequency is shifted back to .
EXAMPLE 10.1
The phase of a polarized plane optical eld is temporally modulated by a sinusoidal variation of
a modulation amplitude 0 and a modulation frequency as E t 0 sin t. What happens
to the polarization of this modulated optical eld? What happens to the magnitude and intensity
of this optical eld? Does this phase modulation result in frequency modulation? What happens
to the frequency of this optical eld in the time domain and in the frequency domain?
Solution:
The modulation is imposed only on the phase of the eld such that
E t ê E expiE t ê E exp i0 sin t :
302
Optical Modulation
Clearly, the polarization vector ê is not affected by the phase modulation; thus, it remains a
constant of time. The eld magnitude jE j is not affected by the phase modulation, either;
therefore, both the eld magnitude and the intensity, which is I / jE j2 , remain constants
of time.
By contrast, this time-varying phase modulation does result in frequency modulation:
t
E 0 cos t:
t
t
In the time domain, we nd that the frequency of this optical eld varies sinusoidally with time
around the center optical carrier frequency as t 0 cos t. To nd the frequency
components in the frequency domain, we use the identity:
exp i0 sin t
J q 0 exp iqt ,
q
where J q is the qth-order Bessel function of the rst kind, which has the property that
J q 1q J q . Therefore, we can express the phase-modulated optical eld as
Et ê jE j(
exp i0 sin t it
)
h
i
X
q iqt
iqt
it
J q 0 e
1 e
ê jE j J 0 0 e
:
q1
It can be seen that in the frequency domain, the sinusoidal phase modulation generates a series
of side bands at the harmonics of the modulation frequency on both the low-frequency and
high-frequency sides of the center optical carrier frequency .
10.2.3 Polarization Modulation

Information can also be encoded on the polarization of an optical eld through polarization
modulation so that the polarization vector is a time-varying function:
E0; t ê t jE j exp iE it ,
(10.8)
where jE j, E , and do not vary with time. In analog polarization modulation, ê t varies
continuously with time; in digital polarization modulation, known as polarization-shift keying
(PolSK), ê t changes abruptly from one polarization to another. In binary polarization-shift
keying (BPolSK), two orthogonal polarization states are used to represent the two binary bits of
0 and 1 for a digital signal. Multiple polarization states can be used to represent multiple
possible bit combinations; in this situation, the polarization states are not all mutually orthogonal because each polarization state has only one corresponding orthogonal polarization state.
Polarization modulation can be achieved through differential phase modulation on two
orthogonally polarized components of an optical eld by using, for example, the electro-optic
303
Pockels effect or the magneto-optic Faraday effect. Any orthonormal set of unit polarization
vectors fê 1 ; ê 2 g on the plane that is normal to the wave propagation direction k^ can be used to
expand the unit polarization vector ê on this plane as a linear superposition of two orthogonal
polarizations:
ê c1 ê 1 c2 ê 2 ,
(10.9)
where c1 and c2 are two complex constants subject to the normalization condition of
c1 c
e 1 ; ê 2 g basis, the unit polarization vector ê that is orthogonal
1 c2 c2 1: On the f^
to the unit polarization vector ê can be expressed as
ê c
e 1 c
e2:
2^
1^
(10.10)
It is clear that fê ; ê g is also an orthonormal basis because ê ê ê ê

1 and
ê ê ê ê 0. Therefore, the two unit polarization vectors ê 1 and ê 2 can be expressed

in terms of the fê ; ê g basis as
ê 1 c
e c2 ê ,
1^
ê 2 c
e c1 ê :
2^
(10.11)
As an example, any polarization state on the xy plane can be represented by the unit vector
ê ^x cos ^y ei sin given in (1.65), which is the linear superposition of the two orthonormal linear polarization unit vectors ^x and ^y with c1 cos and c2 ei sin . In this case,
ê 1 ^x , ê 2 ^y , and ê ^x ei sin ^y cos . As another
example, the linear polarization
p
unit vector ^x can be expressed as ê ^x ê ê = 2 in terms of p
the
linear superposition of
the orthonormal circular polarization unit vectors with c1 c2 1= 2. In this case, ê 1 ê ,
p
ê 2 ê , and ê i^y ê ê = 2.
When the phases of the two orthogonally polarized eld components are differentially
modulated, the polarization vector of the modulated optical wave becomes a function of time:
h
i
ê m t c1 ei1 t ê 1 c2 ei2 t ê 2 c1 ê 1 c2 eit ê 2 ei1 t ,
(10.12)
where
t 2 t 1 t
(10.13)
is the time-varying phase difference due to differential phase modulation between the ê 1 and ê 2
components of the optical eld. By substituting ê 1 and ê 2 of (10.11) into (10.12), we can
express the modulated time-varying unit polarization vector ê m t in terms of ê and ê as

i1 t
i2 t
ê m t c1 c
ê c1 c2 ei1 t c1 c2 ei2 t ê
c2 c
1e
2e
(10.14)

i1 t i1 t
ê c1 c2 1 eit ei1 t ê :
c1 c
e
1 c2 c2 e
It is clear from (10.14) that ê m t ê 6 0 and ê m t 6 ê when c1 c2 6 0 and t 6 2m,
resulting in a polarization change caused by differential phase modulation.
As discussed in Section 1.6, the polarization state of a wave depends only on the phase
difference and the magnitude ratio of the two orthogonally polarized eld components.
Therefore, the polarization state dened by ê m t is determined by the phase difference t
and the magnitude ratio jc1 =c2 j of the ê 1 and ê 2 components, and is independent of the common
304
Optical Modulation
phase factor 1 t. Because the magnitude ratio jc1 =c2 j is not affected by phase modulation, thus
remaining constant, the polarization state can be varied by varying only the phase difference
t . Consequently, polarization modulation of an optical eld can be accomplished through
differential phase modulation on two orthogonally polarized components of the eld.
EXAMPLE 10.2
An optical eld is initially linearly polarized in the x direction. Find two linearly polarized
components of this polarization in the xy plane that are orthogonal to each other. How does the
polarization of this eld change if the two orthogonally polarized components are differentially
phase modulated by a phase difference of =4, =2, , and 2, respectively?
Solution:
In the xy plane, the two linearly polarized orthogonal components of the unit polarization vector
ê ^x can be chosen as
^x ^y
^x ^y
ê 1 p and ê 2 p ,
2
2
p
,which are arbitrarily chosen to be real vectors such that c1 c2 1= 2 and arbitrarily assigned
in the sequence of ê 1 and ê 2 . In the xy plane, the polarization that is orthogonal to ê ^x is
ê ^y . From (10.14), if the two orthogonally polarized components are differentially phase
modulated such that 2 t 1 t t , the polarization of the eld becomes

1 eit
1 eit
ê
ê ei1 t
ê m t
2
2

1 eit
1 eit
^x
^y ei1 t
2
2

t
t
^x i sin
^y ei1 tit=2 :
cos
2
2
The common phase factor 1 t t=2 only changes the phase of the unit polarization
vector ê m t and does not have an effect on the polarization state of the eld. Therefore, we can
ignore this phase factor and consider only the polarization state vector of the differentially
phase-modulated eld:
t
t
^x i sin
^
ê 0m t cos
y:
2
2
We nd different polarization states for different phase differences:
For , ê 0m cos ^x i sin ^y 0:924^x i0:383^y , elliptically polarized;

4
8
8
^x i^y
0
For , ê m cos ^x i sin ^y p , circularly polarized;

2
4
4
2
0
For , ê m cos ^x i sin ^y i^y , linearly polarized parallel to ê ^y ;
2
2
0
For 2, ê m cos ^x i sin ^y ^x , linearly polarized parallel to ê ^x .
305
10.2.4 Amplitude Modulation

One of the most common modulation schemes is amplitude modulation, which encodes information on the magnitude of an optical eld:
E0; t ê jE tj exp iE it,
(10.15)
where ê , E , and do not vary with time. In analog amplitude modulation, jE t j varies
continuously with time; in digital amplitude modulation, known as amplitude-shift keying
(ASK), jE t j changes abruptly from one discrete value to another. In binary ASK, two eld
magnitudes are used, with the binary bit 1 normally represented by a larger eld magnitude and
the bit 0 represented by a smaller magnitude. A special case of binary ASK is on-off keying
(OOK) where the optical eld is turned on at a xed magnitude level for bit 1 and turned off for
bit 0. Multilevel ASK uses multiple discrete eld magnitudes to represent multiple possible bit
combinations for each eld magnitude to encode one possible combination of an equal number
of bits.
Figures 10.3(a) and (b) show the temporal characteristics of the optical eld under analog
modulation and binary ASK, respectively. The magnitude of the carrier eld varies with time
while the frequency and phase stay constant. Amplitude modulation leads to intensity modulation (IM), in which the intensity and the power of an optical wave are modulated because the
intensity and power of the wave are proportional to jE t j2 .
Optical amplitude modulation can be accomplished in many different ways: by direct
modulation on the optical source, as discussed in Section 10.3; by refractive modulation using
any physical mechanism discussed in Section 10.4, followed by proper manipulation of the
optical eld; or by absorptive modulation of a material through which the optical wave
propagates, as discussed in Section 10.5.
Figure 10.3 (a) Analog amplitude modulation. (b) Digital amplitude modulation using two different
discrete eld magnitudes. Both the carrier frequency and phase of the eld stay constant while the magnitude
varies with time.
306
Optical Modulation
Amplitude modulation on an optical eld can be achieved through polarization modulation

by properly selecting a polarization component while ltering out its orthogonal component
after the eld is polarization modulated. As an example, consider the polarization-modulated
optical eld characterized by the time-varying unit polarization vector ê m t expressed in
(10.14). It is clear that by using a polarizer to select either only the ê -polarized component or
only the ê -polarized component, the resulting eld magnitude is modulated. For instance, by
selecting only the ê -polarized component, the output eld has the time-varying magnitude:

h
i

t

it

E 2c1 c2 E sin
,
jE tj c1 c2 1 e
2
(10.16)
where E is the time-independent eld amplitude of the polarization-modulated optical eld. The
intensity of this output eld is modulated as
I t 4jc1 c2 j2 I sin2
t
,
2
(10.17)
where I / jE j2 is the time-independent intensity of the polarization-modulated optical wave.

Though polarization modulation of the optical eld used in the above example is accomplished
by differential phase modulation, the concept of obtaining amplitude modulation by selecting a
polarization component while rejecting its orthogonal component is generally applicable to any
polarization-modulated optical wave.
Optical amplitude modulation can also be achieved through phase modulation to vary the
coupling or interference between different components of an optical wave.
1. By varying the phase mismatch through differential phase modulation on two coupled
modes in a coupler, the coupling efciency can be modulated, as discussed in Section 4.6.
Thus, the eld amplitude of a mode is modulated. This general concept is applicable to any
mode coupler.
2. By varying the interference of two or multiple waves through differential phase modulation,
the superposition of the interfering waves can be amplitude modulated, as discussed in
Section 5.1. This general concept is applicable to any interferometer discussed in Chapter 5.
In analog amplitude modulation, the optical intensity varies continuously with time. To
faithfully encode the analog information on the carrier optical wave, linearity of the modulation
response is desired. However, as the example in (10.17) shows, the response of an amplitude
modulator generally cannot be linear over the whole range of operation. For this reason, the
linearity requirement for analog modulation often limits the modulation depth to a small linear
range of the modulation response.
In digital amplitude modulation, the optical intensity is switched between two or among
multiple discrete levels. In this case, linearity is not required, but clear separation of the discrete
levels is desired. In binary operation, where the switching takes place between a high-intensity
level of I high and a low-intensity level of I low , it is desired that the ratio I low =I high is as small as
possible while I high is sufciently large. In digital amplitude modulation using an external
modulator, the binary states are represented by a high transmittance T high and a low
307
transmittance T low . The ratio of these two levels is dened as the extinction ratio, which is
usually measured in dB:
ER 10 log
I low
T low
10 log
:
I high
T high
(10.18)
A high extinction ratio allows clear separation of the two levels, thus clear identication of the
binary bits. Besides a high extinction ratio, the level of the high transmittance T high has to be
sufciently high for good performance.
10.2.5 Spatial Modulation

Taking the propagation direction to be the z direction without loss of generality, a spatially
modulated optical eld has a time-varying eld pattern of E x; y; 0; t at the xed z 0 location
on the plane that is perpendicular to the propagation direction:
Ex; y; 0; t E x; y; 0; t exp it ê x; y; 0; tjE x; y; 0; t jeiE x;y;0;t eit :
(10.19)
Spatial modulation can be on the eld polarization, with a space- and time-varying polarization
vector ê x; y; 0; t; on the eld magnitude, with a space- and time-varying eld magnitude
jE x; y; 0; t j; or on the phase, with a space- and time-varying eld phase E x; y; 0; t . The
spatial variation can be either a continuous function of x and y, or a digitized function of x and y.
If the spatial variation is expressed in terms of a linear superposition of transverse spatial
normal modes, then
X
Ex; y; 0; t
Av t E^v x; y exp it
(10.20)
v
according to (3.25). Thus, spatial modulation can be described as, and be accomplished
through, the temporal variations of the mode expansion coefcients Av t .
10.2.6 Diffraction Modulation

As discussed in Section 5.2, an optical grating diffracts an incident optical wave into multiple
diffracted beams; the diffraction angle q of the qth-order diffracted beam is determined by the
phase-matching condition given in (5.32):
k sin q k sin i qK
(10.21)
where k n=c is the propagation constant of the optical wave, with n being the refractive
index of the medium; i is the incident angle of the incoming wave; and K 2= is the
wavenumber of the grating, with being the period of the grating. Clearly, the diffraction angle
q , and thus the diffraction pattern, can be varied by varying the refractive index n, the incident
angle i , the grating period , or a combination of these parameters. Many refractive
modulation mechanisms, as discussed in Section 10.4, can be used to modulate the refractive
index of the grating material, thus accomplishing diffraction modulation. The grating period
308
Optical Modulation
can also be modulated if the grating is not a xed structure but is generated by an acoustic wave
through an acousto-optic effect, by a low-frequency electric eld through an electro-optic
effect, or by a periodic optical intensity pattern through optical interference.
10.3 DIRECT MODULATION
..............................................................................................................
The most straightforward way to encode information on an optical wave is to directly modulate
the optical source. This technique is often applied to an LED or a semiconductor laser, both of
which are current-injection devices driven by current sources. Therefore, an LED or a semiconductor laser can be directly modulated by applying the modulation signal to the injection
current, an approach known as direct current modulation. In this approach, the modulation
signal takes the form of a modulating current, which is added to the DC bias current that
supplies electrical power to the device.
Figure 10.4 shows the schematic circuitry of direct current modulation. The LED or
semiconductor laser is biased at a DC injection current level of I 0 and is modulated with a
time-varying modulation current of I m t that carries the modulation signal. Thus the total
current injected into the device is I t I 0 I m t. The output optical power is
Pout t P0 Pm t , where P0 is the constant output optical power at the bias current level of
I 0 and Pm t is the time-varying component of the modulated output optical power responding to
the modulation current I m t . Though the circuitry for direct modulation is the same for an LED
and a semiconductor laser, the characteristics of their modulation responses are very different.
LEDs and semiconductor lasers are both junction diodes that usually have sophisticated
structures for improved performance. In operation as a light source, an LED or semiconductor
laser is injected with a current of I to inject excess electrons and holes, i.e., excess charge
carriers, into an active region of an area A and a thickness d. Taking into consideration the
injection efciency of the charge carriers, the current density J that actually contributes to
carrier injection is related to the total current that is supplied to the device as
J inj
I
,
A
(10.22)
where inj is the carrier injection efciency, which is determined by the device structure. The
injected current creates an excess carrier density of N n n0 p p0 in the active region,
where n0 and p0 are, respectively, the equilibrium electron and hole concentrations in the
absence of current injection, and n and p are the electron and hole concentrations under current
injection.
Figure 10.4 Schematic circuitry of
direct current modulation on an
LED or semiconductor laser.
A resistance in series with the
device is normally used to protect
the device.
309
The excess carriers recombine through radiative and nonradiative mechanisms with a total
spontaneous carrier recombination rate of s and a corresponding spontaneous carrier recombination lifetime of s :
s
1
:
s
(10.23)
The output optical power of an LED is contributed by the spontaneous emission from
spontaneous radiative recombination of the excess carriers. By contrast, the output optical
power of a semiconductor laser comes from the resonant optical eld undergoing stimulated
emission in the laser cavity. A semiconductor laser has a threshold for laser oscillation, but an
LED does not have a turn-on threshold. These fundamental differences lead to very different
modulation characteristics between an LED and a semiconductor laser, as discussed below.
Direct current modulation on an LED or a semiconductor laser is a technique of amplitude
modulation because its objective is the modulation of the output optical power. However, the
time-varying current also causes the refractive index of the LED or laser material to vary with
time; consequently, the phase and frequency of the output optical wave are also varied by the
modulation current. The consequence is an accompanying phase and frequency modulation that
is generally undesirable and difcult to avoid because of the nonlinearity and dispersion in the
variation of the refractive index in response to the modulation current. The temporal variation in
the optical frequency results in frequency chirping in the modulated output optical wave. This
effect is more signicant for direct current modulation on a semiconductor laser than on
an LED.
10.3.1 Light-Emitting Diode

An LED converts electrical energy to optical energy through the spontaneous emission
resulting from spontaneous recombination of the excess carriers. Because spontaneous emission occurs whenever carriers are excited, an LED starts to emit light once current is injected,
i.e., there is no threshold to turn an LED on. Therefore, the output optical power Pout is directly
proportional to AdN= s , which is the total number of excess carriers recombining per second,
and can be expressed as
Pout
e hvAd
N,
inj s
(10.24)
where e is the external quantum efciency, inj is the carrier injection efciency, both
dependent on the structure of the LED, and hv is the photon energy. The temporal variation
of the carrier density in response to the variation in the injection current I is described as
inj
dN
J
N
N
I ,

dt
ed s eAd
s
(10.25)
where e is the electronic charge and J is the injection current density given in (10.22).
The output optical power of an LED as a function of the injection current is known as the
lightcurrent characteristics, or simply the LI characteristics, also called the powercurrent
310
Optical Modulation
characteristics, or simply the PI characteristics. The steady-state solution of (10.25) for N

obtained by setting dN=dt 0 results in the ideal powercurrent relation for an LED in steadystate operation under DC current injection:
Pout e
hv
I,
e
(10.26)
which indicates that the output power of an LED increases linearly with the injection current.
The LI characteristics of a representative LED, shown in Fig. 10.5, are not exactly linear
throughout the entire range of operation, however. These characteristics have several important
features that distinguish an LED from a laser. First, there is no threshold in the LI characteristics of an LED, indicating that an LED is turned on and starts emitting light once it is forward
biased and injected with any amount of current. At moderate current levels, the LI curve of an
LED is indeed quite linear, as indicated by (10.26). This linearity is useful for analog
modulation of an LED. Nonlinearities in the LI relationship are usually found at very low
and very high current levels.
For high-speed applications, a large modulation bandwidth is desired. The intrinsic speed of
an LED is primarily determined by the lifetime of the injected carriers in the active region. For
an LED that is biased at a DC injection current level of I 0 and is modulated at a frequency of
2f with a modulation index of m, we can express the total time-dependent current that is
injected to the LED as
I t I 0 I m t I 0 1 m cos t I 0 mI 0 cos t,
Figure 10.5 Lightcurrent characteristics and direct current modulation of a representative LED.
(10.27)
311
where I m t mI 0 cos t is the modulation current signal, which has an amplitude of

I m mI 0 . The time-varying output optical power Pout t of an LED in response to this
modulation can be found by using (10.24) after solving for N t from (10.25). Note that
the time-varying Pout t cannot be found directly from (10.26) because (10.26) is valid only
for the steady-state operation of an LED when it is only injected with a DC current. In the
linear response regime under the condition that m 1, the output optical power can be
expressed as
Pout t P0 Pm t P0 1 jr j cos t ,
(10.28)
where P0 is the constant output optical power found from (10.26) at the bias current level of I 0 ,
Pm t jr jP0 cos t is the time-varying component of the modulated output power, jr j
is the magnitude of the response to the modulation, and is the phase delay of the response to
the modulation signal. The characteristics of direct current modulation on an LED are illustrated
in Fig. 10.5.
For an LED that is modulated in the linear response regime, the complex response as a
function of the modulation frequency is
r jr jei
m
:
1 i s
(10.29)
The frequency response and the modulation bandwidth of an LED are usually measured in
terms of the electrical power spectrum using a broadband, high-speed photodetector that
converts the output optical power of the LED into an output electrical current of the photodetector. In the linear operating regime of the detector, the detector current is linearly proportional to the optical power of the LED. Therefore, the electrical power spectrum of the detector
output is proportional to jr j2 :
Rf jr f j2
m2
m2
,
1 4 2 f 2 2s 1 f 2 =f 23dB
(10.30)
which has a 3-dB modulation bandwidth of

f 3dB
1
,
2 s
(10.31)
as shown in Fig. 10.6.

The spontaneous carrier lifetime s is normally on the order of a few hundred nanoseconds
to 1 ns for an LED. Therefore, the modulation bandwidth of an LED is typically in the range
of a few megahertz to a few hundred megahertz. A modulation bandwidth up to 1 GHz can
be obtained with a reduction in the internal quantum efciency of an LED by reducing
the carrier lifetime to the subnanosecond range. Aside from this intrinsic response speed
determined by the carrier lifetime, the modulation bandwidth of an LED can be further
limited by the parasitic effects from its electrical contacts and packaging, as well as from its
driving circuitry.
312
Optical Modulation
Figure 10.6 Normalized current-modulation frequency response of an LED measured in terms of the electrical
power spectrum using a photodetector. The spontaneous carrier lifetime is taken to be s 10 ns for this plot.
EXAMPLE 10.3
An LED emitting at a center wavelength of 850 nm has an external quantum efciency of
e 21%. Its spontaneous carrier lifetime is s 10 ns. The LED is biased at a DC injection
current of I 0 20 mA and is modulated at a modulation frequency of f 10 MHz with a
modulation current for a modulation index of m 10%. (a) Find the output power of the LED
at the DC bias point. (b) What is the amplitude of the modulation current? (c) What are the
amplitude of the modulated output power and the phase delay of the response to the current
modulation? (d) Find the 3-dB modulation bandwidth of this LED in terms of its modulation
response in the electrical power spectrum of the photodetector output. (e) At this modulation
frequency, what is the modulation response in the electrical power spectrum of the photodetector used to measure the LED output? What is the normalized modulation response in dB?
Solution:
An LED has no threshold. Therefore, the DC output power is directly proportional to its DC
bias current I 0 , and the modulation index is dened as the ratio of the amplitude I m of the
modulation current to I 0 .
(a) The photon energy at 850 nm is
1239:8
eV 1:46 eV:
850
The DC output power of the LED is found using (10.26):
hv
hv
I 0 0:21 1:46 20 mW 6:13 mW:
e
(b) The amplitude of the modulation current for m 10% is
P0 e
I m mI 0 10% 20 mA 2 mA:
313
(c) From (10.29), we nd

m
0:1
2
jr j q q

8:47 10 ,
2
6
9 2
1 2f s
1 2 10 10 10 10

tan1 2f s tan1 2 10 106 10 109 0:56 rad:
Note that it is always true that jrj < m for an LED at a nonzero modulation frequency. The
amplitude of the modulated output power is
Pm jrjP0 8:47 102 6:13 mW 519 W:
and the phase delay of the modulation response is 0:56 rad.
(d) The 3-dB modulation bandwidth of this LED is, from (10.31),
f 3dB
1
1
Hz 15:9 MHz,
2 s 2 10 109
as seen in Fig. 10.6.

(e) At the modulation frequency of f 10 MHz, the modulation response in the electrical
power spectrum of the photodetector output is, from (10.30),
Rf
m2
0:12
7:2 103 :
1 f 2 =f 23dB 1 10=15:92
Because R0 m2 1 102 , the normalized response is

10 log
Rf
7:2 103
1:43 dB:
10 log
R0
1 102
10.3.2 Semiconductor Laser

For most applications, it is desired that a semiconductor laser oscillate in a single transverse
mode and a single longitudinal mode. Many practical lasers indeed have such a desirable
characteristic. For a single-mode semiconductor laser that is injected with a current of I, the
temporal characteristics of its carrier density N and its intracavity photon density S can be
described by the coupled rate equations:
inj
dN
J
N
N
I gS,
gS
eAd
dt
ed s
s
(10.32)
dS
c S gS,
dt
(10.33)
where e is the electronic charge, s is the spontaneous carrier lifetime, c is the cavity decay rate,
J is the injection current density dened in (10.22), and g is the gain parameter of the gain
region dened in (9.18). The overlap factor appears in the gain term of (10.33) because only
314
Optical Modulation
that fraction of the laser mode volume overlaps with the gain region to receive stimulated
amplication.
The threshold condition for a semiconductor laser is that in (9.20) for any laser:
gth c :
(10.34)
The gain parameter g is a function of the excess carrier density N, which in turn is determined by
the injection current I. The threshold gain parameter gth determines a threshold carrier density
N th at a threshold current density of J th that is supplied by a threshold injection current of I th .
The characteristics of a semiconductor laser in steady-state oscillation above threshold can be
obtained from the steady-state solutions of (10.32) and (10.33) by setting dN=dt dS=dt 0.
It is found that in steady-state oscillation above threshold at an injection current of I > I th , the
carrier density and the gain are clamped at their respective threshold values, N N th and
g gth , while the intracavity photon density builds up for S 6 0. Most of the concepts
developed in Section 9.4 for laser power characteristics are directly applicable to a semiconductor laser. By directly applying the steady-state conditions of g gth c = and
N N th J th s =ed inj s =edAI th to (10.32) to obtain the steady-state solution of S for
dS=dt 0, followed by using the relation J inj =AI from (10.22) and the relation
dA V gain V mode , the CW output power of a semiconductor laser in steady-state oscillation
under DC current injection can be found using (9.29) and can be expressed as a function of the
injection current:
Pout inj
out hv
hv
I I th e I I th ,
c e
e
(10.35)
where e inj out =c is the external quantum efciency of the semiconductor laser.
Figure 10.7 shows the powercurrent characteristics, i.e., the lightcurrent characteristics, of
a representative semiconductor laser. It can be seen from (10.35) that in an ideal situation, the
output power of a semiconductor laser above threshold increases linearly with the injection
current. This characteristic is indeed observed in most semiconductor lasers over a large range
of operating conditions. This linearity is useful for analog modulation of a semiconductor laser
over a large dynamic range. Nonlinearities in the LI characteristics appear at high injection
current levels.
Like an LED, a semiconductor laser can be directly current modulated. Unlike an LED,
however, the modulation speed of a semiconductor laser is not limited by the spontaneous
carrier lifetime s in the active region of the laser. This difference is due to the fact that there is
strong coupling between the carriers and the intracavity laser eld. The effective lifetime of the
carriers in an oscillating laser is much shorter than the spontaneous lifetime because of the
stimulated carrier recombination that takes place in a laser. The modulation speed of a
semiconductor laser is primarily determined by the intracavity photon lifetime and the effective
carrier lifetime. Because both the photon lifetime and the effective carrier lifetime of a
semiconductor laser are generally much shorter than the spontaneous carrier lifetime, a semiconductor laser has a higher modulation speed than an LED. Because the stimulated recombination rate increases with the intracavity photon density, the modulation speed of a
semiconductor laser increases with the laser power.
315
Figure 10.7 Lightcurrent characteristics and direct current modulation of a representative

semiconductor laser.
When a laser is in steady-state oscillation at a DC bias injection current of I 0 > I th in the

absence of modulation, the laser gain and the carrier density are both clamped at their respective
threshold values of gth and N th , but the photon density has a value of S0 corresponding to the
laser output power P0 , which depends on the injection current at the bias point. Under the
dynamical perturbation of a modulation signal, the gain can deviate from gth due to the variations
in the carrier and photon densities caused by the external perturbation. To the rst order, the
dependence of the gain parameter on the carrier and photon densities can be expressed as
g gth gn N N th gp S S0 ,
(10.36)
where gn is the differential gain parameter characterizing the dependence of the gain parameter
on the carrier density and gp is the nonlinear gain parameter characterizing the effect of gain
compression due to the saturation of the gain by intracavity photons. It has been found
empirically that for a given laser, both gn and gp stay quite constant over large ranges of carrier
density and photon density. For most practical purposes, they can be treated as constants over
the operating range of a laser. These parameters are normally measured experimentally though
they can also be calculated theoretically. Note that gn > 0 but gp < 0.
It is convenient to dene a differential carrier relaxation rate, n , and a nonlinear carrier
relaxation rate, p , as
n gn S0 ,
p gp S0 :
(10.37)
In addition, we have the cavity decay rate, c 1= c , and the spontaneous carrier relaxation
rate, s 1= s . These four relaxation rates can be directly measured for a given semiconductor
316
Optical Modulation
laser. They determine the current modulation characteristics of a laser. Note that, for a given
laser, c and s are constants that are independent of the laser power, but n and p are linearly
proportional to the laser power because they are linearly proportional to the photon density, as
seen in (10.37).
Because a semiconductor laser has a threshold, the modulation index m for a laser that is
biased at a DC injection current of I 0 > I th and is modulated at a frequency of 2f is
dened as
I t I 0 I m t I th I 0 I th 1 m cos t I 0 mI 0 I th cos t,
(10.38)
where I m t mI 0 I th cos t is the modulation current signal, which has an amplitude of

I m t mI 0 I th . Note that the modulation index dened in (10.38) for a semiconductor
laser is different from that dened in (10.27) for an LED because a laser has a threshold but an
LED does not have a threshold. In the regime of linear response, the output power of the laser
can be expressed in the same form as that in (10.28) of a directly modulated LED:
Pout t P0 Pm t P0 1 jr j cos t :
(10.39)
The constant output power P0 corresponding to the DC bias current I 0 can be found from
(10.35). However, the time-varying output power Pout t cannot be found directly from (10.35)
because the relation in (10.35) is valid only for the steady-state CW oscillation of a laser that is
injected with a DC current. When the injection current is temporally modulated, the timevarying output optical power of the laser in response to the modulation can be found by using
the relation Pout t out hvV mode St given in (9.29) after solving for the time-varying photon
density St from the coupled equations given in (10.32) and (10.33).
For small-signal modulation of m 1, the complex response function of a laser is
r jrjei
mc n
,
2r ir
2
(10.40)
where r is the relaxation resonance frequency and r is the total carrier relaxation rate for the
relaxation oscillation of the coupling between the carriers and the intracavity laser eld of the
semiconductor laser. They are related to the intrinsic dynamical parameters of the laser as
2r 4 2 f 2r c n s p
(10.41)
r s n p :
(10.42)
and
Because c and s are constants while n and p are linearly proportional to the laser power, r
and f r are proportional to the square root of the laser power, whereas r is a linear function of,
but not proportional to, the laser power. The relation between the relaxation resonance
frequency and the carrier relaxation rate is often characterized by a K factor that is independent
of the laser power:
K
r s
:
f 2r
(10.43)
317
Figure 10.8 Normalized current-modulation frequency response of a semiconductor laser measured in terms of
the electrical power spectrum using a photodetector. The frequency response of a semiconductor laser depends
on the output laser power, with its 3-dB bandwidth increasing approximately with the square root of the output
p
power. These curves are generated with the relations: f r G H z 5 Pout and s ns1 1:5 11 Pout , where
Pout is measured in mW.
The modulation power spectrum of a semiconductor laser is

Rf jr f j2
m2 2c 2n
:
2
16 4 f 2 f 2r 4 2 f 2 2r

As shown in Fig. 10.8, this spectrum has a resonance peak at
1=2
2r
2
f pk f r 2
8
(10.44)
(10.45)
and a 3-dB modulation bandwidth of

f 3dB
1=2
p 1=2 2
2r
1 2
f r p
1:554 f pk :
8 2 2

(10.46)
1=2
Because f r r =2 for most lasers and because f r / P0 , the modulation bandwidth of a

1=2
semiconductor laser increases with the output laser power and scales roughly as f 3dB / P0 .
An intrinsic modulation bandwidth on the order of a few gigahertz is common for a
semiconductor laser. A high-speed semiconductor laser can have a bandwidth larger than
20 GHz. Because the intrinsic modulation bandwidth of a semiconductor laser is signicantly
larger than that of an LED, it is very important to reduce the parasitic effects from electrical
contacts and packaging for high-frequency modulation of a semiconductor laser.
EXAMPLE 10.4
A semiconductor laser emitting at 850 nm has a current injection efciency of inj 60%
and an output coupling rate of out 5:7 1010 s1 . Its spontaneous carrier lifetime is
s 6:67 ns. It has a cavity decay rate of c 2 1011 s1 , a differential carrier relaxation rate
of n 4:9P0 109 s1 , and a nonlinear carrier relaxation rate of p 6:1P0 109 s1 , where
318
Optical Modulation
P0 is the laser output power measured in mW. The laser has a threshold current of I th 12 mA. It
is biased at a DC injection current of I 0 28 mA and is modulated with a modulation current at a
modulation frequency of f 10 GHz and a modulation index of m 10%. (a) Find the output
power of the laser at the DC bias point. (b) What is the amplitude of the modulation current?
(c) Find the relaxation resonance frequency f r and the total carrier relaxation rate r of this laser at
this operating point. What is the value of the K factor? (d) What are the amplitude of the
modulated output power and the phase delay of the response to the current modulation? (e) Find
the 3-dB modulation bandwidth of this laser at this operating point in terms of its modulation
response in the electrical power spectrum of the photodetector output. (f) At this modulation
frequency, what is the modulation response in the electrical power spectrum of the photodetector
used to measure the laser output? What is the normalized modulation response in dB?
Solution:
A laser has a threshold. Therefore, the DC output power is not proportional to its DC bias
current but is proportional to I 0 I th , and the modulation index is dened as the ratio of the
amplitude I m of the modulation current to I 0 I th .
(a) The photon energy at 850 nm is
hv
1239:8
eV 1:46 eV:
850
The DC output power of the laser is found using (10.35):

P0 inj
out hv
5:7 1010
I 0 I th 0:6
1:46 28 12 mW 4:0 mW:
c e
2 1011
(b) The amplitude of the modulation current for m 10% is

I m mI 0 I th 10% 28 12 mA 1:6 mA:
(c) With s 6:67 ns, c 2 1011 s1 , n 4:9P0 109 s1 , and p 6:1P0 109 s1
given, and P0 4:0 mW found in (a), we have
9 1
11 1
10 1
10 1
s 1
s 1:5 10 s , c 2 10 s , n 1:96 10 s , p 2:44 10 s :
Therefore, using (10.41) and (10.42), we nd

fr
1 p
c n s p 10 GHz,
2
r s n p 4:55 1010 s1 :

The K factor is found using (10.43):
K
r s 4:55 1010 1:5 109
s 440 ps:

2
f 2r
10 109
(d) For a modulation frequency of f 10 GHz, we nd that f f r , thus r , because f r

10 GHz as found in (c). Therefore, from (10.40), we nd
jr j
319
mc n mc n 0:1 2 1011 1:96 1010
1:37 101 ,
r
2f r 2 10 109 4:55 1010
1
1 r
rad:
tan
tan
2
2
0
2
r
Note that jr j > m at this modulation frequency because of the response enhancement from
relaxation resonance in a semiconductor laser. The amplitude of the modulated output
power is
Pm jr jP0 1:37 101 4 mW 548 W,
and the phase delay of the modulation response is =2 rad:
(e) The 3-dB modulation bandwidth of this laser is, from (10.46),
1=2
p 1=2 2
2r
f 3dB 1 2
f r p
8 2 2
1=2
p 1=2

45:52
2
1 2
10 p
GHz,
8 2 2
14 GHz

as seen in Fig. 10.8 from the 4 mW curve.

(f) At the modulation frequency of f 10 GHz f r , the modulation response in the electrical
power spectrum of the photodetector output is, from (10.44),
Rf

2
m2 2c 2n
jr f j2 1:37 101 1:88 102 :
2 2
2
4 f r
From (10.46), we have

R0
m2 2c 2n
,
16 4 f 4r
Therefore, for f 10 GHz f r , we nd that

2
Rf 4 2 f 2r 4 2 f 2r 4 2 10 109
2 2 2
2 1:91,
r
R0
f r
4:55 1010
and the normalized response is

10 log
Rf
10 log1:91 2:8 dB:
R0
10.4 REFRACTIVE EXTERNAL MODULATION
..............................................................................................................
The basic principle of refractive modulation is to modulate the real part of a principal dielectric
constant, thus modulating the corresponding principal refractive index, of an optical medium.
320
Optical Modulation
The direct effect is phase modulation on an optical wave that propagates through the medium.
Modulating the real part of a dielectric constant also changes the imaginary part because the real
and imaginary parts are intrinsically related through the KramersKronig relations. This effect
leads to undesirable amplitude modulation that appears as a side effect, which can be minimized
by operating the modulator at an optical carrier frequency that is far away from the transition
resonance frequencies of the material. For this reason, refractive modulation is generally
performed using a material that has little absorption in the spectral region of the modulated
optical wave. As discussed in Section 10.2, any other form of optical modulation can be
accomplished through phase modulation followed by properly manipulating the phasemodulated optical wave.
Refractive modulation through varying the principal refractive indices usually causes differential changes in the principal normal modes of polarization, resulting in induced linear or
circular birefringence, which can be applied to polarization modulation. The induced birefringence that is desired for a specic polarization modulation can usually be achieved by properly
choosing the parameters of the optical wave and the material. Therefore, polarization modulation can often be directly accomplished through proper refractive modulation without indirectly
manipulating a phase-modulated wave.
In principle, any physical mechanism that can cause a change in the refractive index of an
optical medium can be used for refractive modulation. Refractive modulation is most often
implemented through electro-optic modulation using the Pockels effect. It is also implemented
through magneto-optic modulation using the Faraday effect, through acousto-optic modulation
using Bragg diffraction, or through all-optical modulation using the optical-eld-induced
birefringence caused by the third-order nonlinear optical susceptibility. The concepts of these
physical mechanisms are discussed in Sections 2.6 and 2.7. The principles of refractive
modulation based on these physical mechanisms are discussed in the following.
10.4.1 Electro-optic Modulation

Practical electro-optic modulators are based on the Pockels effect, which is the rst-order electrooptic effect, though it exists only in noncentrosymmetric crystals, as discussed in Section 2.6. The
electro-optic Kerr effect, being a second-order effect, is relatively weak, and thus not practically
useful, though it exists in all materials. As seen in Section 2.6, depending on the direction and the
magnitude of the applied electric eld with respect to the principal axes of the crystal, the linear
birefringence induced by the Pockels effect results in changes in the principal indices that might
or might not be accompanied by a rotation of the principal axes. An electro-optically induced
rotation of the principal axes is not required for the functioning of an electro-optic modulator
though it often accompanies the index changes. However, the directions of the principal axes in
the presence of an applied electric eld, whether rotated or not, have to be taken into consideration in the design and operation of an electro-optic modulator. For simplicity without loss of the
general concept, we consider in the following a special case where the electro-optically induced
linear birefringence causes only index changes without rotating the principal axes.
We consider the LiNbO3 crystal, which is the most well-known electro-optic crystal. LiNbO3
is a negative uniaxial crystal of principal indices nx ny no > nz ne . Because of its 3m
symmetry, the r k matrix dened in (2.60) for its Pockels coefcients has only eight
321
nonvanishing elements with four independent values: r13 r 23 , r 12 r 61 r 22 , r 33 , and

r42 r 51 . In order that the electro-optically induced linear birefringence changes only the
values of the principal refractive indices without rotating the principal axes, the electric eld
is applied along the optical axis such that E0x E 0y 0 but E 0z 6 0. In this case, the changes
caused by the Pockels effect are found from (2.60) to be 1 2 r13 E 0z , 3 r33 E 0z ,
and 4 5 6 0, which can be expressed as xx yy r 13 E 0z , zz r33 E 0z ,
and ij 0 for i 6 j by applying the index contraction rule given in (2.59). By using (2.62)
and (2.63), the eld-dependent dielectric permittivity tensor can be found:
0 2
1
no n4o r 13 E 0z
0
0
A:
(10.47)
E 0 0 @
0
0
n2o n4o r13 E 0z
2
4
0
0
ne ne r 33 E 0z
^ ^x , Y^ ^y , and Z^ ^z . The crystal remains
Clearly, the principal axes are not rotated: X
uniaxial with the same optical axis, but the indices of refraction are changed. Since the induced
2
changes are generally so small that jr 13 E 0z j n2
o and jr 33 E 0z j ne , the new principal
indices of refraction can be expressed as
nX nY no
n3o
r 13 E 0z ,
2
nZ ne
n3e
r33 E 0z :
2
(10.48)
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,
^ Y^ , or Z^ , of the crystal that is subjected to a modulation eld. The preferred choice is a
X,
principal axis that has a large electro-optically induced index change but remains in a xed
direction as the magnitude of the modulation electric eld varies. In LiNbO3, this can be
accomplished by applying the electric eld along the z axis, as discussed above and shown in
Figure 10.9. There are two possible arrangements: transverse modulation, which has the
modulation eld perpendicular to the direction of optical wave propagation, as shown in
Fig. 10.9(a), and longitudinal modulation, which has the modulation eld parallel to the
direction of optical wave propagation, as shown in Fig. 10.9(b).
Figure 10.9 (a) LiNbO3 transverse electro-optic phase modulator for an optical wave propagating in the X
direction. (b) LiNbO3 longitudinal electro-optic phase modulator for an optical wave propagating in the Z
direction. In both cases, the modulation eld is applied in the Z direction. The ^x , ^y , and ^z unit vectors represent
the original principal axes of the crystal, and X^ , Y^ , and Z^ represent its new principal axes in the presence of
the modulation voltage.
322
Optical Modulation
Transverse Phase Modulation

We rst consider the situation of the transverse phase modulator shown in Fig. 10.9(a), where
the optical wave propagates in the X direction. In this case, the optical wave can be polarized in
either the Y or Z direction. If it is linearly polarized in the Z direction, its space and time
dependence can be written as

EX; t Z^ E exp ik Z X it Z^ E exp iz it:
(10.49)
For propagation through a crystal that has a length of l, the total phase shift is
n3e
n3e
l
Z
ne l r 33 E 0z l
ne l r 33 V
,
Z k l nZ l
2
2
c
c
c
d
(10.50)
where V E 0z d is the voltage applied to the modulator shown in Fig. 10.9(a).

For sinusoidal modulation of a modulation frequency f =2, the modulation voltage can
be written as
V t V m sin t,
(10.51)
which has a modulation amplitude of V m . The Z-polarized optical eld at the output plane,
X l, of the crystal is phase modulated:
El; t Z^ Eeine l=c expit m sin t,
(10.52)
n3e
l n3
l Vm
r 33 V m e r 33 V m
c 2
d
d V
(10.53)
where
m
is the peak modulated phase shift, known as the phase modulation depth, and
V
d
l
n3e r33
(10.54)
is the modulation voltage that is required for a phase shift of , known as the half-wave voltage,
also denoted as V =2 .
If the optical eld is instead linearly polarized in the Y direction, the phase shift after
propagation through the crystal is
n3o
n3o
l
Y
Y k l nY l
no l r13 E 0z l
no l r 13 V
:
(10.55)
2
2
c
c
c
d
The phase modulation depth for the modulation voltage given in (10.51) is then
m
n3o
l n3
l Vm
r 13 V m o r13 V m
,
c 2
d
d V
(10.56)
where the half-wave voltage for this arrangement is

V
d
:
l
n3o r 13
(10.57)
323
Because no ne but r 33 3:6r13 , it can be seen by comparing (10.57) with (10.54) that for a
desired modulation depth, the modulation voltage required for a Y-polarized optical wave is
about 3.6 times that for a Z-polarized wave.
Longitudinal Phase Modulation
For the longitudinal phase modulator shown in Fig. 10.9(b), an optical wave of any polarization in the XY plane experiences the same amount of phase shift because nX nY no . For a
crystal of a length l as shown in Fig. 10.9(b), we have
n3o
n3o
X Y
no l r 13 E 0z l
no l r 13 V
2
2
c
c
(10.58)
where V E 0z l for the longitudinal modulator. For a sinusoidal modulation voltage as given in
(10.51), the modulation depth of the longitudinal phase modulator is
m
n3o
n3
Vm
,
r 13 V m o r 13 V m
V
c 2
(10.59)
where
V
n3o r 13
(10.60)
Both m and V for longitudinal modulation are independent of the crystal length l.
It is seen that the voltage required for a given modulation depth is independent of the physical
dimensions of the modulator in the case of longitudinal modulation, whereas it is proportional
to d=l in the case of transverse modulation. One advantage of transverse modulation is that the
required modulation voltage can be substantially lowered by reducing the d=l dimensional ratio
of a transverse modulator. Another advantage is that the electrodes of a transverse modulator
can be made using standard techniques and can be patterned if desired, while those of a
longitudinal modulator have to be made of transparent conductors that can be very difcult,
if not impossible, to fabricate in the dimensions of the typical optical waveguide. However, if a
large input and output aperture is desired such that d=l > 1, it becomes advantageous to use
longitudinal modulation rather than transverse modulation.
EXAMPLE 10.5
LiNbO3 is a negative uniaxial crystal, which has nx ny no 2:251 and nz ne 2:170 at
the 850 nm wavelength. It has eight nonvanishing Pockels coefcients, which are r13
r23 8:6 pm V1 , r 12 r61 r 22 3:4 pm V1 , r 33 30:8 pm V1 , and r 42 r51
28 pm V1 . Consider transverse and longitudinal modulation of an optical wave at
850 nm using a LiNbo3 electro-optic modulator in the congurations shown in Figs. 10.9
(a) and (b), respectively. The LiNbo3 modulator has the dimensions of l 3 cm and d 3 mm.
(a) Find the values of the half-wave voltage V for transverse and longitudinal modulation,
respectively, in the case when the optical wave is polarized along the y principal axis. (b) The
324
Optical Modulation
largest Pockels coefcient is r 33 . If this coefcient can be used, what are the values of V for
transverse and longitudinal modulation, respectively?
Solution:
In both congurations shown in Figs. 10.9(a) and (b), the voltage is applied in the direction
along the z principal axis. Therefore, the Pockels coefcients that are useful for the modulation
are r 13 for x-polarized wave, r23 for y-polarized wave, and r 33 for z-polarized wave. Note that
r 13 r 23 8:6 pm V1 and r33 30:8 pm V1 .
(a) For a y-polarized wave, we use r 23 , which is the same as r 13 . For transverse modulation in this
case, the half-wave voltage is that given in (10.57). With l 3 cm and d 3 mm, we nd
V
d
850 109
3 103

V 867 V:
n3o r 13 l 2:2513 8:6 1012 3 102
For longitudinal modulation, the half-wave voltage is that given in (10.60):

V
850 109
V 8:67 kV:
n3o r 13 2:2513 8:6 1012
(b) To use r 33 , the optical wave has to be polarized along the z principal axis while the applied
voltage has to be in this direction as well. This is possible for transverse modulation but is
not possible for longitudinal modulation, as can be seen by examining Figs. 10.9(a) and (b).
For transverse modulation on a z-polarized optical wave in this case, the half-wave voltage
is that given in (10.54). With l 3 cm and d 3 mm, we nd
V
d
850 109
3 103

V 270 V:
n3e r 33 l 2:1703 30:8 1012 3 102
This half-wave voltage is less than one third of that found in (a) for transverse modulation
on a y-polarized optical wave because r 33 is more than three times larger than r 23 .
Polarization Modulation
As discussed in Section 10.2, polarization modulation can be accomplished by differential phase
modulation between two orthogonally polarized eld components. For electro-optic polarization
modulation, the optical wave is not linearly polarized in a direction that is parallel to any of the
principal axes in the presence of the modulation eld. The optical eld can be decomposed into
two linearly polarized normal modes. If the two normal modes see different eld-induced
refractive indices, an electric-eld-dependent phase retardation between the two modes occurs,
resulting in a change of the polarization of the optical wave at the output of the crystal.
The LiNbO3 transverse modulator discussed above becomes a polarization modulator if the
polarization of the input optical eld is not parallel to Y^ or Z^ so that

(10.61)
E0; t Y^ E Y Z^ E Z eit ,
325
Figure 10.10 LiNbO3 transverse electro-optic polarization modulator. The ^x , ^

y , and ^z unit vectors represent the
original principal axes of the crystal, and X^ , Y^ , and Z^ represent its new principal axes in the presence of the
modulation voltage.
where E Y 6 0 and E Z 6 0, as shown in Fig. 10.10. At the output, we nd

Y
Y
Z
El; t Y^ E Y eik l Z^ E Z eik l eit Y^ E Y Z^ E Z ei eik l eit ,
where

3
l
3
2ne no l no r 13 ne r 33 V
k k l
d

(10.62)
(10.63)
is the phase retardation between the Y and Z components. The polarization of the output optical
eld can be electro-optically modulated by a modulation electric eld of E 0z t V t =d that
causes a time-varying phase retardation of t following the time-varying voltage V t.
EXAMPLE 10.6
The phase retardation given in (10.63) between the Y and Z components of the optical eld for
the transverse polarization modulator shown in Fig. 10.10 has a background value that is
independent of the applied voltage V because ne 6 no . This voltage-independent background
phase retardation can be compensated by using a DC bias voltage of V b such that b
2m when V V b . Then (10.63) can be expressed as
b
V Vb
V Vb
2m
:
V
V
In practice, V b can be adjusted to make sure that b 2m. Find the expression for V in the
above relation. Use the parameters of LiNbO3 given in Example 10.5 to nd the value of V at
850 nm for a LiNbO3 polarization modulator of the dimensions of l 3 cm and d 3 mm.
Solution:
The expression for V can be found by taking while ignoring the voltage-independent
background term in (10.63). Thus, we nd that
V
n3o r13
d
:
3
ne r 33 l
Using the parameters given in Example 10.5, we nd that

V
850 109
3 103

V 392 V:
2:2513 8:6 1012 2:1703 30:8 1012 3 102
326
Optical Modulation
Amplitude Modulation
As discussed in Section 10.2, amplitude modulation can be achieved through polarization
modulation by properly selecting a polarization component of the polarization-modulated eld
while ltering out its orthogonal component. This can be done by simply placing a polarization
modulator between a polarizer at the input end and another polarizer, often referred to as an
analyzer, at the output end. The axes of the polarizer and the analyzer are often orthogonally
crossed, though other arrangements are possible. Figure 10.11 shows such an arrangement
using the LiNbO3 polarization modulator discussed above and shown in Fig. 10.10.
Following the discussion in Section 10.2 on polarization modulation and amplitude modulation, here we take
ê ê
ê ê
Y^ Z^
Y^ Z^
ê p , ê p , ê 1 Y^ p , ê 2 Z^ p ,
(10.64)
2
2
2
2
p
with c1 c2 1= 2. The axis of the input polarizer is along ê , and that of the output analyzer
is along ê , as shown in Fig. 10.11. The polarizer ensures that the input optical wave is linearly
polarized in the ê direction, whereas the analyzer passes only the ê component of the optical
wave at the output end. Thus, the input eld is

E
E0; t ê Eeit p Y^ Z^ eit :
2
(10.65)
Then, from (10.62), the eld at the end of the crystal is

Y

Y
E
E
1 ei ê 1 ei ê eik lit ,
El; t p Y^ Z^ ei eik lit
2
2
(10.66)
where is that given in (10.63). Because the analyzer passes only the ê component of the
optical eld, the transmittance of the amplitude modulator is
I out I
1
sin2
1 cos ,
(10.67)
I in
I
2
2
p
which agrees with (10.17) for c1 c2 1= 2.
Electro-optic amplitude modulation can also be accomplished by varying the coupling or
interference between two elds that have differential phase modulation, as discussed in Section
10.2. This concept can be implemented with many different structures, both in free space and in
waveguides. Here we illustrate the concept using a guided-wave electro-optic modulator in the
T
Figure 10.11 Electro-optic amplitude modulator using two cross polarizers at the input and the output of the
LiNbO3 transverse electro-optic polarization modulator shown in Fig. 10.10.
327
Figure 10.12 MachZehnder waveguide

interferometric modulator using
Y junctions fabricated on an x-cut, ypropagating LiNbO3 substrate.
form of the MachZehnder waveguide interferometer, shown in Fig. 10.12. This structure uses
Y-junction couplers as input and output couplers. It is fabricated in an x-cut, y-propagating
LiNbO3 crystal.
In the electrode conguration shown in Fig. 10.12, the modulation voltage is applied to the
central electrode while the outer electrodes are grounded so that the upper arm sees a modulation eld of E0z V=se but the lower arms sees E0z V=se , where se is the separation
between two neighboring electrodes. The modulation electric elds appearing in the two arms
point in opposite directions, resulting in a pushpull operation with equal but opposite phase
shifts in the optical waves propagating through the two arms. For an interferometer that has
identical arms, any other background phase shifts are exactly canceled. Thus the total phase
difference is twice the electro-optically induced phase shift in each arm. If the two arms are
identical single-mode waveguides, the phase difference induced by a modulation voltage V is

V
,
V
(10.68)
where V is the half-wave voltage for a phase difference of between the two arms. For a TElike mode, the transverse optical eld component is primarily the E z component so that
V
se
,
2n3e r 33 TE l
(10.69)
where TE is the overlap factor that accounts for the overlap between the spatial distributions of
the modulation eld and the TE-like mode eld. For a TM-like mode, the transverse optical
eld component is primarily the E x component so that
V
2n3o r 13 TM
se
,
l
(10.70)
where TM is the overlap factor that accounts for the overlap between the spatial distributions of
the modulation eld and the TM-like mode eld.
If both input and output Y junctions of the MachZehnder waveguide interferometer are ideal
3-dB couplers, i.e., the input power is split equally between the two arms and the elds from the
two arms are combined equally for the output, the power transmittance due to interference at the
output between the elds coming from the two arms is
328
Optical Modulation
Pout
1
cos2
1 cos :
Pin
2
2
(10.71)
Thus, electro-optic amplitude modulation can be accomplished through electro-optic phase

modulation to create a differential phase shift of between the two arms.
EXAMPLE 10.7
The x-cut, y-propagating LiNbO3 MachZehnder waveguide interferometer in the pushpull
conguration shown in Fig. 10.12 has identical single-mode waveguides for both arms, which
have connement factors of TE TM 0:5 for 850 nm. The electrodes have an equal
length of l 1 cm and an equal separation of se 10 m. Use the parameters of LiNbO3 given
in Example 10.5 to nd the half-wave voltage of this amplitude modulator for the TE-like mode
at 850 nm. What is the transmittance for an applied voltage of V 1 V?
Solution:
The half-wave voltage of this MachZehnder waveguide interferometer for the TE-like mode is
that given in (10.69) because the optical eld is primarily polarized in the z direction. Using the
LiNbO3 parameters given in Example 10.5, we nd that
V
se
850 109
10 106
V 2:7 V:

2n3e r 33 TE l
2 2:1703 30:8 1012 0:5 1 102
For an applied voltage of V 1 V, the phase difference between the two arms is

:
V 2:7
Therefore, the transmittance is found from (10.71) to be

1
1

T 1 cos
1 cos
69:8%:
2
2
2:7
10.4.2 Magneto-optic Modulation

As discussed in Section 2.6, the optical property of a material can be changed by a magnetization, M 0 , in a ferromagnetic or ferrimagnetic material or by an externally applied static or lowfrequency magnetic eld, H 0 , in a nonmagnetic material. Practical magneto-optic modulators
use the rst-order magneto-optic effect, i.e., the linear magneto-optic effect, because the
second-order magneto-optic effect is relatively weak. The linear magneto-optic effect, in which
the electric permittivity tensor at an optical frequency of is a linear function of M 0 or
H 0 , results in magnetically induced circular birefringence. Then, if the propagation direction is
taken to be along the z direction, the normal modes of propagation are circularly polarized with
329
the unit polarization vectors ê and ê given in (2.18) and the corresponding propagation
constants k and k given in (2.21):
k
,
with n n
c
2n
k
n
;
with n n
c
2n
(10.72)
where n and n are, respectively, the principal refractive indices seen by the ê and ê normal
modes. The parameter quantifying the linear magneto-optic effect on the refractive indices is
dened in (2.78). For an optical wave propagating along the z direction, is a linear function of
the z component M 0z with M 0z M 0z in the case of a magnetization, and H 0z
f 123 H 0z is linearly proportional to the z component H 0z in the case of an applied magnetic eld.
Magneto-optic refractive modulation on an optical wave can be performed through the
dependence of n and n on the magnetization or on the applied magnetic eld by varying
the magnetization or the applied magnetic eld. For magneto-optic phase modulation, the
optical wave has to be a circularly polarized normal mode rather than linearly polarized as in
the case of electro-optic phase modulation. A circularly polarized normal mode, either ê or ê ,
remains in the same polarization state as its phase is modulated when the wave is reected from
or is transmitted through a linear magneto-optic material.
If an optical wave is initially linearly or elliptically polarized, its eld is a superposition of the
two circularly polarized normal modes. This eld then decomposes into two circularly polarized orthogonal components that propagate with different propagation constants of k and k ,
respectively. Magneto-optic modulation on this wave causes differential phase modulation on
the two orthogonal circularly polarized modes, resulting in magneto-optic polarization
modulation on a wave that is initially linearly or elliptically polarized. The polarization change
caused by the linear magneto-optic effect on an optical wave that propagates through a
magneto-optic material is known as the Faraday effect. The polarization change caused by
the linear magneto-optic effect on an optical wave that is reected from the surface of a
magneto-optic material is known as the magneto-optic Kerr effect. Both effects lead to
magneto-optic polarization modulation. The Faraday effect is the mechanism used for optical
isolators and optical circulators. Magneto-optic polarization modulation can be converted into
magneto-optic amplitude modulation using polarizers by following the same principles as
discussed above for electro-optic modulation. The Faraday effect can then be used for
magneto-optic spatial light modulation. The magneto-optic Kerr effect is used for magnetooptic recording.
Of special interest is the Faraday rotation of a linearly polarized optical wave propagating in
a magneto-optic medium. Assume, without loss of generality, that the wave is initially linearly
polarized in the x direction at an arbitrary initial position taken to be z 0:
E
E0; t ^x Eeit p ê ê eit ,
2
(10.73)
p
with E E E= 2. Both circularly polarized components propagate as normal modes with
their respective propagation constants. When the wave propagates a distance of l in the positive
z direction, we have
330
Optical Modulation
El; t ê E exp ik ^z l 0 it ê E exp ik ^z l 0 it

ê E exp ik l it ê E exp ik l it

i

E h ik l
^x e eik l i^y eik l eik l eit

k k
k k
k k
l ^y sin
l exp i
l it :
E ^x cos
2
2
2
(10.74)
The optical eld clearly remains linearly polarized because its x and y components are in phase,
but its plane of polarization is rotated by an angle of
F tan1
E y k k
l n n l
l:
Ex
2
(10.75)
This magnetically induced rotation of the plane of linear polarization is called Faraday
rotation, and this phenomenon is the Faraday effect. It can be shown that the plane of
polarization rotates by the same amount in the same sense if the wave propagates in the
negative z direction for the same distance of l. Therefore, the sense of Faraday rotation is
independent of the direction of wave propagation. A device that provides the function of
Faraday rotation is called a Faraday rotator.
In a paramagnetic or diamagnetic material, which has no internal magnetization, the Faraday
rotation for a linearly polarized wave propagating over a distance of l is linearly proportional to the
externally applied magnetic eld. The Faraday rotation angle in this case is generally expressed as
F VH 0z l,
(10.76)
where
V
f 123 f 123
2cn
n
(10.77)
is the Verdet constant, measured in radians per ampere (rad A1 ). In practice, the Faraday
rotation angle is often expressed as F VB0z l in terms of the magnetic ux; in this case, the
Verdet constant is measured in radians per tesla per meter (rad T1 m1 ).
In a ferromagnetic or ferrimagnetic material, which has an internal magnetization, the total
Faraday rotation angle for an optical wave traveling over a distance of l through such a material
is simply
F F
M 0z
l,
Ms
(10.78)
where M 0z M s is the existing magnetization in the material and M s is the saturation

magnetization of the material. The Faraday rotation can be small if the material is not sufciently magnetized; it is maximized only when the material is fully magnetized to reach its
saturation magnetization. The Faraday rotation is thus characterized by the following specic
Faraday rotation, or rotatory power,
F
M s M s
,
2cn
n
(10.79)
331
which is the amount of rotation per unit length traversed by the optical wave in the material at
the saturation magnetization.
The Faraday effect is nonreciprocal. It has the characteristic that the sense of the Faraday
rotation in a specic material is independent of the direction of wave propagation and is
determined only by the direction of the external magnetic eld, or that of the magnetization in a
ferromagnet or ferrimagnet. The expression of F in (10.76) holds true for propagation in both
the parallel and the antiparallel directions with respect to H 0 , and that of F in (10.79) is also
valid for propagation in both directions with respect to M 0 . In the case when H0 , or M 0 , is not
^ only the longitudinal component of the magnetic
aligned with the wave propagation direction k,
eld, or that of the magnetization, in the k^ or k^ direction counts because the transverse
components that are perpendicular to the wave propagation direction do not contribute to
Faraday rotation. The amount of Faraday rotation is doubled, rather than canceled, when an
optical wave passing through a magneto-optic material is reected to retrace its original path in
the opposite direction back to the starting point. This phenomenon is a consequence of the fact
that the propagation constant of each circularly polarized normal mode is independent of the
wave propagation direction and, therefore, is not changed by reection.
The Faraday rotation is positive when the value of F , or that of F , is positive, meaning that
the rotation is counterclockwise when viewed in the direction against that of H 0 , or that of M 0
when an internal magnetization exists. Therefore, the sense of positive Faraday rotation is the
same as that of the electric current that generates H 0 or the current that can be conceptually
associated with M 0 in the case of a ferromagnet or ferrimagnet. Using the right-hand rule, the
axial vector corresponding to a positive Faraday rotation points in the same direction as that of
the H 0 or M 0 causing the Faraday effect. For negative Faraday rotation, the sense of rotation is
opposite to that of positive Faraday rotation. Figure 10.13 summarizes these concepts.
The nonreciprocity of Faraday rotation is important for optical isolation. Indeed, the Faraday
effect remains the unique physical mechanism for optical isolators and optical circulators which
are necessary components in sophisticated optical systems and networks. The basic structure of
an optical isolator consists of a Faraday rotator that has a total Faraday rotation angle of F
45
and two linear polarizers with axes oriented at 45
with respect to each other, as shown in
Fig. 10.14(a). An optical wave entering the device in the forward direction through the input
polarizer is linearly polarized by this polarizer. The linearly polarized wave emerging from the
Faraday rotator is transmitted by the output polarizer. For reverse isolation, an optical wave of
Figure 10.13 Positive Faraday rotation for an optical wave propagating in (a) a parallel direction and (b) an
antiparallel direction with respect to H 0 or M 0 . The sense of positive rotation is the same as the electric current
that can be associated with H 0 or with M 0 . For negative Faraday rotation, the sense of rotation is just the opposite.
332
Optical Modulation
Figure 10.14 (a) Basic structure and principle of a polarization-dependent optical isolator, which changes the
polarization direction at the output. (b) Two-stage cascaded optical isolator that does not change the
polarization direction at the output.
any polarization entering the Faraday rotator from the output end is linearly polarized by the output
polarizer. Because Faraday rotation is independent of the wave propagation direction, the
backward-propagating wave emerging from the Faraday rotator has a linear polarization that is
orthogonal to the axis of the input polarizer and is thus blocked. Figure 10.14(b) shows a two-stage
cascaded optical isolator that has input and output waves linearly polarized in the same direction.
EXAMPLE 10.8
An optical isolator of the conguration shown in Fig. 10.14(a) consists of a Faraday rotator
made of a TGG crystal, which has a length of l 5 cm. The TGG crystal has a Verdet constant
of V 40 rad T1 m1 at the 1:064 m wavelength of the Nd:YAG laser and
V 190 rad T1 m1 at the 532 nm wavelength. What is the required strength of the
magnetic induction B0z along the wave propagation direction for the isolator to function at each
of the two wavelengths, respectively?
Solution:
For the optical isolator of the conguration shown in Fig. 10.14(a), which has the polarizers
arranged such that the linear polarization rotates in the sense of a positive F , the required
Faraday rotation angle for a single pass is F =4. Therefore, the required magnetic induction
along the wave propagation direction for 1:064 m is
B0z
F
=4
T 393 mT,
Vl 40 5 102
333
and that required for 532 nm is

B0z
F
=4
T 82:7 mT:
Vl 190 5 102
Note that the magnetic induction has to point in the direction opposite to the forwardpropagating direction of the optical wave because the Verdet constant of TGG is negative.
10.4.3 Acousto-optic Modulation

An acoustic wave causes a space- and time-dependent periodic permittivity change in a
medium, as discussed in Section 2.6. For a traveling acoustic wave of the form expressed in
(2.79), which has a wavevector of K and a frequency of , the acousto-optically induced
permittivity change can be generally expressed in the form of (2.88) as
~ sin K r t ,
(10.80)
where the acoustic wavevector K depends on both the polarization and the propagation direction
of the acoustic wave. The wavenumber of the acoustic wave is K 2= 2f =v a , where v a is
the acoustic wave velocity, f =2 is the acoustic wave frequency, and v a =f is the acoustic
wavelength. The space- and time-dependent periodic permittivity change is effectively a timedependent grating, which diffracts an optical wave. In general, as expressed in (2.89), ~ is a
function of the strain and the rotation generated by the acoustic wave in the medium, the elastooptic coefcients of the medium, the mode and direction of the acoustic wave, and the frequency
and polarization of the optical wave, but it is independent of the values of K and . When an
optical wave at a frequency of is incident on this medium, the interaction between the optical
wave and the periodic modulation described by (10.80) can generate diffracted optical waves at
the frequencies of . The diffracted waves at can successively be diffracted to
generate waves at the frequencies of 2. If this process is allowed to cascade, it is possible
to generate a series of diffracted optical waves at the frequencies of q, where q admits both
positive and negative integers and is the order of acousto-optic diffraction:
q q:
(10.81)
The phase-matching condition for the qth-order diffraction is

kq ki qk,
(10.82)
where ki is the wavevector of the incident optical wave.

Acousto-optic modulation is fundamentally diffraction modulation. Because the frequency of
each diffraction order is shifted by an integral multiple of the acoustic frequency, digital
frequency modulation can be accomplished by switching between discrete acoustic frequencies
while analog frequency modulation can be performed using a continuously time-varying
acoustic frequency of t . Because ~ is generally a tensor, even when the medium is an
isotropic material, the polarization of a diffracted wave can be different from that of the incident
wave. Therefore, polarization modulation of a desired polarization change from the incident
334
Optical Modulation
wave to a diffracted wave can be accomplished by properly arranging the parameters of the
acoustic wave. By generating diffracted waves, acousto-optic modulation reduces the power of
the undiffracted optical wave at the original frequency and wavevector ki , thus imposing
amplitude modulation on this optical wave. To encode information through acousto-optic
amplitude modulation, the power of the acoustic wave has to be modulated with a timevarying signal by modulating the amplitude of the acoustic wave.
For the qth-order diffraction to occur, the frequency-shift condition given in (10.81) has to be
strictly obeyed, but the phase-matching condition given in (10.82) does not have to be exactly
satised. As we have learned from Section 4.6, perfect phase matching is necessary for the
maximum efciency, but a small phase mismatch does not completely prohibit the process
though it reduces the efciency. The degree of phase mismatch that can be tolerated in acoustooptic diffraction depends on the length of interaction between the optical wave and the acoustic
wave. The criterion is quantied by the factor:
Q
K 2l
,
k
(10.83)
where K is the wavenumber of the acoustic wave, k is the propagation constant of the optical
wave, and l is the interaction length between the two waves. Based on this criterion, there are
two separate regimes of acousto-optic diffraction.
In the regime of RamanNath diffraction, where Q 1, multiple diffraction orders can take
place simultaneously, as shown in Fig. 10.15. RamanNath diffraction occurs only when the
optical wave propagates in a direction that is normal, or nearly normal, to the propagation
direction K of the acoustic wave. Phase matching in the direction parallel to K is exactly satised
for each diffraction order that occurs, but a phase mismatch in the direction perpendicular to
K can be tolerated because of the short interaction length.
In the regime of Bragg diffraction, where Q 1, the phase-matching condition has to be
satised for a diffraction order to occur in response to an acoustic wave. In practice, it is often
Figure 10.15 (a) Conguration and (b) wavevector diagram for RamanNath diffraction in an isotropic
medium. Phase matching in the x direction determines the propagation angles of the diffracted waves.
Phase mismatch exists only in the z direction.
335
necessary to have Q 4 for clean Bragg diffraction. In its interaction with a traveling acoustic
wave, an incident optical wave, of the zeroth order with a wavevector of ki and a frequency of
, is directly coupled only to the two diffraction orders of q 1 and q 1. It can be seen
from (10.82) that the phase-matching condition for the generation of the diffraction order q 1
at the up-shifted frequency of 1 is
kd k1 ki K,
(10.84)
whereas that for the generation of the diffraction order q 1 at the down-shifted frequency of
1 is
kd k1 ki K:
(10.85)
For Bragg diffraction through the interaction with a given acoustic wave, the angle of
incidence of the incoming optical wave is not arbitrary but is determined by the phasematching condition. The required angle of incidence, i , for the incoming optical wave and
the angle of diffraction, d , at which the diffracted wave appears can be found by solving
(10.84) for the up-shifted diffraction, or (10.85) for the down-shifted diffraction. In the case
when the acoustic medium is an anisotropic crystal, the refractive indices ni and nd that are
respectively seen by the incident and diffracted optical waves can be different because the two
waves might have different polarizations. The solutions of the required incident angle and the
resulting diffraction angle are
1

K 2 k2i k2d
v 2a 2
1 f
2
1 2 2 ni nd ,
sin
2ki K
2ni v a
f
(10.86)
1

K 2 k2d k2i
v 2a 2
1 f
2
1 2 2 nd ni ,
sin
2kd K
2nd v a
f
(10.87)
i sin
d sin
where the upper signs are for up-shifted diffraction and the lower signs are for down-shifted
diffraction. For up-shifted diffraction, i < 0 and d > 0. For down-shifted diffraction, i > 0
and d < 0. Note that i and d are both measured with respect to the z direction, which is
normal to the K vector of the acoustic wave, and each of them can be either positive or
negative, as shown in Fig. 10.15 for the case of a positive i . In the case when the acousto-optic
diffraction takes place in an isotropic medium, ni nd n. Then,
ji j jd j B sin1
f
,
sin1
sin1
2k
2n
2nv a
(10.88)
where B is the Bragg angle. In this case, i B and d B for up-shifted diffraction, and
i B and d B for down-shifted diffraction.
For any diffraction order q to be generated, the phase-matching condition given in (10.82) has
to be satised for q q. In addition, because each diffraction order is directly coupled
only to its neighboring orders, Bragg diffraction at a high order requires the successive
generation of low diffraction orders, thus requiring simultaneous satisfaction of the corresponding phase-matching conditions. Except for some very special cases, these requirements cannot
be fullled. Consequently, only one diffraction order, either q 1 or q 1, is usually
336
Optical Modulation
generated in Bragg diffraction from a traveling acoustic wave. Bragg diffraction occurs in both
isotropic and anisotropic media when the phase-matching condition in (10.84) or (10.85) is
satised. Unlike RamanNath diffraction, the incident optical wave does not have to propagate in
a direction that is normal, or nearly normal, to the direction of propagation of the acoustic wave.
For Bragg diffraction, the optical wave can propagate in any direction, including the K direction
or the K direction, if the phase-matching condition for q 1 or q 1 can be satised.
The characteristics of acousto-optic diffraction from a standing acoustic wave are quite
different. A standing acoustic wave can be considered as a linear superposition of two contrapropagating traveling waves with both K and K simultaneously present for phase matching.
The implication of this situation is two-fold. (1) Both up-shifted and down-shifted frequencies
are simultaneously generated in each phase-matched diffraction direction, and (2) each shifted
optical frequency generated by diffraction can be diffracted back to the direction of the incident
wave with a further shift in frequency. This process cascades. For RamanNath diffraction from
a standing acoustic wave, each of the even spatial orders, including the undiffracted zeroth
order, consists of all of the frequencies up-shifted or down-shifted by the even multiples of ,
whereas each of the odd spatial orders consists of all of the frequencies up-shifted or downshifted by the odd multiples of . For Bragg diffraction from a standing acoustic wave, the
undiffracted beam in the ki direction contains a series of even side bands at 2m, and the
diffracted beam in the kd direction contains the odd side bands at 2m 1.
Traveling-Wave Acousto-optic Modulator
The majority of practical acousto-optic modulators take the conguration of small-angle Bragg
interaction, with the incident optical wavevector ki normal or nearly normal to the acoustic
wavevector K so that the diffraction angle is small. Figure 10.16 shows a traveling-wave
Figure 10.16 Representative solid-state traveling-wave acousto-optic modulator operating in the Bragg regime.
Up-shifted diffraction is illustrated here. For an anisotropic acousto-optic modulator, jd j 6 ji j. For an
isotropic acousto-optic modulator, jd j ji j B sin1 K=2k.
337
acousto-optic modulator operating under small-angle Bragg diffraction. Up-shifted diffraction

takes place in this conguration, where ki K < 0. Down-shifted diffraction occurs if the
optical wave is incident from a direction such that ki K > 0. The diffraction efciency of a
Bragg-type traveling-wave modulator with perfect phase matching can be expressed as
"
1=2 #
M2
2
l ,
(10.89)
PM sin
Pa
2HL
where is the optical wavelength; M 2 is the acousto-optic gure of merit determined by the
properties of the material, the mode of the acoustic wave, and the polarizations of the incident
and diffracted optical waves; H and L are respectively the height and the length of the
transducer that generates the acoustic wave; Pa is the power of the acoustic wave; and l is the
length of interaction between the optical wave and the acoustic wave. In the conguration of
small-angle Bragg interaction shown in Fig. 10.16, l L. In the low-efciency limit, the
diffraction efciency is linearly proportional to the acoustic power:
PM
2 M 2 l2
Pa ,
22 HL
if
PM 1:
(10.90)
EXAMPLE 10.9
A traveling-wave acousto-optic modulator made of silica glass is used to modulate an optical wave at
1:3 m using a longitudinal acoustic wave at an acoustic frequency of f 100 MHz. The silica
glass has a refractive index of n 1:447 at 1:3 m; it has an acoustic wave velocity of v a
5:97 km s1 and a gure of merit of M 2 1:50 1015 m2 W1 for a longitudinal acoustic wave
and an optical wave polarized in a direction that is perpendicular to the propagation direction K
of the acoustic wave. The transducer that generates the acoustic wave has the dimensions of
L 1:5 cm and H 2 mm; it delivers an acoustic power of Pa 500 mW. (a) Does this
modulator operate in the RamanNath regime or the Bragg regime? (b) What is the deection
angle between the diffracted beam and the incident beam? (c) What is the diffraction efciency?
Solution:
The wavelength of the acoustic wave is
v a 5:97 103
m 59:7 m:
f
100 106
For this acousto-optic modulator, the incident angle has to be small so that the interaction length
is approximately l L.
(a) The Q factor is
Q
K 2 l 2l 2 1:3 106 1:5 102
23:8 > 4:

2
k
n2
1:447 59:7 106
Therefore, the modulator works in the Bragg regime.
338
Optical Modulation
(b) The silica glass is isotropic. In the Bragg regime, the phase-matching condition requires
that the angles of incidence and diffraction have the same magnitude but opposite signs:
ji j jd j B sin1
1:3 106
0:43
:
sin1
sin1
2k
2n
2 1:447 59:7 106
Because i and d have opposite signs in both up-shifted and down-shifted diffraction, the
deection angle between the diffracted and incident beams for both cases is
jdef j jd i j 2B 0:86
:
(c) The diffraction efciency is small and can be found using (10.90):
PM

2
2 1:50 1015 1:5 102
2 M 2 l2
2
500 103 1:6%:
Pa

6 2
3
2
2 HL
2 1:3 10
2 10 1:5 10
Standing-Wave Acousto-optic Modulator

A standing-wave acousto-optic modulator provides sinusoidal amplitude modulation at a very
high frequency. It differs from a traveling-wave acousto-optic modulator in many important
aspects, ranging from the device structure to the performance characteristics. Figure 10.17
shows a standing-wave acousto-optic modulator operating under small-angle Bragg diffraction.
In order to create a standing acoustic wave, the acousto-optic cell is made to be a resonant
acoustic cavity. Instead of the angled surface of the acousto-optic cell of a traveling-wave
Figure 10.17 Representative solid-state standing-wave acousto-optic modulator operating in the Bragg regime.
For an anisotropic acousto-optic modulator, jd j 6 ji j. For an isotropic acousto-optic modulator,
jd j ji j B sin1 K=2k.
339
modulator, the surface at the far end across the cell width is made parallel to the near end, which
is attached to the piezoelectric transducer, as seen in Fig. 10.17. With a given cell width W
measured in the direction of the acoustic wave, a standing acoustic wave is formed only when
the acoustic wavelength satises the condition:
Wm ,
2
m integer:
(10.91)
Therefore, the device functions only at the discrete acoustic resonance frequencies of
f m
va
,
2W
m integer,
(10.92)
which are determined by the cell width and the acoustic velocity v a . The diffraction efciency
of a standing-wave acousto-optic modulator with perfect phase matching is
"
#
1=2
M2va
2
Pa
l cos t ,
(10.93)
PM sin
HLWa
where a is the decay rate of the acoustic energy in the acoustic cavity. In the low-efciency
limit, we have
PM
2 M 2 l2 v a
2 M 2 l2 v a
2
P
cos
t
Pa 1 cos 2t ,
a
2 HLWa
22 HLWa
if
PM 1:
(10.94)
Again, l L in the conguration of small-angle Bragg diffraction. As can be seen from (10.94),
the diffracted beam varies with time at a modulation frequency of f m 2f , which is twice the
frequency f of the acoustic wave. Therefore, the undiffracted beam is loss modulated at f m .
EXAMPLE 10.10
A standing-wave acousto-optic modulator made of silica glass is used to modulate an optical
wave at 1:3 m using a longitudinal acoustic wave at an acoustic frequency of
f 100 MHz. The silica glass has a refractive index of n 1:447 at 1:3 m; it has an
acoustic wave velocity of v a 5:97 km s1 and a gure of merit of M 2 1:50 1015 m2 W1
for a longitudinal acoustic wave and an optical wave polarized in a direction perpendicular to
the propagation direction K of the acoustic wave. The transducer that generates the acoustic
wave has the dimensions of L 1:5 cm and H 2 mm; it delivers an acoustic power of
Pa 500 mW. The acoustic cavity has a cell width of W 2 cm and a decay rate of
a 6 104 s1 . (a) Does this modulator operate in the RamanNath regime or the Bragg
regime? (b) What is the deection angle between the diffracted beam and the undiffracted beam?
(c) What is the modulation frequency at which the diffracted and undiffracted beams are
modulated? (d) What is the peak value of the diffraction efciency?
Solution:
The material and the parameters of the optical and acoustic waves are the same as those
described in Example 10.9 for the traveling acousto-optic modulator. Therefore, the answers
to (a) and (b) are the same as those in Example 10.9.
340
Optical Modulation
(a) The modulator works in the Bragg regime because Q 23:8 > 4:
(b) The deection angle between the diffracted and incident beams is
jdef j jd i j 2B 0:86
:
(c) The modulation frequency is twice that of the acoustic frequency:
f m 2f 200 MHz:
(d) The diffraction efciency is found using (10.93). The peak efciency is
pk
PM
"
1=2 #
M 2va
sin
Pa
l
HLWa
"
#
1=2
1:50 105 5:97 103 500 103

2
2
sin

1:5 10
2 103 1:5 102 2 102 6 104
1:3 106
2
15:5%:
10.4.4 All-Optical Refractive Modulation

All-optical modulation is accomplished through a nonlinear optical process based on a thirdorder susceptibility of the form 3 0 0 that causes optical-eld-induced permittivity changes in a nonlinear optical material. In the case when 0 , the process can
involve only one beam, as illustrated in Fig. 10.18(a), or two physically distinguishable beams
of the same frequency, as illustrated in Fig. 10.18(b). In the case when 0 6 , there are always
two optical beams in the interaction, as illustrated in Fig. 10.18(c). All-optical modulation can
be based on either self modulation or cross modulation. In the case of self modulation, only one
optical beam is present, and the modulation on the beam is a function of the characteristics of
the beam itself. In the case of cross modulation, two or more optical beams are present, and the
beam of interest is modulated by one or more other beams that carry the modulation signals. In
either case, no electric, magnetic, or acoustic eld is needed. Therefore, optical modulation and
switching based on nonlinear optical processes are known as all-optical modulation and alloptical switching, respectively.
There are two fundamentally different types of nonlinear optical modulation and switching.
One is the dispersive, or refractive, type, which is based on the optical-eld-induced changes in
the real part of the permittivity of a material. The other is the absorptive type, which relies on an
intensity-dependent absorption coefcient caused by the nonlinear characteristics of the imaginary part of the permittivity. All-optical refractive modulation is discussed here; all-optical
absorptive modulation is discussed in the following section.
In the case of one-beam interaction, we nd by using (2.101) and the intrinsic permutation
3
symmetry of ijkl that
341
Figure 10.18 Third-order processes for eld-induced susceptibility changes: (a) one-beam interaction,
(b) interaction of two beams of the same frequency, and (c) interaction of two beams of different frequencies.
3
Pi 3 0
X
j, k , l
ijkl E j E k E
l :
(10.95)
In the case of two-beam interaction, for either 0 or 0 6 , we have

X 3
3
ijkl E j E k E
Pi 3 0
l
j, k , l
X 3
0
ijkl 0 0 E j E k 0 E
6 0
l
j, k , l
3
(10.96)
and a similar expression for Pi 0 in terms of ijkl 0 0 0 0 and ijkl 0 0

1
. By identifying the total polarization at the frequency as Pi Pi Pi ,

we nd that the total optical-eld-dependent permittivity tensor can be expressed as
(10.97)
ij ; E ij ij ; E,
h
i
1
where ij 0 1 ij represents the eld-independent linear permittivity tensor of
the medium and ij ; E accounts for the optical-eld-dependent change induced by nonlinear optical interaction. For one-beam interaction,
X 3
ijkl E k E
(10.98)
ij ; E 3 0
l :
k, l
For two-beam interaction,
X 3
ij ; E 3 0
ijkl E k E
l
k, l
X 3
(10.99)
0
ijkl 0 0 E k 0 E
6 0
l :
k, l
The eld-dependent permittivity described here is the basis for various forms of all-optical
modulation. Because ij is a tensor, the nonlinear process discussed here generally leads to an
342
Optical Modulation
optical-eld-induced birefringence, known as the optical Kerr effect. The phase of an optical
eld can be modulated by itself through self-phase modulation or by another optical eld
through cross-phase modulation. All-optical polarization modulation can be accomplished
through optical-eld-induced birefringence. Such polarization modulation can be either selfinduced in a one-beam interaction or cross-induced in a two-beam interaction.
The simplest case involves a single linearly polarized optical wave in an isotropic medium
with the optical eld polarized in any xed direction, or in a cubic crystal with the optical eld
polarized along one of the principal axes. Then P3 is parallel to E of the optical eld, and the
3
only susceptibility element that contributes to this interaction is 1111 . Thus,
the permittivity seen by the optical eld is
3
; E 3 0 1111 jEj2
3 1111
I ,
2cn0
(10.100)
where n0 is the linear refractive index of the medium and I is the intensity of the optical
3
beam. We nd from this relation that the real part of 1111 leads to the
intensity-dependent index of refraction:
n n0 n2 I ,
(10.101)
where
30
n2
3 1111
4c 0 n20
(10.102)
is the coefcient of intensity-dependent index change.

The intensity-dependent index of refraction expressed in (10.101) represents the simplest
case of the optical Kerr effect. After the optical wave propagates through such a nonlinear
medium over a distance of l in the z direction, the total phase shift is
x; y; t
2
n0 n2 I x; y; t l 0 K x; y; t :
(10.103)
The intensity-dependent Kerr phase change,

K x; y; t
2
n2 I x; y; t l,
(10.104)
is the space- and time-dependent self-phase modulation because it depends on the intensity of
the optical wave itself. Depending on the material properties, the spatial and temporal proles
of the optical intensity, and the experimental conditions, this all-optical self-phase modulation
leads to the phenomena of self focusing, self defocusing, spectral broadening of optical pulses,
Kerr-lens mode locking, and optical solitons.
EXAMPLE 10.11
At 1:3 m, silica glass has a linear refractive index of n0 1:45 and a nonlinear suscepti30
bility of 1111 1:8 1022 m2 V1 . A laser pulse that has a wavelength of 1:3 m,
a Gaussian pulse shape with a FWHM pulsewidth of t ps 100 fs, and a peak power of
343
Ppk 1 kW propagates through a silica ber that has an effective core radius of a 5 m and a
length of l 1 m. In answering the following questions, ignore the effect of temporal pulse
broadening caused by the dispersion in the ber while the pulse propagates through the 1-m
ber because silica has zero group-velocity dispersion near 1:3 m. (a) Find the value of n2
for silica. (b) Find the optical-eld-induced index change at the peak of the pulse. (c) Find the
self-phase modulation due to the intensity-dependent Kerr phase change. (d) The timedependent phase modulation leads to frequency modulation. Find the percent of frequency
shifts, measured with respect to the original optical frequency, at the two half-width points of
the pulse. Does the frequency shift up or down on the leading and trailing edges of the pulse,
respectively?
Solution:
The Gaussian laser pulse of a peak power Ppk propagating in a ber of an effective core radius a
has a temporal intensity prole of
!
!
t2
P0
t2
I t I 0 exp 4 ln 2 2 2 exp 4 ln 2 2 :
t ps
a
t ps
(a) The value of n2 for silica is found using (10.102):
3 0
n2
3 1111
3 1:8 1022
m2 W1 2:4 1020 m2 W1 :

4c 0 n20 4 3 108 8:85 1012 1:452
(b) The optical-eld-induced index change at the peak of the pulse is

n n2 I pk n2
Ppk
1 103
20
7
2:4

10

3:1 10 :
6 2
a2
5 10
(c) The self-phase modulation due to the intensity-dependent Kerr phase change is
2
n2 I t l
!
2n2 lP0
t2
exp 4 ln 2 2
a2
t ps
K t
2 2:4 1020 1 1 103

t2
exp
4
ln
2

2
t 2ps
1:3 106 5 106
!
t2
1:48 exp 4 ln 2 2 rad,
t ps
!
rad
where tps 100 fs.

(d) Though K has a peak value of only 1:48 rad, the frequency modulation is in fact pretty
large because this self-phase modulation takes place within the pulse duration of
t ps 100 fs. Using (10.7), we nd that
344
Optical Modulation
!
K
n2 lP0 t
n2 lP0 t
t
1 8 ln 2
16 ln 2
:
t
a2 t2ps
ca2 t2ps
On the leading edge of the pulse, t < because t < 0; thus, the frequency shifts
down. On the trailing edge of the pulse, t > because t > 0; thus, the frequency
shifts up. This results in positive chirping, characterized by a frequency that increases with
time. At the two half-width points, t t ps =2, we nd that

n2 lP0
t 1 4 ln 2
ca2 t ps
!
2:4 1020 1 1 103
:
1 4 ln 2

2
3 108 5 106 100 1015
1 2:8%
The frequency shifts down by 2:8% at the half-width point of t tps =2 on the leading
edge of the pulse; it shifts up by 2:8% at the half-width point t t ps =2 on the trailing
edge of the pulse.
10.5 ABSORPTIVE EXTERNAL MODULATION
..............................................................................................................
In contrast to refractive modulation, the basic principle of absorptive modulation is to modulate
the imaginary part of a principal dielectric constant, thus modulating the absorption coefcient,
of an optical medium. By modulating the absorption coefcient, the desired effect of absorptive
modulation is clearly amplitude modulation on an optical wave that propagates through the
medium. Absorptive modulation is usually accompanied by a signicant change in the refractive index, creating undesirable phase modulation that leads to frequency chirping.
As discussed in the preceding section, modulating the real part of a dielectric constant for the
purpose of refractive modulation also changes the imaginary part because they are intrinsically
related through the KramersKronig relations. For the same reason, modulating the imaginary
part for absorptive modulation also changes the real part. For refractive modulation, undesirable
accompanying absorptive modulation can be avoided, or at least minimized, by operating a
modulator in a spectral region far away from any of the transition resonance frequencies of the
material, as discussed in Section 2.4. By contrast, undesirable accompanying refractive modulation is difcult to eliminate in the case of absorptive modulation because the absorption of a
photon takes place through a resonant optical transition at or near the resonance frequency
where the real part of the optical susceptibility varies together with the imaginary part, as seen
in Fig. 2.3.
Unless the optical frequency is tuned exactly at an isolated transition resonance frequency,
where the real part of the resonant susceptibility is zero, the change in the real part of the
resonant susceptibility is directly proportional to the change in the imaginary part. However,
practically it is not realistic to use this possibility in order to avoid the undesirable
345
accompanying refractive modulation for two reasons. (1) The material, such as a semiconductor, that is used for absorptive modulation often has a continuous absorption band rather
than isolated, discrete absorption frequencies. (2) In the case when absorptive modulation is
performed based on a transition between two discrete energy levels, such as the absorption line
of an exciton, the transition resonance frequency shifts under modulation. Because the refractive index near a resonance frequency varies nonlinearly with the optical frequency, as can be
seen in Fig. 2.3, the frequency chirping is often nonlinear and difcult to compensate. As
discussed in Section 10.3, this is also the case for direct modulation. Indeed, direct modulation
is a form of absorptive modulation where the carrier density of a semiconductor gain medium is
modulated. Modulating the gain coefcient is the same as modulating the absorption coefcient
because a gain coefcient is simply a negative absorption coefcient: g .
Absorptive modulation can cause different changes in the imaginary parts of the three
principal dielectric constants of a material, resulting in induced linear or circular dichroism,
which makes the absorption coefcients different for different normal modes of polarization, as
discussed in Section 2.2. Whether induced dichroism occurs or not depends on the properties of
the material used and the physical mechanism responsible for the absorptive modulation.
Induced dichroism has to be avoided to prevent undesirable polarization changes when an
optical wave is amplitude modulated through absorptive modulation. On the other hand,
induced dichroism can be used to accomplish desired polarization modulation at the same time
when an optical wave is amplitude modulated.
In principle, any physical mechanism that can cause a change in the absorption coefcient of
an optical medium can be used for absorptive modulation. Absorptive modulation is most often
implemented with semiconductor materials either using a modulation current, known as current
modulation, to change the carrier density, or using a modulation electric eld, known as
electro-absorption modulation, to change the energy bandgap of a bulk semiconductor, through
the FranzKeldysh effect, or the quantized energy subbands of a quantum-well structure,
through the quantum-conned Stark effect. All-optical absorptive modulation is possible
through the nonlinear optical effect of absorption saturation, or gain saturation in the case of
a gain medium. The principle of current modulation has already been discussed under direct
modulation in Section 10.3. Though current modulation can also be used for external absorptive
modulation, the principle is the same and thus is not further discussed here. The principle of
electro-absorption modulation and that of all-optical absorptive modulation through saturable
absorption are discussed below.
10.5.1 Electro-absorption Modulation

Practical electro-absorption modulators are semiconductor devices based on the FranzKeldysh
effect, for a bulk semiconductor, or based on the quantum-conned Stark effect, for a semiconductor quantum-well structure. Both effects are prominent near the bandgap. In addition, the
electro-absorption change caused by the quantum-conned Stark effect is signicantly
enhanced by excitons.
The absorption coefcient of a semiconductor is primarily determined by band-to-band
absorption, which excites an electron from a valence band to a conduction band, though
346
Optical Modulation
absorption involving impurity states is also possible. Band-to-band absorption creates free
electronhole pairs. It takes place only when the photon energy is larger than the bandgap of
the semiconductor, as shown in Fig. 10.19(a). For a direct-gap bulk semiconductor, such as
GaAs or InP, the absorption coefcient has the following characteristics:
p
/ E g for > E g ,
(10.105)
where Eg is the bandgap of the bulk semiconductor, is the optical frequency, and is the
photon energy. For an indirect-gap semiconductor, such as Si or Ge, band-to-band absorption
near the bandgap is assisted by phonon emission or phonon absorption, thus /
E g Ephonon 2 for > E g Ephonon near the bandgap, where E phonon is the phonon energy.
The electric elds seen by electrons and holes can be respectively expressed in terms of the
spatial gradients of the conduction- and valence-band edges as:
Ee
E c
e
Eh
E v
,
e
(10.106)
where Ec and E v are the conduction- and valence-band edges, respectively, and e is the electronic
charge. In the presence of an applied electric eld, E, the conduction- and valence-band edges
Figure 10.19 (a) Band-to-band absorption of a semiconductor in the absence of an applied electric eld.
(b) Band-to-band absorption of a semiconductor in the presence of an applied electric eld. (c) Change in
the absorption coefcient due to the FranzKeldysh effect for a direct-gap semiconductor.
347
remain parallel to each other and are tilted in the direction of the applied eld such that
Ee Eh E, as shown in Fig. 10.19(b). As a result, the wavefunctions of the electrons in
the conduction band and those of the holes in the valence band penetrate into the bandgap,
creating the probability for an electron and a hole to recombine through quantum mechanical
tunneling at an energy that is lower than the bandgap energy, as illustrated in Fig. 10.19(b). This
effect is known as the FranzKeldysh effect; it can be understood as electric-eld-assisted
absorption or photon-assisted tunneling for band-to-band transition at a photon energy below
Eg . Figure 10.19(c) shows the change in the absorption coefcient due to the FranzKeldysh
effect for a direct-gap semiconductor.
In a quantum-well structure, the electrons and holes are spatially conned within the nite
width, d QW , of the quantum well. This localization leads to the quantization of momentum in
the direction perpendicular to the quantum-well boundaries, resulting in discrete quantized
energy levels associated with the motion of the electrons and holes in this direction, as shown in
Fig. 10.20(a). In the horizontal dimensions, electrons and holes remain free and form energy
bands. As a result, both conduction and valence bands are split into a number of subbands
corresponding to the quantized levels. The minimum photon energy required for band-to-band
absorption in a quantum-well structure is the effective bandgap, EQW
g , of a quantum well, which is
no longer the bandgap Eg of the semiconductor material in the quantum well but is the separation
between the lowest subband of the conduction band and the highest subband of the valence band:
h > E QW
Eg
g
h2
h2
,
2
2
8m
8m
e d QW
h d QW
(10.107)
where h is the Planck constant, and m

e and mh are, respectively, the effective mass of electrons
in the conduction band and that of holes in the valence band.
EXAMPLE 10.12
The bandgap of GaAs at room temperature is E g 1:424 eV, corresponding to the energy of a
photon that has a wavelength of g 870:6 nm. A GaAs/AlGaAs quantum well has GaAs in
the quantum-conned well region; the width of the quantum well is d QW 20 nm. The electron
and hole effective masses for GaAs are m

e 0:067 m0 and mh 0:52 m0 , respectively, where
m0 is the free electron mass. Find the effective bandgap increase caused by the quantum
connement of the quantum well. What are the energy and the corresponding optical wavelength of a photon that can be absorbed by the quantum well?
Solution:
The effective bandgap increase of the quantum well is
h2
h2
2
2
8m
8m
e d QW
h d QW

2

6:626 1034
1
1
1
eV

2
31
9
0:067 0:52
1:6 1019
8 9:11 10
20 20
15:9 meV:
E QW
Eg
g
348
Optical Modulation
For a photon to be absorbed by the quantum well, the photon energy has to be
E g 15:9 meV 1:440 eV,
h > E QW
g
for a corresponding wavelength of
<
1239:8
nm 861 nm:
1:440
When an electric eld is applied on a quantum-well structure, the conduction- and valenceband edges are tilted in the direction of the applied eld, as in the case of a bulk semiconductor.
The lowest quantized conduction subband and the highest quantized valence subband, which
together dene the effective bandgap of the quantum well, are both affected by the applied
electric eld. As shown in Fig. 10.20(b), because of the connement of electrons and holes by
the quantum well, the applied electric eld shifts the electron and hole distributions in the
quantized subbands to opposite sides of the quantum well by distorting their wavefunctions.
Meanwhile, the tilted band edges allow the wavefunctions of the electrons and holes in these
quantized subbands to penetrate into the bandgap, thus lowering the effective bandgap of the
quantum well. This effect is known as the quantum-conned Stark effect.
Figure 10.20(c) shows the change in the absorption coefcient of a quantum-well structure
due to the quantum-conned Stark effect. In the absence of an applied electric eld, band-toband absorption starts at the minimum photon energy required by (10.107). Note that the
quantized energy levels of the quantum well are the band edges of the corresponding subbands;
therefore, band-to-band absorption from the highest valence subband to the lowest conduction
subband continues as the photon energy increases above the effective bandgap. The transition
probability between these quantum-well subbands remains constant until the photon energy
reaches the energy difference between the second conduction subband and the second valence
subband. Therefore, the absorption coefcient varies with the optical frequency as a step
function for photon energies near the effective bandgap EQW
g , as shown in Fig. 10.20(c). In the
presence of an applied electric eld, the quantum-conned Stark effect changes the absorption
coefcient in two ways, shown in Fig. 10.20(c). By lowering the effective bandgap, it shifts the
onset of absorption to a lower photon energy; by shifting the electron and hole distributions, it
reduces the spatial overlap of the electrons in the conduction subband and the holes in the
valence subband, thus reducing the value of the absorption coefcient.
Free excitons in a semiconductor can have a signicant effect on the optical transitions near
the bandgap. An electronhole pair in a semiconductor can be held together by their Coulomb
attraction to form an exciton, like an electronproton pair forming a hydrogen atom. A free
exciton is free to wander around in a semiconductor; its energy is reduced by the energy needed
to hold the electron and hole together so that it is slightly less than the bandgap of the
semiconductor. In a bulk semiconductor, free excitons can form only at very low temperatures
due to their low ionization energies. For this reason, excitonic absorption can be ignored for a
bulk semiconductor at room temperature; no excitonic correction on the electro-absorption due
349
Figure 10.20 (a) Band-to-band absorption in a quantum-well structure in the absence of an applied electric
eld. (b) Band-to-band absorption in a quantum-well structure in the presence of an applied electric eld.
(c) Change in the absorption coefcient due to the quantum-conned Stark effect without accounting for
excitonic absorption. (d) Change in the total absorption coefcient due to the quantum-conned Stark effect
including excitonic absorption.
to the FranzKeldysh effect shown in Fig. 10.19(c) is necessary. By contrast, in a quantum-well

structure, the ionization energies of free excitons are signicantly increased because the
electrons and holes are conned by the quantum wells and are thus localized. Consequently,
even at room temperature, the electro-absorption due to the quantum-conned Stark effect is
signicantly enhanced by excitonic absorption, which causes a strong absorption peak below
the effective bandgap energy EQW
g , as shown in Fig. 10.20(d).
As in the case of an electro-optic modulator, the modulation signal for an electro-absorption
modulator takes the form of a time-varying voltage. Electro-absorption modulators can be made
350
Optical Modulation
either in the waveguide form for easy integration with guided optical waves or in the lumped
form for free-space applications. In any event, the transmittance of an optical wave through an
electro-absorption modulator is
T V
I out
exp V l,
I in
(10.108)
where V is the modulation voltage, V is the voltage-dependent absorption coefcient at the

frequency of the optical wave, and l is the length of the absorption region that the optical wave
travels through. The transmittance T V is a nonlinear function of the modulation voltage V
because V is a nonlinear function of both the voltage V and the optical frequency , as can
be seen in Figs. 10.19(c) and 10.20(d), and T V is also a nonlinear function of V . For this
reason, electro-absorption modulation generally takes the form of digital modulation between
two xed voltages that represent the two binary bits of 0 and 1.
Clearly, an electro-absorption modulator functions only as an intensity modulator, i.e., an
amplitude modulator. Despite this limitation and despite its nonlinearity, electro-absorption
modulation can be easily applied on a semiconductor waveguide structure. Compared to
electro-optic modulation, it can be performed at a much higher speed for a bandwidth up to
tens of GHz and at a low voltage of a few volts. Compared to direct current modulation, it
produces much less frequency chirping because it does not inject carriers into the semiconductor, and it has a bandwidth as large as direct modulation on a fast semiconductor laser.
EXAMPLE 10.13
An electro-absorption modulator is used for binary modulation by switching between two
voltage levels for a maximum transmittance of T high T 0 at V 0 and a minimum transmittance of T low T V at V 6 0. Find the expressions for the maximum and minimum
transmittances and that for the extinction ratio.
Solution:
From (10.108), the maximum and minimum transmittances are simply
T high T 0 exp 0l, T low T V exp V l:
The extinction ratio measured in dB is found using (10.18) as
ER 10 log
T low
4:34V 0l dB:
T high
10.5.2 All-Optical Absorptive Modulation

As mentioned in Section 10.4 regarding all-optical modulation, all-optical absorptive modulation relies on an intensity-dependent absorption coefcient that is connected to the imaginary
351
part of the nonlinear optical susceptibility of a material. From (10.100), we nd that the
imaginary part of the total optical-intensity-dependent susceptibility is
00
100
3 00
3
1111 I :
2cn0 0
(10.109)
Therefore, the imaginary part of 1111 leads to an intensity-dependent change

3 00
in the loss or gain of a medium. As a general rule, the sign of 1111 that is
contributed by a single-photon transition of a resonance frequency at or near is always
the opposite to that of 100 . When 100 > 0, the medium has a linear loss; in this case,
300 < 0 so that the nonlinear susceptibility causes an intensity-dependent reduction of the loss,
resulting in absorption saturation. When 100 < 0, the medium has a gain; then, 300 > 0, and
it causes intensity-dependent gain saturation.
A saturable absorber has an absorption coefcient that decreases with increasing light
intensity, such as that characterized by (10.109) with 100 > 0 and 300 < 0. Note, however,
that the relation in (10.109) originates from taking the leading terms of the power series
expansion of linear and nonlinear polarizations expressed in (2.90). Because absorption saturation necessarily occurs at a resonant transition between two energy levels, the perturbation
approach taken for power series expansion is not valid at a sufciently high intensity. Instead, a
full analysis of the resonant absorption has to be carried out. Such an analysis results in an
intensity-dependent absorption coefcient characterized by the relation:
0
,
1 I=I sat
(10.110)
where 0 is the unsaturated absorption coefcient and I sat is the saturation intensity. The
saturation intensity is a characteristic of the resonant transition that is responsible for the
absorption. For I < I sat , the relation in (10.110) can be expanded:
"
#
2
3
I
I
I
0 1

:
(10.111)
I sat
I sat
I sat
Only when I I sat can be accurately approximated by the rst two terms of this expansion,
resulting in a linear dependence on I like the relation in (10.109). In general, the relation in
(10.110) has to be used because the light intensity encountered in a practical device that uses a
saturable absorber can easily be comparable to or higher than I sat .
The propagation of an optical wave through a saturable absorber that has an absorption
coefcient given in (10.110) is described by
dI
0
I:

1 I=I sat
dz
(10.112)
This equation can be integrated to obtain the relation:

I zeI z=I sat I 0eI 0=I sat e0 z ,
(10.113)
where I 0 is the input light intensity at z 0. The transmittance of an optical wave through a
saturable absorber of a thickness l is
352
Optical Modulation
Figure 10.21 Transmittance of an optical wave through a saturable absorber that has a thickness of l and an
unsaturated absorption coefcient of 0 as a function of the input light intensity normalized to the saturation
intensity. The curves are plotted for different values of 0 l in terms of T 0 e0 l .
I out I l
,
(10.114)
I in
I 0
which is a nonlinear function of the input intensity and can be found by numerically solving
(10.113). The transmittance is plotted in Fig. 10.21 as a function of the input light intensity,
normalized to the saturation intensity, for a few different values of 0 l represented in terms of
T 0 e0 l . As Fig. 10.21 shows, the optical transmittance through a saturable absorber
increases nonlinearly as the input intensity is increased, and it approaches unity at high input
intensities. In a specic application of a saturable absorber, the value of 0 l has to be properly
chosen for a desired difference between the maximum transmittance at high intensities and the
minimum transmittance at low intensities.
Saturable absorbers have many useful applications for self-intensity modulation of optical
beams or optical pulses. A saturable absorber can be used as a spatial light lter, which blocks
low-intensity stray light or background optical noise but transmits a high-intensity signal beam.
It can be used as an optical discriminator, which transmits optical pulses of intensities above a
certain threshold and suppresses those below. A saturable absorber is also commonly used as a
passive Q switch in a Q-switched laser or as a passive mode locker in a mode-locked laser for
the generation of very short laser pulses. The saturable absorber in this kind of application
functions as a passive optical switch in the time domain. It is switched open by the rising
intensity of a laser pulse but closes through its own relaxation after the pulse passes.
T
EXAMPLE 10.14
A saturable absorber is used for all-optical binary modulation by switching between two levels
of the optical intensity for a maximum transmittance of T max at a high input intensity of I in
I high and a minimum transmittance of T min at a low input intensity of I in I low . Find the input
intensity required for a transmittance of T when the saturable absorber has an unsaturated
Problems
353
absorption coefcient of 0 and a thickness of l. What is the minimum value of 0 l that is

required for this modulator to function?
Solution:
Using (10.113) for the output intensity and (10.114) for the transmittance, we have

I l
I 0 I l
I in
1 T 0 l ,
0 l exp
T
exp
I sat
I 0
I sat
where I in I 0 . Therefore, the required input intensity is
0 l ln T
I sat :
1T
A sufciently large value of 0 l is required for I in 0 for both high and low input intensities.
Because I low < I high , the value required for 0 l is found by making sure that I low 0. Thus,
I in
0 l ln T min ) 0 lmin ln T min :

Once the desired values of T max and T min are determined, a proper value of 0 l can be chosen
for the saturable absorber. Then the required input intensities I high and I low can be determined
with a known saturation intensity I sat of the absorber.
Problems
10.1.1 What is the difference between analog optical modulation and digital optical modulation?
How does the nonlinearity in the modulation response limit each type of modulation?
10.1.2 What is the difference between direct modulation and external modulation? What are the
advantages and disadvantages between the two?
10.1.3 What is the basic difference between refractive modulation and absorptive modulation?
Generally speaking, without considering specic physical mechanisms or device structures, which one is expected to have a faster modulation response?
10.2.1 Which modulation scheme is the most fundamental among all of the optical modulation
schemes? Why is it fundamental?
10.2.2 Briey describe how frequency modulation, polarization modulation, amplitude modulation, spatial modulation, and diffraction modulation can each be accomplished through
phase modulation.
10.2.3 Briey describe how information is coded on an optical carrier wave through each
of the following modulation schemes: (a) BPSK and QPSK, (b) BFSK and QFSK,
(c) BPolSK, and (d) OOK.
10.2.4 An optical eld is initially linearly polarized in the direction of the unit polarization vector
p
ê ^x ^y = 2. The two linearly polarized components of the mutually orthogonal x and
354
Optical Modulation
y polarizations are differentially phase modulated for polarization modulation of the optical
eld. Find its orthogonal unit polarization vector ê on the f^x ; ^y g basis. How does the
polarization of this eld change if the two orthogonally polarized components are differentially phase modulated by a phase difference of =4, =2, , and 2, respectively?
10.2.5 An optical eld is initially linearly polarized in the direction of the unit polarization
p

vector ê ^x ^y = 2. The
polarized
p two circularly
p components of the mutually

orthogonal ê ^x i^y = 2 and ê ^x i^y = 2 polarizations are differentially
phase modulated for polarization modulation of the optical eld. Find its orthogonal unit
polarization vector ê on the fê ; ê g basis. How does the polarization of this eld
change if the two orthogonally polarized components are differentially phase modulated
by a phase difference of =4, =2, , and 2, respectively?
10.2.6 Describe two approaches to modulating the output intensity of a waveguide structure by
modulating the phase of a waveguide mode. Give an example for each approach.
10.2.7 An optical wave is normally incident at i 0 on a diffraction grating. A diffraction
order q appears at the diffraction angle of q 30
. If the grating period can be varied
within a range of 10% for diffraction modulation, what is the angular range of
variations for this diffraction order?
10.3.1 For the direct current modulation of an LED with a sinusoidal modulation current as
given in (10.27), show that the output power of the LED has the modulation response
given in (10.28) with the complex response of (10.29) as a function of the modulation
index m and the modulation frequency .
10.3.2 For the LED described in Example 10.3, it is desired that the amplitude of the modulated
output power be Pm 500 W when it is modulated with a modulation index of m
10% at a modulation frequency of f f 3dB 15:9 MHz. This goal can be accomplished by adjusting the bias current I 0 and correspondingly the amplitude I m of the
modulation current. Find the values of I 0 and I m for this purpose.
10.3.3 To increase the 3-dB modulation bandwidth of an LED, the spontaneous carrier lifetime s of
the LED can be prescribed by properly controlling the impurity concentration in the LED. If a
3-dB bandwidth of 50 MHz is desired for an LED, what is the required value for s ? What is
its normalized modulation response measured in dB at a modulation frequency of 20 MHz?
10.3.4 An LED emitting at a center wavelength of 1:3 m has an external quantum
efciency of e 26%. Its spontaneous carrier lifetime is s 3 ns. The LED is biased
at a DC injection current of I 0 10 mA and is modulated at a modulation frequency of
f 40 MHz with a modulation current for a modulation index of m 10%.
(a) Find the output power of the LED at the DC bias point.
(b) What is the amplitude of the modulation current?
(c) What are the amplitude of the modulated output power and the phase delay of the
response to the current modulation?
(d) Find the 3-dB modulation bandwidth of this LED.
(e) At this modulation frequency, what is the modulation response in the electrical
power spectrum of the photodetector that is used to measure the LED output? What
is the normalized modulation response measured in dB?
Problems
355
10.3.5 For the direct current modulation of a semiconductor laser with a sinusoidal modulation
current as given in (10.38), show that the output power of the laser has the modulation
response given in (10.39) with the complex response of (10.40) as a function of the
modulation index m and the modulation frequency .
10.3.6 The semiconductor laser described in Example 10.4 is biased at a DC injection current
level of I 0 so that its output power is P0 5 mW at this DC bias point. It is then
modulated with a modulation current at a modulation frequency of f 10 GHz for a
modulation index of m 10%.
(a) Find the required DC bias injection current level I 0 .
(b) What is the amplitude I m of the modulation current required for a modulation index
of m 10%?
(c) Find the relaxation resonance frequency f r and the total carrier relaxation rate r of
this laser at this operating point. What is the value of the K factor?
(d) What are the amplitude of the modulated output power and the phase delay of the
response to the current modulation at the modulation frequency of f 10 GHz?
(e) Find the 3-dB modulation bandwidth of this laser at this operating point in terms of
its modulation response in the electrical power spectrum of the photodetector.
(f) At the modulation frequency of f 10 GHz, what is the modulation response in the
electrical power spectrum of the photodetector that is used to measure the laser
output? What is the normalized modulation response measured in dB?
10.3.7 The 3-dB bandwidth of a semiconductor laser can be increased by increasing the output
power of the laser at the bias point through increasing the bias injection current.
A GaAs/AlGaAs quantum-well semiconductor laser emitting at 827:6 nm has the
following parameters: cavity decay rate c 2:4 1011 s1 , spontaneous carrier relaxation rate s 1:458 109 s1 , differential carrier relaxation rate n 1:55P0 108 s1 ,
and nonlinear carrier relaxation rate p 2:8P0 108 s1 , where P0 is the laser output
power at the bias point measured in mW.
(a) Find the relaxation resonance frequency f r and the total carrier relaxation rate r of
this laser as functions of the laser output power P0 .
(b) Find the value of the K factor for this laser.
(c) What is the 3-dB modulation bandwidth of the laser when it is biased at an output
power of P0 10 mW?
(d) What is the laser output power at the bias point required for the laser to have a 3-dB
modulation bandwidth of f 3dB 5 GHz?
10.3.8 A semiconductor laser emitting at 1:3 m has an external quantum efciency of
e 21:5%. It has a cavity decay rate of c 5:36 1011 s1 , a spontaneous carrier
relaxation rate of s 5:96 109 s1 , a differential carrier relaxation rate of
n 1:67P0 109 s1 , and a nonlinear carrier relaxation rate of p 4:24P0
109 s1 , where P0 is the laser output power measured in mW. The laser has a threshold
current of I th 18 mA. It is biased at a DC injection current of I 0 50 mA and is
current modulated with a modulation index of m 10% at a modulation frequency of
f 10 GHz.
356
Optical Modulation
(a) Find the output power of the laser at the DC bias point.
(b) What is the amplitude of the modulation current?
(c) Find the relaxation resonance frequency f r and the total carrier relaxation rate r of
this laser at this operating point. What is the value of the K factor?
(d) What are the amplitude of the modulated output power and the phase delay of the
response to the current modulation?
(e) Find the 3-dB modulation bandwidth of this laser at this operating point.
(f) At this modulation frequency, what is the modulation response in the electrical
power spectrum of the photodetector that is used to measure the laser output? What
is the normalized modulation response measured in dB?
10.4.1 LiNbO3 is a negative uniaxial crystal, which has nx ny no 2:222 and nz ne
2:145 at the 1:3 m wavelength. It has eight nonvanishing Pockels coefcients,
which are r13 r 23 8:6 pm V1 , r 12 r61 r 22 3:4 pm V1 , r 33 30:8
pm V1 , and r 42 r 51 28 pm V1 . For electro-optic phase modulation using a

LiNbO3 electro-optic modulator, the half-wave voltage V can be signicantly reduced
by transverse modulation in a waveguide structure while using the largest Pockels
coefcient. How can the lowest possible V be accomplished by properly arranging
the optical wave and the applied voltage with respect to the crystal axes and the
waveguide structure? What is this lowest value of V for a waveguide that has the
dimensions of l 3 mm and d 2 m?
10.4.2 KTP is a biaxial crystal of the mm2 symmetry group, which has nx 1:742,
ny 1:750, and nz 1:832 at the 1:0 m optical wavelength. Its only nonvanishing
Pockels coefcients are r 13 8:8 pm V1 , r 23 13:8 pm V1 , r 33 35 pm V1 ,
r 42 8:8 pm V1 , and r 51 6:9 pm V1 . A KTP electro-optic transverse phase

modulator has the conguration shown in Fig. 10.9(a) with the modulation voltage
applied along the z principal axis. Answer each of the following questions for an
optical wave at 1:0 m that is linearly polarized in the z direction and propagates
in the x direction.
(a) Find the phase modulation depth m as a function of the parameters of KTP, the
dimensions of the modulator, and the peak modulation voltage V m . What is the halfwave voltage required for a modulated phase shift of ?
(b) Find the half-wave voltage V required for a modulated phase shift of for a bulk
modulator that has the dimensions of d 3 mm and l 6 mm.
(c) Find the half-wave voltage V required for a modulated phase shift of for a
waveguide modulator that has the dimensions of d 3 m and l 6 mm.
Figure 10.22 GaAs electro-optic longitudinal modulator.
Problems
357
10.4.3 Consider a GaAs longitudinal electro-optic modulator as shown in Fig. 10.22. GaAs is a
nonbirefringent cubic crystal of the 43m symmetry group. At 1:0 m, it has
nx ny nz no 3:5, and its only nonvanishing Pockels coefcients are
r 41 r 52 r 61 1:2 pm V1 . In the illustration, ^x , ^y , and ^z are the intrinsic principal

^ , Y^ , and Z^ are the new principal
axes of GaAs without an applied voltage, whereas X
axes when a voltage is applied in the z direction as shown.
^ and Y^ ,
(a) For two linearly polarized input optical elds that are polarized along X
respectively, nd the phase changes X and Y in the two elds at the output as
functions of the parameters of GaAs, the dimensions of the modulator, and the
modulation voltage V.
(b) This device can be used as an electro-optic polarization modulator. Describe how
the input eld polarization has to be arranged for the device to function as a voltagecontrolled half-wave plate that rotates the optical eld polarization direction of a
linearly polarized input eld by 90
at the output.
(c) Find the half-wave voltage V required for the modulator to function as a polarization modulator as described in (b) if its dimensions are d 3 mm and l 1 cm.
10.4.4 Answer the questions in Example 10.7 for the TM-like mode instead of the TE-like
mode considered in Example 10.7. What is the lowest voltage required for a transmittance of 50%?
10.4.5 Consider an x-cut, y-propagating KTP MachZehnder waveguide interferometer in
the pushpull conguration as shown in Fig. 10.12 for the LiNbO3 interferometer.
KTP is a biaxial crystal of the mm2 symmetry group, which has nx 1:742, ny
1:750, and nz 1:832 at the 1:0 m optical wavelength. Its only nonvanishing
Pockels coefcients are r 13 8:8 pm V1 , r 23 13:8 pm V1 , r 33 35 pm V1 ,
r 42 8:8 pm V1 , and r 51 6:9 pm V1 . The KTP MachZehnder waveguide interferometer has identical single-mode waveguides for both arms, which have connement
factors of TE TM 0:7 for 1:0 m. The electrodes have an equal length of
l 3 mm and an equal separation of se 8 m.
(a) Find the half-wave voltage of this amplitude modulator for the TE-like mode at
1:0 m. What is the lowest voltage required for a transmittance of 30%?
(b) Answer the questions in (a) for the TM-like mode.
10.4.6 A Faraday rotator consists of a TGG crystal in a magnetic eld that has a ux density
of B0z 0:35 T along the longitudinal axis of the crystal. The Verdet constant of
TGG is V 80 rad T1 m1 at the 750 nm wavelength and V 65 rad T1 m1
at the 800 nm wavelength. If a linearly polarized optical wave at the 750 nm
wavelength is sent through the TGG Faraday rotator, what is the required length of
the crystal for a Faraday rotation angle of 45
in a single pass? In which sense does
the polarization rotate? With this magnetic eld and this crystal length, what is the
Faraday rotation angle in a single pass for a linearly polarized wave at the 800 nm
wavelength?
10.4.7 Ce3+P glass has a Verdet constant of V 94:7 rad T1 m1 at the 500 nm optical
wavelength. A Ce3+P glass rod of a length l 5 cm is placed between two cross
polarizers, which have orthogonally oriented transmission polarization directions. An
358
Optical Modulation
optical beam at the 500 nm wavelength is polarized in the transmission polarization

direction of the input polarizer and is sent through this system.
(a) What is the required magnetic ux density B0 that has to be applied along the
propagation direction of the beam for the beam to be completely transmitted at the
output of the system, i.e., at the output end of the second cross polarizer?
(b) If the magnetic ux density is half of that found in (a), what is the percentage
transmission of the beam through the system?
(c) If the magnetic ux density is twice that found in (a), what is the percentage
transmission of the beam through the system?
(d) If the magnetic ux density remains that found in (a) but the direction of the
magnetic eld is oriented at 45
with respect to the propagation direction of the
beam, what is the percentage transmission of the beam through the system?
10.4.8
Show that as required by the phase-matching condition given in (10.84) for up-shifted
diffraction and that in (10.85) for down-shifted diffraction in the Bragg regime, the
angle of incidence for the incoming optical wave is that given in (10.86) and the angle
of diffraction for the diffracted wave is that given in (10.87) with the upper signs for
up-shifted diffraction and the lower signs for down-shifted diffraction.
10.4.9
LiNbO3 is a negative uniaxial crystal, which has nx ny no 2:222 and nz ne

2:145 at the 1:3 m wavelength and nx ny no 2:291 and nz ne 2:201
at the 632:8 nm wavelength. It is a very good acousto-optic crystal that can be
used at high acoustic frequencies up to 5 GHz. The acoustic wave velocity in LiNbO3
depends on the polarization and propagation directions of the acoustic wave. The
acoustic wave velocity is v La 6:57 km s1 for a longitudinal acoustic wave that
propagates in the 100 crystal direction such that K K L ^x , and it is v Ta 3:59 km s1
for a transverse acoustic wave that propagates in the 001 crystal direction such that
K K T^z . Consider the diffraction of an optical wave by a traveling acoustic wave
through an interaction length of l 5 mm.
(a) The optical wave is linearly polarized along the z principal axis such that E E^z
and the acoustic wave is the longitudinal wave with K K L ^x . For each optical
wavelength, nd the minimum acoustic frequency required for the diffraction to be
in the Bragg regime.
(b) The optical wave is linearly polarized along the x principal axis such that E E^x
and the acoustic wave is the transverse wave with K K T^z . For each optical
wavelength, nd the minimum acoustic frequency required for the diffraction to be
in the Bragg regime.
10.4.10 Silica glass has a refractive index of n 1:452 at 850 nm. It has an acoustic wave
velocity of v a 5:97 km s1 and an acousto-optic gure of merit of M 2 1:50
1015 m2 W1 for a longitudinal acoustic wave and an optical wave polarized in a
direction perpendicular to the propagation direction K of the acoustic wave.
A traveling-wave acousto-optic modulator made of silica glass in the conguration
shown in Fig. 10.16 is used to modulate an optical wave at 850 nm. The transducer
that generates a longitudinal acoustic wave has the dimensions of L 1 cm and
H 3 mm; it delivers an acoustic power of Pa 300 mW.
Problems
359
(a) What is the acoustic frequency required for this modulator to operate in the Bragg
regime?
(b) If the acoustic frequency is chosen to be f 300 MHz, what is the Bragg angle?
What is the deection angle between the diffracted beam and the incident beam?
(c) What is the diffraction efciency?
10.4.11 Silica glass has a refractive index of n 1:452 at 850 nm. It has an acoustic wave
velocity of v a 5:97 km s1 and an acousto-optic gure of merit of M 2 1:50
1015 m2 W1 for a longitudinal acoustic wave and an optical wave polarized in a
direction perpendicular to the propagation direction K of the acoustic wave.
A standing-wave acousto-optic modulator made of silica glass in the conguration
shown in Fig. 10.17 is used to modulate an optical wave at 850 nm. The transducer
that generates a longitudinal acoustic wave has the dimensions of L 1 cm and
H 3 mm; it delivers an acoustic power of Pa 300 mW. The acoustic cavity has
a cell width of W 1 cm and a decay rate of a 9:1 104 s1 .
(a) It is desired that the optical wave is modulated at a modulation frequency of
f m 300 MHz. What is the acoustic frequency required for this purpose?
(b) At the acoustic frequency found in (a), what is the minimum required acoustooptic interaction length l for Bragg diffraction? Does the acousto-optic modulator
satisfy this requirement?
(c) What is the deection angle between the diffracted beam and the undiffracted beam?
(d) What is the peak value of the diffraction efciency?
10.4.12 A laser beam that has a transverse spatial intensity distribution can cause a spatially
varying intensity-dependent Kerr phase change as expressed in (10.104) through selfphase modulation. For a circular beam that propagates along a longitudinal direction
taken to be the z direction, the transverse spatial intensity prole can be expressed as
1=2
I r as a function of the radial variable r x2 y2 .

(a) The radially varying intensity-dependent Kerr phase has the same effect as a thin
lens. The effective focal length f K of the Kerr lens is given by the relation:

1
c d2 K
a
,
(10.115)
fK
dr 2 r0
where a is a correction factor that depends on the prole of the circular optical
beam. Find the relation between f K and the intensity prole I r .
(b) The intensity prole of a fundamental circular Gaussian beam at a xed z
location is

r2
I r I 0 exp 2 2 ,
(10.116)
w
where I 0 is the intensity at the beam center and w is the beam spot size at the given
z location. Find the Kerr focal length f K for the Gaussian beam as a function of the
beam parameters. Express it also in terms of the power P w2 =2 of the beam.
(c) For a circular Gaussian beam, a 1:723. An ultrashort laser pulse at 532 nm
has a peak power of Ppk 10 kW. It has a fundamental circular Gaussian beam
360
Optical Modulation
intensity prole and is focused on a thin silica piece of l 1 mm to a focused spot

size of w 12 m. For silica, the coefcient of intensity-dependent index change
is n2 2:4 1020 m2 W1 . Find the Kerr focal length f K for the Gaussian beam
at its pulse peak.
10.4.13 For a cubic crystal of the 43m symmetry group, the only nonvanishing third-order
3
3
3
3
nonlinear susceptibilities are the elements of the types 1111 , 1122 , 1212 and 1221 ,
3
3
3
3
3
3
2
where 1111 represents xxxx
yyyy
zzzz
and 1122 represents xxyy
yyzz
3
, and so forth. Consider cross-phase modulation on an optical wave at a
zzxx
frequency of by another optical wave at a frequency of 0 , both propagating along
the z principal axis of the crystal. The
optical eld at is initially linearly
pmodulated

polarized as E E ^x ^y = 2, and the modulating optical eld at 0 is
linearly polarized along the x principal axis as E0 E0 ^x . The modulating eld
at 0 is so much stronger than the modulated eld at that the self-phase modulation
of the eld at can be ignored in comparison to the cross-phase modulation by the
eld at 0 .
(a) Find the intensity-dependent refractive indices nx and ny seen by the x and y
components of the modulated eld at as functions of the intensity I 0 of the
modulating eld at 0 .
(b) What is the minimum intensity of the modulating eld required
pfor
the modulated

eld to be linearly polarized in a direction parallel to ^x ^y = 2 after both waves
propagate through the crystal over a distance of l?
10.5.1
The bandgap of GaAs at room temperature is Eg 1:424 eV, corresponding to the

energy of a photon that has a wavelength of g 870:6 nm. A GaAs/AlGaAs quantum
well has GaAs in the quantum-conned well region that has a width of d QW . The
electron and hole effective masses in GaAs are m

e 0:067 m0 and mh 0:52 m0 ,
respectively, where m0 is the free electron mass. It is desired that at room temperature,
the absorption edge of the quantum well be at the optical wavelength of 840 nm
when no voltage is applied to the structure. Find the required width d QW of the
quantum well.
10.5.2
A waveguide electro-absorption modulator for 1:55 m is based on the quantum

Stark effect. It consists of 10 InGaAs/InGaAsP quantum wells. The optical path length
in the waveguide is l 250 m. The voltage-dependent effective absorption coefcient
of the TE waveguide mode has the values of 0 400 m1 , 2 V 5 103 m1 ,
4V 9:6 103 m1 , 6 V 1:34 104 m1 , and 8 V 1:6 104 m1 .
(a) What is the highest possible transmittance T high ? What is the voltage for this
transmittance?
(b) With respect to the highest transmittance found in (a), a low transmittance can be
chosen by applying a specic voltage. Find the low-transmittance value for each
possible voltage listed above other than the high-transmittance voltage found in
(a). Find the extinction ratio, in dB, for each case.
(c) What is the extinction ratio if the device is switched between the two voltages of
V 2 V and V 4 V?
Bibliography
361
10.5.3 An all-optical switch uses a saturable absorber that has an unsaturated absorption
coefcient of 0 , a saturation intensity of I sat , and a thickness of l in the optical path.
It is desired that the transmittance be switched between the two levels of T high 90%
and T low 10% for the high and low input light intensities, respectively.
(a) What is the minimum value of 0 l required for this function?
(b) If the value of 0 l is chosen to be 10% above the minimum value found in (a), i.e.,
0 l 1:10 lmin , what are the required high and low input intensities for T high
90% and T low 10%, respectively?
10.5.4 The temporal intensity prole of a Gaussian optical pulse that has a FWHM pulsewidth
of tps is described as
!
t2
I t I pk exp 4 ln 2 2 ,
(10.117)
t ps
where I pk is the intensity at the temporal pulse peak, which is taken to be at t 0. Such
a Gaussian pulse is passed through a saturable absorber that has an unsaturated absorption coefcient of 0 , a saturation intensity of I sat , and a thickness of l. With a peak
in
intensity of I in
pk 10I sat and a pulsewidth of t ps for the pulse at the input end, it is
found that the transmittance at the pulse peak is T pk 90% such that the pulse at the
in
output end has a peak intensity of I out
pk 0:9I pk 9I sat . The nonlinear response of
the saturable absorber to the temporally varying intensity of the pulse results in a
reduction of the pulsewidth at the output. Find the pulsewidth tout
ps of the output pulse
in
as a percentage of the input pulsewidth tps .
Bibliography
Boyd, R. W., Nonlinear Optics, 3rd edn. Boston, MA: Academic Press, 2008.
Korpel, A., Acousto-Optics, 2nd edn. New York: Marcel Dekker, 1997.
Pollock, C. R. and Lipson, M., Integrated Photonics. Boston, MA: Kluwer, 2003.
Sugano, S. and Kojima, N., eds., Magneto-Optics. Berlin: Springer, 2000.
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
11 - Photodetection pp. 362-395
11
Photodetection
11.1 PHYSICAL PRINCIPLES OF PHOTODETECTION
..............................................................................................................
Photodetection converts an optical signal into a signal of another form. Most photodetectors
convert optical signals into electrical signals that can be further processed or stored. All
photodetectors are square-law detectors that respond to the power or intensity, rather than
the eld amplitude, of an optical signal. The electrical signal generated by an optical signal is
either a photocurrent or a photovoltage that is proportional to the power of the optical signal.
Based on the difference in the conversion mechanisms, there are two classes of photodetectors:
photon detectors and thermal detectors. Photon detectors are quantum detectors based on the
photoelectric effect, which converts a photon into an emitted electron or an electronhole pair; a
photon detector responds to the number of photons absorbed by the detector. Thermal detectors
are based on the photothermal effect, which converts optical energy into heat; a thermal detector
responds to the optical energy, rather than the number of photons, absorbed by the detector.
Because of this fundamental difference, the general characteristics of these two classes of
photodetectors have a number of important differences.
The response of a photon detector is a function of the optical wavelength with a longwavelength cutoff, whereas that of a thermal detector is wavelength independent. A photon
detector can be much more responsive than a thermal detector in a particular spectral region,
which typically falls somewhere within the range from the near ultraviolet to the near infrared. By
comparison, a thermal detector normally covers a wide spectral range from the deep ultraviolet to
the far infrared with a nearly constant response. Photon detectors can be made extremely sensitive.
Some of them have a photon-counting capability that is not possible for a thermal detector.
A photon detector can be designed to have a high response speed capable of following very fast
optical signals. Most thermal detectors are relatively slow in response because the speed of a
thermal detector is limited by thermalization through heat diffusion and by heat dissipation when
the power of an optical signal varies. For these reasons, photon detectors are suitable for detecting
optical signals in photonic systems, whereas thermal detectors are most often used for optical
power measurement or infrared imaging. In this chapter, only the basic principles of photon
detectors are discussed because our major concern is photodetection for photonics applications.
Photon detectors can be classied into two groups: one based on the external photoelectric
effect and the other based on the internal photoelectric effect. Photodetectors based on the
external photoelectric effect are photoemissive devices, such as vacuum photodiodes and
photomultiplier tubes, in which photoelectrons are ejected from the surface of a
photocathode. Photodetectors based on the internal photoelectric effect are semiconductor
devices, in which electronhole pairs are generated through the absorption of incident photons.
363
A host of such devices have been developed, such as photoconductors, junction photodiodes,
many photovoltaic devices, phototransistors, and charge-coupled devices.
In the discussion of photodetection, we consider an input optical signal with an optical power
Ps . The detection system has an electrical response bandwidth of f B to effectively sample
the optical signal within a rectangular time interval of
t
1
:
2B
(11.1)
The total number of photons received by the photodetector within this time interval is
S
Ps
Ps
t
:
hv
2Bhv
(11.2)
If the photodetector has an external quantum efciency of e , the total number of charge carriers
generated in the photodetector by the photoelectric effect upon receiving the photons within the
time interval t is
N e S e
Ps
,
2Bhv
(11.3)
where 0 e 1. Consequently, the photocurrent is

iph
eN
ePs
,
2eBN e
hv
t
(11.4)
where e is the electronic charge. The signal current is is iph for a photodetector that has no
internal gain. The signal current is is Giph for a photodetector that has an internal gain of G.
11.1.1 External Photoelectric Effect

Photoemissive detectors are based on the external photoelectric effect. Photoelectrons are
emitted when the surface of a metal or a semiconductor, known as a photocathode in this
situation, is illuminated with light of a sufcient photon energy. The lowest vacuum energy
level, Evac , for an electron that is free from the connement of a material is higher than the
Fermi level in the material. For both a metal and a semiconductor, the energy barrier between
the lowest vacuum level and the Fermi level is dened as the work function, e Evac EF , of
the material. For a semiconductor, the difference between the lowest vacuum level and the
conduction-band edge is known as the electron afnity, e Evac Ec , of the semiconductor.
The parameters and have the physical property of an electric potential measured in volts.
The work function and the electron afnity are normally measured in electronvolts.
Photoemission from a given material occurs only when the incident photon has an energy
higher than a certain threshold photon energy, Eth , corresponding to an optical wavelength
shorter than a threshold wavelength, th :
hv E th , for th
hc 1:2398
m eV:
E th
Eth
The values of E th and th are the characteristics of a given material.
(11.5)
Figure 11.1 Photon energy requirement for photoemission from the surface of (a) a metal, (b) a nondegenerate
semiconductor, (c) an n-type degenerate semiconductor, and (d) a p-type degenerate semiconductor.
365
1. Metal: In a metal, shown in Fig. 11.1(a), electrons occupy all energy levels below the Fermi
level. The threshold photon energy for the emission of a photoelectron from a metal is
E th e:
(11.6)
2. Nondegenerate semiconductor: In a nondegenerate semiconductor, shown in Fig. 11.1(b),

not all energy levels below the Fermi level, but only those below the valence-band edge, are
occupied by electrons because the Fermi level lies within the bandgap. The threshold photon
energy for photoemission from a nondegenerate semiconductor that has a bandgap of E g is
E th e Eg > e
if > 0:
(11.7)
3. Degenerate semiconductor: In a degenerate semiconductor, the highest level occupied by

electrons is the Fermi level. Therefore, the threshold photon energy for photoemission from a
degenerate semiconductor is the work function, just like that given in (11.6) for a metal. For an
n-type degenerate semiconductor, E th e < e, as shown in Fig. 11.1(c), because the Fermi
level lies in the conduction band. For a p-type degenerate semiconductor, E th e > e Eg ,
as shown in Fig. 11.1(d), because the Fermi level lies in the valence band.
EXAMPLE 11.1
Among all elemental metals, Cs has the lowest work function of 2.14 eV. What is the threshold
wavelength for an optical wave to cause photoemission from a Cs surface? If a Cs surface is
illuminated with a laser beam at the 400 nm wavelength, what is the highest kinetic energy of
the photoemitted electrons?
Solution:
With a work function of e 2:14 eV, the threshold photon energy for photoemission is
Eth e 2:14 eV because Cs is a metal. Therefore, the threshold wavelength is
th
1239:8
1239:8
nm eV
nm 579:3 nm:
E th
2:14
The kinetic energy of a photoemitted electron is T m0 v 2 =2 hv Eth . When a Cs surface is

illuminated with photons at 400 nm, the highest kinetic energy of the photoemitted
electrons is
T max hv E th
1239:8
eV 2:14 eV 959:5 meV:
400
EXAMPLE 11.2
At room temperature, silicon has an electron afnity of e 4:05 eV and a bandgap of
Eg 1:12 eV. (a) The Fermi level of intrinsic Si lies at EF E c 572:8 meV E v
547:2 meV, where E c and E v are the conduction-band and valence-band edges, respectively.
Find the work function of intrinsic Si. What is the threshold photon energy and the threshold
366
Photodetection
wavelength for an optical wave to cause photoemission from its surface? (b) Find the work
function, the threshold photon energy, and the threshold wavelength for a lightly doped n-type
silicon crystal that has a Fermi level at EF E c 200 meV. (c) Find the work function, the
threshold photon energy, and the threshold wavelength for a heavily doped n-type silicon
crystal that has a Fermi level at E F E c 200 meV.
Solution:
The work function of a semiconductor is e Evac E F , and the electron afnity is
e E vac Ec :
(a) The work function of intrinsic Si is
e Evac EF E vac E c 572:8 meV e 547:2 meV 4:5972 eV:
Because the Fermi level lies in the bandgap, the threshold photon energy is not the work
function but is
Eth e Eg 4:05 eV 1:12 eV 5:17 eV > e:
The threshold wavelength is
th
1239:8
1239:8
nm eV
nm 239:8 nm:
E th
5:17
(b) The work function of the lightly doped n-type silicon with E F E c 200 meV is
e Evac E F E vac Ec 200 meV e 200 meV 4:25 eV:
This lightly doped Si is nondegenerate because its Fermi level lies in the bandgap.
Therefore, the threshold photon energy is not the work function but is
Eth e Eg 4:05 eV 1:12 eV 5:17 eV > e:
th
1239:8
1239:8
nm eV
nm 239:8 nm:
E th
5:17
(c) The work function of the heavily doped n-type silicon with E F E c 200 meV is
e Evac E F E vac Ec 200 meV e 200 meV 3:85 eV:
This heavily doped Si is degenerate because its Fermi level lies above the conduction-band
edge. Therefore, the threshold photon energy is the work function:
E th e 3:85 eV < e:
th
1239:8
1239:8
nm eV
nm 322 nm:
E th
3:85
367
The work functions of elemental metals are in the range of 26 eV. The lowest is that of Cs at 2.14
eV. Elemental metals have poor quantum efciencies. Ordinary group IV and IIIV, semiconductors,
including Si, Ge, GaAs, and InP, have work functions typically in the range of 45 eV. Because of
their high threshold photon energies and low quantum efciencies, elemental metals and ordinary
semiconductors are not useful for photocathodes in the visible and infrared spectral regions.
There are two groups of practical photocathodes that have both high quantum efciencies and
low threshold photon energies. One group consists of compounds of alkaline metals and cesiated
silver oxides that are usually labeled using a standard international designation of spectral response
and window type, such as S-1 (AgOCs), S-4 (Cs3Sb), S-10 (AgBiOCs), S-11 (Cs3Sb), S-20
(Na2KCsSb), and S-24 (Na2KSb). These compounds are semiconductors that have low threshold
photon energies in the range of 12 eV because of their small bandgaps and small electron afnities.
Another group consists of negative electron afnity (NEA) photocathodes. An NEA photocathode
is made by depositing a very thin n-type layer on the surface of a p-type semiconductor to cause a
large downward band bending at the surface. The photocathode has a negative effective afnity if
the band bending is sufciently large that the conduction-band edge of the p-type semiconductor
lies above the vacuum level, as shown in Fig. 11.2. Practical NEA photocathodes have been
developed for a few IIIV semiconductors by depositing a thin layer of Cs or Cs2O on the surface;
these include GaAs:Cs2O, InGaAs:Cs, and InAsP:Cs. As can be seen in Fig. 11.2, once an electron
is excited to the conduction band of an NEA photocathode, it has sufcient energy to be emitted by
tunneling through the thin surface layer because E c > E vac . Therefore, the threshold photon energy
for photoemission from an NEA photocathode is simply the bandgap of the semiconductor:
E th E g :
(11.8)
Figure 11.3 shows the spectral responsivity, which is dened in (11.50) in Section 11.3, of
representative photocathodes. The spectral responsivity of a photoemissive device has a longwavelength cutoff determined by the threshold wavelength of the photocathode material and a
Figure 11.2 Energy levels and photoemission in an NEA photocathode.
368
Photodetection
Figure 11.3 Spectral responsivity of representative photocathodes. The quantum efciency is indicated by the
gray curves.
short-wavelength cutoff determined by the window material. The standard international designation
with the letter S, such as S-1, includes both the response of the photocathode material and the
transmission of the window material. Among the practical photocathodes, including alkaline compounds and NEA semiconductors, S-1 has the lowest threshold energy of approximately 1:1 eV,
corresponding to a threshold wavelength around 1:1 m. Currently no photocathode can respond at
wavelengths longer than 1:2 m. Therefore, no photoemissive detectors exist for the infrared at
wavelengths longer than 1:2 m.
11.1.2 Photoconductivity
Photoconductive detectors are based on the phenomenon of photoconductivity. The conductivity of a photoconductor, which is usually a semiconductor but can sometimes be an insulator,
increases with optical illumination due to the photogenerated excess carriers. The conductivity
of a semiconductor that has electron and hole concentrations of n and p, respectively, is
ee n h p,
(11.9)
where e is the electronic charge, and e and h are the electron and hole mobilities, respectively. In the
absence of optical illumination, a semiconductor has a dark conductivity of 0 ee n0 h p0
because the electron and hole concentrations in this situation are the equilibrium concentrations, n0
and p0 , respectively. When a semiconductor is illuminated with light of a sufcient photon energy,
carriers in excess of the equilibrium concentrations are generated. The photoconductivity is the
additional conductivity contributed by these photogenerated excess carriers:
0 ee n h p,
(11.10)
where n n n0 and p p p0 are the photogenerated excess electron and hole concentrations, respectively.
Similar to photoemission, photoconductivity also has a threshold photon energy, E th , and a
corresponding threshold wavelength, th , that are the characteristics of a given photoconductor.
Depending on the processes involved in the photogeneration of free carriers, there are two
369
Figure 11.4 Optical transitions for (a) intrinsic photoconductivity, (b) n-type extrinsic photoconductivity, and
(c) p-type extrinsic photoconductivity.
principal types of photoconductivity. The intrinsic photoconductivity is contributed by the

excess electrons and holes that are generated by band-to-band absorption of the incident
photons, as shown in Fig. 11.4(a). The threshold photon energy of intrinsic photoconductivity
is clearly the bandgap energy of the photoconductor:
Eth E g :
(11.11)
The extrinsic photoconductivity is contributed by carriers that are generated by optical transitions associated with impurity levels within the bandgap of an extrinsic semiconductor. In an ntype extrinsic photoconductor, the impurity levels have an energy of E i E d below the
conduction-band edge; electrons are excited from these donor levels to the conduction band,
as shown in Fig. 11.4(b). In a p-type extrinsic photoconductor, the impurity levels have an
energy of E i E a above the valence-band edge; electrons are excited from the valence band to
these acceptor levels, as shown in Fig. 11.4(c). Thus, the threshold photon energy of extrinsic
photoconductivity for both n-type and p-type photoconductors is
E th E i :
(11.12)
The spectral response of a photoconductor is determined by its threshold wavelength and the
spectral dependence of its absorption coefcient. Photoconductors cover a broad spectral range
from the ultraviolet to the far infrared. In particular, there are many sensitive photoconductors
in the infrared region at wavelengths longer than 1.2 m where no photoemissive detectors
exist. Both direct-gap and indirect-gap semiconductors are used for photoconductors. All group
IV semiconductors, IIIV and IIVI compound semiconductors, and IVVI compound semiconductors can be used as intrinsic photoconductors. Among them, intrinsic silicon photoconductors are the most important photoconductive detectors in the visible and near-infrared
spectral regions at wavelengths shorter than 1.1 m, while intrinsic germanium photoconductors are the most important photoconductive detectors in the near-infrared region at wavelengths
up to 1.8 m. In the mid-infrared region between 2 and 7 m wavelength, one nds intrinsic
photoconductors based on InAs, InSb, PbS, and PbSe. Extrinsic photoconductors are available
for these mid-infrared wavelengths as well as for longer wavelengths well into the far-infrared
region. The most important extrinsic photoconductive detectors are p-type germanium photoconductors, such as Ge:Au, Ge:Hg, Ge:Cd, Ge:Cu, and Ge:Zn. Figure 11.5 shows the specic
detectivity, which is dened in (11.59) in Section 11.3, of representative infrared photoconductive detectors as a function of the optical wavelength.
370
Photodetection
Figure 11.5 Specic detectivity, D , of representative infrared photoconductive detectors as a function of

optical wavelength. The gray curve shows for comparison the ideal D for a background-limited
photoconductor of unity quantum efciency.
EXAMPLE 11.3
At T 300 K, intrinsic Si has the same electron and hole concentration of n0 p0 ni
7:0 1015 m3 in thermal equilibrium. It has an electron mobility of e 0:135 m2 V1 s1
and a hole mobility of h 0:048 m2 V1 s1 . An intrinsic Si crystal that is used as a
photoconductor is uniformly illuminated with an optical beam to generate electronhole pairs
such that the electrons and holes have the same concentration of n p 2:0 1020 m3 . Find
the dark conductivity and the photoconductivity.
Solution:
The dark conductivity is
0 ee n0 p p0
ee p ni
1:6 1019 0:135 0:048 7:0 1015 S m1
1:57 103 S m1 :
Because n p 2:0 1020 m3 is more than four orders of magnitude larger than
n0 p0 ni 7:0 1015 m3 , we nd that n n n0 p p0 n 2:0 1020 m3 .
Therefore, the photoconductivity is
0 ee n p p
ee p n
1:6 1019 0:135 0:048 2:0 1020 S m1
5:856 S m1 :
371
Figure 11.6 Spectral responsivity of representative photodiodes as a function of optical wavelength at 300 K.
The quantum efciency is indicated by the gray curves.
11.1.3 Photodiodes
Every junction diode has a photoresponse that can be utilized for optical detection. Junction
photodiodes are the most commonly used photodetectors in the photonics industry. They can take
many different forms, including semiconductor homojunctions, semiconductor heterojunctions,
and metalsemiconductor junctions. Similar to that of a photoconductor, the photoresponse of a
photodiode results from the photogeneration of electronhole pairs. In contrast to a photoconductor, which can be of either intrinsic or extrinsic type, a photodiode is normally of intrinsic type,
in which electronhole pairs are generated through band-to-band optical absorption. Therefore, the
threshold photon energy of a semiconductor photodiode is the bandgap energy of its active region:
Eth E g :
(11.13)
Junction photodiodes cover a wide spectral range from the ultraviolet to the infrared. All of the
semiconductor materials used for intrinsic photoconductors discussed in the preceding section
can be used for photodiodes with similar spectral characteristics. Figure 11.6 shows the spectral
responsivity, which is dened in (11.50) in Section 11.3, of representative photodiodes as a
function of the optical wavelength at 300 K.
All junction photodiodes share some basic principles and characteristics. Therefore, we consider
a simple pn homojunction photodiode for a general discussion of the common principles and
characteristics. In a semiconductor photodiode, the generation of electronhole pairs by optical
absorption can take place in any of the different regions: the depletion layer, the diffusion regions,
and the homogeneous regions. In the depletion layer of a diode, the immobile space charges create
an internal electric eld that has a polarity in the direction from the n side to the p side, resulting in
an electron energy-band gradient shown in Fig. 11.7. When an electronhole pair is generated in the
depletion layer by photoexcitation, the internal eld sweeps the electron to the n side and the hole to
the p side, as illustrated in Fig. 11.7. This process results in a drift current that ows in the reverse
direction from the cathode on the n side to the anode on the p side. If a photoexcited electronhole
372
Photodetection
Figure 11.7 Photoexcitation and energy-band gradient of a pn photodiode.
pair is generated within one of the diffusion regions at the edges of the depletion layer, the minority
carrier, which is the electron in the p-side diffusion region or the hole in the n-side diffusion region,
can reach the depletion layer by diffusion and then be swept to the other side by the internal eld, as
also illustrated in Fig. 11.7. This process results in a diffusion current that also ows in the reverse
direction. For an electronhole pair generated by absorption of a photon in the p or n homogeneous
region, no current is generated because there is no internal eld to separate the charges and a
minority carrier generated in a homogeneous region cannot diffuse to the depletion layer before
recombining with a majority carrier.
Because photons absorbed in the homogeneous regions do not generate any photocurrent, the
active region of a photodiode consists of only the depletion layer and the diffusion regions. For
a high-performance photodiode, the diffusion current is undesirable and is minimized. Therefore, the active region mainly consists of the depletion layer where a drift photocurrent is
generated. The external quantum efciency, e , of a photodiode is the fraction of total incident
photons absorbed in the active region that actually contributes to the photocurrent.
There are two contributions to the photocurrent in a junction photodiode: a drift current from
photogeneration in the depletion layer and a diffusion current from photogeneration in the
diffusion regions. The homogeneous regions on the two ends of the diode act like blocking
layers for the photogenerated carriers because carriers neither drift nor diffuse through these
regions. Consequently, a junction photodiode acts like a photoconductor with two blocking
contacts, with the external signal current being equal to the photocurrent:
ePs
:
(11.14)
hv
This photocurrent is a reverse current that depends only on the power of the optical signal.
When a bias voltage is applied to the photodiode, the total current of the photodiode is the
combination of the diode current and this reverse photocurrent
is iph e
373
Figure 11.8 Currentvoltage characteristics of a junction photodiode at various power levels of optical
illumination. The basic circuitry and load line are shown for the photodiode (a) in the photoconductive mode
and (b) in the photovoltaic mode.

ePs
,
iV; Ps I 0 eeV=akB T 1 is I 0 eeV=akB T 1 e
hv
(11.15)
which is a function of both the bias voltage V and the optical signal power Ps . The dark
characteristics for Ps 0 are simply those of an unilluminated diode, with I 0 being the
reverse current and a being a device-specic factor that has a value between 1 and 2 for a
realistic diode. Figure 11.8 shows the currentvoltage characteristics of a junction photodiode
at various power levels of optical illumination. According to (11.15), the currentvoltage
characteristics of an illuminated photodiode shift downward from the dark characteristics by
the amount of the photocurrent, which is linearly proportional to the optical power but is
independent of the bias voltage.
As shown in Fig. 11.8, there are two modes of operation for a junction photodiode. The
device functions in the photoconductive mode in the third quadrant of its currentvoltage
characteristics, including the short-circuit condition on the vertical axis for V 0. It functions
in the photovoltaic mode in the fourth quadrant, including the open-circuit condition on the
horizontal axis for i 0. The mode of operation is determined by the external circuitry and the
bias condition.
374
Photodetection
The circuitry for the photoconductive mode, shown in Fig. 11.8(a), normally consists of a
reverse bias voltage of V V r and a load resistance of RL . In this mode of operation, it is
necessary to keep the output voltage, v out , smaller than the bias voltage, V r , so that a reverse
voltage is maintained across the photodiode. This requirement can be fullled if the bias voltage
is sufciently large while the load resistance is smaller than the internal resistance of the
photodiode in reverse bias, as illustrated with the load line in the third quadrant of Fig. 11.8. In
the photoconductive mode under the conditions that RL < Ri and v out < V r , a photodiode has
the following linear response before it saturates:

ePs
v out I 0 is RL I 0 e
RL :
(11.16)
hv
The circuitry for the photovoltaic mode, shown in Fig. 11.8(b), does not require a bias
voltage but requires a large load resistance. In this mode of operation, the photovoltage appears
as a forward bias voltage across the photodiode. As illustrated with the load line in the fourth
quadrant of Fig. 11.8, the load resistance is required to be much larger than the internal
resistance of the photodiode in forward bias, RL Ri , so that the current i owing through
the diode and the load resistance is negligibly small. In the photovoltaic mode under this
condition, the response of the photodiode is not linear but is logarithmic to the optical signal:

akB T
is
akB T
ePs
v out
,
(11.17)
ln 1
ln 1 e
e
e
I0
hvI 0
where a is the realistic diode factor in the diode equation of (11.15).
In the photoconductive mode, electric energy supplied by the bias voltage source is delivered
to the photodiode. In the photovoltaic mode, electric energy generated by the optical signal can
be extracted from the photodiode to the external circuit. Solar cells are basically semiconductor
junction diodes operated in the photovoltaic mode for converting solar energy into electricity.
EXAMPLE 11.4
A Si photodiode has a reverse current of I 0 10 nA and a realistic diode factor of a 1:2 at
T 300 K. For detection of optical signals at the 850 nm wavelength, its external quantum
efciency is e 0:6. It is illuminated with an optical signal at 850 nm that has a power of
Ps 1 mW. (a) If the photodiode is operated in the photoconductive mode with a load resistance of RL 50 and a reverse bias voltage of V r 5 V, what is the output voltage across the
load resistor? Is the bias voltage sufcient for the photodiode to operate in the linear regime at
the input signal level of Ps 1 mW? (b) If the photodiode is operated in the photovoltaic mode
with a very large load resistance, what is the output voltage?
Solution:
The photon energy for 850 nm is
hv
1239:8
eV:
850
The signal current for Ps 1 mW is
is e
375
ePs
850
0:6
1 103 A 411 A:
hv
1239:8
(a) The output voltage in the photoconductive mode with RL 50 is found using (11.16) to be

v out I 0 is RL 10 109 411 106 50 V 20:6 mV:
Because V r =v out 240 1, the bias voltage of V r 5 V is sufcient for the photodiode
to operate in the linear regime at the input power level of Ps 1 mW.
(b) At T 300 K, we have k B T=e 25:9 mV. The output voltage in the photovoltaic mode
with a very large load resistance is found using (11.17) to be

akB T
is
411 106
1:2 25:9 ln 1
mV 330 mV:
v out
ln 1
I0
e
10 109
11.2 PHOTODETECTION NOISE
..............................................................................................................
Noise is one of the most fundamental phenomena in nature. It is ubiquitous. The noise in a
photodetector sets the fundamental limit on the detectivity of the photodetector, thus determining
the usefulness of the photodetector for a particular application. In terms of their physical nature,
there are a few different types of noise relevant for photodetectors. Two types of noise, quantum
noise and thermal noise, originate from the basic physical laws of nature. Quantum noise, described
as shot noise of electrons or photons in electronics and photonics, results from the statistical nature
of a quantum event dictated by the uncertainty principle. Thermal noise, known as Johnson noise or
Nyquist noise in electronics and photonics, is the consequence of thermal uctuations and is directly
associated with thermal radiation. Noise of such fundamental nature can only be minimized but can
never be completely eliminated. The noise in a photodetector can come from a few physical
sources: the photodetector itself, the possible amplier used in conjunction with the photodetector,
and the circuit used to extract the electrical signal from the photodetector.
Noise appears in a signal as random uctuations about the mean value of the signal.
A measured signal s has a mean value of s dened as
X
s
pss,
(11.18)
s
where ps is the probability for the measured signal to have a value of s and the sum is carried
out over all possible values obtained from measuring the signal. This mean value s is the
expected value, or the ensemble average, of the variable s. The variance, or the mean square
deviation, of the signal s is
2s s s 2 s2 s 2 :
The noise in a signal s can be expressed by a random variable sn dened as
(11.19)
376
Photodetection
sn s s:
(11.20)
The noise represented by the random variable sn has a few general characteristics. As can be
clearly seen from (11.20), it has a zero mean value:
sn 0:
(11.21)
From (11.19) and (11.20), we nd that the mean square value of sn is equal to the variance of s:
s2n 2s s2 s 2 :
(11.22)
The mean square value of the noise in a signal is simply the mean square deviation of the signal.
Because sn 0 but s2n 6 0, the average amplitude of the noise vanishes but the power of the
noise does not. Therefore, the magnitude of the noise is not measured by its average value but
rather by its root mean square (rms) value dened as
rmssn s2n
1=2
(11.23)
Noise characterized by random uctuations is incoherent. If two or more independent noise

sources, sn1 , sn2 , , are simultaneously present in a signal s, their combined effect is not found
by adding their amplitudes but is obtained by adding their mean square values, or their powers:
s2n s2n1 s2n2
(11.24)
The total noise from different independent sources then has an rms value of

1=2
1=2
:
rmssn s2n s2n1 s2n2
(11.25)
One important gure of merit for a detection system is the signal-to-noise ratio (SNR or
S/N). It is dened as the ratio of the power of a signal to the power of its noise or, equivalently,
the ratio of the mean square of a signal to the mean square of its noise:
SNR
s2
s2n
s2
, or
2s
SNR 10 log
s2
s2n
10 log
s2
dB:
2s
(11.26)
The SNR dened above is also known as the signal-to-noise power ratio, to be distinguished
from the signal-to-noise current ratio or the signal-to-noise voltage ratio dened as
SNRcurrent
s
s2n
1=2
s
or
s
SNRvoltage
s
s2n
1=2
s
s
(11.27)
for a photocurrent signal or a photovoltage signal, respectively. Without specication, however, the
SNR of a detection system generally refers to the signal-to-noise power ratio dened in (11.26).
In a photodetection system, a signal can take the form of photon number or photon ux as the
input optical signal. It can also take the form of photocurrent or photovoltage as the output
electrical signal. Therefore, the signal s can represent photon number, photon ux, photocurrent,
or photovoltage. The general characteristics discussed above for the noise sn apply to every case.
11.2.1 Shot Noise

The shot noise in a photodetector results from the quantum nature of the photons in the optical
input and that of the charge carriers generated in the photodetector. Due to the quantum
mechanical probabilistic nature of photons, the photons in an optical signal are not distributed
377
uniformly in time but arrive at the photodetector randomly in time. Therefore, both the power
Ps of the optical signal and the number S of photons received in a given time interval t
uctuate randomly around their respective average values of Ps and S. The random uctuations
of the photon numbers are characterized by Poisson statistics. In any given time interval t, the
probability of receiving S photons is given by the Poisson probability distribution:
S
S eS
pS
:
S!
The mean square noise in photon number uctuations can then be calculated as
X

2
S 2n 2S
pS S S S :
(11.28)
(11.29)
This photon contribution to the noise of a photodetector is independent of the physical

properties of the photodetector because it is external to the photodetector. It is the ultimate
lower limit of the noise in an optical detection system. It sets the fundamental limit on the
detectivity of a photodetector.
The photons received by a photodetector are converted to photoelectrons or electronhole
pairs, depending on the type of photodetector, through the photoelectric effect. With a quantum
efciency of e , which has a value between 0 and 1, the number of photoelectrons generated is
only a fraction of that of the photons received by the photodetector. Because a given photon can
only generate either one or no electron, but not a fraction of an electron, the photoelectric
process is clearly quantum mechanical and probabilistic. The shot noise associated with this
process has to be considered if the quantum efciency is less than unity. This effect is fully
accounted for by considering the statistics of the number N of charge carriers that are generated
in the photodetector with a quantum efciency e . The random uctuations of the photogenerated charge carriers are also characterized by the Poisson probability distribution:
N
N eN
pN
,
N!
(11.30)
where N e S as given in (11.3). We nd, through a procedure similar to that used in (11.29),
that the mean square noise in the number of photogenerated carriers is
2
N n 2N N :
(11.31)
Because N < S when e < 1, the noise is actually reduced by an imperfect quantum efciency. This
result seems odd. However, what really counts in a detection system is not the noise alone, but the
SNR. While the noise is reduced by an imperfect quantum efciency of e < 1, the signal is reduced
even more. Consequently, a photodetector that has a poorer quantum efciency has a lower SNR.
We consider here a photodetector that has no internal gain, such that is iph . Using (11.4)
and (11.31), we nd the shot current noise in the photodetector:
i2n, sh 4e2 B2 N 2n 4e2 B2 N 2eBis :
(11.32)
We then nd the mean square current uctuations for the shot noise in a photodetector that
receives an optical power of Ps from an input optical signal:
378
Photodetection
i2n, sh 2eBis 2e e2 B
Ps
:
hv
(11.33)
From this relation, we have

2
i2s is 2eBis :
(11.34)
In practice, there are other sources that also contribute to the shot noise in a photodetector. One
important source is the photons from the background radiation that impinge on the photodetector.
The contribution of this noise source can be minimized by reducing the aperture of the photodetector to the minimum needed to receive the optical signal. It cannot be completely eliminated,
however, because at the very minimum there is still background thermal radiation, which can
only be reduced by reducing the temperature of the environment surrounding the photodetector.
Another important source of shot noise is the dark current of the photodetector. The dark current
is the current in a photodetector when it is not illuminated with any optical input. In a semiconductor device, the dark current is normally caused by thermal generation of electronhole pairs
and by leakage currents due to surface defects of the device. When these additional noise sources
are considered, the total shot noise in a photodetector is given by

i2n, sh 2eBi 2eB is ib id ,
(11.35)
where ib is the photocurrent generated by background radiation and id is the dark current of the
photodetector.
11.2.2 Excess Shot Noise

In a photodetector that has an internal gain, such as a photomultiplier, a photoconductor, or an avalanche
photodiode, both signal and noise are amplied. For a photodetector that has a gain of G, the signal
current, the background radiation current, and the dark current are all amplied by the factor G:
is Giph Ge
ePs
hv
(11.36)
and
ib Gib0 ,
id Gid0 ,
(11.37)
where ib0 and id0 are unamplied background and dark currents, respectively, and ib and id are
amplied currents that can be directly measured externally. The shot noise is also amplied
through a process of random multiplication of the noise electrons. The statistical nature of this
random multiplication process results in an excess noise factor, F, which is a function of the
material, the structure, and the gain of a photodetector. As a consequence, the mean square shot
noise current for a photodetector that has an internal gain can be expressed as

i2n, sh 2eBG2 F iph ib0 id0 2eBGF is ib id ,
(11.38)
2
where the excess noise factor is a function of the gain, dened as F G2 =G . For a photodetector that has no internal gain, we nd that G 1 and F 1; then, the shot noise given in
(11.38) reduces to that in (11.35), as expected.
379
11.2.3 Thermal Noise

Thermal noise results from random thermal motions of the electrons in a conductor. It is
associated with the blackbody radiation of a conductor at the radio or microwave frequency
range of the signal. Because only materials that can absorb and dissipate energy can emit
blackbody radiation, thermal noise is generated only by the resistive components of the
photodetector and its circuit. Capacitive and inductive components do not generate thermal
noise because they neither dissipate nor emit energy.
The energy of the thermal noise generated by a resistive element is independent of the detailed
physical properties of the resistor and is dictated only by the law of blackbody radiation. At a
temperature of T, the thermal noise power in a small frequency interval of df centered at f is
Pn, th f df
4hf
df :
(11.39)
ehf =kB T 1
In the normal operation of most photodetectors, f kB T=h. Thus, the frequency dependence
of the thermal noise power is negligible, resulting in
Pn, th f df 4kB Tdf :
(11.40)
Then, the total thermal noise power for a detection system of a bandwidth B is simply
Pn, th 4kB TB:
(11.41)
For a resistor that has a resistance of R, the thermal noise can be treated as either current or
voltage noise through the relation Pn, th i2n, th R v 2n, th =R. Then, we have
i2n, th
4k B TB
R
(11.42)
and
v 2n, th 4kB TBR:
(11.43)
For an optical detection system, the resistance R is the total equivalent resistance, including the
internal resistance of the photodetector and the load resistance from the circuit, at the output of the
photodetector. For a photodetector that has a current signal, (11.42) is used. In this case, the thermal
noise is determined by the lowest shunt resistance to the photodetector, which is often the load
resistance of the photodetector. The thermal noise can be reduced by increasing this resistance at the
expense of reducing the response speed of the system. For a photodetector that has a voltage signal,
(11.43) is used. In this situation, the thermal noise is determined by the largest series resistance to the
photodetector, which again is often the load resistance of the photodetector. The thermal noise can
now be reduced by decreasing this resistance, but at the expense of reducing the output voltage signal.
11.2.4 Signal-to-Noise Ratio

There are other noise sources, such as the 1=f noise, but they are usually not important for the
normal operation of a photodetector. Therefore, the total noise in a photodetector, whether it has
an internal gain or not, is basically the sum of its shot noise and thermal noise:
i2n i2n, sh i2n, th :
(11.44)
380
Photodetection
A photodetector is said to function in the quantum regime if i2n, sh > i2n, th . A photodetector
operating in the quantum regime is shot-noise limited because shot noise is the primary source
of noise in this regime. A photodetector is in the thermal regime if i2n, th > i2n, sh . A photodetector
operating in the thermal regime is thermal-noise limited because its thermal noise is dominant
compared with shot noise in this regime.
For a photodetector that has no internal gain, the SNR is given by
SNR
i2s
i2n
i2ph
P2s R2

,
2eB iph ib id 4k B TB=R 2eB Ps R ib id 4k B TB=R
(11.45)
where R e e=hv is the responsivity of a photodetector without an internal gain, dened in the
following section. For a photodetector that has an internal gain of G, the SNR is
SNR
i2s
i2n
G2 i2ph
2eBG2 F iph ib0 id0 4k B TB=R
P2s R2

,
2eBGF Ps R ib id 4kB TB=R

(11.46)
where R Ge e=hv is the responsivity of a photodetector with an internal gain of G, also
dened in the following section.
The relations in (11.45) and (11.46) apply to photodetectors that have current signals at the
output. For a photodetector that has an output voltage signal, the SNR is dened as
SNR
v 2s
P2s R2
,
v 2n
v 2n
(11.47)
where R is the responsivity of a photodetector that has an output voltage signal, dened in the
following section.
EXAMPLE 11.5
The Si photodiode described in Example 11.4 is operated in the photoconductive mode with a load
resistance of RL 50 and a reverse bias voltage of V r 5 V. The total equivalent resistance is
R RL 50 . The photodetector has a bandwidth of B 150 MHz. The dark current of the
photodetector is its reverse current, which has the values of I 0 10 nA at T 300 K and I 0
4 nA at T 273 K. When the photodetector is illuminated with an optical signal of Ps 1 mW at
850 nm, a photocurrent of is 411 A is generated. (a) Find the shot noise when the
photodetector is operated at T 300 K and T 273 K, respectively. (b) Find the thermal noise
at T 300 K and T 273 K, respectively. (c) Find the SNR at T 300 K and T 273 K,
respectively. (d) The photocurrent is proportionally reduced to is 41:1 A for an optical signal
of Ps 100 W. Find the SNR for this case at T 300 K and T 273 K, respectively?
Solution:
In this example, two temperatures are considered. Both shot noise and thermal noise vary with
temperature. At T 300 K, id I 0 10 nA and kB T 25:9 meV. At T 273 K, id I 0
4 nA and kB T 23:5 meV.
381
(a) With is is 411 A, the shot noise at T 300 K is

i2n, sh 2eBi 2eB is id

2 1:6 1019 150 106 411 106 10 109 A2
1:97 1017 A2 :
The shot noise at T 273 K is

i2n, sh 2eBi 2eB is id

2 1:6 1019 150 106 411 106 4 109 A2
1:97 1017 A2 :
The shot noise is the same at T 300 K and T 273 K because id is .
(b) With R RL 50 , the thermal noise at T 300 K is
i2n, th
4kB TB 4 25:9 103 1:6 1019 150 106 2

A 4:97 1014 A2 :
50
R
The thermal noise at T 273 K is

i2n, th
4kB TB 4 23:5 103 1:6 1019 150 106 2

A 4:51 1014 A2 :
50
R
The thermal noise at T 300 K is proportionally higher than that at T 273 K.

(c) At T 300 K, the SNR is

2
411 106
i2s
i2s
3:40 106 65:3 dB:

SNR
17
14
2
2
2
1:97

10
4:97

10
in in, sh in, th
At T 273 K, the SNR is
SNR
i2s
i2n
i2s
i2n, sh i2n, th

2
411 106
3:75 106 65:7 dB:

1:97 1017 4:51 1014
(d) When the photocurrent is proportionally reduced to is 41:1 A for an optical signal of
Ps 100 W, the shot noise is reduced but the thermal noise remains unchanged. From (c),
we nd that the SNR is almost entirely determined by the thermal noise because the shot
noise is negligibly small compared to the thermal noise at is 411 A. At is 41:1 A,
the shot noise is even smaller, with i2n, sh 1:97 1018 A2 . At T 300 K, the SNR is

2
41:1 106
i2s
i2s
SNR
3:40 104 45:3 dB:

18
4:97 1014
i2n i2n, th 1:97 10
At T 273 K, the SNR is
SNR
i2s
i2n
i2s
i2n, th
2
41:1 106
3:75 104 45:7 dB:

1:97 1018 4:51 1014
Because the photodetector is thermal-noise limited, the SNR is reduced by two orders of
magnitude, i.e., by 20 dB, when the signal current is reduced by one order.
382
Photodetection
11.3 PHOTODETECTION MEASURES
..............................................................................................................
Several parameters are commonly used to dene the performance characteristics of a photodetector. These parameters can be considered as the gures of merit of a photodetector. They
are used for comparing one photodetector to another and for determining the suitability of a
photodetector for a particular application. In this section, the basic concepts of these parameters
are dened and discussed.
11.3.1 Spectral Response

Because the response of a photon detector is wavelength dependent, a given photodetector is
responsive only within a nite, specic range of the optical spectrum. The spectral range of
response for a photodetector is determined by the material, the structure, and the packaging of
the photodetector. The spectral response of a photodetector is usually specied in terms of the
spectral responsivity or the spectral detectivity of the photodetector. In choosing a photodetector for an application, the match between the spectral content of the optical signal and the
spectral response of the photodetector is the rst thing to be veried.
11.3.2 Quantum Efciency

Quantum efciency is the probability of generating a charge carrier in a photodetector for each
photon that is incident on the photodetector. The external quantum efciency, e , is the
probability for each photon incident from outside the photodetector to generate a charge carrier
that is externally measured, whereas the internal quantum efciency, i , is the probability for
each photon that actually reaches the active region of the photodetector to be absorbed and
converted to an internal charge carrier. The external quantum efciency of a photodetector is
reduced from its internal quantum efciency by the transmission efciency, t , which accounts
for the probability of an externally incident photon to reach the active region of the photodetector, and by the collection efciency, coll , which accounts for the efciency of the
photogenerated electrical carriers being collected into a photocurrent. Thus, we can express
the external quantum efciency of a photodetector as
e coll t i :
(11.48)
As expressed in (11.3), the external quantum efciency can be dened as the ratio of the
number of photogenerated charge carriers, in the form of either photoelectrons or electronhole
pairs, that actually contribute to the photocurrent to the number of incident photons: e N =S.
According to (11.4), the external quantum efciency of a photodetector can then be expressed
in terms of the incident optical power and the photocurrent as
e
iph =e
hviph
:
Ps =hv
ePs
(11.49)
The quantum efciency of a photodetector is a function of the wavelength of the incident

photons because of the spectral response of the photodetector. Its wavelength dependence arises
383
not only from its explicit dependence on the optical frequency v seen in (11.49) but also from
the wavelength dependence of the ratio iph =Ps dened below as the responsivity of the
photodetector.
EXAMPLE 11.6
A photocurrent of 800 A is generated in a Ge photodetector when it is illuminated with an
optical signal of 1 mW power at the 1:55 m wavelength. Find the external quantum efciency
of the photodetector at this wavelength.
Solution:
At 1:55 m, we have
hv 1:2398
V 0:8 V:
e
1:55
With iph 800 A for Ps 1 mW, the external quantum efciency is
e
hviph
800 106
0:8
64%:
ePs
1 103
11.3.3 Responsivity
Responsivity is an important parameter for a photodetector. It allows one to determine the
available output signal of a photodetector for a given input optical signal. The responsivity of a
photodetector is dened as the ratio of the output current or voltage signal to the power of the
input optical signal. For a photodetector that has an output current signal, the responsivity is
dened as
R
is
:
Ps
(11.50)
For a photodetector that has an output voltage signal, the responsivity is dened as
R
vs
:
Ps
(11.51)
Because most of the commonly used photodetectors have output current signals, we consider in
further detail the responsivity of these types of photodetectors in the following. Similar
concepts can be extended to photodetectors that have output voltage signals.
For a photodetector that has no internal gain, the signal current is simply the photocurrent,
is iph . Using (11.49) and (10.50), we nd the expression for its responsivity:
R
iph
e
e :
Ps
hv
(11.52)
384
Photodetection
For a photodetector that has an internal gain, the signal current is amplied by the gain,
is Giph ; then, the responsivity is
R
Giph
e
Ge GR0 ,
Ps
hv
(11.53)
where R0 is the intrinsic responsivity of the photodetector dened as

R0
iph
e
e :
Ps
hv
(11.54)
The responsivity of a photodetector that has no internal gain is simply its intrinsic responsivity,
R R0 , whereas a photodetector that has an internal gain has a responsivity of R GR0 .
The spectral response of a photodetector is usually characterized by the responsivity of the
photodetector as a function of the optical wavelength, R, which is known as the spectral
responsivity. In addition, the responsivity of a photodetector is also a function of the
modulation-signal frequency f . Its frequency dependence, Rf , characterizes the frequency
response of the photodetector, as discussed later.
EXAMPLE 11.7
Find the responsivity at 1:55 m of the Ge photodetector that is described in Example 11.6.
What is the responsivity at 1:3 m if the external quantum efciency remains the same for
both wavelengths?
Solution:
From Example 11.6, iph 800 A for Ps 1 mW at 1:55 m. Thus the responsivity at
1:55 m is
R
iph 800 106
AW1 0:8 A W1 :

Ps
1 103
If the external quantum efciency at 1:3 m remains e 64% as found in Example 11.6,
the responsivity at 1:3 m is
R
iph
e
1:3
e 0:64
A W1 0:67 A W-1 :
Ps
hv
1:2398
11.3.4 Noise Equivalent Power

The noise equivalent power (NEP) of a photodetector is dened as the input power required of
the optical signal for the signal-to-noise ratio to be unity, SNR 1, at the photodetector output.
Then, using the relations in (11.45) and (11.46), the NEP of a photodetector that has an output
current signal, with or without an internal gain, can be dened as
385
1=2
i2
rmsin
NEP n
,
R
R
(11.55)
where i2n is the mean square noise current at an input optical power level for SNR 1 and R is
the responsivity dened in (11.50). Using the relation in (11.47), the NEP of a photodetector
that has an output voltage signal can be dened as
1=2
v2
rmsv n
NEP n
,
R
R
(11.56)
where v 2n is the mean square noise voltage at an input optical power level for SNR 1 and R is
the responsivity dened in (11.51).
For most detection systems at the low input signal level for SNR 1, the shot noise from the
input optical signal is negligible compared to both the shot noise from other sources and the
thermal noise of the photodetector. In this case, the NEP of a photodetector that has an output
current signal but no internal gain can be expressed as

2eib 2eid 4kB T=R

NEP
R
1=2
B1=2 :
(11.57)
The most fundamental limit is the noise contributed by the ubiquitous blackbody radiation in
the background. This background radiation sets the absolute minimum of NEP for a photodetector. It is often the limitation for photodetectors in the mid- and far-infrared spectral
regions, but it is normally not important for photodetectors in the visible and ultraviolet spectral
regions. For most photodetectors responding to optical wavelengths shorter than 3 m, the
noise from background blackbody radiation is dominated by that from the dark current or that
from resistive thermal noise, or both. For such a photodetector, the intrinsic NEP is that dened
by its dark current when the load resistance is sufciently large if the photodetector generates a
photocurrent signal, or when the load resistance is sufciently small if it generates a photovoltage signal. However, in order to reduce its RC time constant, a high-speed photodetector
that has a current signal normally has a small area, thus a small dark current, but it requires a
small load resistance, thus a large thermal noise. Therefore, the NEP of a high-speed photodetector is usually limited by the thermal noise from its external load resistance rather than by
the shot noise from its internal dark current.
Because the mean square noise is proportional to the photodetector bandwidth, i2n / B and
v 2n / B, the NEP of a photodetector is proportional to the square root of the photodetector
bandwidth: NEP / B1=2 . Therefore, the NEP of a photodetector is often specied in terms of
the NEP for a bandwidth of 1 Hz as NEP=B1=2 , in the unit of W Hz1=2 .
EXAMPLE 11.8
A Ge photodiode has a dark current of id 15 A and a negligible background current at
T 300 K. It has a total equivalent resistance of R 2 k, a response bandwidth of
B 5 kHz, and a responsivity of R 0:8 A W1 at 1:55 m. Find its NEP=B1=2 and
NEP at T 300 K for optical signals at 1:55 m.
386
Photodetection
Solution:
At the noise-equivalent power level, the shot noise for this photodetector is contributed only by
the dark current because the background current is negligible. Thus, the shot noise is
i2n, sh 2eBid 2 1:6 1019 15 106 B A2 Hz1 4:8 1024 B A2 Hz1 :
With R 2 k, the thermal noise at T 300 K, for which kB T 25:9 meV, is
i2n, th
4kB TB 4 25:9 103 1:6 1019

B A2 Hz1 8:3 1024 B A2 Hz1 :
R
2 103
The total noise is

i2n i2n, sh i2n, th 4:8 1024 B A2 Hz1 8:3 1024 B A2 Hz1
1:31 1023 B A2 Hz1 :
Therefore,
1=2
NEP
i2n
B1=2 RB1=2

1=2
1:31 1023
W Hz1=2 4:52 pW Hz1=2 :
0:8
With B 5 kHz, the total NEP over the entire bandwidth is

NEP

1=2
NEP
B1=2 4:52 1012 5 103
W 320 pW:
1=2
B
11.3.5 Detectivity
The detectivity characterizes the ability of a photodetector to detect a small optical signal. It is
dened as the inverse of the NEP of the photodetector:
D
1
,
NEP
(11.58)
which has the unit of W1 .

As discussed above, NEP / B1=2 . The shot noise from the input optical signal at the NEP
level is negligible compared to the shot noise from the background radiation current, ib , and that
from the dark current, id , both of which are often proportional to the surface area, A, of a
photodetector. Therefore, when ib and id are the dominant sources of noise for a photodetector,
the intrinsic noise characteristics of the photodetector can be better quantied by normalizing
NEP to AB1=2 . A useful intrinsic parameter of a photodetector is the specic detectivity, D ,
dened as
D
AB1=2
,
NEP
(11.59)
387
which has the unit of m Hz1=2 W1 , often quoted in cm Hz1=2 W1 . Then, for a dark-currentlimited photodetector that has no internal gain, we have
A1=2 R
D
1=2 :
2eid
(11.60)
The specic detectivity D is independent of the area of the photodetector. It is a measure of the
intrinsic detection capability of the material and the structure of the photodetector.
The detectivity of a photodetector is a function of the optical wavelength. The spectral
characteristics of the detectivity, given as D or D , reect the spectral response of a
photodetector. The detectivity is also a function of the modulation frequency f of a signal that is
modulated on the optical carrier.
EXAMPLE 11.9
The Ge photodetector described in Example 11.8 has a circular surface area that has a diameter
of 2r 5 mm. Find its detectivity and specic detectivity.
Solution:
From Example 11.8, we have NEP=B1=2 4:52 pW Hz1=2 and NEP 320 pW for this
photodetector. The detection surface area is
5 103
A r
2
2
2
m2 1:96 105 m2 :
Therefore, the detectivity is

D
1
1
W1 3:13 109 W1 ,
NEP 320 1012
and the specic detectivity is

1=2
1:96 105
AB1=2
D
m Hz1=2 W1 9:8 108 m Hz1=2 W1 :
12
NEP
4:52 10
11.3.6 Linearity and Dynamic Range

Linearity of a photodetector is dened by the linear response of the photodetector, meaning that
its output current or voltage signal is linearly proportional to its input optical signal. Linear
response is required for a photodetector to faithfully convert the modulation waveform of an
input optical signal to an output electrical signal without distortion. When a photodetector has
a linear response, its quantum efciency e and responsivity R dened above are constants
that are independent of the power Ps of the input optical signal. However, every practical
photodetector only has a nite range of linear response, as shown in Fig. 11.9. As the power of
388
Photodetection
Figure 11.9 Typical response characteristics as a function of the power of the input optical signal for (a) a
photodetector with an output current signal and (b) a photodetector with an output voltage signal.
the input optical signal reaches a certain level, the response of a photodetector starts to saturate,
thereby deviating from linearity.
The maximum acceptable power of the input signal is determined by the maximum deviation
from the linear response of a photodetector that can be tolerated in a particular application.
Given the maximum tolerable deviation from linearity to be (for 100%), the saturation signal
power, Psat
s , for the photodetector in the application is the corresponding maximum acceptable
input power. As illustrated in Fig. 11.9, the value of Psat
s can be found from

dis
dv s
1 R or
1 R,
(11.61)

dPs Ps Psat
dPs Ps Psat
s
s
where R is the responsivity of the photodetector in the linear range.
The usefulness of a photodetector for detecting an optical signal is clearly limited by its
saturation, which is quantied by Psat
s , at the large-signal end and by its detectivity, which is
determined by the NEP of the photodetector, at the small-signal end. The range of the input
signal power above the NEP but below Psat
s in the linear-response region is the useful range of
operation for a photodetector. This range is known as the dynamic range (DR) of the
photodetector, as indicated in Fig. 11.9. The dynamic range is usually quantied, in dB, as
DR 10 log
Psat
s
:
NEP
(11.62)
Alternatively, the dynamic range of a photodetector is also frequently stated in terms of the
number of orders of magnitude in the input optical power from the NEP to Psat
s .
EXAMPLE 11.10
The Ge photodetector described in Example 11.8 has NEP 320 pW and a responsivity of
R 0:8 A W1 at 1:55 m. It saturates at a signal current level of 80 mA. Find its
saturation optical power at 1:55 m and its linear dynamic range.
389
Solution:
1
With a saturation signal current of isat
s 80 mA and a responsivity of R 0:8 A W , the
saturation optical power is
Psat
s
isat
80
s
mW 100 mW:
R 0:8
Therefore, the linear dynamic range is

DR 10 log
Psat
100 103
s
dB 85 dB:
10 log
NEP
320 1012
11.3.7 Speed and Frequency Response

The response speed of a photodetector is directly related to its frequency response. It determines the
ability of a photodetector to follow a fast-varying optical signal. To faithfully record an optical
signal, a photodetector must have a speed higher than the fastest temporal variations in the signal or,
equivalently, a frequency response that has a bandwidth covering the entire bandwidth of the signal.
In the time domain, the speed of a photodetector is characterized by the risetime, tr , and the
falltime, t f , of its response to an impulse signal or a square-pulse signal, as shown in Fig. 11.10. The
risetime is dened as the time interval for the response to rise from 10% to 90% of its peak value,
whereas the falltime is dened as the time interval for the response to decay from 90% to 10% of its
peak value. Generally, the overall speed of a photodetector is determined by both its intrinsic
bandwidth and its circuit-limited RC bandwidth. The risetime of the impulse response is determined by the intrinsic bandwidth of a photodetector, and that of the square-pulse response is
determined by the circuit-limited RC bandwidth of the photodetector. The risetime and its
corresponding bandwidth have the relation:
Figure 11.10 Typical responses of a photodetector to (a) an impulse signal and (b) a square-pulse signal.
390
Photodetection
tr
0:35
,
f 3dB
(11.63)
where f 3dB is the 3-dB cutoff frequency dened below.

The frequency response, which is characterized by the frequency dependence of the responsivity Rf at a given optical wavelength, can be obtained by simply taking the Fourier
transform of the impulse response or by registering the response of the photodetector at one
modulation-signal frequency at a time while sweeping this frequency. Note that Rf is the
current or voltage response spectrum of the photodetector because the responsivity of a
photodetector is dened in terms of the output current or voltage signal of the photodetector.
The output electrical power spectrum of the photodetector is R2 f , which denes a 3-dB cutoff
frequency, or 3-dB bandwidth, for a photodetector as
1
R2 f 3dB R2 0:
2
(11.64)
Considering the rectangular time interval of t that is used to dene the bandwidth B, we have
the relation between f 3dB and B of a photodetector:
f 3dB 0:886B
0:443
:
t
(11.65)
The 3-dB bandwidth of a photodetector is a function of the combined effect of a few different
physical factors that determine the speed and the frequency response of the photodetector.
These factors and their relative importance depend on the type of the photodetector.
EXAMPLE 11.11
The Ge photodetector described in Example 11.8 has a response bandwidth of B 5 kHz. Find
its 3-dB cutoff frequency. What is the risetime of the photodetector response to an impulse
signal?
Solution:
The 3-dB cutoff frequency is
f 3dB 0:886B 0:886 5 kHz 4:43 kHz:
The risetime of the photodetector response to an impulse signal is
tr
0:35
0:35
s 79 s:
f 3dB 4:43 103
EXAMPLE 11.12
A photodetector is used to detect an optical pulse that has a pulse duration of 500 ps and a pulse
risetime of 200 ps. What is the minimum bandwidth of the photodetector required to detect the
pulse? What is the minimum bandwidth required for resolving the pulse risetime?
Problems
391
Solution:
The minimum bandwidth required for detecting the pulse duration of t 500 ps is found
using (11.1) or (11.65) as
Bmin
1
1
Hz 1 GHz:
2t 2 500 1012
The minimum bandwidth required for resolving the pulse risetime of tr 200 ps is found using
(11.63) and (11.65) as
Bmin
f min
0:35
0:35
3dB
Hz 1:98 GHz:
0:886 0:886t r 0:886 200 1012
It is clear that a larger bandwidth is needed to resolve the pulse risetime.
Problems
11.1.1 Alkaline metals have low work functions. Besides Cs, which has the lowest work
function of 2.14 eV, as described in Example 11.1, the work functions are 2.29 eV
for K, 2.36 eV for Na, and 2.90 eV for Li. What is the threshold wavelength for an
optical wave to cause photoemission from the surface of each alkaline metal? If the
surface of each metal is illuminated with a laser beam at the 500 nm wavelength, what is
the highest kinetic energy of the photoemitted electrons?
11.1.2 The work function of Ag varies from 4.26 to 4.74 eV, depending on the crystallographic
orientation of the Ag surface. When a specic Ag surface is illuminated with a laser
beam at the 260 nm wavelength, the highest kinetic energy of the photoemitted electrons
is found to be T max 168 meV. What is the work function of this Ag surface?
11.1.3 The work function of Au depends on the crystallographic orientation of the Au surface.
Experimental data on various Au surfaces show threshold wavelengths varying between
226:7 nm and 243:1 nm. Find the work function range of Au.
11.1.4 At room temperature, silicon has an electron afnity of e 4:05 eV and a bandgap of
E g 1:12 eV.
(a) Find the work function, the threshold photon energy, and the threshold wavelength for
a lightly doped p-type silicon crystal that has a Fermi level at EF E v 200 meV.
(b) Find the work function, the threshold photon energy, and the threshold wavelength for
a heavily doped p-type silicon crystal that has a Fermi level at E F E v 200 meV.
11.1.5 At room temperature, GaAs has an electron afnity of e 4:07 eV and a bandgap of
Eg 1:424 eV. The Fermi level of GaAs at room temperature lies within the bandgap at
EF E c 672:2 meV E v 751:8 meV, where E c and E v are the conduction-band
and valence-band edges, respectively. Find the work function of intrinsic GaAs. What is
the threshold photon energy and the threshold wavelength for an optical wave to cause
photoemission from its surface?
392
Photodetection
11.1.6
At room temperature, GaAs has an electron afnity of e 4:07 eV and a bandgap of

E g 1:424 eV.
(a) Find the work function, the threshold photon energy, and the threshold wavelength
for a lightly doped n-type GaAs crystal that has a Fermi level at EF E c 300 meV.
(b) Find the work function, the threshold photon energy, and the threshold wavelength
for a lightly doped p-type GaAs crystal that has a Fermi level at EF E v
300 meV.
11.1.7
At room temperature, GaAs has an electron afnity of e 4:07 eV and a bandgap of

E g 1:424 eV.
(a) Find the work function, the threshold photon energy, and the threshold wavelength for
a heavily doped n-type GaAs crystal that has a Fermi level at EF E c 300 meV.
(b) Find the work function, the threshold photon energy, and the threshold wavelength
for a lightly doped p-type GaAs crystal that has a Fermi level at E F E v 300 meV.
11.1.8
The intrinsic electron and hole concentrations of GaAs in thermal equilibrium at room
temperature are n0 p0 ni 2:33 1012 m1 . It has an electron mobility of e
0:85 m2 V1 s1 and a hole mobility of h 0:04 m2 V1 s1 . An intrinsic GaAs
crystal used as a photoconductor is uniformly illuminated with an optical beam to
generate electronhole pairs for total electron and hole concentrations of
n p 1:0 1020 m3 . Find the dark conductivity and the photoconductivity.
11.1.9
The intrinsic electron and hole concentrations of Ge in thermal equilibrium at room

temperature are n0 p0 ni 1:95 1019 m1 . It has an electron mobility of e
0:39 m2 V1 s1 and a hole mobility of h 0:19 m2 V1 s1 . An intrinsic Ge crystal
used as a photoconductor is uniformly illuminated with an optical beam to generate
electronhole pairs. Find the dark conductivity. What are the required concentrations
of the photogenerated electrons and holes for the photoconductivity to be 20 times the
dark conductivity?
11.1.10 A Si photodiode at T 300 K has a reverse current of I 0 10 nA and a realistic diode

factor of a 1:2. For the detection of optical signals at the 532 nm wavelength, its
external quantum efciency is e 0:7. It is illuminated with an optical signal that has
a power of Ps 200 W at 532 nm.
(a) If the photodiode is operated in the photoconductive mode with a reverse bias
voltage of V r 5 V, what is the required load resistance for the output voltage to
be at least 100 mV?
(b) If the photodiode is operated in the photovoltaic mode with a very large load
resistance, what is the output voltage?
(c) What are the output voltages for Ps 5 mW in the photoconductive mode with the
load resistance found in (a) and in the photovoltaic mode, respectively?
11.1.11 A Ge photodiode has a reverse current of I 0 2 A and a realistic diode factor of a
1:1 at T 300 K. Its external quantum efciency is e 0:54 for an optical signal at
1:55 m. The power of the optical signal varies between 0:5 mW and 5 mW.
(a) The photodiode is operated in the photoconductive mode with a reverse bias
voltage of V r 10 V and a load resistance of RL 50 . What is the range of
Problems
393
the output voltage? What is the range of the signal voltage? Is the bias voltage
sufcient for the photodiode to function in the linear regime for the whole range of
signal powers?
(b) If the photodiode is operated in the photovoltaic mode with a very large load
resistance, what is the range of the signal voltage?
11.2.1 The Si photodiode described in Example 11.5 has a dark current of id 10 nA at T
300 K and a signal current of is 411 A when it is illuminated with an optical signal
of Ps 1 mW at 850 nm. With a total resistance of R RL 50 , it has a
bandwidth of B 150 MHz, and its SNR at T 300 K is limited by thermal noise
with i2n, sh 1:97 1017 A2 and i2n, th 4:97 1014 A2 . Clearly, the SNR can be
increased by reducing the thermal noise, at least until it reaches the level of the shot
noise. How can this be accomplished? Find the parameter changes needed to reduce the
thermal noise to the level of the shot noise. What price has to be paid in doing so?
11.2.2 A large-area Ge photodetector has a dark current of id 10 A at T 300 K and a
signal current of is 400 A when it is illuminated with an optical signal of Ps
500 W at 1:55 m. The total equivalent resistance is R RL 1 k, and the
bandwidth is B 10 kHz. Find the shot noise, the thermal noise, and the SNR of the
photodetector in this operating condition. Which noise source sets the primary limit on
the SNR?
11.2.3 The signal current generated in the Ge photodetector described in Problem 11.2.2 is
proportional to the power of the optical signal. Answer the questions raised in Problem
11.2.2 for (a) an optical power of Ps 5 W generating a photocurrent of is 4 A
and (b) an optical power of Ps 50 W generating a photocurrent of is 40 A.
11.3.1 A photodetector has an InGaAs active layer, which absorbs optical signals to be
detected. The active layer has a bandgap of 0.75 eV. The incoming optical beam has
to pass through an InGaAsP top layer of a higher bandgap of 0.95 eV before reaching
the active layer. What is the optical spectral bandwidth of this photodetector, i.e., the
wavelength range that can be detected by this photodetector?
11.3.2 An uncoated surface of Si has a reectivity of R 32:6% at 850 nm. A Si
photodiode has an active region that absorbs 90% of light at 850 nm that reaches
this region. Almost all photogenerated carriers contribute to the photocurrent.
(a) If the surface of the Si photodiode is not coated, what is the largest possible external
quantum efciency it can have? What is the largest possible photocurrent for an
optical signal at 850 nm that has an input power of 1 mW?
(b) If it is desired that a photocurrent of at least 600 A be generated with an input
power of 1 mW for an optical signal at 850 nm, what is the required external
efciency? How can this efciency be accomplished by properly coating the surface
of the Si photodetector?
11.3.3 The maximum possible external quantum efciency is clearly e 1 for any photodetector. For this reason, the intrinsic responsivity for any photodetector has a maximum
possible value of Rmax
0 , which is a function of only the wavelength of the optical signal.
394
Photodetection
How does the value of Rmax

vary with the wavelength of the optical signal? What is
0
the range of its values for optical signals in the visible spectral region? What are its
values at the three common near-infrared wavelengths of 850 nm, 1:3 m, and 1:55 m
that are used in optical communication systems?
11.3.4
An InGaAs/InP avalanche photodiode (APD) has an internal gain of G, which can be

varied up to about G 20 by varying the bias voltage that is applied to the APD. The
circuitry of the APD has a small load resistance of RL 50 for fast response to highfrequency optical signals. For an optical signal at 1:55 m, the external quantum
efciency of the APD is e 64%.
(a) Find the intrinsic responsivity of the APD at 1:55 m.
(b) By applying a certain bias voltage for an internal gain, a signal voltage of v s
15 mV on the load resistance is observed with an optical signal of Ps 25 W at
1:55 m. Find the responsivity and the gain of the APD at this operating
point.
11.3.5
A Ge photodiode has a negligible background current. Its dark current is id 10 A at

T 300 K and id 20 nA at T 250 K. It has a total equivalent resistance of R
20 k, a response bandwidth of B 1 kHz, and a responsivity of R 0:9 A W1 at
1:55 m.
(a) Find its NEP=B1=2 and NEP at T 300 K for optical signals at 1:55 m.
Which noise source limits the NEP at this temperature?
(b) Find its NEP=B1=2 and NEP at T 250 K for optical signals at 1:55 m.
Which noise source limits the NEP at this temperature?
11.3.6
A Si photodiode has a total equivalent resistance of R 50 and a bandwidth of

B 100 MHz. At T 300 K, it has a negligible background current and a dark
current of id 10 nA. It has a circular surface area that has a diameter of
2r 1 mm. Its responsivity at 850 nm is R 0:52 A W1 . Find its NEP, detectivity, and specic detectivity at 850 nm.
11.3.7
A Si photodiode saturates at a signal photocurrent of isat

s 16 mA. Find the saturation
optical power for an optical signal at 850 nm, where the photodiode has a
responsivity of R 0:45 A W1 . If it has an NEP of 150 nW, what is its linear
dynamic range?
11.3.8
A photodetector has an NEP of 1:6 nW and a linear dynamic range of 67 dB for optical
signals at the 1:3 m wavelength. What is the maximum optical signal power
allowed for the photodetector to respond linearly?
11.3.9
When an optical pulse that has a temporal duration of 1 ps is detected by a photodetector, the electrical response output of the photodetector shows a pulse that has a
risetime of 180 ps. What is the 3-dB cutoff frequency and the electrical response
bandwidth of this photodetector?
11.3.10 A photodetector has a bandwidth of B 8 GHz. What is the duration of the shortest
optical pulse that can be clearly detected using this photodetector? What is the fastest
pulse risetime of an optical pulse that can be resolved by this photodetector?
Bibliography
395
Bibliography
Bhattacharya, P., Semiconductor Optoelectronic Devices, 2nd edn. Englewood Cliffs, NJ: Prentice-Hall, 1997.
Bube, R. H., Photoconductivity of Solids. New York: Wiley, 1960.
Donati, S., Photodetectors: Devices, Circuits, and Applications. Upper Saddle River, NJ: Prentice-Hall, 2000.
Kasap, S. O., Optoelectronics and Photonics: Principles and Practices, 2nd edn. Upper Saddle River, NJ:
Prentice-Hall, 2012.
Nalwa, H. S., ed., Photodetectors and Fiber Optics. San Diego, CA: Academic Press, 2001.
Press, 2007.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
Appendix A - Symbols and Notations pp. 396-402
APPENDIX A
A.1
FIELDS
..............................................................................................................
Field vectors and their scalar magnitudes are represented using a consistent system of symbols
and fonts. All vectors except for unit vectors are represented in bold-face fonts, whereas all
scalar quantities are represents in nonbold fonts. This system is illustrated in the following
using the electric eld as an example.
A.1.1 Real Fields

All real elds are dened in the real space and time domain only. All real eld vectors are
represented in the italic bold capital Roman font, such as
Er; t,
(A.1)
for the real electric eld vector. Other real eld vectors are Hr; t, Dr; t , Br; t , Pr; t ,
M r; t , J r; t , and Sr; t . Except for the current density, all real elds are always represented
in the vector form without separate symbols dened for their scalar magnitudes. The scalar
magnitude of J is represented as J.
A.1.2 Complex Fields

All complex eld vectors are represented in the nonitalic bold capital Roman font. All complex
eld vectors in the real space and time domain are dened in relation to their respective real
eld vectors, such as Er; t dened in (1.40) for the complex electric eld vector:
Er; t Er; t E r; t Er; t complex conjugate:
(A.2)
Other complex eld vectors dened in a similar manner are Hr; t , Dr; t , Br; t , Pr; t ,
Mr; t , and Jr; t. The complex Poynting vector is dened differently, as given in (1.54) and
(1.55):
S E H so that S S S :
(A.3)
The scalar magnitude of a complex eld vector is represented in the nonbold mathematical
capital Roman font, for example E for the magnitude of E:
E Eê ,
(A.4)
where ê is the unit vector of E. Other scalar eld magnitudes represented in a similar manner
are H, D, B, P, and M. No scalar complex current density at an optical frequency is used; the
397
scalar J represents the magnitude of a real current density vector J at DC or a low frequency.
No scalar magnitude of the complex Poynting vector S is used.
A.1.3 Complex Field Amplitudes

The slowly varying amplitude vector of a complex eld vector is represented in the bold capital
script font, such as E for the slowly varying amplitude of E. It is dened as the slow variation of
the eld envelope on its carrier frequency through the relation
Er; t E r; t exp ikr it,
(A.5)
as expressed in (1.52) for the electric eld. Other slowly varying eld amplitude vectors dened
in a similar manner are H, D, B, P, M, and J , but not all of them are used in the text. No
slowly varying eld amplitude is dened for the Poynting vector.
The scalar magnitude of a slowly varying eld amplitude vector is represented in the nonbold
capital script font, for example E for the magnitude of E:
E Eê :
(A.6)
Other scalar magnitudes of slowly varying eld amplitudes represented in a similar manner are
H, D, B, P, M, and J , but not all of them are used in the text.
A.1.4 Mode Fields

Complex mode eld vectors are represented as Ev r; t and Hv r; t with their scalar magnitudes represented as E v r; t and H v r; t , respectively, where the subscript index v represents a
compound mode index such as m or mn for waveguide modes, or mn for Gaussian modes. The
vectorial eld proles of a waveguide mode characterize the transverse spatial distributions of
the mode elds. A waveguide mode eld prole is a function of the transverse spatial coordinates only. The vectorial waveguide mode eld proles are represented as E v x; y and Hv x; y,
or E v ; r and Hv ; r , with their scalar magnitudes represented as E v x; y and Hv x; y, or
E v ; r and Hv ; r . Normalized vectorial mode eld patterns, dened in (3.18), are repre^ v x; y and H
^ v ; r and H
^ v x; y, or E
^ v ; r . Gaussian modes are represented
sented as E
using similar symbols, but they are functions of both transverse and longitudinal spatial
coordinates, x, y, and z, as seen in (3.73).
A.2
VECTORS AND TENSORS
..............................................................................................................
All vectors are represented in bold face, with the exceptions of unit vectors, and their
magnitudes are represented with corresponding symbols in nonbold fonts. A vector is also
represented in the form of a 3 1 column matrix. Besides the eld vectors and their magnitudes
described in the preceding section, we have
k, k; K, K; r, r; u, u; U, U; k, k:
398
Appendix A
All tensors and transformation matrices are represented in bold face or in terms of their
elements with subscript indices. Second-order tensors and transformation matrices are also
represented in the form of 3 3 square matrices. The tensors used include
h i

f ijk ,
pijkl ,
p0ijkl ,
r ijk ,
cijkl ,
h i

R Rij ,
sijkl ,
S Sij ,
ij ,
ij ,
h i
h i
h i
h
i

2
3
2
3
ij , ijk , ijkl , ij , ij :
The transformation matrices used in the text include
Fz; z0 , Rz; 0; l:
A.3
FOURIER-TRANSFORM PAIRS
..............................................................................................................
The same symbol is used for a quantity in real space and its counterpart in the momentum space, or
one in the time domain and its counterpart in the frequency domain. The difference is indicated by
expressing a quantity as a function of r or k, or as a function of t or . Note that the unit of a
quantity is multiplied by a length unit of a meter each time one of the three spatial dimensions is
transformed from the real space to the momentum space, and is multiplied by a time unit of a
second when the quantity is transformed from the time domain to the frequency domain. For
example, the electric eld E r; t in the real space and time domain has the unit of volts per meter
(V m1 ), but E k; t has the unit of volt square meters (V m2 ), E r; has the unit of volt seconds
per meter (V s m1 ), and E k; has the unit of volt second square meters (V s m2 ).
A.4
SPECIAL NOTATIONS
..............................................................................................................
A few special notations are used to label symbols for special meanings.
A.4.1 Unit Vectors and Normalized Quantities

Unit vectors are denoted with a hat on top of a symbol. The unit vectors that appear in the text
are
^ n^, ^r , u^, ^x , ^y , ^z , X
^ , Y^ , Z^ :
ê , k,
Normalized quantities are also denoted with a hat on top of a symbol. The normalized mode
eld proles appear in both vector and scalar forms:
^ v , E^ v , H
^ v, H
^ v:
E
Other normalized quantities that appear in the text include
^ sp , T^ FP , T^ c :
g^ v, P
399
A.4.2 Modied Quantities

A quantity that is modied from the original quantity in some manner is denoted with a tilde on
top of a symbol. Modied quantities that appear in the text include
~ B,
~ ~ , ~ , ~ :
A,
A.4.3 Average Values

The spatial average, temporal average, weighted average, or mean value of a quantity is denoted
with a bar on top of a symbol, such as
i, i2 , k, n, N , N 2 , s, s2 , S, S, S 2 , v 2 , W p , :
A.5
SUBSCRIPTS AND SUPERSCRIPTS
..............................................................................................................
Various fonts and notations are used for subscripts and superscripts. They include numerals, the
mathematical font, the Greek font, coordinate symbols, and the Roman font. Bare numerals,
mathematical font letters, and Greek letters that represent indices or variables are used only for
subscripts. Roman letters and some special notations that have literal meanings can appear
either as subscripts or as superscripts.
A.5.1 Numerals
Bare numerals are used only for subscripts. The following four numbers have special meanings
in a proper context:
0 base value (0 , m0 ), constant value (P0 , S0 ), free-space value ( 0 , 0 ), center value (v0 , 0 ),
unsaturated value (g 0 , g0 ), equilibrium value (n0 , p0 ), beam waist (w0 ), or static eld (E0 , H 0 );
1 parameters for waveguide core (n1 , N 1 , D1 , k1 , h1 ) or
parameters for the lower laser level j1i (E 1 , N 1 , R1 );
2 parameters for waveguide substrate (n2 , N 2 , D2 , k2 , 2 ) or
parameters for the upper laser level j2i (E 2 , N 2 , R2 );
3 parameters for waveguide cover (n3 , N 3 , D3 , k 3 , 3 ) or
parameters for the energy level j3i.
Note that the same symbol can have different meanings in different contexts. For example, n2 in
nonlinear optics also represents the coefcient of intensity-dependent index change dened in
(10.101).
The numbers 1, 2, and 3 are also used as subscripts to represent the orthogonal coordinates of
a general three-dimensional spatial coordinate system. The numbers 1 through 6 are also used
as subscripts representing double indices to label tensor elements under the index contraction
rule dened in (2.59):
xx yy
1 2
zz
3
yz, zy zx, xz xy, yx

4
5
6
400
Appendix A
A numeral in the superscript is always placed in parentheses so that it is never confused with
an exponent. It represents a perturbation order or the order of an interaction process. For
example, 1 is a linear susceptibility, 2 is a second-order nonlinear susceptibility, 3 is a
third-order nonlinear susceptibility, and so forth.
A.5.2 Mathematical and Greek Subscripts

Mathematical and Greek fonts are used only for subscripts. They represent variable indices with
the following well-dened meanings:
a, b, c
i, j, k, l
m, n, p, q
m, n
q
,
, v
general indices or general mode indices;

integers or coordinate indices;
integers or frequency component indices;
transverse mode indices, each labeling a spatial dimension;
longitudinal mode index or diffraction order;
contracted indices representing double coordinate indices;
compound transverse mode indices, each representing a mode.
Some Greek subscripts do not represent indices or variables but express literal meanings.
They include:
, , =2, =4, , =2:
A.5.3 Coordinate Labels

General orthogonal spatial coordinates are labeled as 1, 2, and 3. Specic coordinates include
the rectilinear coordinates x; y; z, the cylindrical coordinates r; ; z, and the spherical
coordinates r; ; . One set of special rectilinear coordinates, X; Y; Z , is used for the new
^ , Y^ , and Z^ of a crystal transformed under the Pockels effect. Two orthogonal
principal axes X
unit vectors, ê and ê dened in (1.75) and (1.78), are used for left-circular and right-circular
polarizations, respectively. Two special symbols are also used to represent directions: ? for
perpendicular and jj for parallel.
Coordinate labels generally appear as subscripts with commonly accepted meanings, with
one exception. This exception takes place when labeling a propagation constant k and the
corresponding wavevector k of an optical eld that has a particular normal mode polarization.
Because kx conventionally represents the x component of the k vector, meaning kx k^x , the
propagation constant of an x-polarized optical eld that can propagate in any direction perpendicular to ^x is represented as kx in order to avoid confusion. To be consistent, the corresponding
wavevector is labeled as kx . Thus, kx nx =c 6 k x , and kx k x k^ where k^ ? ^x . Such superscript coordinate labeling for k and k applies only to the following:
kx ,
kx ,
ky ,
ky ,
kz ,
kz ,
kX ,
kX ,
kY ,
kY ,
kZ ,
kZ ,
k ,
k ,
k ,
k :
401
A.5.4 Roman Labels

All superscript and subscript labels in the Roman font have literal meaning. A given Roman
label can appear either as a subscript or as a superscript, depending on the convenience of
the situation, with exactly the same meaning. Among all subscript and superscript labels, only
the Roman labels have such exibility. With only a few exceptions to avoid confusion, the
conventional rules for abbreviations are largely followed: (1) abbreviations for common words
are in the lower case, with the exceptions of E for TE, M for TM, L for longitudinal or load, and
T for transverse; (2) abbreviations for proper nouns are in the upper case; and (3) acronyms are
all in the upper case. The Roman labels used in this book are listed below.
Label
Meaning
Label
Meaning
absorption, acceptor, acoustic
hole, homogeneous
background, bias, bound
incidence, initial, internal, intrinsic,

impurity
Boltzmann, Bragg, Brewster

in
input
carrier, cavity, center, characteristic,

conduction band, conversion,
coupling, critical, cutoff
ind
induced
inh
inhomogeneous
inj
injection
Kerr
load, longitudinal
magnetization, modulation
coll
collection
cond
conduction
dark, dielectric, diffraction, donor
Doppler
def
deection
TM mode
electrical, electrode, electron,

emission, external, extraordinary
max
maximum
TE mode
min
minimum
ext
external
n-type, noise
fall
nonrad
nonradiative
Faraday, Fermi
ordinary
FP
FabryProt
out
output
FSR
free spectral range
bandgap, gain, gap, glass, grating,

group
p-polarized (TM), p-type,

parallel, phase, plasma, polarization,
pump
ph
photo, photon
402
Appendix A
Label
Meaning
Label
Meaning
pk
peak
sat
saturation
PM
phase matched
sh
shot
ps
pulse
sp
spontaneous, surface plasma
QW
quantum well
ST
ShawlowTownes
radiation, reduced, recombination,

reection, relaxation, reversed, rise
total, transmitted
transverse
Rayleigh
tr
transit, transparency
rad
radiative
th
thermal, threshold
valence band
vac
vacuum
water
res
resonance
RT
round trip
s-polarized (TE), saturation, series,

signal, slope, spontaneous, source,
switching

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
Appendix B - SI Metric System pp. 403-404
APPENDIX B
SI Metric System
Table B.1 SI base units

Quantity
Name
Symbol
Length
meter
Mass
kilogram
kg
Time
second
Electric current
ampere
Temperature
kelvin
Amount of substance
mole
mol
Luminous intensity
candela
cd
Table B.2 SI derived units

Quantity
Name
Symbol
Equivalent
Plane angle
radian
rad
mm
Solid angle
steradian
sr
m2 m
Frequency
hertz
Hz
Force
newton
kg m s
Pressure
pascal
Pa
Nm
Energy
joule
kg m2 s
Power
watt
Js
Electric charge
coulomb
As
Electric potential
volt
J C 1, W A
Magnetic ux
weber
Wb
Vs
Magnetic ux intensity
tesla
Wb m
=1
2
=1
1
2
2
2
404
Appendix B
Table B.2 (cont.)

Quantity
Name
Symbol
Equivalent
Resistance
ohm
VA
Conductance
siemens
A V 1,
Capacitance
farad
CV
Inductance
henry
Wb A
Luminous ux
lumen
lm
cd sr
Illuminance
lux
lx
lm m
Source: Nelson, R. A., Guide for metric practice, Physics Today BG15BG16, August, 2002.
Table B.3 Metric prexes

Name
Symbol
Factor
Exa
1018
Peta
1015
Tera
1012
Giga
109
Mega
106
Kilo
103
Hecto
102
Deca
da
10
Unit
Deci
10
Centi
10
Milli
10
Micro
10
Nano
10
Pico
10
12
Femto
10
15
Atto
10
18
1
1

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
Appendix C - Fundamental Physical Constants pp. 405-405
APPENDIX C
Fundamental Physical Constants
Table C.1 Physical constants

Quantity
Symbol
Value
Unit
Speed of light in free space
2.997 924 58 108
m s1
Magnetic permeability of free space
4 107
1.256 637 061 4 106
H m1
H m1
Electric permittivity of free space
8.854 187 817 1012
F m1
Impedance of free space 0 = 0 1=2
Z0
376.730 313 461
Planck constant
6.626 068 765 2 1034

4.135 667 271 6 1015
Js
eV s
Reduced Planck constant h/2
1.054 571 596 8 1034

6.582 118 892 6 1016
Js
eV s
Elementary charge
1.602 176 462 6 1019
Electron rest mass
m0
9.109 381 887 2 1031
kg
Proton rest mass
mp
1.672 621 581 3 1027
kg
Atomic mass unit
mu
1.660 538 731 3 1027
kg
Boltzmann constant
kB
1.380 650 324 1023

8.617 342 15 105
J K1
eV K1
Thermal energy at T = 300 K
kBT
2.585 202 645 102
eV
Photon constant hc hv
hc
1.239 841 86 106
eV m
Source: Mohr, P. J. and Taylor, B. N., The fundamental physical constants, Physics Today BG6BG13,
August, 2002.

Jia-Ming Liu
Online ISBN: 9781316687109
Chapter
Appendix D - Fourier-Transform Relations pp. 406-408
APPENDIX D
According to the discussion in Chapter 1, we dene the Fourier transform between the time
domain and the frequency domain in terms of the angular frequency as follows
E F fE t g
E t eit dt
(D.1)
E eit d:
(D.2)
and
E t F
1
1
fE g
2
In terms of the real frequency v =2, we have
E v
E t ei2vt dt
(D.3)
E vei2vt dv:
(D.4)
and
E t
The Fourier-transform relations for common functions encountered in the description of

various waveforms are listed in Table D.1. In this table, the Heaviside step function H x is
dened as

1, if x 0,
(D.5)
H x
0 if x <0;
the rectangular function x is dened as

x
1,
0
if jxj 1=2,
if jxj > 1=2;
(D.6)
and the triangular function x is dened as

x
1 jxj,
if jxj 1,
if jxj > 1:
(D.7)
407
Table D.1 Fourier-transform relations

Function form
E(t)
E()
Function form
Gaussian
et
p 2 2 =4
e
Gaussian
sech
secht=
Innite impulse sequence

X t
m
sech=2

X
n
2
n
Complex exponential
ei0 t
2 0
delta
Double-sided exponential
ejt=j
2
1 2 2
Lorentzian
Single-sided exponential
et= H t
1 i
complex Lorentzian
Rectangular
t=
Triangular
t=
Convolution
f tgt
f g
product
Product
f tgt
1
f g
2
convolution
Complex conjugate
f t
f
= 2
sin =2
=2
sech
innite impulse sequence
sinc
sin2 =2
sinc2
=22
The convolution integral is dened as
f x x gx dx
f xgx
f x0 gx x0 dx0 :
(D.8)
Using the Fourier-transform relation between f t gt and f g and that between

f t and f shown in Table D.1, some useful relations can be obtained.
Correlation theorem :
1
f tgt dt
2
Autocorrelation theorem :
(D.9)
1
f t f t dt
2
f gei d,

f j2 ei d,
(D.10)
408
Appendix D
1
f tgt dt
2
Power theorem :
f g d,
1
j f t j dt
2
Parsevals theorem :
(D.11)

f j2 d:
(D.12)
Using (D.3) and (D.4), Parsevals theorem can be written as
jE t j dt
1
jE vj dv
2
jE j2 d:
(D.13)
INDEX
absorption, 23, 35, 224

band-to-band, 345, 369
absorption coefcient, 130, 242243, 250
intensity-dependent, 351
of direct-gap semiconductor, 346
of indirect-gap semiconductor, 346
of waveguide mode, 131
unsaturated, 351
absorption cross section, 238, 250
pump, 254
absorption saturation, 351
absorption transition, 236
absorptive external modulation, 344
absorptive modulation, 297, 305
all-optical, 340, 345, 350
external, 344
AC conductivity, 39
acoustic
frequency, 52
normal mode, 53
longitudinal, 53
quasi-longitudinal, 53
quasi-transverse, 53
transverse, 53
wave, 52
longitudinal, 53
standing, 53
transverse, 53
traveling, 52
wavelength, 52
acousto-optic amplitude modulation, 334
acousto-optic diffraction
Bragg, 334
order, 333
RamanNath, 334
acousto-optic effect, 52
acousto-optic modulation, 297, 320, 333
acousto-optic modulator
standing-wave, 338
traveling-wave, 336
acousto-optic polarization modulation, 333
active region
of photodiode, 372
all-optical absorptive modulation, 340,
345, 350
all-optical dispersive modulation, 340
all-optical modulation, 297, 320, 340, 350
all-optical phase modulation, 342

all-optical polarization modulation, 342
all-optical refractive modulation, 340
all-optical switching, 340
AM, 298, See amplitude modulation
Amperes law, 4
amplication, 23
of optical eld, 241
amplication coefcient, 130, 242
amplication factor
round-trip, 207, 274
amplied spontaneous emission, 269
amplitude modulation, 297298, 305, 309,
320, 326, 344
acousto-optic, 334
analog, 305306
digital, 305306
electro-optic, 326
magneto-optic, 329
amplitude modulator, 326
amplitude-shift keying, 298, 305, See ASK
binary, 305
analog amplitude modulation, 305306
analog frequency modulation, 300, 333
analog modulation, 297298, 306, 310,
314
analog polarization modulation, 302
analyzer, 326
angle
of diffraction, 335
of incidence, 94, 335
of reection, 94
of refraction, 94
angular frequency, 1
anisotropic crystal, 28
anisotropic medium, 24, 77
anisotropy, 24
anomalous dispersion, 36, 122
antiferrimagnetic material, 49
antiferromagnetic material, 49
antiguidance factor, 294
ASK, 298, 305, See amplitude-shift keying
asymmetric coupling, 144, 146, 160
asymmetric waveguide, 118
attenuation
of optical eld, 241
attenuation coefcient, 130, 242

attenuation factor
round-trip, 207
autocorrelation theorem, 407
axial vector, 6
bandgap, 365
of photoconductor, 369
of quantum well, 347
of semiconductor, 346, 365, 371
band-to-band absorption, 345, 369
band-to-band transition, 37
bandwidth, 122
3-dB, 311, 317, 390
gain, 240, 281, 283
modulation. See modulation bandwidth
of detection system, 363, 379
of LED, 311
of photodetector. See photodetector
bandwidth
of semiconductor laser, 317
beam waist, 88
BFSK. See binary frequency-shift keying
biaxial crystal, 29, 77
binary amplitude-shift keying, 305
binary frequency-shift keying, 300
binary phase-shift keying, 300
binary polarization-shift keying, 302
birefringence, 28
circular, 31, 52
electrically induced, 48
linear, 28, 52
magnetically induced, 52
optical-eld-induced, 342
birefringent crystal, 28
blackbody radiation, 235, 379
bleached condition, 257
bottleneck factor, 251, 256257, 260
boundary conditions, 7, 67
BPolSK. See binary polarization-shift
keying
BPSK. See binary phase-shift keying
Bragg angle, 335
Bragg diffraction, 334
down-shifted, 335
up-shifted, 335
410
Index
Brewster angle, 96
Brewster window, 97
bulk modulator, 297
carrier frequency, 3, 12, 19, 123,
300301, 305
carrier injection efciency, 309
carrier relaxation rate
differential, 315
nonlinear, 315
spontaneous, 315
total, 316
carrier wavevector, 19
causality, 22, 44
causality condition, 34
cavity
cold, 214
FabryProt, 204
nesse, 209
folded, 204
laser, 216
length, 206
linear, 204, 206, 212
non-FabryProt, 204
optical, 204
passive, 207
resonance condition, 209, 211
ring, 204, 206, 212
cavity decay rate, 214, 315
of FabryProt cavity, 219
cavity lifetime, 214
centrosymmetric, 46, 58
material, 32, 45, 51, 59
charge carrier
excess, 308, 368
free, 5, 38, 104
number, 363
photogenerated, 363, 368, 376377,
382
positive, 32, 38
charge density, 4
charge-coupled device, 363
chromatic dispersion, 122
circular birefringence, 31
magnetic, 31
magnetically induced, 52, 328
natural, 31
circular dichroism, 31
magnetic, 31
natural, 31
circular polarization, 16, 25
circularly polarized, 14
left, 16
right, 16
codirectional coupling, 154
coherence, 173
spatial, 173
temporal, 173
cold cavity, 214
collection efciency, 382
complex eld, 11
complex eld amplitude, 18
conduction current, 5, 38
conduction electron, 32
conduction susceptibility, 39
conductivity
AC, 39
dark, 368
DC, 39
of photoconductor, 368
of semiconductor, 368
optical, 38
confocal parameter, 88
conservation of charge, 5
conservation of power, 159
constructive interference, 171, 204
complete, 171
partial, 171
continuity equation, 5
contradirectional coupling, 156
convolution integral, 407
correlation theorem, 407
CottonMouton effect, 52
Coulombs law, 5
coupled-mode equations, 143
for multiple substructures, 146
for single structure, 143
for two-mode coupling, 147
codirectional, 154
contradirectional, 156
coupled-mode theory, 141
coupler
3-dB, 182, 327
asymmetric, 149
directional, 147, 149, 182
grating, 151
symmetric, 149
waveguide, 151
Y-junction, 182, 327
coupling coefcient, 143, 146, 161, 164
for multiple-structure coupling, 146
two modes, 148
for periodic structure, 150
for single-structure coupling, 143
qth order, 150
self, 149
coupling efciency, 155, 157, 190, 306
for codirectional coupling, 155
perfectly phase-matched, 161
phase-mismatched, 163
for contradirectional coupling, 157
coupling length, 156

critical angle, 97
critical uorescence power, 268
critical uorescence power density,
268
cross modulation, 340
cross section
absorption, 238, 250
emission, 238, 250
transition, 238
cross-phase modulation, 342
crystal
anisotropic, 28
biaxial, 29
birefringent, 28
cubic, 29
hexagonal, 30
monoclinic, 30
negative uniaxial, 28
orthorhombic, 30
positive uniaxial, 28
structural symmetry, 29
tetragonal, 30
triclinic, 30
trigonal, 30
uniaxal, 28
crystal axis, 29
crystal system, 46, 58
Curie temperature, 49
current
conduction, 5, 38
dark, 378
displacement, 5
induced, 5, 38
current density, 4
current modulation, 345
direct, 308
cutoff frequency
3-dB, 390
of guided mode, 117
of surface plasmon mode, 106
cutoff wavelength
of guided mode, 117
dark conductivity, 368
dark current, 378
DC conductivity, 39
degeneracy
in energy level, 224, 251
degeneracy factor, 224, 236
degenerate semiconductor, 365
n-type, 365
p-type, 365
depletion layer, 371
Index
destructive interference, 171
complete, 171
partial, 172
detectivity, 386
specic, 369, 386
detector
photoconductive, 368
photoemissive, 363
photon, 362
quantum, 362
square-law, 362
thermal, 362
diamagnetic material, 49
dichroism, 28
circular, 31, 52
linear, 28, 52
dielectric constant
principal, 28
tensor, 28
differential carrier relaxation rate, 315
differential gain parameter, 315
differential phase modulation, 302, 324,
326, 329
differential power conversion efciency,
291
diffraction modulation, 297, 299, 307, 333
diffraction order, 184185
reective, 188
transmissive, 185, 188
diffusion current, 372
diffusion region, 371
digital amplitude modulation, 305306
digital frequency modulation, 300, 333
digital modulation, 297298, 350
digital polarization modulation, 302
direct current-modulation, 308
direct modulation, 297, 299, 305, 308, 345
direct-gap semiconductor, 249, 346, 369
directional coupler, 147, 149, 182
asymmetric, 149
symmetric, 149
two-channel, 149
discrete energy level, 32
dispersion, 122
anomalous, 36, 122
chromatic, 122
frequency, 23
group-velocity, 124
coefcient, 124
effective, 126
negative, 124
positive, 124
intermode, 122, 127
intramode, 122, 127
material, 122, 126
modal, 115, 117, 122, 127

mode-order, 128
momentum, 23
normal, 36, 122
phase-velocity, 122
polarization, 117, 122, 127
polarization-mode, 128
waveguide, 122, 126
displacement current, 5
distributed feedback, 204
distributed feedback laser, 275
distributed loss, 218, 276
divergence angle, 86, 88
Doppler broadening, 231
Doppler effect, 230
double refraction, 98
double-slit interference, 176
drift current, 371372
Drude model, 38, 62
dynamic range of photodetector, 387388
effective group index, 126
effective group-velocity dispersion, 126
effective mass, 32, 38, 347
effective population inversion, 250
effective refractive index, 126
effective waveguide thickness
for guided TE mode, 114
for guided TM mode, 115
EH mode, 69
Einstein A coefcient, 226, 234
Einstein B coefcient, 234
elastic wave, 52
elasto-optic coefcient, 53
electric displacement, 4
electric eld, 4
complex, 12, 18, 169
electric permittivity, 4, 7
of free space, 4
tensor, 7
electric polarization, 4, 7
electric susceptibility, 7
tensor, 7
electric-dipole approximation, 58
electric-dipole interaction, 59
electric-dipole operator, 33
electro-absorption modulation, 345
electro-absorption modulator, 345, 349
electromagnetic eld, 4
electron
bound, 32
conduction, 32
free, 32
valence, 32
electron afnity, 363
411
electron mobility, 368

electro-optic amplitude modulation, 326
electro-optic coefcient
linear, 45
quadratic, 45
electro-optic effect, 44
rst order, 45
linear, 45
quadratic, 45
second order, 45
electro-optic Kerr coefcient, 45, 59
electro-optic Kerr effect, 45
electro-optic modulation, 297, 320
electro-optic modulator, 320
electro-optic phase modulation, 321, 328
electro-optic polarization modulation, 324
elliptic polarization, 14, 25
elliptically polarized, 14
ellipticity, 14
emission
spontaneous, 224
stimulated, 35, 224
emission cross section, 238, 250
homogeneously broadened medium,
239
inhomogeneously broadened medium,
239
pump, 254
energy band, 32
energy density
of optical radiation, 234
energy level, 249
ground, 255
lower, 224
upper, 224
vacuum, 363
envelope, 19, 123
Er:ber, 239
etalon, 191
evanescent radiation mode, 111
excess noise factor, 378
excess shot noise, 378
exciton, 345
free, 348
external modulation, 297
absorptive, 344
refractive, 319
external modulator, 306
external photoelectric effect, 362363
external quantum efciency
of LED, 309
of photodetector, 363, 382
of photodiode, 372
external reection, 96
extinction ratio, 307
extraordinary index, 28
412
Index
extraordinary wave, 81
extrinsic photoconductivity, 369
extrinsic photoconductor
n-type, 369
p-type, 369
extrinsic semiconductor, 369
frequency-shift keying, 298, See FSK

binary. See BFSK
quadrature. See QFSK
Fresnel equations, 95
FSK, 298, See frequency-shift keying
fundamental mode, 109, 118, 121
FabryProt cavity, 204, 216

stability criterion, 216
FabryProt etalon, 191
FabryProt interferometer, 178, 191, 204,
209
FabryProt laser, 275
falltime, 389
Faraday effect, 52, 329330
Faraday rotation, 329330
sense, 331
specic, 330
Faraday rotator, 330
Faradays law, 4
fast axis, 80
Fermi level, 363, 365
ferrimagnetic material, 49
ferromagnetic material, 49
eld amplitude
complex, 18
scalar, 18
vectorial, 12, 18
lling factor
of gain medium, 206
nesse, 194, 209
of lossless FabryProt interferometer,
193, 219
of lossy FabryProt cavity, 219
of optical cavity, 209
at interface, 190
uorescence, 226
uorescence lifetime, 227, 239, 250
FM, 298, See frequency modulation
folded cavity, 204
forward-coupling matrix, 154, 161
Fourier-transform relations, 13, 406
four-level system, 256, 259260
ideal, 256
FranzKeldysh effect, 345346
free spectral range, 194, 210
frequency chirping, 129, 309
negative, 129
positive, 129
frequency dispersion, 23
frequency modulation, 297298, 300, 309
analog, 300, 333
digital, 300, 333
frequency response
of LED, 311
of photodetector, 389
gain bandwidth, 240, 281, 283

gain coefcient, 130, 242, 244, 250, 259
small-signal, 263
unsaturated, 259, 263
gain compression, 315
gain factor
round-trip, 207, 274
gain lling factor, 206
gain parameter, 286
differential, 315
nonlinear, 315
of gain medium, 286
of laser mode, 286
saturated, 287
unsaturated, 287
gain saturation, 264, 281282, 351
Gausss law, 5
Gaussian beam, 87
complex radius of curvature, 90
confocal parameter, 88
divergence angle, 86, 88
radius of curvature, 88
Rayleigh range, 88
spot size, 86, 88
waist, 88
Gaussian lineshape, 231
Gaussian mode, 86, 211
basis for linear expansion, 89, 92
orthonormality relation, 87
propagation constant, 87
graded-index waveguide, 108
grating, 151, 183
at interface, 187
one-dimensional, 151
period, 153
sinusoidal, 152
square-function, 153
surface, 189
transmission, 183
transmissive diffraction, 183184
grating waveguide coupler, 151
ground level, 255
group index, 124
effective for waveguide mode, 126
group IV semiconductor, 367
group velocity, 122123
group-velocity dispersion, 124
coefcient, 124
negative, 124
positive, 124
guided mode, 109, 112
cutoff condition, 117, 121
cutoff frequency, 117
cutoff wavelength, 117
fundamental, 109, 118, 121
high-order, 109
number, 109, 118, 121
half-wave plate, 80
half-wave voltage, 322
harmonic eld, 11
HE mode, 69
Heaviside step function, 406
helicity
negative, 17
positive, 16
Helmholtz equation, 86
Hermite polynomial, 90
HermiteGaussian function, 89
HermiteGaussian mode, 90
high-order mode, 109
hole mobility, 368
homogeneous broadening, 225
homogeneous region, 371
homogeneous system, 225
homogeneously broadened laser, 280
hybrid mode, 69, 108
IM, 305, See intensity modulation
impulse response, 389
index contraction, 45
index ellipsoid, 82
index of refraction, 76
complex, 129
extraordinary, 28
for extraordinary wave, 82
for ordinary wave, 82
intensity-dependent, 342
ordinary, 28
principal, 28
indirect-gap semiconductor, 346, 369
induced transition, 224
induced transition rate, 235
inhomogeneous broadening, 230
inhomogeneous system, 230
inhomogeneously broadened laser,
282
intensity, 2
light, 1213
of normal mode, 71
of optical eld, 130
intensity modulation, 305
Index
intensity-dependent absorption coefcient,
351
intensity-dependent index change, 342
intensity-dependent index of refraction,
342
interaction length, 161, 164165, 334
interband transition, 37
interface mode, 92
interference
constructive, 171, 204
complete, 171
partial, 171
destructive, 171
complete, 171
partial, 172
double-slit, 176
of two elds, 169
interference lter, 199
interference fringe, 173, 175
bright, 175
dark, 175
interferometer
FabryProt, 178, 191
MachZehnder, 178, 180
Michelson, 178
intermode dispersion, 122, 127
internal gain
internal photoelectric effect, 362
internal quantum efciency
of LED, 311
internal reection, 96
total, 97, 99
intraband transition, 38
intracavity energy growth rate, 286
threshold, 286
intracavity photon density, 286
intracavity photon growth rate, 286
intramode dispersion, 122, 127
intrinsic photoconductivity, 369
intrinsic responsivity, 384
inversion symmetry, 45
irradiance, 12
isotropic material, 46
isotropic medium, 24, 75
Johnson noise, 375
junction photodiode, 363, 371
photoconductive mode, 373374
photovoltaic mode, 373
K factor, 316
Kerr effect
electro-optic, 45
magneto-optic, 52, 329
optical, 342
Kerr-lens mode locking, 342

KramersKronig relations, 44
LaguerreGaussian function, 90
LaguerreGaussian mode, 90
laser, 274
homogeneously broadened, 280
inhomogeneously broadened, 282
longitudinal mode, 280, 282
transverse mode, 280, 282
laser amplier, 265
laser cavity, 216
laser level
lower, 250, 253, 255
upper, 250, 253, 255
laser linewidth, 283
laser oscillation, 274
condition, 274
gain condition, 275
phase condition, 275
laser oscillator, 274
laser power, 285
laser system
four-level, 256, 259260
ideal four-level, 256
quasi-two-level, 254, 259
three-level, 255, 259260
two-level, 253
laser threshold, 276
laser transition, 218, 275
LED, 297, 308309, See light-emitting
diode
left-circular polarization, 16
LI characteristics
of LED, 309
lifetime
cavity, 214
uorescence, 227
of energy level, 226
photon, 214
saturation, 259
spontaneous radiative, 226, 236
lifetime broadening, 226
light intensity, 1213
lightcurrent characteristics
of LED, 309
light-emitting diode, 297, 309, See LED
line lter, 199
linear birefringence, 28, 48
linear cavity, 204, 206, 212
linear dichroism, 28, 48
413
linear medium, 24
linear optical property, 29, 46
linear polarization, 15, 25, 55
linear susceptibility, 22, 56
linearity of photodetector, 387
linearly polarized, 14
lineshape
Gaussian, 231
Lorentzian, 35, 226
Voigt, 234
lineshape function, 225, 230
linewidth, 194
Doppler broadening, 231
homogeneous broadening, 226
inhomogeneous broadening, 231
laser mode, 283
lifetime broadening, 227
longitudinal mode, 210
mixed broadening, 234
natural broadening, 227
linewidth enhancement factor, 294
longitudinal mode, 207, 210, 280, 282
frequency spacing, 210, 219
of optical resonator, 207, 210
longitudinal mode frequency, 210
longitudinal mode width, 210
longitudinal modulation, 297, 321, 323
longitudinal modulator, 323
longitudinal phase modulation, 323
longitudinal phase modulator, 323
Lorentz model, 33, 61
Lorentz reciprocity theorem, 26, 142
Lorentzian lineshape, 35, 226
loss parameter, 286
output-coupling, 288
total, 288
lower energy state, 32
lower laser level, 250, 253, 255
MachZehnder interferometer, 178, 180
waveguide, 327
magnetic eld, 4
magnetic induction, 4
magnetic material, 26
magnetic permeability, 4
of free space, 4
magnetic polarization, 4
magnetically disordered, 50
magnetically ordered, 49
magnetization, 4, 7
spontaneous, 50
magneto-optic amplitude modulation,
329
414
Index
magneto-optic effect, 49
rst-order, 50, 328
linear, 50, 328
quadratic, 50
second-order, 50, 328
magneto-optic Kerr effect, 52, 329
magneto-optic modulation, 297, 320, 328
magneto-optic modulator, 328
magneto-optic phase modulation, 329
material dispersion, 122, 126
Maxwells equations, 45, 11, 142
for plane wave normal mode, 73
for wave propagation, 68
in terms of mode eld components, 68
Maxwellian velocity distribution, 231
mean square noise, 377
mean square noise current, 385
mean square noise voltage, 385
metallic waveguide, 69
Michelson interferometer, 178
MichelsonMorley experiment, 178
mixed broadening, 233
modal dispersion, 115, 117, 122, 127
mode coupling, 141
asymmetric, 144, 146, 160
between two modes, 147
codirectional, 154
contradirectional, 156
multiple-structure, 145
order, 150
phase-matched, 149, 161
single-structure, 142
symmetric, 143, 160
mode eld
normalized, 71, 141
prole, 67
mode index, 67
mode locker, 352
mode number, 109
mode power, 131, 142
mode volume, 287
mode-order dispersion, 128
mode-pulling effect, 277
modulation
absorptive, 297, 305
all-optical, 340, 345, 350
external, 344
acousto-optic, 297, 320, 333
all-optical, 297, 320, 340, 342
absorptive, 340
dispersive, 340
refractive, 340
amplitude, 297298, 305, 309, 320,
326, 344
acousto-optic, 334
analog, 305
digital, 305306
electro-optic, 326
magneto-optic, 329
analog, 297298, 306, 310, 314
current, 345
diffraction, 297, 299, 307, 333
digital, 297298, 350
direct, 297, 299, 305, 308, 345
electro-absorption, 345
electro-optic, 297, 320
external, 297
absorptive, 344
refractive, 319
frequency, 297298, 300, 309
analog, 300, 333
digital, 300, 333
longitudinal, 297, 321, 323
magneto-optic, 297, 320, 328
optical, 297
phase, 297299, 306, 320, 344
analog, 299
differential, 302, 324, 326, 329
digital, 299
longitudinal, 323
magneto-optic, 329
transverse, 322
polarization, 297, 302, 306, 324
acousto-optic, 333
analog, 302
digital, 302
electro-optic, 324
magneto-optic, 329
refractive, 297, 300, 305, 307
all-optical, 340
electro-optic, 320
external, 319
magneto-optic, 329
small-signal, 316
spatial, 297, 299, 307
transverse, 297, 321, 323
modulation bandwidth
3-dB, 311, 317
of LED, 311
modulation current, 308, 311, 316
modulation depth, 306, 322
modulation frequency, 311, 316, 339
modulation index, 310, 316
modulation power spectrum
of LED, 311
modulation response, 308
of LED, 311
modulation scheme, 297298
modulation signal, 308
modulator
acousto-optic
standing-wave, 338
traveling-wave, 336
amplitude, 326
bulk, 297
electro-absorption, 345, 349
electro-optic, 320
external, 306
longitudinal, 323
magneto-optic, 328
phase, 322323
polarization, 326
transverse, 322
waveguide, 297
momentum dispersion, 23
momentum relaxation time, 38
monochromatic optical wave, 13, 20, 73
multimode waveguide, 118
natural broadening, 227
Nd:glass, 233
Nd:YAG, 233
Nd:YAG laser, 258
NEA, 367, See negative electron afnity
Nel temperature, 49
negative electron afnity, 367
NEP, 384, See noise equivalent power
noise, 225
Johnson, 375
mean square, 377, 385
Nyquist, 375
quantum, 375
shot, 375376
thermal, 375, 379
noise equivalent power, 384
noncentrosymmetric, 46, 58
nondegenerate semiconductor, 365
nonlinear carrier relaxation rate, 315
nonlinear gain parameter, 315
nonlinear optical amplier, 265
nonlinear optical property, 29
nonlinear optical susceptibility, 55
nonlinear polarization, 55
second-order, 55
third-order, 55
nonlinear susceptibility, 56
nth-order, 56
second-order, 59
third-order, 59
nonmagnetic material, 26
nonplanar optical structure, 66
dielectric, 70
nonplanar waveguide, 66, 108
dielectric, 69, 108
nonradiative relaxation rate, 226
Index
nonreciprocal, 5152, 331
nonreciprocal medium, 26
normal dispersion, 36, 122
normal incidence, 96
normal mode
acoustic, 53
basis for linear expansion, 72, 141
coupling, 141
degenerate, 67
extraordinary wave, 81
eld prole, 67
Gaussian, 86
guided, 109, 112
fundamental, 109, 121
high-order, 109
number, 109
hybrid, 69
index, 67
intensity distribution, 71
interface, 92
of planar interface, 98
of propagation, 66, 73
ordinary wave, 81
orthogonality relation, 71
orthonormality relation, 71
plane wave, 73, 78, 83
power, 71
principal polarization, 25
radiation, 98
substrate radiation, 111112
substratecover radiation, 111, 113
super, 145
surface plasmon, 104
transverse electric, TE, 69
transverse electromagnetic, TEM, 69
transverse magnetic, TM, 69
waveguide, 107
normal mode eld pattern, 67
normal state, 32
normalized frequency and waveguide
thickness, 112
normalized guide index, 112
normalized mode eld, 71, 141
normalized transmittance, 193, 208
of FabryProt interferometer, 193
of optical cavity, 208
Nyquist noise, 375
on-off keying, 305
OOK, 305, See on-off keying
optical activity, 30
natural, 30
optical amplication, 129, 225, 244, 265
optical amplier, 265
optical anisotropy, 24, 29
optical attenuation, 129, 225, 244

optical axis, 28, 77, 81
optical carrier, 3, 12, 19, 123
optical cavity, 204, 209
optical conductivity, 38
optical connement, 108
optical discriminator, 352
optical energy, 8
density, 9
optical feedback, 204, 274
optical eld
angular frequency, 20
complex, 12, 18, 169
frequency, 18
harmonic, 12
magnitude, 18
phase, 18, 169
polarization, 18
real amplitude, 19
scalar complex amplitude, 18
vectorial complex amplitude, 18
wavevector, 18
optical-eld-induced birefringence, 342
optical frequency, 1
optical gain, 23, 129
optical gain coefcient, 259
optical grating, 183, 307
optical interference, 169
optical interferometer, 178
optical Kerr effect, 342
optical loss, 23, 129
distributed, 218
optical medium
anisotropic, 24
isotropic, 24
linear, 24
lossless, 26
lossy, 26
nonmagnetic, 26
nonreciprocal, 26
optically active, 26
reciprocal, 26
optical modulation, 297
optical noise, 225
optical nonlinearity, 55
optical path length, 176
round-trip, 206
optical power, 8, 131, 363
optical property
linear, 29, 46
nonlinear, 29
optical pumping, 254
optical resonance, 204
optical resonator, 204
optical soliton, 342
optical spectrum analyzer, 195
optical switching, 298
415
optical thin lm, 196

optical transition, 224
optical wave
monochromatic, 13, 20
plane, 13, 19
optical wavelength
in free space, 1
in homogeneous medium, 76
order of coupling, 150
order of diffraction, 184
ordinary index, 28
ordinary wave, 81
orientation, 14
orthogonal polarizations, 17
orthogonality relation
of normal mode, 71
orthonormality relation
of Gaussian mode, 87
of normal mode, 71
output-coupling loss parameter, 288
output-coupling rate, 288
overlap coefcient, 146
overlap factor, 206, 327
p polarized, polarized, 95
p wave, wave, 95
parallel polarization, 95
paramagnetic material, 49
paraxial approximation, 87
Parsevals theorem, 408
passive cavity, 207
perfect phase matching, 153, 161162,
165, 334
periodic index modulation, 151
periodic perturbation, 150
periodic structural corrugation, 151
permittivity, 7, 39, 122
acousto-optically induced change, 333
electric eld-dependent, 44
frequency domain, 13, 23
magnetic eld-dependent, 44
magnetization-dependent, 50
momentum space, 13, 23
of gain medium, 218
optical, 22
optical eld-dependent, 341
photoelastic, 54
principal, 27
real space, 7, 23
rotation eld-dependent, 54
scalar, 66
strain eld-dependent, 54
tensor, 7
time domain, 7, 23
total, 39
perpendicular polarization, 95
perturbing polarization, 142
416
Index
phase matched, 156, 158, 160

phase matching, 160, 165, 183, 187,
334, 336
perfect, 153, 161162, 165, 334
phase mismatch, 149, 151, 160, 163, 165,
185, 306
phase modulation, 297299, 306, 320, 344
all-optical, 342
analog, 299
cross, 342
depth, 322
differential, 302, 324, 326, 329
digital, 299
longitudinal, 323
magneto-optic, 329
self, 342
transverse, 322
phase relaxation, 250
phase relaxation rate, 225
phase retardation, 324
phase velocity, 122
phase-matched coupling, 149, 161
phase-matching condition, 160, 186, 188,
190191, 307, 333
for down-shifted Bragg diffraction,
335
for up-shifted Bragg diffraction,
335
phase-mismatched coupling, 163
phase-shift keying, 298, See PSK
binary. See BPSK
quadrature. See QPSK
phase-velocity dispersion, 122
photocathode, 362363
photoconductive detector, 368
photoconductive mode, 373374
photoconductivity, 368
extrinsic, 369
intrinsic, 369
photoconductor, 363
extrinsic, 369
intrinsic, 369
photocurrent, 363, 372
reverse, 372
photodetector, 362
photodetector bandwidth, 385
intrinsic, 389
RC, 389
photodiode, 371
junction, 363, 371
vacuum, 362
photoelastic coefcient, 53
photoelastic effect, 53
dynamic, 53
photoelastic permittivity tensor, 54
photoelectric effect, 362

external, 362363
internal, 362
photoelectron, 363
photoemission, 363
from degenerate semiconductor, 365
from metal, 365
from NEA photocathode, 367
from nondegenerate semiconductor, 365
photoemissive detector, 363
photoemissive device, 362
photomultiplier tube, 362
photon, 1
energy, 1
ux, 2
ux density, 2
momentum, 1
number, 363
speed, 1
photon detector, 362
photon lifetime, 214
of FabryProt cavity, 219
photothermal effect, 362
phototransistor, 363
photovoltaic device, 363
photovoltaic mode, 373
PI characteristics
of LED, 310
planar dielectric waveguide, 108
planar interface, 66, 92
planar optical structure, 66
dielectric, 70
planar waveguide, 66, 108
dielectric, 69
metallic, 69
Plancks formula, 235
plane of incidence, 94
plane polarized, 14
plane wave, 13, 19, 73
normal mode, 73, 78, 83
basis for linear expansion, 74, 76, 78,
84
plasma frequency, 40, 104
surface, 106
PM, 298, See phase modulation
pn homojunction, 371
Pockels coefcient, 45, 59
Pockels effect, 45
point group, 46, 58
Poisson probability distribution, 377
polar semiconductor, 29
polar vector, 6
polarization, 4, 7
circular, 16, 25
elliptic, 14, 25
ellipticity, 14
left-circular, 16
linear, 15, 25, 55

nonlinear, 55
nth-order, 56
second-order, 55
third-order, 55
of optical eld, 13
orientation, 14
orthogonality relation, 17, 74
principal state, 25
right-circular, 16
state, 16
unit vector, 17
polarization dispersion, 117, 122, 127
polarization modulation, 297, 302,
306, 324
acousto-optic, 333
all-optical, 342
analog, 302
digital, 302
electro-optic, 324
magneto-optic, 329
polarization modulator, 326
polarization-mode dispersion, 128
polarization-shift keying, 302, See PolSK
binary. See BPolSK
polarizer, 326
polarizing beam splitter, 85
PolSK, 302, See polarization-shift keying
population
density, 32, 249, 253
distribution, 33
population decay rate, 33
population difference, 34, 244
population inversion, 33, 242, 249, 251,
255
condition, 252
effective, 250
effective condition, 252
population relaxation, 250
power, 2
of normal mode, 71, 142
power conversion efciency, 291
differential, 291
power density, 10
power gain, 207, 265
small-signal, 265
unsaturated, 265
power theorem, 408
powercurrent characteristics
of LED, 309
Poynting vector, 9, 71, 73
complex, 12
time-averaged, 12
principal axis, 27, 29, 78
principal dielectric axis, 27, 46, 48
principal dielectric constant, 28
Index
principal dielectric susceptibility, 28
principal index of refraction, 28, 46, 48
principal normal mode, 25, 48, 78
principal permittivity, 27, 78
principal polarization state, 25
principle of detailed balance, 236
probability density function, 230
propagation constant, 19, 67, 75
complex, 129, 131
of extraordinary wave, 82
of Gaussian normal mode, 87
of guided mode, 109, 113
of normal mode, 67
of ordinary wave, 82
of principal normal mode, 28, 78
of substrate radiation mode, 111, 113
of substratecover radiation mode, 111,
113
of waveguide mode, 126, 131
propagation direction, 19, 76
PSK, 298299, See phase-shift keying
pump power
threshold, 287
transparency, 287
pumping, 249, 254
pumping rate, 250, 253
transparency, 257
pumping ratio, 287
pumping requirement, 254256
minimum, 255256, 260
pumping technique, 249
pushpull operation, 327
Q switch, 352
QFSK. See quadrature frequency-shift
keying
QPSK. See quadrature phase-shift keying
quadrature frequency-shift keying, 300
quadrature phase-shift keying, 300
quality factor, 214
of cold cavity, 214
quantum detector, 362
quantum efciency, 368, 371, 377
external, 309, 363, 382
internal, 382
pump, 262
quantum noise, 375
quantum regime, 380
quantum-conned Stark effect, 345
quantum-well structure, 347
quarter-wave plate, 80
quasi-two-level system, 254, 259
radiation mode, 98
of planar interface, 98
one-sided, 99
substrate, 111
substratecover, 111
two-sided, 101
radiative relaxation rate, 226
RamanNath diffraction, 334
rate equations, 250
for semiconductor laser, 313
Rayleigh range, 88
reality condition, 24, 56
reciprocal, 5152
reciprocal medium, 26
rectangular function, 406
reectance, 9596
reection, 93
external, 96
internal, 96
partial, 101
total, 97, 99
reection coefcient, 95
of contradirectional coupling, 158
reection-type polarizer, 97
reectivity, 95
of contradirectional coupling, 158
refraction, 93
double, 98
refractive external modulation, 319
refractive index, 19, 76, See index of
refraction
effective for waveguide mode, 112,
126
refractive modulation, 297, 300, 305,
307
all-optical, 340
electro-optic, 320
external, 319
magneto-optic, 329
relative impermeability tensor, 45, 53
relaxation oscillation, 316
relaxation rate, 32
carrier. See carrier relaxation rate
nonradiative, 226
optical polarization, 33
phase, 225, 250
population, 33, 250
radiative, 226
susceptibility, 33
total, 226
relaxation resonance frequency, 316
resonance condition
of optical cavity, 209, 211
for Gaussian mode, 212
for waveguide mode, 212
transverse
for waveguide mode, 109
resonance frequency, 32, 204
417
of cold FabryProt cavity, 218

of optical cavity, 204, 211
transition, 33, 224225
resonant interaction, 32
resonant laser transition, 218, 275
resonant optical cavity, 204
resonant optical susceptibility, 32, 243
resonant susceptibility tensor, 33
resonant transition, 224
response speed of a photodetector, 389
responsivity, 380, 383
intrinsic, 384
spectral, 367, 371, 384
reverse current, 372
reverse-coupling matrix, 157, 161
right-circular polarization, 16
ring cavity, 204, 206, 212
ring laser, 275
risetime, 389
rotating-wave approximation, 34
rotation tensor, 53
rotation-optic coefcient, 53
rotatory power, 330
round-trip amplication factor, 207, 274
round-trip gain factor, 207, 274
mode-dependent, 275
of cold cavity, 214
of cold FabryProt cavity, 218
round-trip optical path length, 206, 212
round-trip phase shift, 192193, 207, 209,
274
mode-dependent, 211, 217, 275
of Gaussian mode, 218
round-trip time, 205, 210
ruby laser, 258
s polarized, polarized, 95
s wave, wave, 95
saturable absorber, 351
saturated gain parameter, 287
saturation intensity, 259, 351
saturation lifetime, 259260
saturation output power, 288
saturation photon density, 287
saturation power, 265
scalar eld amplitude, 18
SchawlowTownes limit, 284
SchawlowTownes relation, 284
self defocusing, 342
self focusing, 342
self modulation, 340
self-phase modulation, 342
semiconductor
degenerate, 365
direct-gap, 249, 346, 369
extrinsic, 369
group IV, 367
418
Index
semiconductor (cont.)
IIIV, 29, 367
indirect-gap, 346, 369
nondegenerate, 365
polar, 29
semiconductor laser, 308, 313
semiconductor photodiode, 371
shot noise, 375376
excess, 378
of photodetector with internal gain, 378
shot-noise limited, 380
signal current, 372
signal-to-noise ratio, 376, 379
current, 376
power, 376
voltage, 376
single-mode waveguide, 118, 121
slab waveguide, 111
symmetric, 119
slope efciency, 291
slow axis, 80
slowly varying amplitude, 56
small-signal gain coefcient, 263
small-signal modulation, 316
small-signal power gain, 265
Snells law, 95
SNR, 376, See signal-to-noise ratio
space inversion, 6, 51, 59
transformation, 6, 46
spatial beam walk-off, 84
spatial light lter, 352
spatial modulation, 297, 299, 307
spatially coherent, 173
spatially incoherent, 173
specic detectivity, 369, 386
of photoconductive detector, 369
specic Faraday rotation, 330
spectral broadening, 225, 342
dephasing, 228
Doppler, 231
homogeneous, 225
inhomogeneous, 230
lifetime, 226
mixed, 233
natural, 227
spectral detectivity, 382
spectral hole burning, 282
spectral intensity distribution, 234
spectral lineshape, 225
spectral response, 367, 382
spectral responsivity, 371, 382, 384
of photocathode, 367
of photodiode, 371
speed of light
in free space, 1, 75
in homogeneous medium, 76
spontaneous carrier recombination
lifetime, 309
spontaneous carrier recombination rate,
309
spontaneous carrier relaxation rate, 315
spontaneous emission, 224, 267
amplied, 269
spontaneous emission factor, 284
spontaneous emission power, 268
spontaneous emission power density, 268
spontaneous emission rate, 234
spontaneous emission spectrum, 234
spontaneous magnetization, 50
spontaneous radiative lifetime, 226, 236
spontaneous transition, 236
spontaneous transition rate, 235
spot size, 86
square-law detector, 362
square-pulse response, 389
stability criterion
for FabryProt cavity, 216
standing wave, 182
step-index waveguide, 108
planar, 111
stimulated emission, 35, 224
stimulated-emission transition, 236
strain
shear, 53
tensile, 53
strain tensor, 53
strain-optic coefcient, 53
substrate radiation mode, 111112
substratecover radiation mode, 111, 113
super mode, 145
super structure, 145
surface grating, 189
surface plasma frequency, 106
surface plasmon mode, 104
dispersion curve, 106
susceptibility, 7, 122
conduction, 39
electric eld-dependent, 44
frequency domain, 7, 13, 23
linear, 22, 56
magnetic eld-dependent, 44
magnetization-dependent, 50
momentum space, 7, 13, 23
nonlinear, 56
nth-order, 56
second-order, 59
third-order, 59
nonlinear optical, 55
optical, 22
principal, 28
real space, 23
resonant, 32, 243
rotation eld-dependent, 54
strain eld-dependent, 54
tensor, 7, 33
time domain, 23
switching
all-optical, 340
optical, 298
symmetric coupling, 143, 160
symmetric waveguide, 112, 118119, 121
slab, 119
TE mode, 69, 76, 114, 119, 121
TE polarization, 95
TEM mode, 69, 87, 108
TEM wave, 75
temporal beat, 173
temporally coherent, 173
temporally incoherent, 173
thermal detector, 362
thermal equilibrium, 32, 235, 249
thermal noise, 375, 379
thermal regime, 380
thermal-noise limited, 380
three-level system, 255, 259260
threshold carrier density, 314
threshold condition, 286
for semiconductor laser, 314
threshold current density, 314
threshold gain coefcient, 276, 280
threshold gain parameter, 286, 314
threshold injection current, 314
threshold intracavity energy growth rate,
286
threshold photon energy, 363, 368
of extrinsic photoconductivity, 369
of intrinsic photoconductivity, 369
of photoemission
from degenerate semiconductor, 365
from metal, 365
from NEA photocathode, 367
from nondegenerate semiconductor,
365
of semiconductor photodiode, 371
threshold pump power, 276, 287
threshold pumping level, 276
threshold wavelength, 363, 368
Ti:sapphire, 239
time reversal, 6, 51
transformation, 6, 50
TM mode, 69, 76, 108, 115, 119, 121
TM polarization, 95
Index
total carrier relaxation rate, 316
total internal reection, 97, 99
total relaxation rate, 226
transition
absorption, 236
band-to-band, 37
energy, 236
induced, 224
interband, 37
intraband, 38
laser, 218, 275
optical, 224
resonance frequency, 33, 224
resonant, 224
spontaneous, 236
stimulated-emission, 236
transition cross section, 238
transition rate, 234
absorption, 234
induced, 235
induced downward, 234
spontaneous, 235
spontaneous emission, 234
stimulated emission, 234
upward transition, 234
transition resonance frequency, 33, 224
225
transmission coefcient, 95
transmission efciency, 382
transmission grating, 183
transmission line, 69
transmissive diffraction grating, 183184
transmissivity, 95
transmittance, 9596
normalized, 193, 208
transparency, 257, 260
transparency population density, 257
transparency pump power, 276, 287
transparency pumping rate, 257
transverse electric mode. See TE mode
transverse electromagnetic mode. See
TEM mode
transverse magnetic mode. See TM mode

transverse mode, 211, 280, 282
of optical resonator, 211
transverse modulation, 297, 321, 323
transverse modulator, 322323
transverse phase modulation, 322
transverse phase modulator, 322
triangular function, 406
two-level system, 253
quasi, 254, 259
uniaxial crystal, 28, 77, 81
negative, 28
positive, 28
uniform perturbation, 149
unit polarization vector, 17
unsaturated absorption coefcient, 351
unsaturated gain coefcient, 259260, 263
unsaturated gain parameter, 286
unsaturated power gain, 265
upper energy state, 32
upper laser level, 250, 253, 255
population, 257, 267
transparency population density, 257
V number, 112
vacuum energy level, 363
vacuum photodiode, 362
valence electron, 32
vectorial eld amplitude, 18
Verdet constant, 330
Voigt lineshape, 234
walk-off angle, 84
wave equation, 1011
for plane wave normal mode, 74
wavefront, 19, 73
waveguide
asymmetric, 118
asymmetry factor, 112
cladding, 108
core, 108
dielectric, 108
419
evanescent radiation mode, 111

graded-index, 108
guided mode, 109, 112
cutoff condition, 117, 121
fundamental, 109, 118, 121
high-order, 109
number, 109, 118, 121
TE, 114, 119, 121
TM, 115, 119, 121
metallic, 69
mode, 107
effective refractive index, 112
multimode, 118
nonplanar, 66, 108
dielectric, 69
planar, 66, 108
cover, 108
dielectric, 69
lm, 108
metallic, 69
step-index, 111
substrate, 108
single-mode, 118, 121
slab, 111
symmetric, 119
step-index, 108
substrate radiation mode, 111112
substratecover radiation mode, 111,
113
symmetric, 112, 118119, 121
V number, 112
weakly guiding, 117
waveguide dispersion, 122, 126
waveguide modulator, 297
wavelength
acoustic, 52
optical, 1, 76
wavenumber, 19, 76, 130
wavevector, 1, 73
weakly guiding waveguide, 117
work function, 363, 367

Principles of Photonics

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

Principles of Photonics

Uploaded by

Copyright:

Available Formats

Principles of Photonics

ENDORSEMENTS FOR LIU, PRINCIPLES OF PHOTONICS

A pedagogical tour-de-force. Professor Liu covers the principles of photonics with

University Printing House, Cambridge CB2 8BS, United Kingdom

To Vida and Janelle

Basic Concepts of Optical Fields

Optical Properties of Materials

Optical Wave Propagation

7 Optical Absorption and Emission

11.1 Physical Principles of Photodetection

Symbols and Notations

Cambridge Books Online

PARTIAL LIST OF SYMBOLS

round-trip intracavity eld amplication factor

asymmetry factors for TE and TM modes

confocal parameter of Gaussian beam

normalized guide index

linewidth enhancement factor

real magnetic induction in the time domain

complex magnetic induction

speed of light in free space

overlap coefcient between modes v and

quadratic magneto-optic coefcient

thickness or distance; d g , dQW

beam spot size diameter, d 2w, d 0 2w0

effective waveguide thicknesses for TE and TM modes

Partial List of Symbols

group-velocity dispersion; D D=c

real electric displacement in the time domain

complex electric displacement; De , Do , D , D

slowly varying amplitudes of D and D; De , D0

unit vector of electric eld polarization; ê e , ê o , ê , ê

energies of levels j1i and j2i

conduction-band and valence-band edges

threshold photon energy

real electric eld in the time domain

static or low-frequency electric eld

electric elds seen by electrons and holes

complex electric eld

complex electric eld of mode v

slowly varying amplitudes of E and E; E e , E o , E , E

complex electric eld prole of mode v

normalized electric mode eld distribution, E v Av E^ v

acoustic or modulation frequency, f =2

3-dB modulation bandwidth or cutoff frequency

Kerr focal length

linear magneto-optic coefcient, Faraday coefcient

Partial List of Symbols

relaxation resonance frequency, f r r =2

excess noise factor

nesse of interferometer or optical cavity

forward-coupling matrix for codirectional coupling

gain coefcient, amplication coefcient

unsaturated gain coefcient

threshold gain coefcient

unsaturated gain parameter

differential gain parameter

nonlinear gain parameter

threshold gain parameter

cavity round-trip eld gain; Gc , Gmn , Gcmn

photodetector current gain

optical amplier power gain, G0 for unsaturated gain

Plancks constant, h=2

transverse oscillation parameters of mode eld